targets/gba: implement interrupt handler

Make it clearer where the stack is located.

.circleci/config.yml

@@ -18,7 +18,6 @@ commands:
             wget -O - https://apt.llvm.org/llvm-snapshot.gpg.key|sudo apt-key add -
             sudo apt-get update
             sudo apt-get install \
-                python3 \
                 llvm<<parameters.llvm>>-dev \
                 clang<<parameters.llvm>> \
                 libclang<<parameters.llvm>>-dev \
@@ -40,78 +39,42 @@ commands:
             sudo tar -C /usr/local -xf node-v10.15.1-linux-x64.tar.xz
             sudo ln -s /usr/local/node-v10.15.1-linux-x64/bin/node /usr/bin/node
             rm node-v10.15.1-linux-x64.tar.xz
-  dep:
-    steps:
-      - run:
-          name: "Install Go dependencies"
-          command: |
-            curl https://raw.githubusercontent.com/golang/dep/master/install.sh | sh
-            dep ensure --vendor-only
   llvm-source-linux:
     steps:
       - restore_cache:
           keys:
-            - llvm-source-8-v2
+            - llvm-source-9-v0
       - run:
           name: "Fetch LLVM source"
           command: make llvm-source
       - save_cache:
-          key: llvm-source-8-v2
+          key: llvm-source-9-v0
           paths:
-            - llvm
-  smoketest:
-    steps:
-      - smoketest-no-avr
-      - run: tinygo build -size short -o test.elf -target=arduino             examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=digispark           examples/blinky1
-  smoketest-no-avr:
-    steps:
-      - run: tinygo build -size short -o test.elf -target=pca10040            examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=pca10040            examples/blinky2
-      - run: tinygo build             -o blinky2                              examples/blinky2 # TODO: re-enable -size flag with MachO support
-      - run: tinygo build -size short -o test.elf -target=pca10040            examples/test
-      - run: tinygo build -size short -o test.elf -target=microbit            examples/echo
-      - run: tinygo build -size short -o test.elf -target=nrf52840-mdk        examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=pca10031            examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=bluepill            examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=reelboard           examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=reelboard           examples/blinky2
-      - run: tinygo build -size short -o test.elf -target=pca10056            examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=pca10056            examples/blinky2
-      - run: tinygo build -size short -o test.elf -target=itsybitsy-m0        examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=circuitplay-express examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=stm32f4disco        examples/blinky1
-      - run: tinygo build -size short -o test.elf -target=stm32f4disco        examples/blinky2
-      - run: tinygo build -size short -o test.elf -target=circuitplay-express examples/i2s
-      - run: tinygo build             -o wasm.wasm -target=wasm               examples/wasm/export
-      - run: tinygo build             -o wasm.wasm -target=wasm               examples/wasm/main
+            - llvm-project
   test-linux:
-    parameters:
-      llvm:
-        type: string
     steps:
       - checkout
       - submodules
       - apt-dependencies:
-          llvm: <<parameters.llvm>>
+          llvm: "-9"
       - install-node
       - restore_cache:
           keys:
-            - go-cache-{{ checksum "Gopkg.lock" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
-            - go-cache-{{ checksum "Gopkg.lock" }}
+            - go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
+            - go-cache-v2-{{ checksum "go.mod" }}
       - llvm-source-linux
-      - dep
       - run: go install .
-      - run: go test -v
+      - run: go test -v ./cgo ./compileopts ./interp ./transform .
       - run: make gen-device -j4
-      - smoketest
+      - run: make smoketest
       - save_cache:
-          key: go-cache-{{ checksum "Gopkg.lock" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
+          key: go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
           paths:
             - ~/.cache/go-build
             - ~/.cache/tinygo
+            - /go/pkg/mod
       - run: make fmt-check
-  build-linux:
+  assert-test-linux:
     steps:
       - checkout
       - submodules
@@ -119,7 +82,6 @@ commands:
           name: "Install apt dependencies"
           command: |
             sudo apt-get install \
-                python3 \
                 gcc-arm-linux-gnueabihf \
                 binutils-arm-none-eabi \
                 libc6-dev-armel-cross \
@@ -132,12 +94,70 @@ commands:
       - install-node
       - restore_cache:
           keys:
-            - go-cache-{{ checksum "Gopkg.lock" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
-            - go-cache-{{ checksum "Gopkg.lock" }}
+            - go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
+            - go-cache-v2-{{ checksum "go.mod" }}
       - llvm-source-linux
       - restore_cache:
           keys:
-            - llvm-build-8-linux-v4
+            - llvm-build-9-linux-v0-assert
+      - run:
+          name: "Build LLVM"
+          command: |
+            if [ ! -f llvm-build/lib/liblldELF.a ]
+            then
+              # install dependencies
+              sudo apt-get install cmake clang ninja-build
+              # make build faster
+              export CC=clang
+              export CXX=clang++
+              # hack ninja to use less jobs
+              echo -e '#!/bin/sh\n/usr/bin/ninja -j3 "$@"' > /go/bin/ninja
+              chmod +x /go/bin/ninja
+              # build!
+              make ASSERT=1 llvm-build
+            fi
+      - save_cache:
+          key: llvm-build-9-linux-v0-assert
+          paths:
+            llvm-build
+      - run: make ASSERT=1
+      - run:
+          name: "Test TinyGo"
+          command: make ASSERT=1 test
+      - save_cache:
+          key: go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
+          paths:
+            - ~/.cache/go-build
+            - ~/.cache/tinygo
+            - /go/pkg/mod
+      - run: make gen-device -j4
+      - run: make smoketest TINYGO=build/tinygo
+  build-linux:
+    steps:
+      - checkout
+      - submodules
+      - run:
+          name: "Install apt dependencies"
+          command: |
+            sudo apt-get install \
+                gcc-arm-linux-gnueabihf \
+                binutils-arm-none-eabi \
+                libc6-dev-armel-cross \
+                gcc-aarch64-linux-gnu \
+                libc6-dev-arm64-cross \
+                qemu-system-arm \
+                qemu-user \
+                gcc-avr \
+                avr-libc
+      - install-node
+      - restore_cache:
+          keys:
+            - go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
+            - go-cache-v2-{{ checksum "go.mod" }}
+      - llvm-source-linux
+      - restore_cache:
+          keys:
+            - llvm-build-9-linux-v0
       - run:
           name: "Build LLVM"
           command: |
@@ -155,16 +175,9 @@ commands:
               make llvm-build
             fi
       - save_cache:
-          key: llvm-build-8-linux-v4
+          key: llvm-build-9-linux-v0
           paths:
             llvm-build
-      - run:
-          name: "Create LLVM symlinks"
-          command: |
-            ln -s $PWD/llvm-build/bin/clang-8 /go/bin/clang-8
-            ln -s $PWD/llvm-build/bin/ld.lld  /go/bin/ld.lld-8
-            ln -s $PWD/llvm-build/bin/wasm-ld /go/bin/wasm-ld-8
-      - dep
       - run:
           name: "Test TinyGo"
           command: make test
@@ -176,10 +189,11 @@ commands:
       - store_artifacts:
           path: /tmp/tinygo.linux-amd64.tar.gz
       - save_cache:
-          key: go-cache-{{ checksum "Gopkg.lock" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
+          key: go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
           paths:
             - ~/.cache/go-build
             - ~/.cache/tinygo
+            - /go/pkg/mod
       - run:
           name: "Extract release tarball"
           command: |
@@ -187,7 +201,7 @@ commands:
             tar -C ~/lib -xf /tmp/tinygo.linux-amd64.tar.gz
             ln -s ~/lib/tinygo/bin/tinygo /go/bin/tinygo
             tinygo version
-      - smoketest
+      - run: make smoketest
   build-macos:
     steps:
       - checkout
@@ -195,20 +209,27 @@ commands:
       - run:
           name: "Install dependencies"
           command: |
-            HOMEBREW_NO_AUTO_UPDATE=1 brew install go dep qemu
+            curl https://dl.google.com/go/go1.13.darwin-amd64.tar.gz -o go1.13.darwin-amd64.tar.gz
+            sudo tar -C /usr/local -xzf go1.13.darwin-amd64.tar.gz
+            ln -s /usr/local/go/bin/go /usr/local/bin/go
+            HOMEBREW_NO_AUTO_UPDATE=1 brew install qemu
       - restore_cache:
           keys:
-            - llvm-source-8-macos-v2
+            - go-cache-macos-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
+            - go-cache-macos-v2-{{ checksum "go.mod" }}
+      - restore_cache:
+          keys:
+            - llvm-source-9-macos-v0
       - run:
           name: "Fetch LLVM source"
           command: make llvm-source
       - save_cache:
-          key: llvm-source-8-macos-v2
+          key: llvm-source-9-macos-v0
           paths:
-            - llvm
+            - llvm-project
       - restore_cache:
           keys:
-            - llvm-build-8-macos-v3
+            - llvm-build-9-macos-v0
       - run:
           name: "Build LLVM"
           command: |
@@ -220,16 +241,9 @@ commands:
               make llvm-build
             fi
       - save_cache:
-          key: llvm-build-8-macos-v3
+          key: llvm-build-9-macos-v0
           paths:
             llvm-build
-      - run:
-          name: "Create LLVM symlinks"
-          command: |
-            ln -s $PWD/llvm-build/bin/clang-8 /usr/local/bin/clang-8
-      - run:
-          name: "Install Go dependencies"
-          command: dep ensure --vendor-only
       - run:
           name: "Test TinyGo"
           command: make test
@@ -247,34 +261,49 @@ commands:
             tar -C /usr/local/opt -xf /tmp/tinygo.darwin-amd64.tar.gz
             ln -s /usr/local/opt/tinygo/bin/tinygo /usr/local/bin/tinygo
             tinygo version
-      - smoketest-no-avr
+      - run:
+          name: "Download SiFive GNU toolchain"
+          command: |
+            curl -O https://static.dev.sifive.com/dev-tools/riscv64-unknown-elf-gcc-8.2.0-2019.05.3-x86_64-apple-darwin.tar.gz
+            sudo tar -C /usr/local --strip-components=1 -xf riscv64-unknown-elf-gcc-8.2.0-2019.05.3-x86_64-apple-darwin.tar.gz
+      - run: make smoketest AVR=0
+      - save_cache:
+          key: go-cache-macos-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
+          paths:
+            - ~/.cache/go-build
+            - ~/.cache/tinygo
+            - /go/pkg/mod
 
 
 jobs:
-  test-llvm8-go111:
+  test-llvm9-go111:
     docker:
-      - image: circleci/golang:1.11
-    working_directory: /go/src/github.com/tinygo-org/tinygo
+      - image: circleci/golang:1.11-stretch
     steps:
-      - test-linux:
-          llvm: "-8"
-  test-llvm8-go112:
+      - test-linux
+  test-llvm9-go112:
     docker:
-      - image: circleci/golang:1.12
-    working_directory: /go/src/github.com/tinygo-org/tinygo
+      - image: circleci/golang:1.12-stretch
     steps:
-      - test-linux:
-          llvm: "-8"
+      - test-linux
+  test-llvm9-go113:
+    docker:
+      - image: circleci/golang:1.13-stretch
+    steps:
+      - test-linux
+  assert-test-linux:
+    docker:
+      - image: circleci/golang:1.13-stretch
+    steps:
+      - assert-test-linux
   build-linux:
     docker:
-      - image: circleci/golang:1.12
-    working_directory: /go/src/github.com/tinygo-org/tinygo
+      - image: circleci/golang:1.13-stretch
     steps:
       - build-linux
   build-macos:
     macos:
       xcode: "10.1.0"
-    working_directory: ~/go/src/github.com/tinygo-org/tinygo
     steps:
       - build-macos
 
@@ -284,7 +313,9 @@ jobs:
 workflows:
   test-all:
     jobs:
-      - test-llvm8-go111
-      - test-llvm8-go112
+      - test-llvm9-go111
+      - test-llvm9-go112
+      - test-llvm9-go113
       - build-linux
       - build-macos
+      - assert-test-linux

.gitignore

@@ -5,10 +5,12 @@ src/device/avr/*.ld
 src/device/avr/*.s
 src/device/nrf/*.go
 src/device/nrf/*.s
-src/device/stm32/*.go
-src/device/stm32/*.s
 src/device/sam/*.go
 src/device/sam/*.s
+src/device/sifive/*.go
+src/device/sifive/*.s
+src/device/stm32/*.go
+src/device/stm32/*.s
 vendor
-llvm
 llvm-build
+llvm-project

.gitmodules

@@ -9,7 +9,7 @@
 	url = https://github.com/avr-rust/avr-mcu.git
 [submodule "lib/cmsis-svd"]
 	path = lib/cmsis-svd
-	url = https://github.com/posborne/cmsis-svd
+	url = https://github.com/tinygo-org/cmsis-svd
 [submodule "lib/compiler-rt"]
 	path = lib/compiler-rt
 	url = https://github.com/llvm-mirror/compiler-rt.git

BUILDING.md

@@ -16,7 +16,6 @@ LLVM, Clang and LLD are quite light on dependencies, requiring only standard
 build tools to be built. Go is of course necessary to build TinyGo itself.
 
   * Go (1.11+)
-  * [dep](https://golang.github.io/dep/)
   * Standard build tools (gcc/clang)
   * git
   * CMake
@@ -27,16 +26,20 @@ on a different system like Mac.
 
 ## Download the source
 
-The first step is to download the TinyGo sources. Then, inside the directory,
-perform these steps:
+The first step is to download the TinyGo sources (use `--recursive` if you clone
+the git repository). Then, inside the directory, download the LLVM source:
 
-    dep ensure -vendor-only # download Go dependencies
-    make llvm-source        # download LLVM
+    make llvm-source
 
 You can also store LLVM outside of the TinyGo root directory by setting the
 `LLVM_BUILDDIR`, `CLANG_SRC` and `LLD_SRC` make variables, but that is not
 covered by this guide.
 
+TinyGo uses Go modules, so if you clone TinyGo inside your GOPATH (and are using
+Go below 1.13), make sure that Go modules are enabled:
+
+    export GO111MODULE=on
+
 ## Build LLVM, Clang, LLD
 
 Before starting the build, you may want to set the following environment

CHANGELOG.md

@@ -1,3 +1,270 @@
+0.11.0
+---
+* **command line**
+  - add support for QEMU in `gdb` subcommand
+  - use builtin Clang when building statically, dropping the clang-9 dependency
+  - search for default serial port on both macOS and Linux
+  - windows: support `tinygo flash` directly by using win32 wmi
+* **compiler**
+  - add location information to the IR checker
+  - make reflection sidetables constant globals
+  - improve error locations in goroutine lowering
+  - interp: improve support for maps with string keys
+  - interp: add runtime fallback for mapassign operations
+* **standard library**
+  - `machine`: add support for `SPI.Tx()` on play.tinygo.org
+  - `machine`: rename `CPU_FREQUENCY` to `CPUFrequency()`
+* **targets**
+  - `adafruit-pybadge`: add Adafruit Pybadge
+  - `arduino-nano33`: allow simulation on play.tinygo.org
+  - `arduino-nano33`: fix default SPI pin numbers to be D13/D11/D12
+  - `circuitplay-express`: allow simulation on play.tinygo.org
+  - `hifive1-qemu`: add target for testing RISC-V bare metal in QEMU
+  - `riscv`: fix heap corruption due to changes in LLVM 9
+  - `riscv`: add support for compiler-rt
+  - `qemu`: rename to `cortex-m-qemu`
+
+0.10.0
+---
+* **command line**
+  - halt GDB after flashing with `gdb` subcommand
+  - fix a crash when using `-ocd-output`
+  - add `info` subcommand
+  - add `-programmer` flag
+* **builder**
+  - macos: use llvm@8 instead of just llvm in paths
+  - add `linkerscript` key to target JSON files
+  - write a symbol table when writing out the compiler-rt lib
+  - make Clang header detection more robust
+  - switch to LLVM 9
+* **compiler**
+  - fix interface miscompilation with reflect
+  - fix miscompile of static goroutine calls to closures
+  - fix `todo: store` panic
+  - fix incorrect starting value for optimized allocations in a loop
+  - optimize coroutines on non-Cortex-M targets
+  - fix crash for programs which have heap allocations but never hit the GC
+  - add support for async interface calls
+  - fix inserting non-const values in a const global
+  - interp: improve error reporting
+  - interp: implement comparing ptrtoint to 0
+* **cgo**
+  - improve diagnostics
+  - implement the constant parser (for `#define`) as a real parser
+  - rename reserved field names such as `type`
+  - avoid `"unsafe" imported but not used` error
+  - include all enums in the CGo Go AST
+  - add support for nested structs and unions
+  - implement `#cgo CFLAGS`
+* **standard library**
+  - `reflect`: add implementation of array alignment
+  - `runtime`: improve scheduler performance when no goroutines are queued
+  - `runtime`: add blocking select
+  - `runtime`: implement interface equality in non-trivial cases
+  - `runtime`: add AdjustTimeOffset to update current time
+  - `runtime`: only implement CountString for required platforms
+  - `runtime`: use MSP/PSP registers for scheduling on Cortex-M
+* **targets**
+  - `arm`: add system timer registers
+  - `atmega`: add port C GPIO support
+  - `atsamd21`: correct handling of pins >= 32
+  - `atsamd21`: i2s initialization fixes
+  - `atsamd51`: fix clock init code
+  - `atsamd51`: correct initialization for RTC
+  - `atsamd51`: fix pin function selection
+  - `atsamd51`: pin method cleanup
+  - `atsamd51`: allow setting pin mode for each of the SPI pins
+  - `atsamd51`: correct channel init and pin map for ADC based on ItsyBitsy-M4
+  - `feather-m4`: add Adafruit Feather M4 board
+  - `hifive1b`: add support for SPI1
+  - `hifive1b`: fix compiling in simulation
+  - `linux`: fix time on arm32
+  - `metro-m4`: add support for Adafruit Metro M4 Express Airlift board
+  - `metro-m4`: fixes for UART2
+  - `pinetime-devkit0`: add support for the PineTime dev kit
+  - `x9pro`: add support for this smartwatch
+  - `pca10040-s132v6`: add support for SoftDevice
+  - `pca10056-s140v7`: add support for SoftDevice
+  - `arduino-nano33`: added SPI1 connected to NINA-W102 chip on Arduino Nano 33 IOT
+
+0.9.0
+---
+* **command line**
+  - implement 1200-baud UART bootloader reset when flashing boards that support
+    it
+  - flash using mass-storage device for boards that support it
+  - implement `tinygo env`
+  - add support for Windows (but not yet producing Windows binaries)
+  - add Go version to `tinygo env`
+  - update SVD files for up-to-date peripheral interfaces
+* **compiler**
+  - add `//go:align` pragma
+  - fix bug related to type aliases
+  - add support for buffered channels
+  - remove incorrect reflect optimization
+  - implement copying slices in init interpretation
+  - add support for constant indices with a named type
+  - add support for recursive types like linked lists
+  - fix miscompile of function nil panics
+  - fix bug related to goroutines
+* **standard library**
+  - `machine`: do not check for nil slices in `SPI.Tx`
+  - `reflectlite`: add support for Go 1.13
+  - `runtime`: implement `internal/bytealg.CountString`
+  - `sync`: properly handle nil `New` func in `sync.Pool`
+* **targets**
+  - `arduino`: fix .bss section initialization
+  - `fe310`: implement `Pin.Get`
+  - `gameboy-advance`: support directly outputting .gba files
+  - `samd`: reduce code size by avoiding reflection
+  - `samd21`: do not hardcode pin numbers for peripherals
+  - `stm32f103`: avoid issue with `time.Sleep` less than 200µs
+
+0.8.0
+---
+* **command line**
+  - fix parsing of beta Go versions
+  - check the major/minor installed version of Go before compiling
+  - validate `-target` flag better to not panic on an invalid target
+* **compiler**
+  - implement full slice expression: `s[:2:4]`
+  - fix a crash when storing a linked list in an interface
+  - fix comparing struct types by making type IDs more unique
+  - fix some bugs in IR generation
+  - add support for linked lists in reflect data
+  - implement `[]rune` to string conversion
+  - implement support for `go` on func values
+* **standard library**
+  - `reflect`: add support for named types
+  - `reflect`: add support for `t.Bits()`
+  - `reflect`: add basic support for `t.AssignableTo()`
+  - `reflect`: implement `t.Align()`
+  - `reflect`: add support for struct types
+  - `reflect`: fix bug in `v.IsNil` and `v.Pointer` for addressable values
+  - `reflect`: implement support for array types
+  - `reflect`: implement `t.Comparable()`
+  - `runtime`: implement stack-based scheduler
+  - `runtime`: fix bug in the sleep queue of the scheduler
+  - `runtime`: implement `memcpy` for Cortex-M
+  - `testing`: implement stub `testing.B` struct
+  - `testing`: add common test logging methods such as Errorf/Fatalf/Printf
+* **targets**
+  - `386`: add support for linux/386 syscalls
+  - `atsamd21`: make SPI pins configurable so that multiple SPI ports can be
+    used
+  - `atsamd21`: correct issue with invalid first reading coming from ADC
+  - `atsamd21`: add support for reset-to-bootloader using 1200baud over USB-CDC
+  - `atsamd21`: make pin selection more flexible for peripherals
+  - `atsamd21`: fix minimum delay in `time.Sleep`
+  - `atsamd51`: fix minimum delay in `time.Sleep`
+  - `nrf`: improve SPI write-only speed, by making use of double buffering
+  - `stm32f103`: fix SPI frequency selection
+  - `stm32f103`: add machine.Pin.Get method for reading GPIO values
+  - `stm32f103`: allow board specific UART usage
+  - `nucleo-f103rb`: add support for NUCLEO-F103RB board
+  - `itsybitsy-m4`: add support for this board with a SAMD51 family chip
+  - `cortex-m`: add support for `arm.SystemReset()`
+  - `gameboy-advance`: add initial support for the GameBoy Advance
+  - `wasm`: add `//go:wasm-module` magic comment to set the wasm module name
+  - `wasm`: add syscall/js.valueSetIndex support
+  - `wasm`: add syscall/js.valueInvoke support
+
+0.7.1
+---
+* **targets**
+  - `atsamd21`: add support for the `-port` flag in the flash subcommand
+
+0.7.0
+---
+* **command line**
+  - try more locations to find Clang built-in headers
+  - add support for `tinygo test`
+  - build current directory if no package is specified
+  - support custom .json target spec with `-target` flag
+  - use zversion.go to detect version of GOROOT version
+  - make initial heap size configurable for some targets (currently WebAssembly
+    only)
+* **cgo**
+  - add support for bitfields using generated getters and setters
+  - add support for anonymous structs
+* **compiler**
+  - show an error instead of panicking on duplicate function definitions
+  - allow packages like github.com/tinygo-org/tinygo/src/\* by aliasing it
+  - remove `//go:volatile` support  
+    It has been replaced with the runtime/volatile package.
+  - allow poiners in map keys
+  - support non-constant syscall numbers
+  - implement non-blocking selects
+  - add support for the `-tags` flag
+  - add support for `string` to `[]rune` conversion
+  - implement a portable conservative garbage collector (with support for wasm)
+  - add the `//go:noinline` pragma
+* **standard library**
+  - `os`: add `os.Exit` and `syscall.Exit`
+  - `os`: add several stubs
+  - `runtime`: fix heap corruption in conservative GC
+  - `runtime`: add support for math intrinsics where supported, massively
+    speeding up some benchmarks
+  - `testing`: add basic support for testing
+* **targets**
+  - add support for a generic target that calls `__tinygo_*` functions for
+    peripheral access
+  - `arduino-nano33`: add support for this board
+  - `hifive1`: add support for this RISC-V board
+  - `reelboard`: add e-paper pins
+  - `reelboard`: add `PowerSupplyActive` to enable voltage for on-board devices
+  - `wasm`: put the stack at the start of linear memory, to detect stack
+    overflows
+
+0.6.0
+---
+* **command line**
+  - some portability improvements
+  - make `$GOROOT` more robust and configurable
+  - check for Clang at the Homebrew install location as fallback
+* **compiler driver**
+  - support multiple variations of LLVM commands, for non-Debian distributions
+* **compiler**
+  - improve code quality in multiple ways
+  - make panic configurable, adding trap on panic
+  - refactor many internal parts of the compiler
+  - print all errors encountered during compilation
+  - implement calling function values of a named type
+  - implement returning values from blocking functions
+  - allow larger-than-int values to be sent across a channel
+  - implement complex arithmetic
+  - improve hashmap support
+  - add debuginfo for function arguments
+  - insert nil checks on stores (increasing code size)
+  - implement volatile operations as compiler builtins
+  - add `//go:inline` pragma
+  - add build tags for the Go stdlib version
+* **cgo**
+  - implement `char`, `enum` and `void*` types
+  - support `#include` for builtin headers
+  - improve typedef/struct/enum support
+  - only include symbols that are necessary, for broader support
+  - mark external function args as `nocapture`
+  - implement support for some `#define` constants
+  - implement support for multiple CGo files in a single package
+- **standard library**
+  - `machine`: remove microbit matrix (moved to drivers repository)
+  - `machine`: refactor pins to use `Pin` type instead of `GPIO`
+  - `runtime`: print more interface types on panic, including `error`
+* **targets**
+  - `arm`: print an error on HardFault (including stack overflows)
+  - `atsamd21`: fix a bug in the ADC peripheral
+  - `atsamd21`: add support for I2S
+  - `feather-m0`: add support for this board
+  - `nrf51`: fix a bug in I2C
+  - `stm32f103xx`: fix a bug in I2C
+  - `syscall`: implement `Exit` on unix
+  - `trinket-m0`: add support for this board
+  - `wasm`: make _main_ example smaller
+  - `wasm`: don't cache wasm file in the server, for ease of debugging
+  - `wasm`: work around bug #41508 that caused a deadlock while linking
+  - `wasm`: add support for `js.FuncOf`
+
 0.5.0
 ---
 - **compiler driver**

CODE-OF-CONDUCT.md

@@ -0,0 +1,76 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+In the interest of fostering an open and welcoming environment, we as
+contributors and maintainers pledge to make participation in our project and
+our community a harassment-free experience for everyone, regardless of age, body
+size, disability, ethnicity, sex characteristics, gender identity and expression,
+level of experience, education, socio-economic status, nationality, personal
+appearance, race, religion, or sexual identity and orientation.
+
+## Our Standards
+
+Examples of behavior that contributes to creating a positive environment
+include:
+
+* Using welcoming and inclusive language
+* Being respectful of differing viewpoints and experiences
+* Gracefully accepting constructive criticism
+* Focusing on what is best for the community
+* Showing empathy towards other community members
+
+Examples of unacceptable behavior by participants include:
+
+* The use of sexualized language or imagery and unwelcome sexual attention or
+  advances
+* Trolling, insulting/derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or electronic
+  address, without explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Our Responsibilities
+
+Project maintainers are responsible for clarifying the standards of acceptable
+behavior and are expected to take appropriate and fair corrective action in
+response to any instances of unacceptable behavior.
+
+Project maintainers have the right and responsibility to remove, edit, or
+reject comments, commits, code, wiki edits, issues, and other contributions
+that are not aligned to this Code of Conduct, or to ban temporarily or
+permanently any contributor for other behaviors that they deem inappropriate,
+threatening, offensive, or harmful.
+
+## Scope
+
+This Code of Conduct applies within all project spaces, and it also applies when
+an individual is representing the project or its community in public spaces.
+Examples of representing a project or community include using an official
+project e-mail address, posting via an official social media account, or acting
+as an appointed representative at an online or offline event. Representation of
+a project may be further defined and clarified by project maintainers.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported by contacting the project team at [conduct@tinygo.org](mailto:conduct@tinygo.org). All
+complaints will be reviewed and investigated and will result in a response that
+is deemed necessary and appropriate to the circumstances. The project team is
+obligated to maintain confidentiality with regard to the reporter of an incident.
+Further details of specific enforcement policies may be posted separately.
+
+Project maintainers who do not follow or enforce the Code of Conduct in good
+faith may face temporary or permanent repercussions as determined by other
+members of the project's leadership.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 1.4,
+available at https://www.contributor-covenant.org/version/1/4/code-of-conduct.html
+
+[homepage]: https://www.contributor-covenant.org
+
+For answers to common questions about this code of conduct, see
+https://www.contributor-covenant.org/faq

CONTRIBUTING.md

@@ -16,12 +16,16 @@ Please open a Github issue with your problem, and we will be happy to assist.
 
 We probably have not implemented it yet. Please take a look at our [Roadmap](https://github.com/tinygo-org/tinygo/wiki/Roadmap). Your pull request adding the functionality to TinyGo would be greatly appreciated.
 
+Please open a Github issue. We want to help, and also make sure that there is no duplications of efforts. Sometimes what you need is already being worked on by someone else.
+
 A long tail of small (and large) language features haven't been implemented yet. In almost all cases, the compiler will show a `todo:` error from `compiler/compiler.go` when you try to use it. You can try implementing it, or open a bug report with a small code sample that fails to compile.
 
 ### Some specific hardware you want to use does not appear to be in TinyGo
 
 As above, we probably have not implemented it yet. Your contribution adding the hardware support to TinyGo would be greatly appreciated.
 
+Please start by opening a Github issue. We want to help you to help us to help you.
+
 Lots of targets/boards are still unsupported. Adding an architecture often requires a few compiler changes, but if the architecture is supported you can try implementing support for a new chip or board in `src/runtime`. For details, see [this wiki entry on adding archs/chips/boards](https://github.com/tinygo-org/tinygo/wiki/Adding-a-new-board).
 
 Microcontrollers have lots of peripherals (I2C, SPI, ADC, etc.) and many don't have an implementation yet in the `machine` package. Adding support for new peripherals is very useful.

CONTRIBUTORS

@@ -0,0 +1,17 @@
+# This is the official list of TinyGo authors for copyright purposes.
+#
+# This file is not actively maintained.
+# To be included, send a change adding the individual or
+# company who owns a contribution's copyright.
+#
+# Names should be added to this file as one of
+#     Organization's name
+#     Individual's name <submission email address>
+#     Individual's name <submission email address> <email2> <emailN>
+#
+# Please keep the list sorted.
+
+Ayke van Laethem <aykevanlaethem@gmail.com>
+Daniel Esteban <conejo@conejo.me>
+Loon, LLC.
+Ron Evans <ron@hybridgroup.com>

Dockerfile

@@ -1,52 +1,48 @@
-# TinyGo base stage just installs LLVM 8 and the TinyGo compiler itself.
-FROM golang:latest AS tinygo-base
+# TinyGo base stage installs Go 1.13, LLVM 9 and the TinyGo compiler itself.
+FROM golang:1.13 AS tinygo-base
 
 RUN wget -O- https://apt.llvm.org/llvm-snapshot.gpg.key| apt-key add - && \
-    echo "deb http://apt.llvm.org/stretch/ llvm-toolchain-stretch-8 main" >> /etc/apt/sources.list && \
+    echo "deb http://apt.llvm.org/buster/ llvm-toolchain-buster-9 main" >> /etc/apt/sources.list && \
     apt-get update && \
-    apt-get install -y llvm-8-dev libclang-8-dev git
+    apt-get install -y llvm-9-dev libclang-9-dev git
 
-RUN wget -O- https://raw.githubusercontent.com/golang/dep/master/install.sh | sh
-
-COPY . /go/src/github.com/tinygo-org/tinygo
+COPY . /tinygo
 
 # remove submodules directories and re-init them to fix any hard-coded paths
 # after copying the tinygo directory in the previous step.
-RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
+RUN cd /tinygo/ && \
     rm -rf ./lib/* && \
     git submodule update --init --recursive --force
 
-RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
-    dep ensure --vendor-only && \
-    go install /go/src/github.com/tinygo-org/tinygo/
+RUN cd /tinygo/ && \
+    go install /tinygo/
 
 # tinygo-wasm stage installs the needed dependencies to compile TinyGo programs for WASM.
 FROM tinygo-base AS tinygo-wasm
 
 COPY --from=tinygo-base /go/bin/tinygo /go/bin/tinygo
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/src /go/src/github.com/tinygo-org/tinygo/src
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/targets /go/src/github.com/tinygo-org/tinygo/targets
+COPY --from=tinygo-base /tinygo/src /tinygo/src
+COPY --from=tinygo-base /tinygo/targets /tinygo/targets
 
 RUN wget -O- https://apt.llvm.org/llvm-snapshot.gpg.key| apt-key add - && \
-    echo "deb http://apt.llvm.org/stretch/ llvm-toolchain-stretch-8 main" >> /etc/apt/sources.list && \
+    echo "deb http://apt.llvm.org/buster/ llvm-toolchain-buster-9 main" >> /etc/apt/sources.list && \
     apt-get update && \
-    apt-get install -y libllvm8 lld-8
+    apt-get install -y libllvm9 lld-9
 
 # tinygo-avr stage installs the needed dependencies to compile TinyGo programs for AVR microcontrollers.
 FROM tinygo-base AS tinygo-avr
 
 COPY --from=tinygo-base /go/bin/tinygo /go/bin/tinygo
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/src /go/src/github.com/tinygo-org/tinygo/src
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/targets /go/src/github.com/tinygo-org/tinygo/targets
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/Makefile /go/src/github.com/tinygo-org/tinygo/
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/tools /go/src/github.com/tinygo-org/tinygo/tools
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/lib /go/src/github.com/tinygo-org/tinygo/lib
+COPY --from=tinygo-base /tinygo/src /tinygo/src
+COPY --from=tinygo-base /tinygo/targets /tinygo/targets
+COPY --from=tinygo-base /tinygo/Makefile /tinygo/
+COPY --from=tinygo-base /tinygo/tools /tinygo/tools
+COPY --from=tinygo-base /tinygo/lib /tinygo/lib
 
-RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
+RUN cd /tinygo/ && \
     apt-get update && \
-    apt-get install -y apt-utils python3 make binutils-avr gcc-avr avr-libc && \
+    apt-get install -y apt-utils make binutils-avr gcc-avr avr-libc && \
     make gen-device-avr && \
-    apt-get remove -y python3 make && \
     apt-get autoremove -y && \
     apt-get clean
 
@@ -54,33 +50,27 @@ RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
 FROM tinygo-base AS tinygo-arm
 
 COPY --from=tinygo-base /go/bin/tinygo /go/bin/tinygo
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/src /go/src/github.com/tinygo-org/tinygo/src
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/targets /go/src/github.com/tinygo-org/tinygo/targets
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/Makefile /go/src/github.com/tinygo-org/tinygo/
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/tools /go/src/github.com/tinygo-org/tinygo/tools
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/lib /go/src/github.com/tinygo-org/tinygo/lib
+COPY --from=tinygo-base /tinygo/src /tinygo/src
+COPY --from=tinygo-base /tinygo/targets /tinygo/targets
+COPY --from=tinygo-base /tinygo/Makefile /tinygo/
+COPY --from=tinygo-base /tinygo/tools /tinygo/tools
+COPY --from=tinygo-base /tinygo/lib /tinygo/lib
 
-RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
+RUN cd /tinygo/ && \
     apt-get update && \
-    apt-get install -y apt-utils python3 make clang-8 && \
-    make gen-device-nrf && make gen-device-stm32 && \
-    apt-get remove -y python3 make && \
-    apt-get autoremove -y && \
-    apt-get clean
+    apt-get install -y apt-utils make clang-9 && \
+    make gen-device-nrf && make gen-device-stm32
 
 # tinygo-all stage installs the needed dependencies to compile TinyGo programs for all platforms.
 FROM tinygo-wasm AS tinygo-all
 
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/Makefile /go/src/github.com/tinygo-org/tinygo/
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/tools /go/src/github.com/tinygo-org/tinygo/tools
-COPY --from=tinygo-base /go/src/github.com/tinygo-org/tinygo/lib /go/src/github.com/tinygo-org/tinygo/lib
+COPY --from=tinygo-base /tinygo/Makefile /tinygo/
+COPY --from=tinygo-base /tinygo/tools /tinygo/tools
+COPY --from=tinygo-base /tinygo/lib /tinygo/lib
 
-RUN cd /go/src/github.com/tinygo-org/tinygo/ && \
+RUN cd /tinygo/ && \
     apt-get update && \
-    apt-get install -y apt-utils python3 make clang-8 binutils-avr gcc-avr avr-libc && \
-    make gen-device && \
-    apt-get remove -y python3 make && \
-    apt-get autoremove -y && \
-    apt-get clean
+    apt-get install -y apt-utils make clang-9 binutils-avr gcc-avr avr-libc && \
+    make gen-device
 
 CMD ["tinygo"]

Gopkg.lock

@@ -1,51 +0,0 @@
-# This file is autogenerated, do not edit; changes may be undone by the next 'dep ensure'.
-
-
-[[projects]]
-  branch = "master"
-  digest = "1:06519a2ec1d59040eaccec40206f9d0b59dc662db2a032f974d6d6b9a2bcb839"
-  name = "github.com/blakesmith/ar"
-  packages = ["."]
-  pruneopts = "UT"
-  revision = "8bd4349a67f2533b078dbc524689d15dba0f4659"
-
-[[projects]]
-  branch = "master"
-  digest = "1:00b45e06c7843541372fc17d982242bd6adfc2fc382b6f2e9ef9ce53d87a50b9"
-  name = "github.com/marcinbor85/gohex"
-  packages = ["."]
-  pruneopts = "UT"
-  revision = "7a43cd876e46e0f6ddc553f10f91731a78e6e949"
-
-[[projects]]
-  branch = "master"
-  digest = "1:ba70784a3deee74c0ca3c87bcac3c2f93d3b2d27d8f237b768c358b45ba47da8"
-  name = "golang.org/x/tools"
-  packages = [
-    "go/ast/astutil",
-    "go/ssa",
-    "go/types/typeutil",
-  ]
-  pruneopts = "UT"
-  revision = "8dcc6e70cdefe9a82236b6e195e4f4e2108fcb9f"
-
-[[projects]]
-  branch = "llvm8"
-  digest = "1:bf5539bdf6b3cc3ec1e45926db05d81180da11ce722fa1edcce3f0b4e1967da5"
-  name = "tinygo.org/x/go-llvm"
-  packages = ["."]
-  pruneopts = "UT"
-  revision = "7707ae5d1261a8929edea7336c8087ca8b520d8d"
-
-[solve-meta]
-  analyzer-name = "dep"
-  analyzer-version = 1
-  input-imports = [
-    "github.com/blakesmith/ar",
-    "github.com/marcinbor85/gohex",
-    "golang.org/x/tools/go/ast/astutil",
-    "golang.org/x/tools/go/ssa",
-    "tinygo.org/x/go-llvm",
-  ]
-  solver-name = "gps-cdcl"
-  solver-version = 1

Gopkg.toml

@@ -1,11 +0,0 @@
-[[constraint]]
-  branch = "llvm8"
-  name = "tinygo.org/x/go-llvm"
-
-[[constraint]]
-  branch = "master"
-  name = "golang.org/x/tools"
-
-[prune]
-  go-tests = true
-  unused-packages = true

LICENSE

@@ -3,6 +3,9 @@ Copyright (c) 2018-2019 TinyGo Authors. All rights reserved.
 TinyGo includes portions of the Go standard library.
 Copyright (c) 2009-2019 The Go Authors. All rights reserved.
 
+TinyGo includes portions of LLVM, which is under the Apache License v2.0 with
+LLVM Exceptions. See https://llvm.org/LICENSE.txt for license information.
+
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are
 met:

Makefile

@@ -1,90 +1,263 @@
 
 # aliases
 all: tinygo
-tinygo: build/tinygo
-
-.PHONY: all tinygo build/tinygo test llvm-build llvm-source clean fmt gen-device gen-device-nrf gen-device-avr
 
 # Default build and source directories, as created by `make llvm-build`.
 LLVM_BUILDDIR ?= llvm-build
-CLANG_SRC ?= llvm/tools/clang
-LLD_SRC ?= llvm/tools/lld
+CLANG_SRC ?= llvm-project/clang
+LLD_SRC ?= llvm-project/lld
 
-LLVM_COMPONENTS = all-targets analysis asmparser asmprinter bitreader bitwriter codegen core coroutines debuginfodwarf executionengine instrumentation interpreter ipo irreader linker lto mc mcjit objcarcopts option profiledata scalaropts support target
+# Go binary and GOROOT to select
+GO ?= go
+export GOROOT = $(shell $(GO) env GOROOT)
 
-UNAME_S := $(shell uname -s)
-ifeq ($(UNAME_S),Linux)
+# md5sum binary
+MD5SUM = md5sum
+
+# tinygo binary for tests
+TINYGO ?= tinygo
+
+# Use CCACHE for LLVM if possible
+ifneq (, $(shell which ccache))
+    LLVM_OPTION += '-DLLVM_CCACHE_BUILD=ON'
+endif
+
+# Allow enabling LLVM assertions
+ifeq (1, $(ASSERT))
+    LLVM_OPTION += '-DLLVM_ENABLE_ASSERTIONS=ON'
+else
+    LLVM_OPTION += '-DLLVM_ENABLE_ASSERTIONS=OFF'
+endif
+
+.PHONY: all tinygo test $(LLVM_BUILDDIR) llvm-source clean fmt gen-device gen-device-nrf gen-device-avr
+
+LLVM_COMPONENTS = all-targets analysis asmparser asmprinter bitreader bitwriter codegen core coroutines coverage debuginfodwarf executionengine instrumentation interpreter ipo irreader linker lto mc mcjit objcarcopts option profiledata scalaropts support target
+
+ifeq ($(OS),Windows_NT)
+    EXE = .exe
+    START_GROUP = -Wl,--start-group
+    END_GROUP = -Wl,--end-group
+
+    # LLVM compiled using MinGW on Windows appears to have problems with threads.
+    # Without this flag, linking results in errors like these:
+    #     libLLVMSupport.a(Threading.cpp.obj):Threading.cpp:(.text+0x55): undefined reference to `std::thread::hardware_concurrency()'
+    LLVM_OPTION += -DLLVM_ENABLE_THREADS=OFF
+
+    CGO_LDFLAGS += -static -static-libgcc -static-libstdc++
+    CGO_LDFLAGS_EXTRA += -lversion
+
+    # Build libclang manually because the CMake-based build system on Windows
+    # doesn't allow building libclang as a static library.
+    LIBCLANG_PATH = $(abspath build/libclang-custom.a)
+    LIBCLANG_FILES = $(abspath $(wildcard $(LLVM_BUILDDIR)/tools/clang/tools/libclang/CMakeFiles/libclang.dir/*.cpp.obj))
+
+    # Add the libclang dependency to the tinygo binary target.
+tinygo: $(LIBCLANG_PATH)
+test: $(LIBCLANG_PATH)
+    # Build libclang.
+$(LIBCLANG_PATH): $(LIBCLANG_FILES)
+	@mkdir -p build
+	ar rcs $(LIBCLANG_PATH) $^
+
+else ifeq ($(shell uname -s),Darwin)
+    MD5SUM = md5
+    LIBCLANG_PATH = $(abspath $(LLVM_BUILDDIR))/lib/libclang.a
+else ifeq ($(shell uname -s),FreeBSD)
+    MD5SUM = md5
+    LIBCLANG_PATH = $(abspath $(LLVM_BUILDDIR))/lib/libclang.a
+    START_GROUP = -Wl,--start-group
+    END_GROUP = -Wl,--end-group
+else
+    LIBCLANG_PATH = $(abspath $(LLVM_BUILDDIR))/lib/libclang.a
     START_GROUP = -Wl,--start-group
     END_GROUP = -Wl,--end-group
 endif
 
-CLANG_LIBS = $(START_GROUP) $(abspath $(LLVM_BUILDDIR))/lib/libclang.a -lclangAnalysis -lclangARCMigrate -lclangAST -lclangASTMatchers -lclangBasic -lclangCodeGen -lclangCrossTU -lclangDriver -lclangDynamicASTMatchers -lclangEdit -lclangFormat -lclangFrontend -lclangFrontendTool -lclangHandleCXX -lclangHandleLLVM -lclangIndex -lclangLex -lclangParse -lclangRewrite -lclangRewriteFrontend -lclangSema -lclangSerialization -lclangStaticAnalyzerCheckers -lclangStaticAnalyzerCore -lclangStaticAnalyzerFrontend -lclangTooling -lclangToolingASTDiff -lclangToolingCore -lclangToolingInclusions -lclangToolingRefactor $(END_GROUP) -lstdc++
+CLANG_LIBS = $(START_GROUP) -lclangAnalysis -lclangARCMigrate -lclangAST -lclangASTMatchers -lclangBasic -lclangCodeGen -lclangCrossTU -lclangDriver -lclangDynamicASTMatchers -lclangEdit -lclangFormat -lclangFrontend -lclangFrontendTool -lclangHandleCXX -lclangHandleLLVM -lclangIndex -lclangLex -lclangParse -lclangRewrite -lclangRewriteFrontend -lclangSema -lclangSerialization -lclangStaticAnalyzerCheckers -lclangStaticAnalyzerCore -lclangStaticAnalyzerFrontend -lclangTooling -lclangToolingASTDiff -lclangToolingCore -lclangToolingInclusions $(END_GROUP) -lstdc++
 
 LLD_LIBS = $(START_GROUP) -llldCOFF -llldCommon -llldCore -llldDriver -llldELF -llldMachO -llldMinGW -llldReaderWriter -llldWasm -llldYAML $(END_GROUP)
 
 
 # For static linking.
-CGO_CPPFLAGS=$(shell $(LLVM_BUILDDIR)/bin/llvm-config --cppflags) -I$(abspath $(CLANG_SRC))/include -I$(abspath $(LLD_SRC))/include
-CGO_CXXFLAGS=-std=c++11
-CGO_LDFLAGS=-L$(LLVM_BUILDDIR)/lib $(CLANG_LIBS) $(LLD_LIBS) $(shell $(LLVM_BUILDDIR)/bin/llvm-config --ldflags --libs --system-libs $(LLVM_COMPONENTS))
+ifneq ("$(wildcard $(LLVM_BUILDDIR)/bin/llvm-config*)","")
+    CGO_CPPFLAGS=$(shell $(LLVM_BUILDDIR)/bin/llvm-config --cppflags) -I$(abspath $(LLVM_BUILDDIR))/tools/clang/include -I$(abspath $(CLANG_SRC))/include -I$(abspath $(LLD_SRC))/include
+    CGO_CXXFLAGS=-std=c++11
+    CGO_LDFLAGS+=$(LIBCLANG_PATH) -std=c++11 -L$(abspath $(LLVM_BUILDDIR)/lib) $(CLANG_LIBS) $(LLD_LIBS) $(shell $(LLVM_BUILDDIR)/bin/llvm-config --ldflags --libs --system-libs $(LLVM_COMPONENTS)) -lstdc++ $(CGO_LDFLAGS_EXTRA)
+endif
 
 
 clean:
 	@rm -rf build
 
-FMT_PATHS = ./*.go cgo compiler interp ir loader src/device/arm src/examples src/machine src/os src/reflect src/runtime src/sync src/syscall
+FMT_PATHS = ./*.go builder cgo compiler interp ir loader src/device/arm src/examples src/machine src/os src/reflect src/runtime src/sync src/syscall src/internal/reflectlite transform
 fmt:
 	@gofmt -l -w $(FMT_PATHS)
 fmt-check:
 	@unformatted=$$(gofmt -l $(FMT_PATHS)); [ -z "$$unformatted" ] && exit 0; echo "Unformatted:"; for fn in $$unformatted; do echo "  $$fn"; done; exit 1
 
-gen-device: gen-device-avr gen-device-nrf gen-device-sam gen-device-stm32
+
+gen-device: gen-device-avr gen-device-nrf gen-device-sam gen-device-sifive gen-device-stm32
 
 gen-device-avr:
-	./tools/gen-device-avr.py lib/avr/packs/atmega src/device/avr/
-	./tools/gen-device-avr.py lib/avr/packs/tiny src/device/avr/
-	go fmt ./src/device/avr
+	$(GO) build -o ./build/gen-device-avr ./tools/gen-device-avr/
+	./build/gen-device-avr lib/avr/packs/atmega src/device/avr/
+	./build/gen-device-avr lib/avr/packs/tiny src/device/avr/
+	@GO111MODULE=off $(GO) fmt ./src/device/avr
 
-gen-device-nrf:
-	./tools/gen-device-svd.py lib/nrfx/mdk/ src/device/nrf/ --source=https://github.com/NordicSemiconductor/nrfx/tree/master/mdk
-	go fmt ./src/device/nrf
+build/gen-device-svd: ./tools/gen-device-svd/*.go
+	$(GO) build -o $@ ./tools/gen-device-svd/
 
-gen-device-sam:
-	./tools/gen-device-svd.py lib/cmsis-svd/data/Atmel/ src/device/sam/ --source=https://github.com/posborne/cmsis-svd/tree/master/data/Atmel
-	go fmt ./src/device/sam
+gen-device-nrf: build/gen-device-svd
+	./build/gen-device-svd -source=https://github.com/NordicSemiconductor/nrfx/tree/master/mdk lib/nrfx/mdk/ src/device/nrf/
+	GO111MODULE=off $(GO) fmt ./src/device/nrf
 
-gen-device-stm32:
-	./tools/gen-device-svd.py lib/cmsis-svd/data/STMicro/ src/device/stm32/ --source=https://github.com/posborne/cmsis-svd/tree/master/data/STMicro
-	go fmt ./src/device/stm32
+gen-device-sam: build/gen-device-svd
+	./build/gen-device-svd -source=https://github.com/posborne/cmsis-svd/tree/master/data/Atmel lib/cmsis-svd/data/Atmel/ src/device/sam/
+	GO111MODULE=off $(GO) fmt ./src/device/sam
+
+gen-device-sifive: build/gen-device-svd
+	./build/gen-device-svd -source=https://github.com/posborne/cmsis-svd/tree/master/data/SiFive-Community lib/cmsis-svd/data/SiFive-Community/ src/device/sifive/
+	GO111MODULE=off $(GO) fmt ./src/device/sifive
+
+gen-device-stm32: build/gen-device-svd
+	./build/gen-device-svd -source=https://github.com/posborne/cmsis-svd/tree/master/data/STMicro lib/cmsis-svd/data/STMicro/ src/device/stm32/
+	GO111MODULE=off $(GO) fmt ./src/device/stm32
 
 
 # Get LLVM sources.
-llvm/README.txt:
-	git clone -b release_80 https://github.com/llvm-mirror/llvm.git llvm
-llvm/tools/clang/README.txt:
-	git clone -b release_80 https://github.com/llvm-mirror/clang.git llvm/tools/clang
-llvm/tools/lld/README.md:
-	git clone -b release_80 https://github.com/llvm-mirror/lld.git llvm/tools/lld
-llvm-source: llvm/README.txt llvm/tools/clang/README.txt llvm/tools/lld/README.md
+llvm-project/README.md:
+	git clone -b release/9.x https://github.com/llvm/llvm-project
+llvm-source: llvm-project/README.md
 
 # Configure LLVM.
-llvm-build/build.ninja: llvm-source
-	mkdir -p llvm-build; cd llvm-build; cmake -G Ninja ../llvm "-DLLVM_TARGETS_TO_BUILD=X86;ARM;AArch64;WebAssembly" "-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=AVR" -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_ASSERTIONS=OFF -DLIBCLANG_BUILD_STATIC=ON -DLLVM_ENABLE_TERMINFO=OFF -DLLVM_ENABLE_ZLIB=OFF
+TINYGO_SOURCE_DIR=$(shell pwd)
+$(LLVM_BUILDDIR)/build.ninja: llvm-source
+	mkdir -p $(LLVM_BUILDDIR); cd $(LLVM_BUILDDIR); cmake -G Ninja $(TINYGO_SOURCE_DIR)/llvm-project/llvm "-DLLVM_TARGETS_TO_BUILD=X86;ARM;AArch64;RISCV;WebAssembly" "-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=AVR" -DCMAKE_BUILD_TYPE=Release -DLIBCLANG_BUILD_STATIC=ON -DLLVM_ENABLE_TERMINFO=OFF -DLLVM_ENABLE_ZLIB=OFF -DLLVM_ENABLE_PROJECTS="clang;lld" -DLLVM_TOOL_CLANG_TOOLS_EXTRA_BUILD=OFF $(LLVM_OPTION)
 
 # Build LLVM.
-llvm-build: llvm-build/build.ninja
-	cd llvm-build; ninja
+$(LLVM_BUILDDIR): $(LLVM_BUILDDIR)/build.ninja
+	cd $(LLVM_BUILDDIR); ninja
 
 
 # Build the Go compiler.
-build/tinygo:
-	@if [ ! -f llvm-build/bin/llvm-config ]; then echo "Fetch and build LLVM first by running:\n  make llvm-source\n  make llvm-build"; exit 1; fi
-	CGO_CPPFLAGS="$(CGO_CPPFLAGS)" CGO_CXXFLAGS="$(CGO_CXXFLAGS)" CGO_LDFLAGS="$(CGO_LDFLAGS)" go build -o build/tinygo -tags byollvm .
+tinygo:
+	@if [ ! -f "$(LLVM_BUILDDIR)/bin/llvm-config" ]; then echo "Fetch and build LLVM first by running:"; echo "  make llvm-source"; echo "  make $(LLVM_BUILDDIR)"; exit 1; fi
+	CGO_CPPFLAGS="$(CGO_CPPFLAGS)" CGO_CXXFLAGS="$(CGO_CXXFLAGS)" CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GO) build -o build/tinygo$(EXE) -tags byollvm .
 
 test:
-	CGO_CPPFLAGS="$(CGO_CPPFLAGS)" CGO_CXXFLAGS="$(CGO_CXXFLAGS)" CGO_LDFLAGS="$(CGO_LDFLAGS)" go test -v -tags byollvm .
+	CGO_CPPFLAGS="$(CGO_CPPFLAGS)" CGO_CXXFLAGS="$(CGO_CXXFLAGS)" CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GO) test -v -tags byollvm ./cgo ./compileopts ./interp ./transform .
 
-release: build/tinygo gen-device
+tinygo-test:
+	cd tests/tinygotest && tinygo test
+
+.PHONY: smoketest
+smoketest:
+	$(TINYGO) version
+	# test all examples
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/adc
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/blinkm
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/blinky2
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/button
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/button2
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/echo
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=circuitplay-express examples/i2s
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/mcp3008
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=microbit            examples/microbit-blink
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/pwm
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/serial
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/systick
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/test
+	@$(MD5SUM) test.hex
+	# test simulated boards on play.tinygo.org
+	$(TINYGO) build             -o test.wasm -tags=arduino              examples/blinky1
+	@$(MD5SUM) test.wasm
+	$(TINYGO) build             -o test.wasm -tags=hifive1b             examples/blinky1
+	@$(MD5SUM) test.wasm
+	$(TINYGO) build             -o test.wasm -tags=reelboard            examples/blinky1
+	@$(MD5SUM) test.wasm
+	$(TINYGO) build             -o test.wasm -tags=pca10040             examples/blinky2
+	@$(MD5SUM) test.wasm
+	$(TINYGO) build             -o test.wasm -tags=pca10056             examples/blinky2
+	@$(MD5SUM) test.wasm
+	$(TINYGO) build             -o test.wasm -tags=circuitplay_express  examples/blinky1
+	@$(MD5SUM) test.wasm
+	# test all targets/boards
+	$(TINYGO) build -size short -o test.hex -target=pca10040-s132v6     examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=microbit            examples/echo
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=nrf52840-mdk        examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10031            examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=bluepill            examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=reelboard           examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=reelboard           examples/blinky2
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10056            examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10056            examples/blinky2
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=itsybitsy-m0        examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=feather-m0          examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=trinket-m0          examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=circuitplay-express examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=stm32f4disco        examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=stm32f4disco        examples/blinky2
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=circuitplay-express examples/i2s
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.gba -target=gameboy-advance     examples/gba-display
+	@$(MD5SUM) test.gba
+	$(TINYGO) build -size short -o test.hex -target=itsybitsy-m4        examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=feather-m4          examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pybadge             examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=metro-m4-airlift    examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=nucleo-f103rb       examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pinetime-devkit0    examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=x9pro               examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=pca10056-s140v7     examples/blinky1
+	@$(MD5SUM) test.hex
+ifneq ($(AVR), 0)
+	$(TINYGO) build -size short -o test.hex -target=arduino             examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=digispark           examples/blinky1
+	@$(MD5SUM) test.hex
+endif
+	$(TINYGO) build -size short -o test.hex -target=hifive1b            examples/blinky1
+	@$(MD5SUM) test.hex
+	$(TINYGO) build             -o wasm.wasm -target=wasm               examples/wasm/export
+	$(TINYGO) build             -o wasm.wasm -target=wasm               examples/wasm/main
+
+release: tinygo gen-device
 	@mkdir -p build/release/tinygo/bin
 	@mkdir -p build/release/tinygo/lib/clang/include
 	@mkdir -p build/release/tinygo/lib/CMSIS/CMSIS
@@ -94,7 +267,7 @@ release: build/tinygo gen-device
 	@mkdir -p build/release/tinygo/pkg/armv7m-none-eabi
 	@mkdir -p build/release/tinygo/pkg/armv7em-none-eabi
 	@echo copying source files
-	@cp -p  build/tinygo                 build/release/tinygo/bin
+	@cp -p  build/tinygo$(EXE)           build/release/tinygo/bin
 	@cp -p $(abspath $(CLANG_SRC))/lib/Headers/*.h build/release/tinygo/lib/clang/include
 	@cp -rp lib/CMSIS/CMSIS/Include      build/release/tinygo/lib/CMSIS/CMSIS
 	@cp -rp lib/CMSIS/README.md          build/release/tinygo/lib/CMSIS

README.md

@@ -1,8 +1,8 @@
 # TinyGo - Go compiler for small places
 
-[![CircleCI](https://circleci.com/gh/tinygo-org/tinygo/tree/dev.svg?style=svg)](https://circleci.com/gh/tinygo-org/tinygo/tree/dev)
+[![CircleCI](https://circleci.com/gh/tinygo-org/tinygo/tree/dev.svg?style=svg)](https://circleci.com/gh/tinygo-org/tinygo/tree/dev) [![Build Status](https://dev.azure.com/tinygo/tinygo/_apis/build/status/tinygo-CI?branchName=dev)](https://dev.azure.com/tinygo/tinygo/_build/latest?definitionId=1&branchName=dev)
 
-TinyGo is a Go compiler intended for use in small places such as microcontrollers, WebAssembly (WASM), and command-line tools.
+TinyGo is a Go compiler intended for use in small places such as microcontrollers, WebAssembly (Wasm), and command-line tools.
 
 It reuses libraries used by the [Go language tools](https://golang.org/pkg/go/) alongside [LLVM](http://llvm.org) to provide an alternative way to compile programs written in the Go programming language.
 
@@ -17,8 +17,8 @@ import (
 )
 
 func main() {
-    led := machine.GPIO{machine.LED}
-    led.Configure(machine.GPIOConfig{Mode: machine.GPIO_OUTPUT})
+    led := machine.LED
+    led.Configure(machine.PinConfig{Mode: machine.PinOutput})
     for {
         led.Low()
         time.Sleep(time.Millisecond * 1000)
@@ -43,19 +43,32 @@ See the [getting started instructions](https://tinygo.org/getting-started/) for
 
 You can compile TinyGo programs for microcontrollers, WebAssembly and Linux.
 
-The following microcontroller boards are currently supported:
+The following 22 microcontroller boards are currently supported:
 
 * [Adafruit Circuit Playground Express](https://www.adafruit.com/product/3333)
+* [Adafruit Feather M0](https://www.adafruit.com/product/2772)
+* [Adafruit Feather M4](https://www.adafruit.com/product/3857)
 * [Adafruit ItsyBitsy M0](https://www.adafruit.com/product/3727)
+* [Adafruit ItsyBitsy M4](https://www.adafruit.com/product/3800)
+* [Adafruit Metro M4 Express Airlift](https://www.adafruit.com/product/4000)
+* [Adafruit PyBadge](https://www.adafruit.com/product/4200)
+* [Adafruit Trinket M0](https://www.adafruit.com/product/3500)
+* [Arduino Nano33 IoT](https://store.arduino.cc/nano-33-iot)
 * [Arduino Uno](https://store.arduino.cc/arduino-uno-rev3)
-* [BBC:Microbit](https://microbit.org/)
-* [ST Micro STM32F103XX "Bluepill"](http://wiki.stm32duino.com/index.php?title=Blue_Pill)
+* [BBC micro:bit](https://microbit.org/)
 * [Digispark](http://digistump.com/products/1)
+* [Game Boy Advance](https://en.wikipedia.org/wiki/Game_Boy_Advance)
+* [Makerdiary nRF52840-MDK](https://wiki.makerdiary.com/nrf52840-mdk/)
 * [Nordic Semiconductor PCA10031](https://www.nordicsemi.com/eng/Products/nRF51-Dongle)
 * [Nordic Semiconductor PCA10040](https://www.nordicsemi.com/eng/Products/Bluetooth-low-energy/nRF52-DK)
 * [Nordic Semiconductor PCA10056](https://www.nordicsemi.com/Software-and-Tools/Development-Kits/nRF52840-DK)
-* [Makerdiary nRF52840-MDK](https://wiki.makerdiary.com/nrf52840-mdk/)
 * [Phytec reel board](https://www.phytec.eu/product-eu/internet-of-things/reelboard/)
+* [PineTime DevKit](https://www.pine64.org/pinetime/)
+* [SiFIve HiFive1](https://www.sifive.com/boards/hifive1)
+* [ST Micro "Nucleo F103RB"](https://www.st.com/en/evaluation-tools/nucleo-f103rb.html)
+* [ST Micro STM32F103XX "Bluepill"](http://wiki.stm32duino.com/index.php?title=Blue_Pill)
+* [ST Micro STM32F407 "Discovery"](https://www.st.com/en/evaluation-tools/stm32f4discovery.html)
+* [X9 Pro smartwatch](https://github.com/curtpw/nRF5x-device-reverse-engineering/tree/master/X9-nrf52832-activity-tracker/)
 
 For more information, see [this list of boards](https://tinygo.org/microcontrollers/). Pull requests for additional support are welcome!
 
@@ -114,3 +127,5 @@ The original reasoning was: if [Python](https://micropython.org/) can run on mic
 ## License
 
 This project is licensed under the BSD 3-clause license, just like the [Go project](https://golang.org/LICENSE) itself.
+
+Some code has been copied from the LLVM project and is therefore licensed under [a variant of the Apache 2.0 license](http://releases.llvm.org/9.0.0/LICENSE.TXT). This has been clearly indicated in the header of these files.

azure-pipelines.yml

@@ -0,0 +1,74 @@
+# Avoid lengthy LLVM rebuilds on each newly pushed branch. Pull requests will
+# be built anyway.
+trigger:
+- master
+- dev
+
+jobs:
+- job: Build
+  timeoutInMinutes: 240 # 4h
+  pool:
+    vmImage: 'VS2017-Win2016'
+  steps:
+    - checkout: self
+    - task: CacheBeta@0
+      displayName: Cache LLVM source
+      inputs:
+        key: llvm-source-9-windows-v0
+        path: llvm-project
+    - task: Bash@3
+      displayName: Download LLVM source
+      inputs:
+        targetType: inline
+        script: make llvm-source
+    - task: CacheBeta@0
+      displayName: Cache LLVM build
+      inputs:
+        key: llvm-build-9-windows-v0
+        path: llvm-build
+    - task: Bash@3
+      displayName: Build LLVM
+      inputs:
+        targetType: inline
+        script: |
+          if [ ! -f llvm-build/lib/liblldELF.a ]
+          then
+            choco install ninja
+            # LLVM 9 cannot be built with MinGW 8.
+            # For details: https://reviews.llvm.org/D70266
+            choco uninstall mingw
+            choco install mingw --version=7.3.0
+            make llvm-build
+          fi
+    - task: Bash@3
+      displayName: Install QEMU
+      inputs:
+        targetType: inline
+        script: choco install qemu
+    - task: Bash@3
+      displayName: Test TinyGo
+      inputs:
+        targetType: inline
+        script: |
+          export PATH="/c/Go1.13/bin:$PATH:./llvm-build/bin:/c/Program Files/qemu"
+          unset GOROOT
+          make test
+    - task: Bash@3
+      displayName: Build TinyGo release tarball
+      inputs:
+        targetType: inline
+        script: |
+          export PATH="/c/Go1.13/bin:$PATH:./llvm-build/bin:/c/Program Files/qemu"
+          unset GOROOT
+          make release -j4
+    - publish: $(System.DefaultWorkingDirectory)/build/release.tar.gz
+      displayName: Publish tarball as artifact
+      artifact: tinygo.windows-amd64.tar.gz
+    - task: Bash@3
+      displayName: Smoke tests
+      inputs:
+        targetType: inline
+        script: |
+          export PATH="/c/Go1.13/bin:$PATH:./llvm-build/bin:/c/Program Files/qemu"
+          unset GOROOT
+          make smoketest TINYGO=build/tinygo AVR=0

builder/ar.go

@@ -0,0 +1,182 @@
+package builder
+
+import (
+	"bytes"
+	"debug/elf"
+	"encoding/binary"
+	"errors"
+	"io"
+	"os"
+	"path/filepath"
+	"time"
+
+	"github.com/blakesmith/ar"
+)
+
+// makeArchive creates an arcive for static linking from a list of object files
+// given as a parameter. It is equivalent to the following command:
+//
+//     ar -rcs <archivePath> <objs...>
+func makeArchive(archivePath string, objs []string) error {
+	// Open the archive file.
+	arfile, err := os.Create(archivePath)
+	if err != nil {
+		return err
+	}
+	defer arfile.Close()
+	arwriter := ar.NewWriter(arfile)
+	err = arwriter.WriteGlobalHeader()
+	if err != nil {
+		return &os.PathError{"write ar header", archivePath, err}
+	}
+
+	// Open all object files and read the symbols for the symbol table.
+	symbolTable := []struct {
+		name      string // symbol name
+		fileIndex int    // index into objfiles
+	}{}
+	objfiles := make([]struct {
+		file          *os.File
+		archiveOffset int32
+	}, len(objs))
+	for i, objpath := range objs {
+		objfile, err := os.Open(objpath)
+		if err != nil {
+			return err
+		}
+		objfiles[i].file = objfile
+
+		// Read the symbols and add them to the symbol table.
+		dbg, err := elf.NewFile(objfile)
+		if err != nil {
+			return err
+		}
+		symbols, err := dbg.Symbols()
+		if err != nil {
+			return err
+		}
+		for _, symbol := range symbols {
+			bind := elf.ST_BIND(symbol.Info)
+			if bind != elf.STB_GLOBAL && bind != elf.STB_WEAK {
+				// Don't include local symbols (STB_LOCAL).
+				continue
+			}
+			if elf.ST_TYPE(symbol.Info) != elf.STT_FUNC {
+				// Not a function.
+				// TODO: perhaps globals variables should also be included?
+				continue
+			}
+			// Include in archive.
+			symbolTable = append(symbolTable, struct {
+				name      string
+				fileIndex int
+			}{symbol.Name, i})
+		}
+	}
+
+	// Create the symbol table buffer.
+	// For some (sparse) details on the file format:
+	// https://en.wikipedia.org/wiki/Ar_(Unix)#System_V_(or_GNU)_variant
+	buf := &bytes.Buffer{}
+	binary.Write(buf, binary.BigEndian, int32(len(symbolTable)))
+	for range symbolTable {
+		// This is a placeholder index, it will be updated after all files have
+		// been written to the archive (see the end of this function).
+		err = binary.Write(buf, binary.BigEndian, int32(0))
+		if err != nil {
+			return err
+		}
+	}
+	for _, sym := range symbolTable {
+		_, err := buf.Write([]byte(sym.name + "\x00"))
+		if err != nil {
+			return err
+		}
+	}
+	for buf.Len()%2 != 0 {
+		// The symbol table must be aligned.
+		// This appears to be required by lld.
+		buf.WriteByte(0)
+	}
+
+	// Write the symbol table.
+	err = arwriter.WriteHeader(&ar.Header{
+		Name:    "/",
+		ModTime: time.Unix(0, 0),
+		Uid:     0,
+		Gid:     0,
+		Mode:    0,
+		Size:    int64(buf.Len()),
+	})
+	if err != nil {
+		return err
+	}
+
+	// Keep track of the start of the symbol table.
+	symbolTableStart, err := arfile.Seek(0, os.SEEK_CUR)
+	if err != nil {
+		return err
+	}
+
+	// Write symbol table contents.
+	_, err = arfile.Write(buf.Bytes())
+	if err != nil {
+		return err
+	}
+
+	// Add all object files to the archive.
+	for i, objfile := range objfiles {
+		// Store the start index, for when we'll update the symbol table with
+		// the correct file start indices.
+		offset, err := arfile.Seek(0, os.SEEK_CUR)
+		if err != nil {
+			return err
+		}
+		if int64(int32(offset)) != offset {
+			return errors.New("large archives (4GB+) not supported: " + archivePath)
+		}
+		objfiles[i].archiveOffset = int32(offset)
+
+		// Write the file header.
+		st, err := objfile.file.Stat()
+		if err != nil {
+			return err
+		}
+		err = arwriter.WriteHeader(&ar.Header{
+			Name:    filepath.Base(objfile.file.Name()),
+			ModTime: time.Unix(0, 0),
+			Uid:     0,
+			Gid:     0,
+			Mode:    0644,
+			Size:    st.Size(),
+		})
+		if err != nil {
+			return err
+		}
+
+		// Copy the file contents into the archive.
+		n, err := io.Copy(arwriter, objfile.file)
+		if err != nil {
+			return err
+		}
+		if n != st.Size() {
+			return errors.New("file modified during ar creation: " + archivePath)
+		}
+
+		// File is not needed anymore.
+		objfile.file.Close()
+	}
+
+	// Create symbol indices.
+	indicesBuf := &bytes.Buffer{}
+	for _, sym := range symbolTable {
+		err = binary.Write(indicesBuf, binary.BigEndian, objfiles[sym.fileIndex].archiveOffset)
+		if err != nil {
+			return err
+		}
+	}
+
+	// Overwrite placeholder indices.
+	_, err = arfile.WriteAt(indicesBuf.Bytes(), symbolTableStart+4)
+	return err
+}

builder/build.go

@@ -0,0 +1,217 @@
+// Package builder is the compiler driver of TinyGo. It takes in a package name
+// and an output path, and outputs an executable. It manages the entire
+// compilation pipeline in between.
+package builder
+
+import (
+	"errors"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+
+	"github.com/tinygo-org/tinygo/compileopts"
+	"github.com/tinygo-org/tinygo/compiler"
+	"github.com/tinygo-org/tinygo/goenv"
+	"github.com/tinygo-org/tinygo/interp"
+)
+
+// Build performs a single package to executable Go build. It takes in a package
+// name, an output path, and set of compile options and from that it manages the
+// whole compilation process.
+//
+// The error value may be of type *MultiError. Callers will likely want to check
+// for this case and print such errors individually.
+func Build(pkgName, outpath string, config *compileopts.Config, action func(string) error) error {
+	c, err := compiler.NewCompiler(pkgName, config)
+	if err != nil {
+		return err
+	}
+
+	// Compile Go code to IR.
+	errs := c.Compile(pkgName)
+	if len(errs) != 0 {
+		return newMultiError(errs)
+	}
+	if config.Options.PrintIR {
+		fmt.Println("; Generated LLVM IR:")
+		fmt.Println(c.IR())
+	}
+	if err := c.Verify(); err != nil {
+		return errors.New("verification error after IR construction")
+	}
+
+	err = interp.Run(c.Module(), config.DumpSSA())
+	if err != nil {
+		return err
+	}
+	if err := c.Verify(); err != nil {
+		return errors.New("verification error after interpreting runtime.initAll")
+	}
+
+	if config.GOOS() != "darwin" {
+		c.ApplyFunctionSections() // -ffunction-sections
+	}
+
+	// Browsers cannot handle external functions that have type i64 because it
+	// cannot be represented exactly in JavaScript (JS only has doubles). To
+	// keep functions interoperable, pass int64 types as pointers to
+	// stack-allocated values.
+	// Use -wasm-abi=generic to disable this behaviour.
+	if config.Options.WasmAbi == "js" && strings.HasPrefix(config.Triple(), "wasm") {
+		err := c.ExternalInt64AsPtr()
+		if err != nil {
+			return err
+		}
+	}
+
+	// Optimization levels here are roughly the same as Clang, but probably not
+	// exactly.
+	errs = nil
+	switch config.Options.Opt {
+	case "none:", "0":
+		errs = c.Optimize(0, 0, 0) // -O0
+	case "1":
+		errs = c.Optimize(1, 0, 0) // -O1
+	case "2":
+		errs = c.Optimize(2, 0, 225) // -O2
+	case "s":
+		errs = c.Optimize(2, 1, 225) // -Os
+	case "z":
+		errs = c.Optimize(2, 2, 5) // -Oz, default
+	default:
+		errs = []error{errors.New("unknown optimization level: -opt=" + config.Options.Opt)}
+	}
+	if len(errs) > 0 {
+		return newMultiError(errs)
+	}
+	if err := c.Verify(); err != nil {
+		return errors.New("verification failure after LLVM optimization passes")
+	}
+
+	// On the AVR, pointers can point either to flash or to RAM, but we don't
+	// know. As a temporary fix, load all global variables in RAM.
+	// In the future, there should be a compiler pass that determines which
+	// pointers are flash and which are in RAM so that pointers can have a
+	// correct address space parameter (address space 1 is for flash).
+	if strings.HasPrefix(config.Triple(), "avr") {
+		c.NonConstGlobals()
+		if err := c.Verify(); err != nil {
+			return errors.New("verification error after making all globals non-constant on AVR")
+		}
+	}
+
+	// Generate output.
+	outext := filepath.Ext(outpath)
+	switch outext {
+	case ".o":
+		return c.EmitObject(outpath)
+	case ".bc":
+		return c.EmitBitcode(outpath)
+	case ".ll":
+		return c.EmitText(outpath)
+	default:
+		// Act as a compiler driver.
+
+		// Create a temporary directory for intermediary files.
+		dir, err := ioutil.TempDir("", "tinygo")
+		if err != nil {
+			return err
+		}
+		defer os.RemoveAll(dir)
+
+		// Write the object file.
+		objfile := filepath.Join(dir, "main.o")
+		err = c.EmitObject(objfile)
+		if err != nil {
+			return err
+		}
+
+		// Load builtins library from the cache, possibly compiling it on the
+		// fly.
+		var librt string
+		if config.Target.RTLib == "compiler-rt" {
+			librt, err = loadBuiltins(config.Triple())
+			if err != nil {
+				return err
+			}
+		}
+
+		// Prepare link command.
+		executable := filepath.Join(dir, "main")
+		tmppath := executable // final file
+		ldflags := append(config.LDFlags(), "-o", executable, objfile)
+		if config.Target.RTLib == "compiler-rt" {
+			ldflags = append(ldflags, librt)
+		}
+
+		// Compile extra files.
+		root := goenv.Get("TINYGOROOT")
+		for i, path := range config.ExtraFiles() {
+			abspath := filepath.Join(root, path)
+			outpath := filepath.Join(dir, "extra-"+strconv.Itoa(i)+"-"+filepath.Base(path)+".o")
+			err := runCCompiler(config.Target.Compiler, append(config.CFlags(), "-c", "-o", outpath, abspath)...)
+			if err != nil {
+				return &commandError{"failed to build", path, err}
+			}
+			ldflags = append(ldflags, outpath)
+		}
+
+		// Compile C files in packages.
+		for i, pkg := range c.Packages() {
+			for _, file := range pkg.CFiles {
+				path := filepath.Join(pkg.Package.Dir, file)
+				outpath := filepath.Join(dir, "pkg"+strconv.Itoa(i)+"-"+file+".o")
+				err := runCCompiler(config.Target.Compiler, append(config.CFlags(), "-c", "-o", outpath, path)...)
+				if err != nil {
+					return &commandError{"failed to build", path, err}
+				}
+				ldflags = append(ldflags, outpath)
+			}
+		}
+
+		// Link the object files together.
+		err = link(config.Target.Linker, ldflags...)
+		if err != nil {
+			return &commandError{"failed to link", executable, err}
+		}
+
+		if config.Options.PrintSizes == "short" || config.Options.PrintSizes == "full" {
+			sizes, err := loadProgramSize(executable)
+			if err != nil {
+				return err
+			}
+			if config.Options.PrintSizes == "short" {
+				fmt.Printf("   code    data     bss |   flash     ram\n")
+				fmt.Printf("%7d %7d %7d | %7d %7d\n", sizes.Code, sizes.Data, sizes.BSS, sizes.Code+sizes.Data, sizes.Data+sizes.BSS)
+			} else {
+				fmt.Printf("   code  rodata    data     bss |   flash     ram | package\n")
+				for _, name := range sizes.sortedPackageNames() {
+					pkgSize := sizes.Packages[name]
+					fmt.Printf("%7d %7d %7d %7d | %7d %7d | %s\n", pkgSize.Code, pkgSize.ROData, pkgSize.Data, pkgSize.BSS, pkgSize.Flash(), pkgSize.RAM(), name)
+				}
+				fmt.Printf("%7d %7d %7d %7d | %7d %7d | (sum)\n", sizes.Sum.Code, sizes.Sum.ROData, sizes.Sum.Data, sizes.Sum.BSS, sizes.Sum.Flash(), sizes.Sum.RAM())
+				fmt.Printf("%7d       - %7d %7d | %7d %7d | (all)\n", sizes.Code, sizes.Data, sizes.BSS, sizes.Code+sizes.Data, sizes.Data+sizes.BSS)
+			}
+		}
+
+		// Get an Intel .hex file or .bin file from the .elf file.
+		if outext == ".hex" || outext == ".bin" || outext == ".gba" {
+			tmppath = filepath.Join(dir, "main"+outext)
+			err := objcopy(executable, tmppath)
+			if err != nil {
+				return err
+			}
+		} else if outext == ".uf2" {
+			// Get UF2 from the .elf file.
+			tmppath = filepath.Join(dir, "main"+outext)
+			err := convertELFFileToUF2File(executable, tmppath)
+			if err != nil {
+				return err
+			}
+		}
+		return action(tmppath)
+	}
+}

builder/buildcache.go

@@ -1,20 +1,13 @@
-package main
+package builder
 
 import (
 	"io"
 	"os"
 	"path/filepath"
 	"time"
-)
 
-// Get the cache directory, usually ~/.cache/tinygo
-func cacheDir() string {
-	dir, err := os.UserCacheDir()
-	if err != nil {
-		panic("could not find cache dir: " + err.Error())
-	}
-	return filepath.Join(dir, "tinygo")
-}
+	"github.com/tinygo-org/tinygo/goenv"
+)
 
 // Return the newest timestamp of all the file paths passed in. Used to check
 // for stale caches.
@@ -41,8 +34,7 @@ func cacheTimestamp(paths []string) (time.Time, error) {
 // TODO: the configKey is currently ignored. It is supposed to be used as extra
 // data for the cache key, like the compiler version and arguments.
 func cacheLoad(name, configKey string, sourceFiles []string) (string, error) {
-	dir := cacheDir()
-	cachepath := filepath.Join(dir, name)
+	cachepath := filepath.Join(goenv.Get("GOCACHE"), name)
 	cacheStat, err := os.Stat(cachepath)
 	if os.IsNotExist(err) {
 		return "", nil // does not exist
@@ -76,7 +68,7 @@ func cacheStore(tmppath, name, configKey string, sourceFiles []string) (string,
 
 	// TODO: check the config key
 
-	dir := cacheDir()
+	dir := goenv.Get("GOCACHE")
 	err := os.MkdirAll(dir, 0777)
 	if err != nil {
 		return "", err
@@ -117,10 +109,10 @@ func moveFile(src, dst string) error {
 		return err
 	}
 
-	err = os.Rename(dst+".tmp", dst)
+	err = outf.Close()
 	if err != nil {
 		return err
 	}
 
-	return outf.Close()
+	return os.Rename(dst+".tmp", dst)
 }

builder/builtins.go

@@ -1,15 +1,12 @@
-package main
+package builder
 
 import (
-	"errors"
-	"io"
 	"io/ioutil"
 	"os"
 	"path/filepath"
 	"strings"
-	"time"
 
-	"github.com/blakesmith/ar"
+	"github.com/tinygo-org/tinygo/goenv"
 )
 
 // These are the GENERIC_SOURCES according to CMakeList.txt.
@@ -169,13 +166,13 @@ func builtinFiles(target string) []string {
 
 // builtinsDir returns the directory where the sources for compiler-rt are kept.
 func builtinsDir() string {
-	return filepath.Join(sourceDir(), "lib", "compiler-rt", "lib", "builtins")
+	return filepath.Join(goenv.Get("TINYGOROOT"), "lib", "compiler-rt", "lib", "builtins")
 }
 
 // Get the builtins archive, possibly generating it as needed.
 func loadBuiltins(target string) (path string, err error) {
 	// Try to load a precompiled compiler-rt library.
-	precompiledPath := filepath.Join(sourceDir(), "pkg", target, "compiler-rt.a")
+	precompiledPath := filepath.Join(goenv.Get("TINYGOROOT"), "pkg", target, "compiler-rt.a")
 	if _, err := os.Stat(precompiledPath); err == nil {
 		// Found a precompiled compiler-rt for this OS/architecture. Return the
 		// path directly.
@@ -196,7 +193,7 @@ func loadBuiltins(target string) (path string, err error) {
 	}
 
 	var cachepath string
-	err = compileBuiltins(target, func(path string) error {
+	err = CompileBuiltins(target, func(path string) error {
 		path, err := cacheStore(path, outfile, commands["clang"][0], srcs)
 		cachepath = path
 		return err
@@ -204,11 +201,11 @@ func loadBuiltins(target string) (path string, err error) {
 	return cachepath, err
 }
 
-// compileBuiltins compiles builtins from compiler-rt into a static library.
+// CompileBuiltins compiles builtins from compiler-rt into a static library.
 // When it succeeds, it will call the callback with the resulting path. The path
 // will be removed after callback returns. If callback returns an error, this is
 // passed through to the return value of this function.
-func compileBuiltins(target string, callback func(path string) error) error {
+func CompileBuiltins(target string, callback func(path string) error) error {
 	builtinsDir := builtinsDir()
 
 	builtins := builtinFiles(target)
@@ -239,56 +236,25 @@ func compileBuiltins(target string, callback func(path string) error) error {
 		// Note: -fdebug-prefix-map is necessary to make the output archive
 		// reproducible. Otherwise the temporary directory is stored in the
 		// archive itself, which varies each run.
-		err := execCommand(commands["clang"], "-c", "-Oz", "-g", "-Werror", "-Wall", "-std=c11", "-fshort-enums", "-nostdlibinc", "-ffunction-sections", "-fdata-sections", "--target="+target, "-fdebug-prefix-map="+dir+"="+remapDir, "-o", objpath, srcpath)
+		args := []string{"-c", "-Oz", "-g", "-Werror", "-Wall", "-std=c11", "-fshort-enums", "-nostdlibinc", "-ffunction-sections", "-fdata-sections", "-Wno-macro-redefined", "--target=" + target, "-fdebug-prefix-map=" + dir + "=" + remapDir}
+		if strings.HasPrefix(target, "riscv32-") {
+			args = append(args, "-march=rv32imac", "-mabi=ilp32", "-fforce-enable-int128")
+		}
+		err := runCCompiler("clang", append(args, "-o", objpath, srcpath)...)
 		if err != nil {
 			return &commandError{"failed to build", srcpath, err}
 		}
 	}
 
-	// Put all builtins in an archive to link as a static library.
-	// Note: this does not create a symbol index, but ld.lld doesn't seem to
-	// care.
+	// Put all the object files in a single archive. This archive file will be
+	// used to statically link compiler-rt.
 	arpath := filepath.Join(dir, "librt.a")
-	arfile, err := os.Create(arpath)
+	err = makeArchive(arpath, objs)
 	if err != nil {
 		return err
 	}
-	defer arfile.Close()
-	arwriter := ar.NewWriter(arfile)
-	err = arwriter.WriteGlobalHeader()
-	if err != nil {
-		return &os.PathError{"write ar header", arpath, err}
-	}
-	for _, objpath := range objs {
-		name := filepath.Base(objpath)
-		objfile, err := os.Open(objpath)
-		if err != nil {
-			return err
-		}
-		defer objfile.Close()
-		st, err := objfile.Stat()
-		if err != nil {
-			return err
-		}
-		arwriter.WriteHeader(&ar.Header{
-			Name:    name,
-			ModTime: time.Unix(0, 0),
-			Uid:     0,
-			Gid:     0,
-			Mode:    0644,
-			Size:    st.Size(),
-		})
-		n, err := io.Copy(arwriter, objfile)
-		if err != nil {
-			return err
-		}
-		if n != st.Size() {
-			return errors.New("file modified during ar creation: " + arpath)
-		}
-	}
 
 	// Give the caller the resulting file. The callback must copy the file,
 	// because after it returns the temporary directory will be removed.
-	arfile.Close()
 	return callback(arpath)
 }

builder/cc1as.cpp

@@ -0,0 +1,504 @@
+// +build byollvm
+
+//===-- cc1as.cpp - Clang Assembler  --------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This is the entry point to the clang -cc1as functionality, which implements
+// the direct interface to the LLVM MC based assembler.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/Diagnostic.h"
+#include "clang/Basic/DiagnosticOptions.h"
+#include "clang/Driver/DriverDiagnostic.h"
+#include "clang/Driver/Options.h"
+#include "clang/Frontend/FrontendDiagnostic.h"
+#include "clang/Frontend/TextDiagnosticPrinter.h"
+#include "clang/Frontend/Utils.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/StringSwitch.h"
+#include "llvm/ADT/Triple.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/MC/MCAsmBackend.h"
+#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCCodeEmitter.h"
+#include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCInstrInfo.h"
+#include "llvm/MC/MCObjectFileInfo.h"
+#include "llvm/MC/MCObjectWriter.h"
+#include "llvm/MC/MCParser/MCAsmParser.h"
+#include "llvm/MC/MCParser/MCTargetAsmParser.h"
+#include "llvm/MC/MCRegisterInfo.h"
+#include "llvm/MC/MCSectionMachO.h"
+#include "llvm/MC/MCStreamer.h"
+#include "llvm/MC/MCSubtargetInfo.h"
+#include "llvm/MC/MCTargetOptions.h"
+#include "llvm/Option/Arg.h"
+#include "llvm/Option/ArgList.h"
+#include "llvm/Option/OptTable.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/FormattedStream.h"
+#include "llvm/Support/Host.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Path.h"
+#include "llvm/Support/Signals.h"
+#include "llvm/Support/SourceMgr.h"
+#include "llvm/Support/TargetRegistry.h"
+#include "llvm/Support/TargetSelect.h"
+#include "llvm/Support/Timer.h"
+#include "llvm/Support/raw_ostream.h"
+#include <memory>
+#include <system_error>
+using namespace clang;
+using namespace clang::driver;
+using namespace clang::driver::options;
+using namespace llvm;
+using namespace llvm::opt;
+
+#include "cc1as.h"
+
+bool AssemblerInvocation::CreateFromArgs(AssemblerInvocation &Opts,
+                                         ArrayRef<const char *> Argv,
+                                         DiagnosticsEngine &Diags) {
+  bool Success = true;
+
+  // Parse the arguments.
+  std::unique_ptr<OptTable> OptTbl(createDriverOptTable());
+
+  const unsigned IncludedFlagsBitmask = options::CC1AsOption;
+  unsigned MissingArgIndex, MissingArgCount;
+  InputArgList Args = OptTbl->ParseArgs(Argv, MissingArgIndex, MissingArgCount,
+                                        IncludedFlagsBitmask);
+
+  // Check for missing argument error.
+  if (MissingArgCount) {
+    Diags.Report(diag::err_drv_missing_argument)
+        << Args.getArgString(MissingArgIndex) << MissingArgCount;
+    Success = false;
+  }
+
+  // Issue errors on unknown arguments.
+  for (const Arg *A : Args.filtered(OPT_UNKNOWN)) {
+    auto ArgString = A->getAsString(Args);
+    std::string Nearest;
+    if (OptTbl->findNearest(ArgString, Nearest, IncludedFlagsBitmask) > 1)
+      Diags.Report(diag::err_drv_unknown_argument) << ArgString;
+    else
+      Diags.Report(diag::err_drv_unknown_argument_with_suggestion)
+          << ArgString << Nearest;
+    Success = false;
+  }
+
+  // Construct the invocation.
+
+  // Target Options
+  Opts.Triple = llvm::Triple::normalize(Args.getLastArgValue(OPT_triple));
+  Opts.CPU = Args.getLastArgValue(OPT_target_cpu);
+  Opts.Features = Args.getAllArgValues(OPT_target_feature);
+
+  // Use the default target triple if unspecified.
+  if (Opts.Triple.empty())
+    Opts.Triple = llvm::sys::getDefaultTargetTriple();
+
+  // Language Options
+  Opts.IncludePaths = Args.getAllArgValues(OPT_I);
+  Opts.NoInitialTextSection = Args.hasArg(OPT_n);
+  Opts.SaveTemporaryLabels = Args.hasArg(OPT_msave_temp_labels);
+  // Any DebugInfoKind implies GenDwarfForAssembly.
+  Opts.GenDwarfForAssembly = Args.hasArg(OPT_debug_info_kind_EQ);
+
+  if (const Arg *A = Args.getLastArg(OPT_compress_debug_sections,
+                                     OPT_compress_debug_sections_EQ)) {
+    if (A->getOption().getID() == OPT_compress_debug_sections) {
+      // TODO: be more clever about the compression type auto-detection
+      Opts.CompressDebugSections = llvm::DebugCompressionType::GNU;
+    } else {
+      Opts.CompressDebugSections =
+          llvm::StringSwitch<llvm::DebugCompressionType>(A->getValue())
+              .Case("none", llvm::DebugCompressionType::None)
+              .Case("zlib", llvm::DebugCompressionType::Z)
+              .Case("zlib-gnu", llvm::DebugCompressionType::GNU)
+              .Default(llvm::DebugCompressionType::None);
+    }
+  }
+
+  Opts.RelaxELFRelocations = Args.hasArg(OPT_mrelax_relocations);
+  Opts.DwarfVersion = getLastArgIntValue(Args, OPT_dwarf_version_EQ, 2, Diags);
+  Opts.DwarfDebugFlags = Args.getLastArgValue(OPT_dwarf_debug_flags);
+  Opts.DwarfDebugProducer = Args.getLastArgValue(OPT_dwarf_debug_producer);
+  Opts.DebugCompilationDir = Args.getLastArgValue(OPT_fdebug_compilation_dir);
+  Opts.MainFileName = Args.getLastArgValue(OPT_main_file_name);
+
+  for (const auto &Arg : Args.getAllArgValues(OPT_fdebug_prefix_map_EQ))
+    Opts.DebugPrefixMap.insert(StringRef(Arg).split('='));
+
+  // Frontend Options
+  if (Args.hasArg(OPT_INPUT)) {
+    bool First = true;
+    for (const Arg *A : Args.filtered(OPT_INPUT)) {
+      if (First) {
+        Opts.InputFile = A->getValue();
+        First = false;
+      } else {
+        Diags.Report(diag::err_drv_unknown_argument) << A->getAsString(Args);
+        Success = false;
+      }
+    }
+  }
+  Opts.LLVMArgs = Args.getAllArgValues(OPT_mllvm);
+  Opts.OutputPath = Args.getLastArgValue(OPT_o);
+  Opts.SplitDwarfOutput = Args.getLastArgValue(OPT_split_dwarf_output);
+  if (Arg *A = Args.getLastArg(OPT_filetype)) {
+    StringRef Name = A->getValue();
+    unsigned OutputType = StringSwitch<unsigned>(Name)
+      .Case("asm", FT_Asm)
+      .Case("null", FT_Null)
+      .Case("obj", FT_Obj)
+      .Default(~0U);
+    if (OutputType == ~0U) {
+      Diags.Report(diag::err_drv_invalid_value) << A->getAsString(Args) << Name;
+      Success = false;
+    } else
+      Opts.OutputType = FileType(OutputType);
+  }
+  Opts.ShowHelp = Args.hasArg(OPT_help);
+  Opts.ShowVersion = Args.hasArg(OPT_version);
+
+  // Transliterate Options
+  Opts.OutputAsmVariant =
+      getLastArgIntValue(Args, OPT_output_asm_variant, 0, Diags);
+  Opts.ShowEncoding = Args.hasArg(OPT_show_encoding);
+  Opts.ShowInst = Args.hasArg(OPT_show_inst);
+
+  // Assemble Options
+  Opts.RelaxAll = Args.hasArg(OPT_mrelax_all);
+  Opts.NoExecStack = Args.hasArg(OPT_mno_exec_stack);
+  Opts.FatalWarnings = Args.hasArg(OPT_massembler_fatal_warnings);
+  Opts.RelocationModel = Args.getLastArgValue(OPT_mrelocation_model, "pic");
+  Opts.TargetABI = Args.getLastArgValue(OPT_target_abi);
+  Opts.IncrementalLinkerCompatible =
+      Args.hasArg(OPT_mincremental_linker_compatible);
+  Opts.SymbolDefs = Args.getAllArgValues(OPT_defsym);
+
+  // EmbedBitcode Option. If -fembed-bitcode is enabled, set the flag.
+  // EmbedBitcode behaves the same for all embed options for assembly files.
+  if (auto *A = Args.getLastArg(OPT_fembed_bitcode_EQ)) {
+    Opts.EmbedBitcode = llvm::StringSwitch<unsigned>(A->getValue())
+                            .Case("all", 1)
+                            .Case("bitcode", 1)
+                            .Case("marker", 1)
+                            .Default(0);
+  }
+
+  return Success;
+}
+
+static std::unique_ptr<raw_fd_ostream>
+getOutputStream(StringRef Path, DiagnosticsEngine &Diags, bool Binary) {
+  // Make sure that the Out file gets unlinked from the disk if we get a
+  // SIGINT.
+  if (Path != "-")
+    sys::RemoveFileOnSignal(Path);
+
+  std::error_code EC;
+  auto Out = llvm::make_unique<raw_fd_ostream>(
+      Path, EC, (Binary ? sys::fs::F_None : sys::fs::F_Text));
+  if (EC) {
+    Diags.Report(diag::err_fe_unable_to_open_output) << Path << EC.message();
+    return nullptr;
+  }
+
+  return Out;
+}
+
+bool ExecuteAssembler(AssemblerInvocation &Opts, DiagnosticsEngine &Diags) {
+  // Get the target specific parser.
+  std::string Error;
+  const Target *TheTarget = TargetRegistry::lookupTarget(Opts.Triple, Error);
+  if (!TheTarget)
+    return Diags.Report(diag::err_target_unknown_triple) << Opts.Triple;
+
+  ErrorOr<std::unique_ptr<MemoryBuffer>> Buffer =
+      MemoryBuffer::getFileOrSTDIN(Opts.InputFile);
+
+  if (std::error_code EC = Buffer.getError()) {
+    Error = EC.message();
+    return Diags.Report(diag::err_fe_error_reading) << Opts.InputFile;
+  }
+
+  SourceMgr SrcMgr;
+
+  // Tell SrcMgr about this buffer, which is what the parser will pick up.
+  unsigned BufferIndex = SrcMgr.AddNewSourceBuffer(std::move(*Buffer), SMLoc());
+
+  // Record the location of the include directories so that the lexer can find
+  // it later.
+  SrcMgr.setIncludeDirs(Opts.IncludePaths);
+
+  std::unique_ptr<MCRegisterInfo> MRI(TheTarget->createMCRegInfo(Opts.Triple));
+  assert(MRI && "Unable to create target register info!");
+
+  std::unique_ptr<MCAsmInfo> MAI(TheTarget->createMCAsmInfo(*MRI, Opts.Triple));
+  assert(MAI && "Unable to create target asm info!");
+
+  // Ensure MCAsmInfo initialization occurs before any use, otherwise sections
+  // may be created with a combination of default and explicit settings.
+  MAI->setCompressDebugSections(Opts.CompressDebugSections);
+
+  MAI->setRelaxELFRelocations(Opts.RelaxELFRelocations);
+
+  bool IsBinary = Opts.OutputType == AssemblerInvocation::FT_Obj;
+  if (Opts.OutputPath.empty())
+    Opts.OutputPath = "-";
+  std::unique_ptr<raw_fd_ostream> FDOS =
+      getOutputStream(Opts.OutputPath, Diags, IsBinary);
+  if (!FDOS)
+    return true;
+  std::unique_ptr<raw_fd_ostream> DwoOS;
+  if (!Opts.SplitDwarfOutput.empty())
+    DwoOS = getOutputStream(Opts.SplitDwarfOutput, Diags, IsBinary);
+
+  // FIXME: This is not pretty. MCContext has a ptr to MCObjectFileInfo and
+  // MCObjectFileInfo needs a MCContext reference in order to initialize itself.
+  std::unique_ptr<MCObjectFileInfo> MOFI(new MCObjectFileInfo());
+
+  MCContext Ctx(MAI.get(), MRI.get(), MOFI.get(), &SrcMgr);
+
+  bool PIC = false;
+  if (Opts.RelocationModel == "static") {
+    PIC = false;
+  } else if (Opts.RelocationModel == "pic") {
+    PIC = true;
+  } else {
+    assert(Opts.RelocationModel == "dynamic-no-pic" &&
+           "Invalid PIC model!");
+    PIC = false;
+  }
+
+  MOFI->InitMCObjectFileInfo(Triple(Opts.Triple), PIC, Ctx);
+  if (Opts.SaveTemporaryLabels)
+    Ctx.setAllowTemporaryLabels(false);
+  if (Opts.GenDwarfForAssembly)
+    Ctx.setGenDwarfForAssembly(true);
+  if (!Opts.DwarfDebugFlags.empty())
+    Ctx.setDwarfDebugFlags(StringRef(Opts.DwarfDebugFlags));
+  if (!Opts.DwarfDebugProducer.empty())
+    Ctx.setDwarfDebugProducer(StringRef(Opts.DwarfDebugProducer));
+  if (!Opts.DebugCompilationDir.empty())
+    Ctx.setCompilationDir(Opts.DebugCompilationDir);
+  else {
+    // If no compilation dir is set, try to use the current directory.
+    SmallString<128> CWD;
+    if (!sys::fs::current_path(CWD))
+      Ctx.setCompilationDir(CWD);
+  }
+  if (!Opts.DebugPrefixMap.empty())
+    for (const auto &KV : Opts.DebugPrefixMap)
+      Ctx.addDebugPrefixMapEntry(KV.first, KV.second);
+  if (!Opts.MainFileName.empty())
+    Ctx.setMainFileName(StringRef(Opts.MainFileName));
+  Ctx.setDwarfVersion(Opts.DwarfVersion);
+  if (Opts.GenDwarfForAssembly)
+    Ctx.setGenDwarfRootFile(Opts.InputFile,
+                            SrcMgr.getMemoryBuffer(BufferIndex)->getBuffer());
+
+  // Build up the feature string from the target feature list.
+  std::string FS;
+  if (!Opts.Features.empty()) {
+    FS = Opts.Features[0];
+    for (unsigned i = 1, e = Opts.Features.size(); i != e; ++i)
+      FS += "," + Opts.Features[i];
+  }
+
+  std::unique_ptr<MCStreamer> Str;
+
+  std::unique_ptr<MCInstrInfo> MCII(TheTarget->createMCInstrInfo());
+  std::unique_ptr<MCSubtargetInfo> STI(
+      TheTarget->createMCSubtargetInfo(Opts.Triple, Opts.CPU, FS));
+
+  raw_pwrite_stream *Out = FDOS.get();
+  std::unique_ptr<buffer_ostream> BOS;
+
+  MCTargetOptions MCOptions;
+  MCOptions.ABIName = Opts.TargetABI;
+
+  // FIXME: There is a bit of code duplication with addPassesToEmitFile.
+  if (Opts.OutputType == AssemblerInvocation::FT_Asm) {
+    MCInstPrinter *IP = TheTarget->createMCInstPrinter(
+        llvm::Triple(Opts.Triple), Opts.OutputAsmVariant, *MAI, *MCII, *MRI);
+
+    std::unique_ptr<MCCodeEmitter> CE;
+    if (Opts.ShowEncoding)
+      CE.reset(TheTarget->createMCCodeEmitter(*MCII, *MRI, Ctx));
+    std::unique_ptr<MCAsmBackend> MAB(
+        TheTarget->createMCAsmBackend(*STI, *MRI, MCOptions));
+
+    auto FOut = llvm::make_unique<formatted_raw_ostream>(*Out);
+    Str.reset(TheTarget->createAsmStreamer(
+        Ctx, std::move(FOut), /*asmverbose*/ true,
+        /*useDwarfDirectory*/ true, IP, std::move(CE), std::move(MAB),
+        Opts.ShowInst));
+  } else if (Opts.OutputType == AssemblerInvocation::FT_Null) {
+    Str.reset(createNullStreamer(Ctx));
+  } else {
+    assert(Opts.OutputType == AssemblerInvocation::FT_Obj &&
+           "Invalid file type!");
+    if (!FDOS->supportsSeeking()) {
+      BOS = make_unique<buffer_ostream>(*FDOS);
+      Out = BOS.get();
+    }
+
+    std::unique_ptr<MCCodeEmitter> CE(
+        TheTarget->createMCCodeEmitter(*MCII, *MRI, Ctx));
+    std::unique_ptr<MCAsmBackend> MAB(
+        TheTarget->createMCAsmBackend(*STI, *MRI, MCOptions));
+    std::unique_ptr<MCObjectWriter> OW =
+        DwoOS ? MAB->createDwoObjectWriter(*Out, *DwoOS)
+              : MAB->createObjectWriter(*Out);
+
+    Triple T(Opts.Triple);
+    Str.reset(TheTarget->createMCObjectStreamer(
+        T, Ctx, std::move(MAB), std::move(OW), std::move(CE), *STI,
+        Opts.RelaxAll, Opts.IncrementalLinkerCompatible,
+        /*DWARFMustBeAtTheEnd*/ true));
+    Str.get()->InitSections(Opts.NoExecStack);
+  }
+
+  // When -fembed-bitcode is passed to clang_as, a 1-byte marker
+  // is emitted in __LLVM,__asm section if the object file is MachO format.
+  if (Opts.EmbedBitcode && Ctx.getObjectFileInfo()->getObjectFileType() ==
+                               MCObjectFileInfo::IsMachO) {
+    MCSection *AsmLabel = Ctx.getMachOSection(
+        "__LLVM", "__asm", MachO::S_REGULAR, 4, SectionKind::getReadOnly());
+    Str.get()->SwitchSection(AsmLabel);
+    Str.get()->EmitZeros(1);
+  }
+
+  // Assembly to object compilation should leverage assembly info.
+  Str->setUseAssemblerInfoForParsing(true);
+
+  bool Failed = false;
+
+  std::unique_ptr<MCAsmParser> Parser(
+      createMCAsmParser(SrcMgr, Ctx, *Str.get(), *MAI));
+
+  // FIXME: init MCTargetOptions from sanitizer flags here.
+  std::unique_ptr<MCTargetAsmParser> TAP(
+      TheTarget->createMCAsmParser(*STI, *Parser, *MCII, MCOptions));
+  if (!TAP)
+    Failed = Diags.Report(diag::err_target_unknown_triple) << Opts.Triple;
+
+  // Set values for symbols, if any.
+  for (auto &S : Opts.SymbolDefs) {
+    auto Pair = StringRef(S).split('=');
+    auto Sym = Pair.first;
+    auto Val = Pair.second;
+    int64_t Value;
+    // We have already error checked this in the driver.
+    Val.getAsInteger(0, Value);
+    Ctx.setSymbolValue(Parser->getStreamer(), Sym, Value);
+  }
+
+  if (!Failed) {
+    Parser->setTargetParser(*TAP.get());
+    Failed = Parser->Run(Opts.NoInitialTextSection);
+  }
+
+  // Close Streamer first.
+  // It might have a reference to the output stream.
+  Str.reset();
+  // Close the output stream early.
+  BOS.reset();
+  FDOS.reset();
+
+  // Delete output file if there were errors.
+  if (Failed) {
+    if (Opts.OutputPath != "-")
+      sys::fs::remove(Opts.OutputPath);
+    if (!Opts.SplitDwarfOutput.empty() && Opts.SplitDwarfOutput != "-")
+      sys::fs::remove(Opts.SplitDwarfOutput);
+  }
+
+  return Failed;
+}
+
+static void LLVMErrorHandler(void *UserData, const std::string &Message,
+                             bool GenCrashDiag) {
+  DiagnosticsEngine &Diags = *static_cast<DiagnosticsEngine*>(UserData);
+
+  Diags.Report(diag::err_fe_error_backend) << Message;
+
+  // We cannot recover from llvm errors.
+  exit(1);
+}
+
+int cc1as_main(ArrayRef<const char *> Argv, const char *Argv0, void *MainAddr) {
+  // Initialize targets and assembly printers/parsers.
+  InitializeAllTargetInfos();
+  InitializeAllTargetMCs();
+  InitializeAllAsmParsers();
+
+  // Construct our diagnostic client.
+  IntrusiveRefCntPtr<DiagnosticOptions> DiagOpts = new DiagnosticOptions();
+  TextDiagnosticPrinter *DiagClient
+    = new TextDiagnosticPrinter(errs(), &*DiagOpts);
+  DiagClient->setPrefix("clang -cc1as");
+  IntrusiveRefCntPtr<DiagnosticIDs> DiagID(new DiagnosticIDs());
+  DiagnosticsEngine Diags(DiagID, &*DiagOpts, DiagClient);
+
+  // Set an error handler, so that any LLVM backend diagnostics go through our
+  // error handler.
+  ScopedFatalErrorHandler FatalErrorHandler
+    (LLVMErrorHandler, static_cast<void*>(&Diags));
+
+  // Parse the arguments.
+  AssemblerInvocation Asm;
+  if (!AssemblerInvocation::CreateFromArgs(Asm, Argv, Diags))
+    return 1;
+
+  if (Asm.ShowHelp) {
+    std::unique_ptr<OptTable> Opts(driver::createDriverOptTable());
+    Opts->PrintHelp(llvm::outs(), "clang -cc1as [options] file...",
+                    "Clang Integrated Assembler",
+                    /*Include=*/driver::options::CC1AsOption, /*Exclude=*/0,
+                    /*ShowAllAliases=*/false);
+    return 0;
+  }
+
+  // Honor -version.
+  //
+  // FIXME: Use a better -version message?
+  if (Asm.ShowVersion) {
+    llvm::cl::PrintVersionMessage();
+    return 0;
+  }
+
+  // Honor -mllvm.
+  //
+  // FIXME: Remove this, one day.
+  if (!Asm.LLVMArgs.empty()) {
+    unsigned NumArgs = Asm.LLVMArgs.size();
+    auto Args = llvm::make_unique<const char*[]>(NumArgs + 2);
+    Args[0] = "clang (LLVM option parsing)";
+    for (unsigned i = 0; i != NumArgs; ++i)
+      Args[i + 1] = Asm.LLVMArgs[i].c_str();
+    Args[NumArgs + 1] = nullptr;
+    llvm::cl::ParseCommandLineOptions(NumArgs + 1, Args.get());
+  }
+
+  // Execute the invocation, unless there were parsing errors.
+  bool Failed = Diags.hasErrorOccurred() || ExecuteAssembler(Asm, Diags);
+
+  // If any timers were active but haven't been destroyed yet, print their
+  // results now.
+  TimerGroup::printAll(errs());
+
+  return !!Failed;
+}

builder/cc1as.h

@@ -0,0 +1,120 @@
+//===-- cc1as.h - Clang Assembler  ----------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This is the entry point to the clang -cc1as functionality, which implements
+// the direct interface to the LLVM MC based assembler.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ADT/ArrayRef.h"
+
+/// Helper class for representing a single invocation of the assembler.
+struct AssemblerInvocation {
+  /// @name Target Options
+  /// @{
+
+  /// The name of the target triple to assemble for.
+  std::string Triple;
+
+  /// If given, the name of the target CPU to determine which instructions
+  /// are legal.
+  std::string CPU;
+
+  /// The list of target specific features to enable or disable -- this should
+  /// be a list of strings starting with '+' or '-'.
+  std::vector<std::string> Features;
+
+  /// The list of symbol definitions.
+  std::vector<std::string> SymbolDefs;
+
+  /// @}
+  /// @name Language Options
+  /// @{
+
+  std::vector<std::string> IncludePaths;
+  unsigned NoInitialTextSection : 1;
+  unsigned SaveTemporaryLabels : 1;
+  unsigned GenDwarfForAssembly : 1;
+  unsigned RelaxELFRelocations : 1;
+  unsigned DwarfVersion;
+  std::string DwarfDebugFlags;
+  std::string DwarfDebugProducer;
+  std::string DebugCompilationDir;
+  std::map<const std::string, const std::string> DebugPrefixMap;
+  llvm::DebugCompressionType CompressDebugSections =
+      llvm::DebugCompressionType::None;
+  std::string MainFileName;
+  std::string SplitDwarfOutput;
+
+  /// @}
+  /// @name Frontend Options
+  /// @{
+
+  std::string InputFile;
+  std::vector<std::string> LLVMArgs;
+  std::string OutputPath;
+  enum FileType {
+    FT_Asm,  ///< Assembly (.s) output, transliterate mode.
+    FT_Null, ///< No output, for timing purposes.
+    FT_Obj   ///< Object file output.
+  };
+  FileType OutputType;
+  unsigned ShowHelp : 1;
+  unsigned ShowVersion : 1;
+
+  /// @}
+  /// @name Transliterate Options
+  /// @{
+
+  unsigned OutputAsmVariant;
+  unsigned ShowEncoding : 1;
+  unsigned ShowInst : 1;
+
+  /// @}
+  /// @name Assembler Options
+  /// @{
+
+  unsigned RelaxAll : 1;
+  unsigned NoExecStack : 1;
+  unsigned FatalWarnings : 1;
+  unsigned IncrementalLinkerCompatible : 1;
+  unsigned EmbedBitcode : 1;
+
+  /// The name of the relocation model to use.
+  std::string RelocationModel;
+
+  /// The ABI targeted by the backend. Specified using -target-abi. Empty
+  /// otherwise.
+  std::string TargetABI;
+
+  /// @}
+
+public:
+  AssemblerInvocation() {
+    Triple = "";
+    NoInitialTextSection = 0;
+    InputFile = "-";
+    OutputPath = "-";
+    OutputType = FT_Asm;
+    OutputAsmVariant = 0;
+    ShowInst = 0;
+    ShowEncoding = 0;
+    RelaxAll = 0;
+    NoExecStack = 0;
+    FatalWarnings = 0;
+    IncrementalLinkerCompatible = 0;
+    DwarfVersion = 0;
+    EmbedBitcode = 0;
+  }
+
+  static bool CreateFromArgs(AssemblerInvocation &Res,
+                             ArrayRef<const char *> Argv,
+                             DiagnosticsEngine &Diags);
+};
+
+bool ExecuteAssembler(AssemblerInvocation &Opts, DiagnosticsEngine &Diags);

builder/clang.cpp

@@ -0,0 +1,97 @@
+// +build byollvm
+
+#include <clang/Basic/DiagnosticOptions.h>
+#include <clang/CodeGen/CodeGenAction.h>
+#include <clang/Driver/Compilation.h>
+#include <clang/Driver/Driver.h>
+#include <clang/Frontend/CompilerInstance.h>
+#include <clang/Frontend/CompilerInvocation.h>
+#include <clang/Frontend/FrontendDiagnostic.h>
+#include <clang/Frontend/TextDiagnosticPrinter.h>
+#include <clang/FrontendTool/Utils.h>
+#include <llvm/ADT/IntrusiveRefCntPtr.h>
+#include <llvm/Option/Option.h>
+
+using namespace llvm;
+using namespace clang;
+
+#include "cc1as.h"
+
+// This file provides C wrappers for the builtin tools cc1 and cc1as
+// provided by Clang, and calls them as the driver would call them.
+
+extern "C" {
+
+bool tinygo_clang_driver(int argc, char **argv) {
+	std::vector<const char*> args(argv, argv + argc);
+
+	// The compiler invocation needs a DiagnosticsEngine so it can report problems
+	llvm::IntrusiveRefCntPtr<clang::DiagnosticOptions> DiagOpts = new clang::DiagnosticOptions();
+	clang::TextDiagnosticPrinter DiagnosticPrinter(llvm::errs(), &*DiagOpts);
+	clang::DiagnosticsEngine Diags(llvm::IntrusiveRefCntPtr<clang::DiagnosticIDs>(new clang::DiagnosticIDs()), &*DiagOpts, &DiagnosticPrinter, false);
+
+	// Create the clang driver
+	clang::driver::Driver TheDriver(args[0], llvm::sys::getDefaultTargetTriple(), Diags);
+
+	// Create the set of actions to perform
+	std::unique_ptr<clang::driver::Compilation> C(TheDriver.BuildCompilation(args));
+	if (!C) {
+		return false;
+	}
+	const clang::driver::JobList &Jobs = C->getJobs();
+
+	// There may be more than one job, for example for .S files
+	// (preprocessor + assembler).
+	for (auto Cmd : Jobs) {
+		// Select the tool: cc1 or cc1as.
+		const llvm::opt::ArgStringList &CCArgs = Cmd.getArguments();
+
+		if (strcmp(*CCArgs.data(), "-cc1") == 0) {
+			// This is the C frontend.
+			// Initialize a compiler invocation object from the clang (-cc1) arguments.
+			std::unique_ptr<clang::CompilerInstance> Clang(new clang::CompilerInstance());
+			bool success = clang::CompilerInvocation::CreateFromArgs(
+					Clang->getInvocation(),
+					const_cast<const char **>(CCArgs.data()),
+					const_cast<const char **>(CCArgs.data()) + CCArgs.size(),
+					Diags);
+			if (!success) {
+				return false;
+			}
+
+			// Create the actual diagnostics engine.
+			Clang->createDiagnostics();
+			if (!Clang->hasDiagnostics()) {
+				return false;
+			}
+
+			// Execute the frontend actions.
+			success = ExecuteCompilerInvocation(Clang.get());
+			if (!success) {
+				return false;
+			}
+
+		} else if (strcmp(*CCArgs.data(), "-cc1as") == 0) {
+			// This is the assembler frontend. Parse the arguments.
+			AssemblerInvocation Asm;
+			if (!AssemblerInvocation::CreateFromArgs(Asm, llvm::ArrayRef<const char*>(CCArgs).slice(1), Diags))
+				return false;
+
+			// Execute the invocation, unless there were parsing errors.
+			bool failed = Diags.hasErrorOccurred() || ExecuteAssembler(Asm, Diags);
+			if (failed) {
+				return false;
+			}
+
+		} else {
+			// Unknown tool, print the tool and exit.
+			fprintf(stderr, "unknown tool: %s\n", *CCArgs.data());
+			return false;
+		}
+	}
+
+	// Commands executed successfully.
+	return true;
+}
+
+} // extern "C"

builder/commands.go

@@ -0,0 +1,61 @@
+package builder
+
+import (
+	"errors"
+	"os"
+	"os/exec"
+	"runtime"
+	"strings"
+)
+
+// Commands lists command alternatives for various operating systems. These
+// commands may have a slightly different name across operating systems and
+// distributions or may not even exist in $PATH, in which case absolute paths
+// may be used.
+var commands = map[string][]string{
+	"clang":   {"clang-9"},
+	"ld.lld":  {"ld.lld-9", "ld.lld"},
+	"wasm-ld": {"wasm-ld-9", "wasm-ld"},
+}
+
+func init() {
+	// Add the path to a Homebrew-installed LLVM 9 for ease of use (no need to
+	// manually set $PATH).
+	if runtime.GOOS == "darwin" {
+		commands["clang"] = append(commands["clang"], "/usr/local/opt/llvm@9/bin/clang-9")
+		commands["ld.lld"] = append(commands["ld.lld"], "/usr/local/opt/llvm@9/bin/ld.lld")
+		commands["wasm-ld"] = append(commands["wasm-ld"], "/usr/local/opt/llvm@9/bin/wasm-ld")
+	}
+	// Add the path for when LLVM was installed with the installer from
+	// llvm.org, which by default doesn't add LLVM to the $PATH environment
+	// variable.
+	if runtime.GOOS == "windows" {
+		commands["clang"] = append(commands["clang"], "clang", "C:\\Program Files\\LLVM\\bin\\clang.exe")
+		commands["ld.lld"] = append(commands["ld.lld"], "lld", "C:\\Program Files\\LLVM\\bin\\lld.exe")
+		commands["wasm-ld"] = append(commands["wasm-ld"], "C:\\Program Files\\LLVM\\bin\\wasm-ld.exe")
+	}
+	// Add the path to the llvm90 installed from ports
+	if runtime.GOOS == "freebsd" {
+		commands["clang"] = append(commands["clang"], "/usr/local/llvm90/bin/clang-9")
+		commands["ld.lld"] = append(commands["ld.lld"], "/usr/local/llvm90/bin/ld.lld")
+		commands["wasm-ld"] = append(commands["wasm-ld"], "/usr/local/llvm90/bin/wasm-ld")
+	}
+}
+
+func execCommand(cmdNames []string, args ...string) error {
+	for _, cmdName := range cmdNames {
+		cmd := exec.Command(cmdName, args...)
+		cmd.Stdout = os.Stdout
+		cmd.Stderr = os.Stderr
+		err := cmd.Run()
+		if err != nil {
+			if err, ok := err.(*exec.Error); ok && (err.Err == exec.ErrNotFound || err.Err.Error() == "file does not exist") {
+				// this command was not found, try the next
+				continue
+			}
+			return err
+		}
+		return nil
+	}
+	return errors.New("none of these commands were found in your $PATH: " + strings.Join(cmdNames, " "))
+}

builder/compiler-builtin.go

@@ -0,0 +1,60 @@
+// +build byollvm
+
+package builder
+
+import (
+	"errors"
+	"os"
+	"os/exec"
+	"unsafe"
+
+	"github.com/tinygo-org/tinygo/goenv"
+)
+
+/*
+#cgo CXXFLAGS: -fno-rtti
+#include <stdbool.h>
+#include <stdlib.h>
+bool tinygo_clang_driver(int argc, char **argv);
+*/
+import "C"
+
+// runCCompiler invokes a C compiler with the given arguments.
+//
+// This version invokes the built-in Clang when trying to run the Clang compiler.
+func runCCompiler(command string, flags ...string) error {
+	switch command {
+	case "clang":
+		// Compile this with the internal Clang compiler.
+		headerPath := getClangHeaderPath(goenv.Get("TINYGOROOT"))
+		if headerPath == "" {
+			return errors.New("could not locate Clang headers")
+		}
+		flags = append(flags, "-I"+headerPath)
+		flags = append([]string{"tinygo:" + command}, flags...)
+		var cflag *C.char
+		buf := C.calloc(C.size_t(len(flags)), C.size_t(unsafe.Sizeof(cflag)))
+		cflags := (*[1 << 10]*C.char)(unsafe.Pointer(buf))[:len(flags):len(flags)]
+		for i, flag := range flags {
+			cflag := C.CString(flag)
+			cflags[i] = cflag
+			defer C.free(unsafe.Pointer(cflag))
+		}
+		ok := C.tinygo_clang_driver(C.int(len(flags)), (**C.char)(buf))
+		if !ok {
+			return errors.New("failed to compile using built-in clang")
+		}
+		return nil
+	default:
+		// Running some other compiler. Maybe it has been defined in the
+		// commands map (unlikely).
+		if cmdNames, ok := commands[command]; ok {
+			return execCommand(cmdNames, flags...)
+		}
+		// Alternatively, run the compiler directly.
+		cmd := exec.Command(command, flags...)
+		cmd.Stdout = os.Stdout
+		cmd.Stderr = os.Stderr
+		return cmd.Run()
+	}
+}

builder/compiler-external.go

@@ -0,0 +1,24 @@
+// +build !byollvm
+
+package builder
+
+// This file provides a way to a C compiler as an external command. See also:
+// clang-external.go
+
+import (
+	"os"
+	"os/exec"
+)
+
+// runCCompiler invokes a C compiler with the given arguments.
+//
+// This version always runs the compiler as an external command.
+func runCCompiler(command string, flags ...string) error {
+	if cmdNames, ok := commands[command]; ok {
+		return execCommand(cmdNames, flags...)
+	}
+	cmd := exec.Command(command, flags...)
+	cmd.Stdout = os.Stdout
+	cmd.Stderr = os.Stderr
+	return cmd.Run()
+}

builder/config.go

@@ -0,0 +1,39 @@
+package builder
+
+import (
+	"errors"
+	"fmt"
+
+	"github.com/tinygo-org/tinygo/compileopts"
+	"github.com/tinygo-org/tinygo/goenv"
+)
+
+// NewConfig builds a new Config object from a set of compiler options. It also
+// loads some information from the environment while doing that. For example, it
+// uses the currently active GOPATH (from the goenv package) to determine the Go
+// version to use.
+func NewConfig(options *compileopts.Options) (*compileopts.Config, error) {
+	spec, err := compileopts.LoadTarget(options.Target)
+	if err != nil {
+		return nil, err
+	}
+	goroot := goenv.Get("GOROOT")
+	if goroot == "" {
+		return nil, errors.New("cannot locate $GOROOT, please set it manually")
+	}
+	major, minor, err := getGorootVersion(goroot)
+	if err != nil {
+		return nil, fmt.Errorf("could not read version from GOROOT (%v): %v", goroot, err)
+	}
+	if major != 1 || (minor != 11 && minor != 12 && minor != 13) {
+		return nil, fmt.Errorf("requires go version 1.11, 1.12, or 1.13, got go%d.%d", major, minor)
+	}
+	clangHeaderPath := getClangHeaderPath(goenv.Get("TINYGOROOT"))
+	return &compileopts.Config{
+		Options:        options,
+		Target:         spec,
+		GoMinorVersion: minor,
+		ClangHeaders:   clangHeaderPath,
+		TestConfig:     options.TestConfig,
+	}, nil
+}

builder/env.go

@@ -0,0 +1,127 @@
+package builder
+
+import (
+	"errors"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"regexp"
+	"sort"
+	"strings"
+)
+
+// getGorootVersion returns the major and minor version for a given GOROOT path.
+// If the goroot cannot be determined, (0, 0) is returned.
+func getGorootVersion(goroot string) (major, minor int, err error) {
+	s, err := GorootVersionString(goroot)
+	if err != nil {
+		return 0, 0, err
+	}
+
+	if s == "" || s[:2] != "go" {
+		return 0, 0, errors.New("could not parse Go version: version does not start with 'go' prefix")
+	}
+
+	parts := strings.Split(s[2:], ".")
+	if len(parts) < 2 {
+		return 0, 0, errors.New("could not parse Go version: version has less than two parts")
+	}
+
+	// Ignore the errors, we don't really handle errors here anyway.
+	var trailing string
+	n, err := fmt.Sscanf(s, "go%d.%d%s", &major, &minor, &trailing)
+	if n == 2 && err == io.EOF {
+		// Means there were no trailing characters (i.e., not an alpha/beta)
+		err = nil
+	}
+	if err != nil {
+		return 0, 0, fmt.Errorf("failed to parse version: %s", err)
+	}
+	return
+}
+
+// GorootVersionString returns the version string as reported by the Go
+// toolchain for the given GOROOT path. It is usually of the form `go1.x.y` but
+// can have some variations (for beta releases, for example).
+func GorootVersionString(goroot string) (string, error) {
+	if data, err := ioutil.ReadFile(filepath.Join(
+		goroot, "src", "runtime", "internal", "sys", "zversion.go")); err == nil {
+
+		r := regexp.MustCompile("const TheVersion = `(.*)`")
+		matches := r.FindSubmatch(data)
+		if len(matches) != 2 {
+			return "", errors.New("Invalid go version output:\n" + string(data))
+		}
+
+		return string(matches[1]), nil
+
+	} else if data, err := ioutil.ReadFile(filepath.Join(goroot, "VERSION")); err == nil {
+		return string(data), nil
+
+	} else {
+		return "", err
+	}
+}
+
+// getClangHeaderPath returns the path to the built-in Clang headers. It tries
+// multiple locations, which should make it find the directory when installed in
+// various ways.
+func getClangHeaderPath(TINYGOROOT string) string {
+	// Check whether we're running from the source directory.
+	path := filepath.Join(TINYGOROOT, "llvm-project", "clang", "lib", "Headers")
+	if _, err := os.Stat(path); !os.IsNotExist(err) {
+		return path
+	}
+
+	// Check whether we're running from the installation directory.
+	path = filepath.Join(TINYGOROOT, "lib", "clang", "include")
+	if _, err := os.Stat(path); !os.IsNotExist(err) {
+		return path
+	}
+
+	// It looks like we are built with a system-installed LLVM. Do a last
+	// attempt: try to use Clang headers relative to the clang binary.
+	for _, cmdName := range commands["clang"] {
+		binpath, err := exec.LookPath(cmdName)
+		if err == nil {
+			// This should be the command that will also be used by
+			// execCommand. To avoid inconsistencies, make sure we use the
+			// headers relative to this command.
+			binpath, err = filepath.EvalSymlinks(binpath)
+			if err != nil {
+				// Unexpected.
+				return ""
+			}
+			// Example executable:
+			//     /usr/lib/llvm-9/bin/clang
+			// Example include path:
+			//     /usr/lib/llvm-9/lib/clang/9.0.1/include/
+			llvmRoot := filepath.Dir(filepath.Dir(binpath))
+			clangVersionRoot := filepath.Join(llvmRoot, "lib", "clang")
+			dirs, err := ioutil.ReadDir(clangVersionRoot)
+			if err != nil {
+				// Unexpected.
+				continue
+			}
+			dirnames := make([]string, len(dirs))
+			for i, d := range dirs {
+				dirnames[i] = d.Name()
+			}
+			sort.Strings(dirnames)
+			// Check for the highest version first.
+			for i := len(dirnames) - 1; i >= 0; i-- {
+				path := filepath.Join(clangVersionRoot, dirnames[i], "include")
+				_, err := os.Stat(filepath.Join(path, "stdint.h"))
+				if err == nil {
+					return path
+				}
+			}
+		}
+	}
+
+	// Could not find it.
+	return ""
+}

builder/error.go

@@ -0,0 +1,39 @@
+package builder
+
+// MultiError is a list of multiple errors (actually: diagnostics) returned
+// during LLVM IR generation.
+type MultiError struct {
+	Errs []error
+}
+
+func (e *MultiError) Error() string {
+	// Return the first error, to conform to the error interface. Clients should
+	// really do a type-assertion on *MultiError.
+	return e.Errs[0].Error()
+}
+
+// newMultiError returns a *MultiError if there is more than one error, or
+// returns that error directly when there is only one. Passing an empty slice
+// will lead to a panic.
+func newMultiError(errs []error) error {
+	switch len(errs) {
+	case 0:
+		panic("attempted to create empty MultiError")
+	case 1:
+		return errs[0]
+	default:
+		return &MultiError{errs}
+	}
+}
+
+// commandError is an error type to wrap os/exec.Command errors. This provides
+// some more information regarding what went wrong while running a command.
+type commandError struct {
+	Msg  string
+	File string
+	Err  error
+}
+
+func (e *commandError) Error() string {
+	return e.Msg + " " + e.File + ": " + e.Err.Error()
+}

builder/linker-builtin.go

@@ -1,6 +1,6 @@
 // +build byollvm
 
-package main
+package builder
 
 // This file provides a Link() function that uses the bundled lld if possible.
 
@@ -9,6 +9,8 @@ import (
 	"os"
 	"os/exec"
 	"unsafe"
+
+	"github.com/tinygo-org/tinygo/goenv"
 )
 
 /*
@@ -19,10 +21,10 @@ bool tinygo_link_wasm(int argc, char **argv);
 */
 import "C"
 
-// Link invokes a linker with the given name and flags.
+// link invokes a linker with the given name and flags.
 //
 // This version uses the built-in linker when trying to use lld.
-func Link(linker string, flags ...string) error {
+func link(linker string, flags ...string) error {
 	switch linker {
 	case "ld.lld":
 		flags = append([]string{"tinygo:" + linker}, flags...)
@@ -63,6 +65,7 @@ func Link(linker string, flags ...string) error {
 		cmd := exec.Command(linker, flags...)
 		cmd.Stdout = os.Stdout
 		cmd.Stderr = os.Stderr
+		cmd.Dir = goenv.Get("TINYGOROOT")
 		return cmd.Run()
 	}
 }

builder/linker-external.go

@@ -1,6 +1,6 @@
 // +build !byollvm
 
-package main
+package builder
 
 // This file provides a Link() function that always runs an external command. It
 // is provided for when tinygo is built without linking to liblld.
@@ -8,17 +8,20 @@ package main
 import (
 	"os"
 	"os/exec"
+
+	"github.com/tinygo-org/tinygo/goenv"
 )
 
-// Link invokes a linker with the given name and arguments.
+// link invokes a linker with the given name and arguments.
 //
 // This version always runs the linker as an external command.
-func Link(linker string, flags ...string) error {
+func link(linker string, flags ...string) error {
 	if cmdNames, ok := commands[linker]; ok {
 		return execCommand(cmdNames, flags...)
 	}
 	cmd := exec.Command(linker, flags...)
 	cmd.Stdout = os.Stdout
 	cmd.Stderr = os.Stderr
+	cmd.Dir = goenv.Get("TINYGOROOT")
 	return cmd.Run()
 }

builder/objcopy.go

@@ -1,4 +1,4 @@
-package main
+package builder
 
 import (
 	"debug/elf"
@@ -10,31 +10,31 @@ import (
 	"github.com/marcinbor85/gohex"
 )
 
-// ObjcopyError is an error returned by functions that act like objcopy.
-type ObjcopyError struct {
+// objcopyError is an error returned by functions that act like objcopy.
+type objcopyError struct {
 	Op  string
 	Err error
 }
 
-func (e ObjcopyError) Error() string {
+func (e objcopyError) Error() string {
 	if e.Err == nil {
 		return e.Op
 	}
 	return e.Op + ": " + e.Err.Error()
 }
 
-type ProgSlice []*elf.Prog
+type progSlice []*elf.Prog
 
-func (s ProgSlice) Len() int           { return len(s) }
-func (s ProgSlice) Less(i, j int) bool { return s[i].Paddr < s[j].Paddr }
-func (s ProgSlice) Swap(i, j int)      { s[i], s[j] = s[j], s[i] }
+func (s progSlice) Len() int           { return len(s) }
+func (s progSlice) Less(i, j int) bool { return s[i].Paddr < s[j].Paddr }
+func (s progSlice) Swap(i, j int)      { s[i], s[j] = s[j], s[i] }
 
-// ExtractROM extracts a firmware image and the first load address from the
+// extractROM extracts a firmware image and the first load address from the
 // given ELF file. It tries to emulate the behavior of objcopy.
-func ExtractROM(path string) (uint64, []byte, error) {
+func extractROM(path string) (uint64, []byte, error) {
 	f, err := elf.Open(path)
 	if err != nil {
-		return 0, nil, ObjcopyError{"failed to open ELF file to extract text segment", err}
+		return 0, nil, objcopyError{"failed to open ELF file to extract text segment", err}
 	}
 	defer f.Close()
 
@@ -56,7 +56,7 @@ func ExtractROM(path string) (uint64, []byte, error) {
 		}
 	}
 
-	progs := make(ProgSlice, 0, 2)
+	progs := make(progSlice, 0, 2)
 	for _, prog := range f.Progs {
 		if prog.Type != elf.PT_LOAD || prog.Filesz == 0 {
 			continue
@@ -64,18 +64,18 @@ func ExtractROM(path string) (uint64, []byte, error) {
 		progs = append(progs, prog)
 	}
 	if len(progs) == 0 {
-		return 0, nil, ObjcopyError{"file does not contain ROM segments: " + path, nil}
+		return 0, nil, objcopyError{"file does not contain ROM segments: " + path, nil}
 	}
 	sort.Sort(progs)
 
 	var rom []byte
 	for _, prog := range progs {
 		if prog.Paddr != progs[0].Paddr+uint64(len(rom)) {
-			return 0, nil, ObjcopyError{"ROM segments are non-contiguous: " + path, nil}
+			return 0, nil, objcopyError{"ROM segments are non-contiguous: " + path, nil}
 		}
 		data, err := ioutil.ReadAll(prog.Open())
 		if err != nil {
-			return 0, nil, ObjcopyError{"failed to extract segment from ELF file: " + path, err}
+			return 0, nil, objcopyError{"failed to extract segment from ELF file: " + path, err}
 		}
 		rom = append(rom, data...)
 	}
@@ -91,9 +91,9 @@ func ExtractROM(path string) (uint64, []byte, error) {
 	}
 }
 
-// Objcopy converts an ELF file to a different (simpler) output file format:
+// objcopy converts an ELF file to a different (simpler) output file format:
 // .bin or .hex. It extracts only the .text section.
-func Objcopy(infile, outfile string) error {
+func objcopy(infile, outfile string) error {
 	f, err := os.OpenFile(outfile, os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0666)
 	if err != nil {
 		return err
@@ -101,13 +101,17 @@ func Objcopy(infile, outfile string) error {
 	defer f.Close()
 
 	// Read the .text segment.
-	addr, data, err := ExtractROM(infile)
+	addr, data, err := extractROM(infile)
 	if err != nil {
 		return err
 	}
 
 	// Write to the file, in the correct format.
 	switch filepath.Ext(outfile) {
+	case ".gba":
+		// The address is not stored in a .gba file.
+		_, err := f.Write(data)
+		return err
 	case ".bin":
 		// The address is not stored in a .bin file (therefore you
 		// should use .hex files in most cases).
@@ -117,7 +121,7 @@ func Objcopy(infile, outfile string) error {
 		mem := gohex.NewMemory()
 		err := mem.AddBinary(uint32(addr), data)
 		if err != nil {
-			return ObjcopyError{"failed to create .hex file", err}
+			return objcopyError{"failed to create .hex file", err}
 		}
 		mem.DumpIntelHex(f, 16) // TODO: handle error
 		return nil

builder/sizes.go

@@ -1,4 +1,4 @@
-package main
+package builder
 
 import (
 	"debug/elf"
@@ -6,18 +6,18 @@ import (
 	"strings"
 )
 
-// Statistics about code size in a program.
-type ProgramSize struct {
-	Packages map[string]*PackageSize
-	Sum      *PackageSize
+// programSize contains size statistics per package of a compiled program.
+type programSize struct {
+	Packages map[string]*packageSize
+	Sum      *packageSize
 	Code     uint64
 	Data     uint64
 	BSS      uint64
 }
 
-// Return the list of package names (ProgramSize.Packages) sorted
-// alphabetically.
-func (ps *ProgramSize) SortedPackageNames() []string {
+// sortedPackageNames returns the list of package names (ProgramSize.Packages)
+// sorted alphabetically.
+func (ps *programSize) sortedPackageNames() []string {
 	names := make([]string, 0, len(ps.Packages))
 	for name := range ps.Packages {
 		names = append(names, name)
@@ -26,8 +26,9 @@ func (ps *ProgramSize) SortedPackageNames() []string {
 	return names
 }
 
-// The size of a package, calculated from the linked object file.
-type PackageSize struct {
+// packageSize contains the size of a package, calculated from the linked object
+// file.
+type packageSize struct {
 	Code   uint64
 	ROData uint64
 	Data   uint64
@@ -35,12 +36,12 @@ type PackageSize struct {
 }
 
 // Flash usage in regular microcontrollers.
-func (ps *PackageSize) Flash() uint64 {
+func (ps *packageSize) Flash() uint64 {
 	return ps.Code + ps.ROData + ps.Data
 }
 
 // Static RAM usage in regular microcontrollers.
-func (ps *PackageSize) RAM() uint64 {
+func (ps *packageSize) RAM() uint64 {
 	return ps.Data + ps.BSS
 }
 
@@ -64,8 +65,9 @@ func (l symbolList) Swap(i, j int) {
 	l[i], l[j] = l[j], l[i]
 }
 
-// Calculate program/data size breakdown of each package for a given ELF file.
-func Sizes(path string) (*ProgramSize, error) {
+// loadProgramSize calculate a program/data size breakdown of each package for a
+// given ELF file.
+func loadProgramSize(path string) (*programSize, error) {
 	file, err := elf.Open(path)
 	if err != nil {
 		return nil, err
@@ -115,7 +117,7 @@ func Sizes(path string) (*ProgramSize, error) {
 	}
 	sort.Sort(symbolList(symbols))
 
-	sizes := map[string]*PackageSize{}
+	sizes := map[string]*packageSize{}
 	var lastSymbolValue uint64
 	for _, symbol := range symbols {
 		symType := elf.ST_TYPE(symbol.Info)
@@ -129,7 +131,7 @@ func Sizes(path string) (*ProgramSize, error) {
 		}
 		pkgSize := sizes[pkgName]
 		if pkgSize == nil {
-			pkgSize = &PackageSize{}
+			pkgSize = &packageSize{}
 			sizes[pkgName] = pkgSize
 		}
 		if lastSymbolValue != symbol.Value || lastSymbolValue == 0 {
@@ -148,7 +150,7 @@ func Sizes(path string) (*ProgramSize, error) {
 		lastSymbolValue = symbol.Value
 	}
 
-	sum := &PackageSize{}
+	sum := &packageSize{}
 	for _, pkg := range sizes {
 		sum.Code += pkg.Code
 		sum.ROData += pkg.ROData
@@ -156,5 +158,5 @@ func Sizes(path string) (*ProgramSize, error) {
 		sum.BSS += pkg.BSS
 	}
 
-	return &ProgramSize{Packages: sizes, Code: sumCode, Data: sumData, BSS: sumBSS, Sum: sum}, nil
+	return &programSize{Packages: sizes, Code: sumCode, Data: sumData, BSS: sumBSS, Sum: sum}, nil
 }

builder/uf2.go

@@ -1,10 +1,11 @@
-// Converts firmware files from BIN to UF2 format before flashing.
+package builder
+
+// This file converts firmware files from BIN to UF2 format before flashing.
 //
 // For more information about the UF2 firmware file format, please see:
 // https://github.com/Microsoft/uf2
 //
 //
-package main
 
 import (
 	"bytes"
@@ -12,24 +13,24 @@ import (
 	"io/ioutil"
 )
 
-// ConvertELFFileToUF2File converts an ELF file to a UF2 file.
-func ConvertELFFileToUF2File(infile, outfile string) error {
+// convertELFFileToUF2File converts an ELF file to a UF2 file.
+func convertELFFileToUF2File(infile, outfile string) error {
 	// Read the .text segment.
-	_, data, err := ExtractROM(infile)
+	targetAddress, data, err := extractROM(infile)
 	if err != nil {
 		return err
 	}
 
-	output, _ := ConvertBinToUF2(data)
+	output, _ := convertBinToUF2(data, uint32(targetAddress))
 	return ioutil.WriteFile(outfile, output, 0644)
 }
 
-// ConvertBinToUF2 converts the binary bytes in input to UF2 formatted data.
-func ConvertBinToUF2(input []byte) ([]byte, int) {
+// convertBinToUF2 converts the binary bytes in input to UF2 formatted data.
+func convertBinToUF2(input []byte, targetAddr uint32) ([]byte, int) {
 	blocks := split(input, 256)
 	output := make([]byte, 0)
 
-	bl := NewUF2Block()
+	bl := newUF2Block(targetAddr)
 	bl.SetNumBlocks(len(blocks))
 
 	for i := 0; i < len(blocks); i++ {
@@ -44,14 +45,13 @@ func ConvertBinToUF2(input []byte) ([]byte, int) {
 }
 
 const (
-	uf2MagicStart0  = 0x0A324655 // "UF2\n"
-	uf2MagicStart1  = 0x9E5D5157 // Randomly selected
-	uf2MagicEnd     = 0x0AB16F30 // Ditto
-	uf2StartAddress = 0x2000
+	uf2MagicStart0 = 0x0A324655 // "UF2\n"
+	uf2MagicStart1 = 0x9E5D5157 // Randomly selected
+	uf2MagicEnd    = 0x0AB16F30 // Ditto
 )
 
-// UF2Block is the structure used for each UF2 code block sent to device.
-type UF2Block struct {
+// uf2Block is the structure used for each UF2 code block sent to device.
+type uf2Block struct {
 	magicStart0 uint32
 	magicStart1 uint32
 	flags       uint32
@@ -64,12 +64,12 @@ type UF2Block struct {
 	magicEnd    uint32
 }
 
-// NewUF2Block returns a new UF2Block struct that has been correctly populated
-func NewUF2Block() *UF2Block {
-	return &UF2Block{magicStart0: uf2MagicStart0,
+// newUF2Block returns a new uf2Block struct that has been correctly populated
+func newUF2Block(targetAddr uint32) *uf2Block {
+	return &uf2Block{magicStart0: uf2MagicStart0,
 		magicStart1: uf2MagicStart1,
 		magicEnd:    uf2MagicEnd,
-		targetAddr:  uf2StartAddress,
+		targetAddr:  targetAddr,
 		flags:       0x0,
 		familyID:    0x0,
 		payloadSize: 256,
@@ -77,8 +77,8 @@ func NewUF2Block() *UF2Block {
 	}
 }
 
-// Bytes converts the UF2Block to a slice of bytes that can be written to file.
-func (b *UF2Block) Bytes() []byte {
+// Bytes converts the uf2Block to a slice of bytes that can be written to file.
+func (b *uf2Block) Bytes() []byte {
 	buf := bytes.NewBuffer(make([]byte, 0, 512))
 	binary.Write(buf, binary.LittleEndian, b.magicStart0)
 	binary.Write(buf, binary.LittleEndian, b.magicStart1)
@@ -95,23 +95,23 @@ func (b *UF2Block) Bytes() []byte {
 }
 
 // IncrementAddress moves the target address pointer forward by count bytes.
-func (b *UF2Block) IncrementAddress(count uint32) {
+func (b *uf2Block) IncrementAddress(count uint32) {
 	b.targetAddr += b.payloadSize
 }
 
 // SetData sets the data to be used for the current block.
-func (b *UF2Block) SetData(d []byte) {
+func (b *uf2Block) SetData(d []byte) {
 	b.data = make([]byte, 476)
 	copy(b.data[:], d)
 }
 
 // SetBlockNo sets the current block number to be used.
-func (b *UF2Block) SetBlockNo(bn int) {
+func (b *uf2Block) SetBlockNo(bn int) {
 	b.blockNo = uint32(bn)
 }
 
 // SetNumBlocks sets the total number of blocks for this UF2 file.
-func (b *UF2Block) SetNumBlocks(total int) {
+func (b *uf2Block) SetNumBlocks(total int) {
 	b.numBlocks = uint32(total)
 }
 

cgo/cgo_test.go

@@ -0,0 +1,129 @@
+package cgo
+
+import (
+	"bytes"
+	"flag"
+	"fmt"
+	"go/ast"
+	"go/format"
+	"go/parser"
+	"go/token"
+	"go/types"
+	"io/ioutil"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"testing"
+)
+
+// Pass -update to go test to update the output of the test files.
+var flagUpdate = flag.Bool("update", false, "Update images based on test output.")
+
+func TestCGo(t *testing.T) {
+	var cflags = []string{"--target=armv6m-none-eabi"}
+
+	for _, name := range []string{"basic", "errors", "types", "flags", "const"} {
+		name := name // avoid a race condition
+		t.Run(name, func(t *testing.T) {
+			// Read the AST in memory.
+			path := filepath.Join("testdata", name+".go")
+			fset := token.NewFileSet()
+			f, err := parser.ParseFile(fset, path, nil, parser.ParseComments)
+			if err != nil {
+				t.Fatal("could not parse Go source file:", err)
+			}
+
+			// Process the AST with CGo.
+			cgoAST, cgoErrors := Process([]*ast.File{f}, "testdata", fset, cflags)
+
+			// Check the AST for type errors.
+			var typecheckErrors []error
+			config := types.Config{
+				Error: func(err error) {
+					typecheckErrors = append(typecheckErrors, err)
+				},
+				Importer: simpleImporter{},
+				Sizes:    types.SizesFor("gccgo", "arm"),
+			}
+			_, err = config.Check("", fset, []*ast.File{f, cgoAST}, nil)
+			if err != nil && len(typecheckErrors) == 0 {
+				// Only report errors when no type errors are found (an
+				// unexpected condition).
+				t.Error(err)
+			}
+
+			// Store the (formatted) output in a buffer. Format it, so it
+			// becomes easier to read (and will hopefully change less with CGo
+			// changes).
+			buf := &bytes.Buffer{}
+			if len(cgoErrors) != 0 {
+				buf.WriteString("// CGo errors:\n")
+				for _, err := range cgoErrors {
+					buf.WriteString(formatDiagnostic(err))
+				}
+				buf.WriteString("\n")
+			}
+			if len(typecheckErrors) != 0 {
+				buf.WriteString("// Type checking errors after CGo processing:\n")
+				for _, err := range typecheckErrors {
+					buf.WriteString(formatDiagnostic(err))
+				}
+				buf.WriteString("\n")
+			}
+			err = format.Node(buf, fset, cgoAST)
+			if err != nil {
+				t.Errorf("could not write out CGo AST: %v", err)
+			}
+			actual := strings.Replace(string(buf.Bytes()), "\r\n", "\n", -1)
+
+			// Read the file with the expected output, to compare against.
+			outfile := filepath.Join("testdata", name+".out.go")
+			expectedBytes, err := ioutil.ReadFile(outfile)
+			if err != nil {
+				t.Fatalf("could not read expected output: %v", err)
+			}
+			expected := strings.Replace(string(expectedBytes), "\r\n", "\n", -1)
+
+			// Check whether the output is as expected.
+			if expected != actual {
+				// It is not. Test failed.
+				if *flagUpdate {
+					// Update the file with the expected data.
+					err := ioutil.WriteFile(outfile, []byte(actual), 0666)
+					if err != nil {
+						t.Error("could not write updated output file:", err)
+					}
+					return
+				}
+				t.Errorf("output did not match:\n%s", string(actual))
+			}
+		})
+	}
+}
+
+// simpleImporter implements the types.Importer interface, but only allows
+// importing the unsafe package.
+type simpleImporter struct {
+}
+
+// Import implements the Importer interface. For testing usage only: it only
+// supports importing the unsafe package.
+func (i simpleImporter) Import(path string) (*types.Package, error) {
+	switch path {
+	case "unsafe":
+		return types.Unsafe, nil
+	default:
+		return nil, fmt.Errorf("importer not implemented for package %s", path)
+	}
+}
+
+// formatDiagnostics formats the error message to be an indented comment. It
+// also fixes Windows path name issues (backward slashes).
+func formatDiagnostic(err error) string {
+	msg := err.Error()
+	if runtime.GOOS == "windows" {
+		// Fix Windows path slashes.
+		msg = strings.Replace(msg, "testdata\\", "testdata/", -1)
+	}
+	return "//     " + msg + "\n"
+}

cgo/const.go

@@ -0,0 +1,218 @@
+package cgo
+
+// This file implements a parser of a subset of the C language, just enough to
+// parse common #define statements to Go constant expressions.
+
+import (
+	"fmt"
+	"go/ast"
+	"go/scanner"
+	"go/token"
+	"strings"
+)
+
+// parseConst parses the given string as a C constant.
+func parseConst(pos token.Pos, fset *token.FileSet, value string) (ast.Expr, *scanner.Error) {
+	t := newTokenizer(pos, fset, value)
+	expr, err := parseConstExpr(t)
+	if t.token != token.EOF {
+		return nil, &scanner.Error{
+			Pos: t.fset.Position(t.pos),
+			Msg: "unexpected token " + t.token.String(),
+		}
+	}
+	return expr, err
+}
+
+// parseConstExpr parses a stream of C tokens to a Go expression.
+func parseConstExpr(t *tokenizer) (ast.Expr, *scanner.Error) {
+	switch t.token {
+	case token.LPAREN:
+		lparen := t.pos
+		t.Next()
+		x, err := parseConstExpr(t)
+		if err != nil {
+			return nil, err
+		}
+		if t.token != token.RPAREN {
+			return nil, unexpectedToken(t, token.RPAREN)
+		}
+		expr := &ast.ParenExpr{
+			Lparen: lparen,
+			X:      x,
+			Rparen: t.pos,
+		}
+		t.Next()
+		return expr, nil
+	case token.INT, token.FLOAT, token.STRING, token.CHAR:
+		expr := &ast.BasicLit{
+			ValuePos: t.pos,
+			Kind:     t.token,
+			Value:    t.value,
+		}
+		t.Next()
+		return expr, nil
+	case token.IDENT:
+		expr := &ast.Ident{
+			NamePos: t.pos,
+			Name:    "C." + t.value,
+		}
+		t.Next()
+		return expr, nil
+	case token.EOF:
+		return nil, &scanner.Error{
+			Pos: t.fset.Position(t.pos),
+			Msg: "empty constant",
+		}
+	default:
+		return nil, &scanner.Error{
+			Pos: t.fset.Position(t.pos),
+			Msg: fmt.Sprintf("unexpected token %s", t.token),
+		}
+	}
+}
+
+// unexpectedToken returns an error of the form "unexpected token FOO, expected
+// BAR".
+func unexpectedToken(t *tokenizer, expected token.Token) *scanner.Error {
+	return &scanner.Error{
+		Pos: t.fset.Position(t.pos),
+		Msg: fmt.Sprintf("unexpected token %s, expected %s", t.token, expected),
+	}
+}
+
+// tokenizer reads C source code and converts it to Go tokens.
+type tokenizer struct {
+	pos   token.Pos
+	fset  *token.FileSet
+	token token.Token
+	value string
+	buf   string
+}
+
+// newTokenizer initializes a new tokenizer, positioned at the first token in
+// the string.
+func newTokenizer(start token.Pos, fset *token.FileSet, buf string) *tokenizer {
+	t := &tokenizer{
+		pos:   start,
+		fset:  fset,
+		buf:   buf,
+		token: token.ILLEGAL,
+	}
+	t.Next() // Parse the first token.
+	return t
+}
+
+// Next consumes the next token in the stream. There is no return value, read
+// the next token from the pos, token and value properties.
+func (t *tokenizer) Next() {
+	t.pos += token.Pos(len(t.value))
+	for {
+		if len(t.buf) == 0 {
+			t.token = token.EOF
+			return
+		}
+		c := t.buf[0]
+		switch {
+		case c == ' ' || c == '\f' || c == '\n' || c == '\r' || c == '\t' || c == '\v':
+			// Skip whitespace.
+			// Based on this source, not sure whether it represents C whitespace:
+			// https://en.cppreference.com/w/cpp/string/byte/isspace
+			t.pos++
+			t.buf = t.buf[1:]
+		case c == '(' || c == ')':
+			// Single-character tokens.
+			switch c {
+			case '(':
+				t.token = token.LPAREN
+			case ')':
+				t.token = token.RPAREN
+			}
+			t.value = t.buf[:1]
+			t.buf = t.buf[1:]
+			return
+		case c >= '0' && c <= '9':
+			// Numeric constant (int, float, etc.).
+			// Find the last non-numeric character.
+			tokenLen := len(t.buf)
+			hasDot := false
+			for i, c := range t.buf {
+				if c == '.' {
+					hasDot = true
+				}
+				if c >= '0' && c <= '9' || c == '.' || c == '_' || c >= 'a' && c <= 'z' || c >= 'A' && c <= 'Z' {
+					tokenLen = i + 1
+				} else {
+					break
+				}
+			}
+			t.value = t.buf[:tokenLen]
+			t.buf = t.buf[tokenLen:]
+			if hasDot {
+				// Integer constants are more complicated than this but this is
+				// a close approximation.
+				// https://en.cppreference.com/w/cpp/language/integer_literal
+				t.token = token.FLOAT
+				t.value = strings.TrimRight(t.value, "f")
+			} else {
+				t.token = token.INT
+				t.value = strings.TrimRight(t.value, "uUlL")
+			}
+			return
+		case c >= 'A' && c <= 'Z' || c >= 'a' && c <= 'z' || c == '_':
+			// Identifier. Find all remaining tokens that are part of this
+			// identifier.
+			tokenLen := len(t.buf)
+			for i, c := range t.buf {
+				if c >= '0' && c <= '9' || c >= 'A' && c <= 'Z' || c >= 'a' && c <= 'z' || c == '_' {
+					tokenLen = i + 1
+				} else {
+					break
+				}
+			}
+			t.value = t.buf[:tokenLen]
+			t.buf = t.buf[tokenLen:]
+			t.token = token.IDENT
+			return
+		case c == '"':
+			// String constant. Find the first '"' character that is not
+			// preceded by a backslash.
+			escape := false
+			tokenLen := len(t.buf)
+			for i, c := range t.buf {
+				if i != 0 && c == '"' && !escape {
+					tokenLen = i + 1
+					break
+				}
+				if !escape {
+					escape = c == '\\'
+				}
+			}
+			t.token = token.STRING
+			t.value = t.buf[:tokenLen]
+			t.buf = t.buf[tokenLen:]
+			return
+		case c == '\'':
+			// Char (rune) constant. Find the first '\'' character that is not
+			// preceded by a backslash.
+			escape := false
+			tokenLen := len(t.buf)
+			for i, c := range t.buf {
+				if i != 0 && c == '\'' && !escape {
+					tokenLen = i + 1
+					break
+				}
+				if !escape {
+					escape = c == '\\'
+				}
+			}
+			t.token = token.CHAR
+			t.value = t.buf[:tokenLen]
+			t.buf = t.buf[tokenLen:]
+			return
+		default:
+			t.token = token.ILLEGAL
+			return
+		}
+	}
+}

cgo/const_test.go

@@ -0,0 +1,60 @@
+package cgo
+
+import (
+	"bytes"
+	"go/format"
+	"go/token"
+	"strings"
+	"testing"
+)
+
+func TestParseConst(t *testing.T) {
+	// Test converting a C constant to a Go constant.
+	for _, tc := range []struct {
+		C  string
+		Go string
+	}{
+		{`5`, `5`},
+		{`(5)`, `(5)`},
+		{`(((5)))`, `(5)`},
+		{`)`, `error: 1:1: unexpected token )`},
+		{`5)`, `error: 1:2: unexpected token )`},
+		{"  \t)", `error: 1:4: unexpected token )`},
+		{`5.8f`, `5.8`},
+		{`foo`, `C.foo`},
+		{``, `error: 1:1: empty constant`}, // empty constants not allowed in Go
+		{`"foo"`, `"foo"`},
+		{`"a\\n"`, `"a\\n"`},
+		{`"a\n"`, `"a\n"`},
+		{`"a\""`, `"a\""`},
+		{`'a'`, `'a'`},
+		{`0b10`, `0b10`},
+		{`0x1234_5678`, `0x1234_5678`},
+		{`5 5`, `error: 1:3: unexpected token INT`}, // test for a bugfix
+	} {
+		fset := token.NewFileSet()
+		startPos := fset.AddFile("", -1, 1000).Pos(0)
+		expr, err := parseConst(startPos, fset, tc.C)
+		s := "<invalid>"
+		if err != nil {
+			if !strings.HasPrefix(tc.Go, "error: ") {
+				t.Errorf("expected value %#v for C constant %#v but got error %#v", tc.Go, tc.C, err.Error())
+				continue
+			}
+			s = "error: " + err.Error()
+		} else if expr != nil {
+			// Serialize the Go constant to a string, for more readable test
+			// cases.
+			buf := &bytes.Buffer{}
+			err := format.Node(buf, fset, expr)
+			if err != nil {
+				t.Errorf("could not format expr from C constant %#v: %v", tc.C, err)
+				continue
+			}
+			s = buf.String()
+		}
+		if s != tc.Go {
+			t.Errorf("C constant %#v was parsed to %#v while expecting %#v", tc.C, s, tc.Go)
+		}
+	}
+}

cgo/libclang.go

@@ -15,7 +15,7 @@ import (
 )
 
 /*
-#include <clang-c/Index.h> // if this fails, install libclang-8-dev
+#include <clang-c/Index.h> // if this fails, install libclang-9-dev
 #include <stdlib.h>
 #include <stdint.h>
 
@@ -49,9 +49,13 @@ GoCXCursor tinygo_clang_Cursor_getArgument(GoCXCursor c, unsigned i);
 CXSourceLocation tinygo_clang_getCursorLocation(GoCXCursor c);
 CXSourceRange tinygo_clang_getCursorExtent(GoCXCursor c);
 CXTranslationUnit tinygo_clang_Cursor_getTranslationUnit(GoCXCursor c);
+long long tinygo_clang_getEnumConstantDeclValue(GoCXCursor c);
+CXType tinygo_clang_getEnumDeclIntegerType(GoCXCursor c);
+unsigned tinygo_clang_Cursor_isBitField(GoCXCursor c);
 
 int tinygo_clang_globals_visitor(GoCXCursor c, GoCXCursor parent, CXClientData client_data);
 int tinygo_clang_struct_visitor(GoCXCursor c, GoCXCursor parent, CXClientData client_data);
+int tinygo_clang_enum_visitor(GoCXCursor c, GoCXCursor parent, CXClientData client_data);
 */
 import "C"
 
@@ -105,7 +109,8 @@ func (p *cgoPackage) parseFragment(fragment string, cflags []string, posFilename
 		C.CXTranslationUnit_DetailedPreprocessingRecord,
 		&unit)
 	if errCode != 0 {
-		panic("loader: failed to parse source with libclang")
+		// This is probably a bug in the usage of libclang.
+		panic("cgo: failed to parse source with libclang")
 	}
 	defer C.clang_disposeTranslationUnit(unit)
 
@@ -114,27 +119,8 @@ func (p *cgoPackage) parseFragment(fragment string, cflags []string, posFilename
 			spelling := getString(C.clang_getDiagnosticSpelling(diagnostic))
 			severity := diagnosticSeverity[C.clang_getDiagnosticSeverity(diagnostic)]
 			location := C.clang_getDiagnosticLocation(diagnostic)
-			var libclangFilename C.CXString
-			var line C.unsigned
-			var column C.unsigned
-			C.clang_getPresumedLocation(location, &libclangFilename, &line, &column)
-			filename := getString(libclangFilename)
-			if filepath.IsAbs(filename) {
-				// Relative paths for readability, like other Go parser errors.
-				relpath, err := filepath.Rel(p.dir, filename)
-				if err == nil {
-					filename = relpath
-				}
-			}
-			p.errors = append(p.errors, &scanner.Error{
-				Pos: token.Position{
-					Filename: filename,
-					Offset:   0, // not provided by clang_getPresumedLocation
-					Line:     int(line),
-					Column:   int(column),
-				},
-				Msg: severity + ": " + spelling,
-			})
+			pos := p.getClangLocationPosition(location, unit)
+			p.addError(pos, severity+": "+spelling)
 		}
 		for i := 0; i < numDiagnostics; i++ {
 			diagnostic := C.clang_getDiagnostic(unit, C.uint(i))
@@ -146,7 +132,6 @@ func (p *cgoPackage) parseFragment(fragment string, cflags []string, posFilename
 				addDiagnostic(C.clang_getDiagnosticInSet(diagnostics, C.uint(j)))
 			}
 		}
-		return
 	}
 
 	ref := storedRefs.Put(p)
@@ -233,14 +218,17 @@ func tinygo_clang_globals_visitor(c, parent C.GoCXCursor, client_data C.CXClient
 		var startOffset, endOffset C.unsigned
 		C.clang_getExpansionLocation(start, &file, nil, nil, &startOffset)
 		if file == nil {
-			panic("could not find file where macro is defined")
+			p.addError(pos, "internal error: could not find file where macro is defined")
+			break
 		}
 		C.clang_getExpansionLocation(end, &endFile, nil, nil, &endOffset)
 		if file != endFile {
-			panic("expected start and end location of a #define to be in the same file")
+			p.addError(pos, "internal error: expected start and end location of a macro to be in the same file")
+			break
 		}
 		if startOffset > endOffset {
-			panic("startOffset > endOffset")
+			p.addError(pos, "internal error: start offset of macro is after end offset")
+			break
 		}
 
 		// read file contents and extract the relevant byte range
@@ -248,59 +236,29 @@ func tinygo_clang_globals_visitor(c, parent C.GoCXCursor, client_data C.CXClient
 		var size C.size_t
 		sourcePtr := C.clang_getFileContents(tu, file, &size)
 		if endOffset >= C.uint(size) {
-			panic("endOffset lies after end of file")
+			p.addError(pos, "internal error: end offset of macro lies after end of file")
+			break
 		}
 		source := string(((*[1 << 28]byte)(unsafe.Pointer(sourcePtr)))[startOffset:endOffset:endOffset])
 		if !strings.HasPrefix(source, name) {
-			panic(fmt.Sprintf("expected #define value to start with %#v, got %#v", name, source))
+			p.addError(pos, fmt.Sprintf("internal error: expected macro value to start with %#v, got %#v", name, source))
+			break
 		}
-		value := strings.TrimSpace(source[len(name):])
-		for len(value) != 0 && value[0] == '(' && value[len(value)-1] == ')' {
-			value = strings.TrimSpace(value[1 : len(value)-1])
+		value := source[len(name):]
+		// Try to convert this #define into a Go constant expression.
+		expr, err := parseConst(pos+token.Pos(len(name)), p.fset, value)
+		if err != nil {
+			p.errors = append(p.errors, err)
 		}
-		if len(value) == 0 {
-			// Pretend it doesn't exist at all.
-			return C.CXChildVisit_Continue
-		}
-		// For information about integer literals:
-		// https://en.cppreference.com/w/cpp/language/integer_literal
-		if value[0] == '"' {
-			// string constant
-			p.constants[name] = constantInfo{&ast.BasicLit{pos, token.STRING, value}, pos}
-			return C.CXChildVisit_Continue
-		}
-		if value[0] == '\'' {
-			// char constant
-			p.constants[name] = constantInfo{&ast.BasicLit{pos, token.CHAR, value}, pos}
-			return C.CXChildVisit_Continue
-		}
-		// assume it's a number (int or float)
-		value = strings.Replace(value, "'", "", -1) // remove ' chars
-		value = strings.TrimRight(value, "lu")      // remove llu suffixes etc.
-		// find the first non-number
-		nonnum := byte(0)
-		for i := 0; i < len(value); i++ {
-			if value[i] < '0' || value[i] > '9' {
-				nonnum = value[i]
-				break
-			}
-		}
-		// determine number type based on the first non-number
-		switch nonnum {
-		case 0:
-			// no non-number found, must be an integer
-			p.constants[name] = constantInfo{&ast.BasicLit{pos, token.INT, value}, pos}
-		case 'x', 'X':
-			// hex integer constant
-			// TODO: may also be a floating point number per C++17.
-			p.constants[name] = constantInfo{&ast.BasicLit{pos, token.INT, value}, pos}
-		case '.', 'e':
-			// float constant
-			value = strings.TrimRight(value, "fFlL")
-			p.constants[name] = constantInfo{&ast.BasicLit{pos, token.FLOAT, value}, pos}
-		default:
-			// unknown type, ignore
+		if expr != nil {
+			// Parsing was successful.
+			p.constants[name] = constantInfo{expr, pos}
 		}
+	case C.CXCursor_EnumDecl:
+		// Visit all enums, because the fields may be used even when the enum
+		// type itself is not.
+		typ := C.tinygo_clang_getCursorType(c)
+		p.makeASTType(typ, pos)
 	}
 	return C.CXChildVisit_Continue
 }
@@ -312,11 +270,16 @@ func getString(clangString C.CXString) (s string) {
 	return
 }
 
-// getCursorPosition returns a usable token.Pos from a libclang cursor. If the
-// file for this cursor has not been seen before, it is read from libclang
-// (which already has the file in memory) and added to the token.FileSet.
+// getCursorPosition returns a usable token.Pos from a libclang cursor.
 func (p *cgoPackage) getCursorPosition(cursor C.GoCXCursor) token.Pos {
-	location := C.tinygo_clang_getCursorLocation(cursor)
+	return p.getClangLocationPosition(C.tinygo_clang_getCursorLocation(cursor), C.tinygo_clang_Cursor_getTranslationUnit(cursor))
+}
+
+// getClangLocationPosition returns a usable token.Pos based on a libclang
+// location and translation unit. If the file for this cursor has not been seen
+// before, it is read from libclang (which already has the file in memory) and
+// added to the token.FileSet.
+func (p *cgoPackage) getClangLocationPosition(location C.CXSourceLocation, tu C.CXTranslationUnit) token.Pos {
 	var file C.CXFile
 	var line C.unsigned
 	var column C.unsigned
@@ -330,7 +293,6 @@ func (p *cgoPackage) getCursorPosition(cursor C.GoCXCursor) token.Pos {
 	if _, ok := p.tokenFiles[filename]; !ok {
 		// File has not been seen before in this package, add line information
 		// now by reading the file from libclang.
-		tu := C.tinygo_clang_Cursor_getTranslationUnit(cursor)
 		var size C.size_t
 		sourcePtr := C.clang_getFileContents(tu, file, &size)
 		source := ((*[1 << 28]byte)(unsafe.Pointer(sourcePtr)))[:size:size]
@@ -344,7 +306,62 @@ func (p *cgoPackage) getCursorPosition(cursor C.GoCXCursor) token.Pos {
 		f.SetLines(lines)
 		p.tokenFiles[filename] = f
 	}
-	return p.tokenFiles[filename].Pos(int(offset))
+	positionFile := p.tokenFiles[filename]
+
+	// Check for alternative line/column information (set with a line directive).
+	var filename2String C.CXString
+	var line2 C.unsigned
+	var column2 C.unsigned
+	C.clang_getPresumedLocation(location, &filename2String, &line2, &column2)
+	filename2 := getString(filename2String)
+	if filename2 != filename || line2 != line || column2 != column {
+		// The location was changed with a preprocessor directive.
+		// TODO: this only works for locations that are added in order. Adding
+		// line/column info to a file that already has line/column info after
+		// the given offset is ignored.
+		positionFile.AddLineColumnInfo(int(offset), filename2, int(line2), int(column2))
+	}
+
+	return positionFile.Pos(int(offset))
+}
+
+// addError is a utility function to add an error to the list of errors. It will
+// convert the token position to a line/column position first, and call
+// addErrorAt.
+func (p *cgoPackage) addError(pos token.Pos, msg string) {
+	p.addErrorAt(p.fset.PositionFor(pos, true), msg)
+}
+
+// addErrorAfter is like addError, but adds the text `after` to the source
+// location.
+func (p *cgoPackage) addErrorAfter(pos token.Pos, after, msg string) {
+	position := p.fset.PositionFor(pos, true)
+	lines := strings.Split(after, "\n")
+	if len(lines) != 1 {
+		// Adjust lines.
+		// For why we can't just do pos+token.Pos(len(after)), see:
+		// https://github.com/golang/go/issues/35803
+		position.Line += len(lines) - 1
+		position.Column = len(lines[len(lines)-1]) + 1
+	} else {
+		position.Column += len(after)
+	}
+	p.addErrorAt(position, msg)
+}
+
+// addErrorAt is a utility function to add an error to the list of errors.
+func (p *cgoPackage) addErrorAt(position token.Position, msg string) {
+	if filepath.IsAbs(position.Filename) {
+		// Relative paths for readability, like other Go parser errors.
+		relpath, err := filepath.Rel(p.dir, position.Filename)
+		if err == nil {
+			position.Filename = relpath
+		}
+	}
+	p.errors = append(p.errors, scanner.Error{
+		Pos: position,
+		Msg: msg,
+	})
 }
 
 // makeASTType return the ast.Expr for the given libclang type. In other words,
@@ -455,7 +472,7 @@ func (p *cgoPackage) makeASTType(typ C.CXType, pos token.Pos) ast.Expr {
 						// This happens for some very special purpose architectures
 						// (DSPs etc.) that are not currently targeted.
 						// https://www.embecosm.com/2017/04/18/non-8-bit-char-support-in-clang-and-llvm/
-						panic("unknown char width")
+						p.addError(pos, fmt.Sprintf("unknown char width: %d", typeSize))
 					}
 					switch underlyingType.kind {
 					case C.CXType_Char_S:
@@ -501,90 +518,86 @@ func (p *cgoPackage) makeASTType(typ C.CXType, pos token.Pos) ast.Expr {
 		switch underlying.kind {
 		case C.CXType_Record:
 			return p.makeASTType(underlying, pos)
+		case C.CXType_Enum:
+			return p.makeASTType(underlying, pos)
 		default:
-			panic("unknown elaborated type")
+			typeKindSpelling := getString(C.clang_getTypeKindSpelling(underlying.kind))
+			p.addError(pos, fmt.Sprintf("unknown elaborated type (libclang type kind %s)", typeKindSpelling))
+			typeName = "<unknown>"
 		}
 	case C.CXType_Record:
 		cursor := C.tinygo_clang_getTypeDeclaration(typ)
 		name := getString(C.tinygo_clang_getCursorSpelling(cursor))
-		var cgoName string
+		var cgoRecordPrefix string
 		switch C.tinygo_clang_getCursorKind(cursor) {
 		case C.CXCursor_StructDecl:
-			cgoName = "struct_" + name
+			cgoRecordPrefix = "struct_"
 		case C.CXCursor_UnionDecl:
-			cgoName = "union_" + name
+			cgoRecordPrefix = "union_"
 		default:
-			panic("unknown record declaration")
+			// makeASTRecordType will create an appropriate error.
+			cgoRecordPrefix = "record_"
 		}
-		if _, ok := p.elaboratedTypes[cgoName]; !ok {
-			p.elaboratedTypes[cgoName] = nil // predeclare (to avoid endless recursion)
-			fieldList := &ast.FieldList{
-				Opening: pos,
-				Closing: pos,
+		if name == "" {
+			// Anonymous record, probably inside a typedef.
+			typeInfo := p.makeASTRecordType(cursor, pos)
+			if typeInfo.bitfields != nil || typeInfo.unionSize != 0 {
+				// This record is a union or is a struct with bitfields, so we
+				// have to declare it as a named type (for getters/setters to
+				// work).
+				p.anonStructNum++
+				cgoName := cgoRecordPrefix + strconv.Itoa(p.anonStructNum)
+				p.elaboratedTypes[cgoName] = typeInfo
+				return &ast.Ident{
+					NamePos: pos,
+					Name:    "C." + cgoName,
+				}
 			}
-			ref := storedRefs.Put(struct {
-				fieldList *ast.FieldList
-				pkg       *cgoPackage
-			}{fieldList, p})
+			return typeInfo.typeExpr
+		} else {
+			cgoName := cgoRecordPrefix + name
+			if _, ok := p.elaboratedTypes[cgoName]; !ok {
+				p.elaboratedTypes[cgoName] = nil // predeclare (to avoid endless recursion)
+				p.elaboratedTypes[cgoName] = p.makeASTRecordType(cursor, pos)
+			}
+			return &ast.Ident{
+				NamePos: pos,
+				Name:    "C." + cgoName,
+			}
+		}
+	case C.CXType_Enum:
+		cursor := C.tinygo_clang_getTypeDeclaration(typ)
+		name := getString(C.tinygo_clang_getCursorSpelling(cursor))
+		underlying := C.tinygo_clang_getEnumDeclIntegerType(cursor)
+		if name == "" {
+			// anonymous enum
+			ref := storedRefs.Put(p)
 			defer storedRefs.Remove(ref)
-			C.tinygo_clang_visitChildren(cursor, C.CXCursorVisitor(C.tinygo_clang_struct_visitor), C.CXClientData(ref))
-			switch C.tinygo_clang_getCursorKind(cursor) {
-			case C.CXCursor_StructDecl:
-				p.elaboratedTypes[cgoName] = &elaboratedTypeInfo{
-					typeExpr: &ast.StructType{
-						Struct: pos,
-						Fields: fieldList,
-					},
-					pos: pos,
+			C.tinygo_clang_visitChildren(cursor, C.CXCursorVisitor(C.tinygo_clang_enum_visitor), C.CXClientData(ref))
+			return p.makeASTType(underlying, pos)
+		} else {
+			// named enum
+			if _, ok := p.enums[name]; !ok {
+				ref := storedRefs.Put(p)
+				defer storedRefs.Remove(ref)
+				C.tinygo_clang_visitChildren(cursor, C.CXCursorVisitor(C.tinygo_clang_enum_visitor), C.CXClientData(ref))
+				p.enums[name] = enumInfo{
+					typeExpr: p.makeASTType(underlying, pos),
+					pos:      pos,
 				}
-			case C.CXCursor_UnionDecl:
-				if len(fieldList.List) > 1 {
-					// Insert a special field at the front (of zero width) as a
-					// marker that this is struct is actually a union. This is done
-					// by giving the field a name that cannot be expressed directly
-					// in Go.
-					// Other parts of the compiler look at the first element in a
-					// struct (of size > 2) to know whether this is a union.
-					// Note that we don't have to insert it for single-element
-					// unions as they're basically equivalent to a struct.
-					unionMarker := &ast.Field{
-						Type: &ast.StructType{
-							Struct: pos,
-						},
-					}
-					unionMarker.Names = []*ast.Ident{
-						&ast.Ident{
-							NamePos: pos,
-							Name:    "C union",
-							Obj: &ast.Object{
-								Kind: ast.Var,
-								Name: "C union",
-								Decl: unionMarker,
-							},
-						},
-					}
-					fieldList.List = append([]*ast.Field{unionMarker}, fieldList.List...)
-				}
-				p.elaboratedTypes[cgoName] = &elaboratedTypeInfo{
-					typeExpr: &ast.StructType{
-						Struct: pos,
-						Fields: fieldList,
-					},
-					pos: pos,
-				}
-			default:
-				panic("unreachable")
 			}
-		}
-		return &ast.Ident{
-			NamePos: pos,
-			Name:    "C." + cgoName,
+			return &ast.Ident{
+				NamePos: pos,
+				Name:    "C.enum_" + name,
+			}
 		}
 	}
 	if typeName == "" {
-		// Fallback, probably incorrect but at least the error points to an odd
-		// type name.
-		typeName = "C." + getString(C.clang_getTypeSpelling(typ))
+		// Report this as an error.
+		typeSpelling := getString(C.clang_getTypeSpelling(typ))
+		typeKindSpelling := getString(C.clang_getTypeKindSpelling(typ.kind))
+		p.addError(pos, fmt.Sprintf("unknown C type: %v (libclang type kind %s)", typeSpelling, typeKindSpelling))
+		typeName = "C.<unknown>"
 	}
 	return &ast.Ident{
 		NamePos: pos,
@@ -592,25 +605,151 @@ func (p *cgoPackage) makeASTType(typ C.CXType, pos token.Pos) ast.Expr {
 	}
 }
 
+// makeASTRecordType parses a C record (struct or union) and translates it into
+// a Go struct type.
+func (p *cgoPackage) makeASTRecordType(cursor C.GoCXCursor, pos token.Pos) *elaboratedTypeInfo {
+	fieldList := &ast.FieldList{
+		Opening: pos,
+		Closing: pos,
+	}
+	var bitfieldList []bitfieldInfo
+	inBitfield := false
+	bitfieldNum := 0
+	ref := storedRefs.Put(struct {
+		fieldList    *ast.FieldList
+		pkg          *cgoPackage
+		inBitfield   *bool
+		bitfieldNum  *int
+		bitfieldList *[]bitfieldInfo
+	}{fieldList, p, &inBitfield, &bitfieldNum, &bitfieldList})
+	defer storedRefs.Remove(ref)
+	C.tinygo_clang_visitChildren(cursor, C.CXCursorVisitor(C.tinygo_clang_struct_visitor), C.CXClientData(ref))
+	renameFieldKeywords(fieldList)
+	switch C.tinygo_clang_getCursorKind(cursor) {
+	case C.CXCursor_StructDecl:
+		return &elaboratedTypeInfo{
+			typeExpr: &ast.StructType{
+				Struct: pos,
+				Fields: fieldList,
+			},
+			pos:       pos,
+			bitfields: bitfieldList,
+		}
+	case C.CXCursor_UnionDecl:
+		typeInfo := &elaboratedTypeInfo{
+			typeExpr: &ast.StructType{
+				Struct: pos,
+				Fields: fieldList,
+			},
+			pos:       pos,
+			bitfields: bitfieldList,
+		}
+		if len(fieldList.List) <= 1 {
+			// Useless union, treat it as a regular struct.
+			return typeInfo
+		}
+		if bitfieldList != nil {
+			// This is valid C... but please don't do this.
+			p.addError(pos, "bitfield in a union is not supported")
+		}
+		typ := C.tinygo_clang_getCursorType(cursor)
+		alignInBytes := int64(C.clang_Type_getAlignOf(typ))
+		sizeInBytes := int64(C.clang_Type_getSizeOf(typ))
+		if sizeInBytes == 0 {
+			p.addError(pos, "zero-length union is not supported")
+		}
+		typeInfo.unionSize = sizeInBytes
+		typeInfo.unionAlign = alignInBytes
+		return typeInfo
+	default:
+		cursorKind := C.tinygo_clang_getCursorKind(cursor)
+		cursorKindSpelling := getString(C.clang_getCursorKindSpelling(cursorKind))
+		p.addError(pos, fmt.Sprintf("expected StructDecl or UnionDecl, not %s", cursorKindSpelling))
+		return &elaboratedTypeInfo{
+			typeExpr: &ast.StructType{
+				Struct: pos,
+			},
+			pos: pos,
+		}
+	}
+}
+
 //export tinygo_clang_struct_visitor
 func tinygo_clang_struct_visitor(c, parent C.GoCXCursor, client_data C.CXClientData) C.int {
 	passed := storedRefs.Get(unsafe.Pointer(client_data)).(struct {
-		fieldList *ast.FieldList
-		pkg       *cgoPackage
+		fieldList    *ast.FieldList
+		pkg          *cgoPackage
+		inBitfield   *bool
+		bitfieldNum  *int
+		bitfieldList *[]bitfieldInfo
 	})
 	fieldList := passed.fieldList
 	p := passed.pkg
-	if C.tinygo_clang_getCursorKind(c) != C.CXCursor_FieldDecl {
-		panic("expected field inside cursor")
+	inBitfield := passed.inBitfield
+	bitfieldNum := passed.bitfieldNum
+	bitfieldList := passed.bitfieldList
+	pos := p.getCursorPosition(c)
+	switch cursorKind := C.tinygo_clang_getCursorKind(c); cursorKind {
+	case C.CXCursor_FieldDecl:
+		// Expected. This is a regular field.
+	case C.CXCursor_StructDecl, C.CXCursor_UnionDecl:
+		// Ignore. The next field will be the struct/union itself.
+		return C.CXChildVisit_Continue
+	default:
+		cursorKindSpelling := getString(C.clang_getCursorKindSpelling(cursorKind))
+		p.addError(pos, fmt.Sprintf("expected FieldDecl in struct or union, not %s", cursorKindSpelling))
+		return C.CXChildVisit_Continue
 	}
 	name := getString(C.tinygo_clang_getCursorSpelling(c))
+	if name == "" {
+		// Assume this is a bitfield of 0 bits.
+		// Warning: this is not necessarily true!
+		return C.CXChildVisit_Continue
+	}
 	typ := C.tinygo_clang_getCursorType(c)
 	field := &ast.Field{
 		Type: p.makeASTType(typ, p.getCursorPosition(c)),
 	}
+	offsetof := int64(C.clang_Type_getOffsetOf(C.tinygo_clang_getCursorType(parent), C.CString(name)))
+	alignOf := int64(C.clang_Type_getAlignOf(typ) * 8)
+	bitfieldOffset := offsetof % alignOf
+	if bitfieldOffset != 0 {
+		if C.tinygo_clang_Cursor_isBitField(c) != 1 {
+			p.addError(pos, "expected a bitfield")
+			return C.CXChildVisit_Continue
+		}
+		if !*inBitfield {
+			*bitfieldNum++
+		}
+		bitfieldName := "__bitfield_" + strconv.Itoa(*bitfieldNum)
+		prevField := fieldList.List[len(fieldList.List)-1]
+		if !*inBitfield {
+			// The previous element also was a bitfield, but wasn't noticed
+			// then. Add it now.
+			*inBitfield = true
+			*bitfieldList = append(*bitfieldList, bitfieldInfo{
+				field:    prevField,
+				name:     prevField.Names[0].Name,
+				startBit: 0,
+				pos:      prevField.Names[0].NamePos,
+			})
+			prevField.Names[0].Name = bitfieldName
+			prevField.Names[0].Obj.Name = bitfieldName
+		}
+		prevBitfield := &(*bitfieldList)[len(*bitfieldList)-1]
+		prevBitfield.endBit = bitfieldOffset
+		*bitfieldList = append(*bitfieldList, bitfieldInfo{
+			field:    prevField,
+			name:     name,
+			startBit: bitfieldOffset,
+			pos:      pos,
+		})
+		return C.CXChildVisit_Continue
+	}
+	*inBitfield = false
 	field.Names = []*ast.Ident{
 		&ast.Ident{
-			NamePos: p.getCursorPosition(c),
+			NamePos: pos,
 			Name:    name,
 			Obj: &ast.Object{
 				Kind: ast.Var,
@@ -622,3 +761,16 @@ func tinygo_clang_struct_visitor(c, parent C.GoCXCursor, client_data C.CXClientD
 	fieldList.List = append(fieldList.List, field)
 	return C.CXChildVisit_Continue
 }
+
+//export tinygo_clang_enum_visitor
+func tinygo_clang_enum_visitor(c, parent C.GoCXCursor, client_data C.CXClientData) C.int {
+	p := storedRefs.Get(unsafe.Pointer(client_data)).(*cgoPackage)
+	name := getString(C.tinygo_clang_getCursorSpelling(c))
+	pos := p.getCursorPosition(c)
+	value := C.tinygo_clang_getEnumConstantDeclValue(c)
+	p.constants[name] = constantInfo{
+		expr: &ast.BasicLit{pos, token.INT, strconv.FormatInt(int64(value), 10)},
+		pos:  pos,
+	}
+	return C.CXChildVisit_Continue
+}

cgo/libclang_config.go

@@ -3,9 +3,11 @@
 package cgo
 
 /*
-#cgo linux  CFLAGS: -I/usr/lib/llvm-8/include
-#cgo darwin CFLAGS: -I/usr/local/opt/llvm/include
-#cgo linux  LDFLAGS: -L/usr/lib/llvm-8/lib -lclang
-#cgo darwin LDFLAGS: -L/usr/local/opt/llvm/lib -lclang -lffi
+#cgo linux  CFLAGS: -I/usr/lib/llvm-9/include
+#cgo darwin CFLAGS: -I/usr/local/opt/llvm@9/include
+#cgo freebsd CFLAGS: -I/usr/local/llvm90/include
+#cgo linux  LDFLAGS: -L/usr/lib/llvm-9/lib -lclang
+#cgo darwin LDFLAGS: -L/usr/local/opt/llvm@9/lib -lclang -lffi
+#cgo freebsd LDFLAGS: -L/usr/local/llvm90/lib -lclang
 */
 import "C"

cgo/libclang_stubs.c

@@ -3,7 +3,7 @@
 // are slightly different from the ones defined in libclang.go, but they
 // should be ABI compatible.
 
-#include <clang-c/Index.h> // if this fails, install libclang-8-dev
+#include <clang-c/Index.h> // if this fails, install libclang-9-dev
 
 CXCursor tinygo_clang_getTranslationUnitCursor(CXTranslationUnit tu) {
 	return clang_getTranslationUnitCursor(tu);
@@ -56,3 +56,15 @@ CXSourceRange tinygo_clang_getCursorExtent(CXCursor c) {
 CXTranslationUnit tinygo_clang_Cursor_getTranslationUnit(CXCursor c) {
 	return clang_Cursor_getTranslationUnit(c);
 }
+
+long long tinygo_clang_getEnumConstantDeclValue(CXCursor c) {
+	return clang_getEnumConstantDeclValue(c);
+}
+
+CXType tinygo_clang_getEnumDeclIntegerType(CXCursor c) {
+	return clang_getEnumDeclIntegerType(c);
+}
+
+unsigned tinygo_clang_Cursor_isBitField(CXCursor c) {
+	return clang_Cursor_isBitField(c);
+}

cgo/security.go

@@ -0,0 +1,301 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// This file has been copied from the Go 1.13 release tree.
+
+// Checking of compiler and linker flags.
+// We must avoid flags like -fplugin=, which can allow
+// arbitrary code execution during the build.
+// Do not make changes here without carefully
+// considering the implications.
+// (That's why the code is isolated in a file named security.go.)
+//
+// Note that -Wl,foo means split foo on commas and pass to
+// the linker, so that -Wl,-foo,bar means pass -foo bar to
+// the linker. Similarly -Wa,foo for the assembler and so on.
+// If any of these are permitted, the wildcard portion must
+// disallow commas.
+//
+// Note also that GNU binutils accept any argument @foo
+// as meaning "read more flags from the file foo", so we must
+// guard against any command-line argument beginning with @,
+// even things like "-I @foo".
+// We use safeArg (which is even more conservative)
+// to reject these.
+//
+// Even worse, gcc -I@foo (one arg) turns into cc1 -I @foo (two args),
+// so although gcc doesn't expand the @foo, cc1 will.
+// So out of paranoia, we reject @ at the beginning of every
+// flag argument that might be split into its own argument.
+
+package cgo
+
+import (
+	"fmt"
+	"os"
+	"regexp"
+	"strings"
+	"unicode/utf8"
+)
+
+var re = regexp.MustCompile
+
+var validCompilerFlags = []*regexp.Regexp{
+	re(`-D([A-Za-z_].*)`),
+	re(`-F([^@\-].*)`),
+	re(`-I([^@\-].*)`),
+	re(`-O`),
+	re(`-O([^@\-].*)`),
+	re(`-W`),
+	re(`-W([^@,]+)`), // -Wall but not -Wa,-foo.
+	re(`-Wa,-mbig-obj`),
+	re(`-Wp,-D([A-Za-z_].*)`),
+	re(`-ansi`),
+	re(`-f(no-)?asynchronous-unwind-tables`),
+	re(`-f(no-)?blocks`),
+	re(`-f(no-)builtin-[a-zA-Z0-9_]*`),
+	re(`-f(no-)?common`),
+	re(`-f(no-)?constant-cfstrings`),
+	re(`-fdiagnostics-show-note-include-stack`),
+	re(`-f(no-)?eliminate-unused-debug-types`),
+	re(`-f(no-)?exceptions`),
+	re(`-f(no-)?fast-math`),
+	re(`-f(no-)?inline-functions`),
+	re(`-finput-charset=([^@\-].*)`),
+	re(`-f(no-)?fat-lto-objects`),
+	re(`-f(no-)?keep-inline-dllexport`),
+	re(`-f(no-)?lto`),
+	re(`-fmacro-backtrace-limit=(.+)`),
+	re(`-fmessage-length=(.+)`),
+	re(`-f(no-)?modules`),
+	re(`-f(no-)?objc-arc`),
+	re(`-f(no-)?objc-nonfragile-abi`),
+	re(`-f(no-)?objc-legacy-dispatch`),
+	re(`-f(no-)?omit-frame-pointer`),
+	re(`-f(no-)?openmp(-simd)?`),
+	re(`-f(no-)?permissive`),
+	re(`-f(no-)?(pic|PIC|pie|PIE)`),
+	re(`-f(no-)?plt`),
+	re(`-f(no-)?rtti`),
+	re(`-f(no-)?split-stack`),
+	re(`-f(no-)?stack-(.+)`),
+	re(`-f(no-)?strict-aliasing`),
+	re(`-f(un)signed-char`),
+	re(`-f(no-)?use-linker-plugin`), // safe if -B is not used; we don't permit -B
+	re(`-f(no-)?visibility-inlines-hidden`),
+	re(`-fsanitize=(.+)`),
+	re(`-ftemplate-depth-(.+)`),
+	re(`-fvisibility=(.+)`),
+	re(`-g([^@\-].*)?`),
+	re(`-m32`),
+	re(`-m64`),
+	re(`-m(abi|arch|cpu|fpu|tune)=([^@\-].*)`),
+	re(`-m(no-)?v?aes`),
+	re(`-marm`),
+	re(`-m(no-)?avx[0-9a-z]*`),
+	re(`-mfloat-abi=([^@\-].*)`),
+	re(`-mfpmath=[0-9a-z,+]*`),
+	re(`-m(no-)?avx[0-9a-z.]*`),
+	re(`-m(no-)?ms-bitfields`),
+	re(`-m(no-)?stack-(.+)`),
+	re(`-mmacosx-(.+)`),
+	re(`-mios-simulator-version-min=(.+)`),
+	re(`-miphoneos-version-min=(.+)`),
+	re(`-mtvos-simulator-version-min=(.+)`),
+	re(`-mtvos-version-min=(.+)`),
+	re(`-mwatchos-simulator-version-min=(.+)`),
+	re(`-mwatchos-version-min=(.+)`),
+	re(`-mnop-fun-dllimport`),
+	re(`-m(no-)?sse[0-9.]*`),
+	re(`-m(no-)?ssse3`),
+	re(`-mthumb(-interwork)?`),
+	re(`-mthreads`),
+	re(`-mwindows`),
+	re(`--param=ssp-buffer-size=[0-9]*`),
+	re(`-pedantic(-errors)?`),
+	re(`-pipe`),
+	re(`-pthread`),
+	re(`-?-std=([^@\-].*)`),
+	re(`-?-stdlib=([^@\-].*)`),
+	re(`--sysroot=([^@\-].*)`),
+	re(`-w`),
+	re(`-x([^@\-].*)`),
+	re(`-v`),
+}
+
+var validCompilerFlagsWithNextArg = []string{
+	"-arch",
+	"-D",
+	"-I",
+	"-framework",
+	"-isysroot",
+	"-isystem",
+	"--sysroot",
+	"-target",
+	"-x",
+}
+
+var validLinkerFlags = []*regexp.Regexp{
+	re(`-F([^@\-].*)`),
+	re(`-l([^@\-].*)`),
+	re(`-L([^@\-].*)`),
+	re(`-O`),
+	re(`-O([^@\-].*)`),
+	re(`-f(no-)?(pic|PIC|pie|PIE)`),
+	re(`-f(no-)?openmp(-simd)?`),
+	re(`-fsanitize=([^@\-].*)`),
+	re(`-flat_namespace`),
+	re(`-g([^@\-].*)?`),
+	re(`-headerpad_max_install_names`),
+	re(`-m(abi|arch|cpu|fpu|tune)=([^@\-].*)`),
+	re(`-mfloat-abi=([^@\-].*)`),
+	re(`-mmacosx-(.+)`),
+	re(`-mios-simulator-version-min=(.+)`),
+	re(`-miphoneos-version-min=(.+)`),
+	re(`-mthreads`),
+	re(`-mwindows`),
+	re(`-(pic|PIC|pie|PIE)`),
+	re(`-pthread`),
+	re(`-rdynamic`),
+	re(`-shared`),
+	re(`-?-static([-a-z0-9+]*)`),
+	re(`-?-stdlib=([^@\-].*)`),
+	re(`-v`),
+
+	// Note that any wildcards in -Wl need to exclude comma,
+	// since -Wl splits its argument at commas and passes
+	// them all to the linker uninterpreted. Allowing comma
+	// in a wildcard would allow tunnelling arbitrary additional
+	// linker arguments through one of these.
+	re(`-Wl,--(no-)?allow-multiple-definition`),
+	re(`-Wl,--(no-)?allow-shlib-undefined`),
+	re(`-Wl,--(no-)?as-needed`),
+	re(`-Wl,-Bdynamic`),
+	re(`-Wl,-berok`),
+	re(`-Wl,-Bstatic`),
+	re(`-WL,-O([^@,\-][^,]*)?`),
+	re(`-Wl,-d[ny]`),
+	re(`-Wl,--disable-new-dtags`),
+	re(`-Wl,-e[=,][a-zA-Z0-9]*`),
+	re(`-Wl,--enable-new-dtags`),
+	re(`-Wl,--end-group`),
+	re(`-Wl,--(no-)?export-dynamic`),
+	re(`-Wl,-framework,[^,@\-][^,]+`),
+	re(`-Wl,-headerpad_max_install_names`),
+	re(`-Wl,--no-undefined`),
+	re(`-Wl,-R([^@\-][^,@]*$)`),
+	re(`-Wl,--just-symbols[=,]([^,@\-][^,@]+)`),
+	re(`-Wl,-rpath(-link)?[=,]([^,@\-][^,]+)`),
+	re(`-Wl,-s`),
+	re(`-Wl,-search_paths_first`),
+	re(`-Wl,-sectcreate,([^,@\-][^,]+),([^,@\-][^,]+),([^,@\-][^,]+)`),
+	re(`-Wl,--start-group`),
+	re(`-Wl,-?-static`),
+	re(`-Wl,-?-subsystem,(native|windows|console|posix|xbox)`),
+	re(`-Wl,-syslibroot[=,]([^,@\-][^,]+)`),
+	re(`-Wl,-undefined[=,]([^,@\-][^,]+)`),
+	re(`-Wl,-?-unresolved-symbols=[^,]+`),
+	re(`-Wl,--(no-)?warn-([^,]+)`),
+	re(`-Wl,-z,(no)?execstack`),
+	re(`-Wl,-z,relro`),
+
+	re(`[a-zA-Z0-9_/].*\.(a|o|obj|dll|dylib|so)`), // direct linker inputs: x.o or libfoo.so (but not -foo.o or @foo.o)
+	re(`\./.*\.(a|o|obj|dll|dylib|so)`),
+}
+
+var validLinkerFlagsWithNextArg = []string{
+	"-arch",
+	"-F",
+	"-l",
+	"-L",
+	"-framework",
+	"-isysroot",
+	"--sysroot",
+	"-target",
+	"-Wl,-framework",
+	"-Wl,-rpath",
+	"-Wl,-R",
+	"-Wl,--just-symbols",
+	"-Wl,-undefined",
+}
+
+func checkCompilerFlags(name string, list []string) error {
+	return checkFlags(name, list, validCompilerFlags, validCompilerFlagsWithNextArg)
+}
+
+func checkLinkerFlags(name string, list []string) error {
+	return checkFlags(name, list, validLinkerFlags, validLinkerFlagsWithNextArg)
+}
+
+func checkFlags(name string, list []string, valid []*regexp.Regexp, validNext []string) error {
+	// Let users override rules with $CGO_CFLAGS_ALLOW, $CGO_CFLAGS_DISALLOW, etc.
+	var (
+		allow    *regexp.Regexp
+		disallow *regexp.Regexp
+	)
+	if env := os.Getenv("CGO_" + name + "_ALLOW"); env != "" {
+		r, err := regexp.Compile(env)
+		if err != nil {
+			return fmt.Errorf("parsing $CGO_%s_ALLOW: %v", name, err)
+		}
+		allow = r
+	}
+	if env := os.Getenv("CGO_" + name + "_DISALLOW"); env != "" {
+		r, err := regexp.Compile(env)
+		if err != nil {
+			return fmt.Errorf("parsing $CGO_%s_DISALLOW: %v", name, err)
+		}
+		disallow = r
+	}
+
+Args:
+	for i := 0; i < len(list); i++ {
+		arg := list[i]
+		if disallow != nil && disallow.FindString(arg) == arg {
+			goto Bad
+		}
+		if allow != nil && allow.FindString(arg) == arg {
+			continue Args
+		}
+		for _, re := range valid {
+			if re.FindString(arg) == arg { // must be complete match
+				continue Args
+			}
+		}
+		for _, x := range validNext {
+			if arg == x {
+				if i+1 < len(list) && safeArg(list[i+1]) {
+					i++
+					continue Args
+				}
+
+				// Permit -Wl,-framework -Wl,name.
+				if i+1 < len(list) &&
+					strings.HasPrefix(arg, "-Wl,") &&
+					strings.HasPrefix(list[i+1], "-Wl,") &&
+					safeArg(list[i+1][4:]) &&
+					!strings.Contains(list[i+1][4:], ",") {
+					i++
+					continue Args
+				}
+
+				if i+1 < len(list) {
+					return fmt.Errorf("invalid flag: %s %s (see https://golang.org/s/invalidflag)", arg, list[i+1])
+				}
+				return fmt.Errorf("invalid flag: %s without argument (see https://golang.org/s/invalidflag)", arg)
+			}
+		}
+	Bad:
+		return fmt.Errorf("invalid flag: %s", arg)
+	}
+	return nil
+}
+
+func safeArg(name string) bool {
+	if name == "" {
+		return false
+	}
+	c := name[0]
+	return '0' <= c && c <= '9' || 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' || c == '.' || c == '_' || c == '/' || c >= utf8.RuneSelf
+}

cgo/security_test.go

@@ -0,0 +1,260 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// This file has been copied from the Go 1.13 release tree.
+
+package cgo
+
+import (
+	"os"
+	"testing"
+)
+
+var goodCompilerFlags = [][]string{
+	{"-DFOO"},
+	{"-Dfoo=bar"},
+	{"-F/Qt"},
+	{"-I/"},
+	{"-I/etc/passwd"},
+	{"-I."},
+	{"-O"},
+	{"-O2"},
+	{"-Osmall"},
+	{"-W"},
+	{"-Wall"},
+	{"-fobjc-arc"},
+	{"-fno-objc-arc"},
+	{"-fomit-frame-pointer"},
+	{"-fno-omit-frame-pointer"},
+	{"-fpic"},
+	{"-fno-pic"},
+	{"-fPIC"},
+	{"-fno-PIC"},
+	{"-fpie"},
+	{"-fno-pie"},
+	{"-fPIE"},
+	{"-fno-PIE"},
+	{"-fsplit-stack"},
+	{"-fno-split-stack"},
+	{"-fstack-xxx"},
+	{"-fno-stack-xxx"},
+	{"-fsanitize=hands"},
+	{"-g"},
+	{"-ggdb"},
+	{"-march=souza"},
+	{"-mcpu=123"},
+	{"-mfpu=123"},
+	{"-mtune=happybirthday"},
+	{"-mstack-overflow"},
+	{"-mno-stack-overflow"},
+	{"-mmacosx-version"},
+	{"-mnop-fun-dllimport"},
+	{"-pthread"},
+	{"-std=c99"},
+	{"-xc"},
+	{"-D", "FOO"},
+	{"-D", "foo=bar"},
+	{"-I", "."},
+	{"-I", "/etc/passwd"},
+	{"-I", "世界"},
+	{"-framework", "Chocolate"},
+	{"-x", "c"},
+	{"-v"},
+}
+
+var badCompilerFlags = [][]string{
+	{"-D@X"},
+	{"-D-X"},
+	{"-F@dir"},
+	{"-F-dir"},
+	{"-I@dir"},
+	{"-I-dir"},
+	{"-O@1"},
+	{"-Wa,-foo"},
+	{"-W@foo"},
+	{"-g@gdb"},
+	{"-g-gdb"},
+	{"-march=@dawn"},
+	{"-march=-dawn"},
+	{"-std=@c99"},
+	{"-std=-c99"},
+	{"-x@c"},
+	{"-x-c"},
+	{"-D", "@foo"},
+	{"-D", "-foo"},
+	{"-I", "@foo"},
+	{"-I", "-foo"},
+	{"-framework", "-Caffeine"},
+	{"-framework", "@Home"},
+	{"-x", "--c"},
+	{"-x", "@obj"},
+}
+
+func TestCheckCompilerFlags(t *testing.T) {
+	for _, f := range goodCompilerFlags {
+		if err := checkCompilerFlags("test", f); err != nil {
+			t.Errorf("unexpected error for %q: %v", f, err)
+		}
+	}
+	for _, f := range badCompilerFlags {
+		if err := checkCompilerFlags("test", f); err == nil {
+			t.Errorf("missing error for %q", f)
+		}
+	}
+}
+
+var goodLinkerFlags = [][]string{
+	{"-Fbar"},
+	{"-lbar"},
+	{"-Lbar"},
+	{"-fpic"},
+	{"-fno-pic"},
+	{"-fPIC"},
+	{"-fno-PIC"},
+	{"-fpie"},
+	{"-fno-pie"},
+	{"-fPIE"},
+	{"-fno-PIE"},
+	{"-fsanitize=hands"},
+	{"-g"},
+	{"-ggdb"},
+	{"-march=souza"},
+	{"-mcpu=123"},
+	{"-mfpu=123"},
+	{"-mtune=happybirthday"},
+	{"-pic"},
+	{"-pthread"},
+	{"-Wl,-rpath,foo"},
+	{"-Wl,-rpath,$ORIGIN/foo"},
+	{"-Wl,-R", "/foo"},
+	{"-Wl,-R", "foo"},
+	{"-Wl,-R,foo"},
+	{"-Wl,--just-symbols=foo"},
+	{"-Wl,--just-symbols,foo"},
+	{"-Wl,--warn-error"},
+	{"-Wl,--no-warn-error"},
+	{"foo.so"},
+	{"_世界.dll"},
+	{"./x.o"},
+	{"libcgosotest.dylib"},
+	{"-F", "framework"},
+	{"-l", "."},
+	{"-l", "/etc/passwd"},
+	{"-l", "世界"},
+	{"-L", "framework"},
+	{"-framework", "Chocolate"},
+	{"-v"},
+	{"-Wl,-framework", "-Wl,Chocolate"},
+	{"-Wl,-framework,Chocolate"},
+	{"-Wl,-unresolved-symbols=ignore-all"},
+}
+
+var badLinkerFlags = [][]string{
+	{"-DFOO"},
+	{"-Dfoo=bar"},
+	{"-W"},
+	{"-Wall"},
+	{"-fobjc-arc"},
+	{"-fno-objc-arc"},
+	{"-fomit-frame-pointer"},
+	{"-fno-omit-frame-pointer"},
+	{"-fsplit-stack"},
+	{"-fno-split-stack"},
+	{"-fstack-xxx"},
+	{"-fno-stack-xxx"},
+	{"-mstack-overflow"},
+	{"-mno-stack-overflow"},
+	{"-mnop-fun-dllimport"},
+	{"-std=c99"},
+	{"-xc"},
+	{"-D", "FOO"},
+	{"-D", "foo=bar"},
+	{"-I", "FOO"},
+	{"-L", "@foo"},
+	{"-L", "-foo"},
+	{"-x", "c"},
+	{"-D@X"},
+	{"-D-X"},
+	{"-I@dir"},
+	{"-I-dir"},
+	{"-O@1"},
+	{"-Wa,-foo"},
+	{"-W@foo"},
+	{"-g@gdb"},
+	{"-g-gdb"},
+	{"-march=@dawn"},
+	{"-march=-dawn"},
+	{"-std=@c99"},
+	{"-std=-c99"},
+	{"-x@c"},
+	{"-x-c"},
+	{"-D", "@foo"},
+	{"-D", "-foo"},
+	{"-I", "@foo"},
+	{"-I", "-foo"},
+	{"-l", "@foo"},
+	{"-l", "-foo"},
+	{"-framework", "-Caffeine"},
+	{"-framework", "@Home"},
+	{"-Wl,-framework,-Caffeine"},
+	{"-Wl,-framework", "-Wl,@Home"},
+	{"-Wl,-framework", "@Home"},
+	{"-Wl,-framework,Chocolate,@Home"},
+	{"-x", "--c"},
+	{"-x", "@obj"},
+	{"-Wl,-rpath,@foo"},
+	{"-Wl,-R,foo,bar"},
+	{"-Wl,-R,@foo"},
+	{"-Wl,--just-symbols,@foo"},
+	{"../x.o"},
+}
+
+func TestCheckLinkerFlags(t *testing.T) {
+	for _, f := range goodLinkerFlags {
+		if err := checkLinkerFlags("test", f); err != nil {
+			t.Errorf("unexpected error for %q: %v", f, err)
+		}
+	}
+	for _, f := range badLinkerFlags {
+		if err := checkLinkerFlags("test", f); err == nil {
+			t.Errorf("missing error for %q", f)
+		}
+	}
+}
+
+func TestCheckFlagAllowDisallow(t *testing.T) {
+	if err := checkCompilerFlags("TEST", []string{"-disallow"}); err == nil {
+		t.Fatalf("missing error for -disallow")
+	}
+	os.Setenv("CGO_TEST_ALLOW", "-disallo")
+	if err := checkCompilerFlags("TEST", []string{"-disallow"}); err == nil {
+		t.Fatalf("missing error for -disallow with CGO_TEST_ALLOW=-disallo")
+	}
+	os.Setenv("CGO_TEST_ALLOW", "-disallow")
+	if err := checkCompilerFlags("TEST", []string{"-disallow"}); err != nil {
+		t.Fatalf("unexpected error for -disallow with CGO_TEST_ALLOW=-disallow: %v", err)
+	}
+	os.Unsetenv("CGO_TEST_ALLOW")
+
+	if err := checkCompilerFlags("TEST", []string{"-Wall"}); err != nil {
+		t.Fatalf("unexpected error for -Wall: %v", err)
+	}
+	os.Setenv("CGO_TEST_DISALLOW", "-Wall")
+	if err := checkCompilerFlags("TEST", []string{"-Wall"}); err == nil {
+		t.Fatalf("missing error for -Wall with CGO_TEST_DISALLOW=-Wall")
+	}
+	os.Setenv("CGO_TEST_ALLOW", "-Wall") // disallow wins
+	if err := checkCompilerFlags("TEST", []string{"-Wall"}); err == nil {
+		t.Fatalf("missing error for -Wall with CGO_TEST_DISALLOW=-Wall and CGO_TEST_ALLOW=-Wall")
+	}
+
+	os.Setenv("CGO_TEST_ALLOW", "-fplugin.*")
+	os.Setenv("CGO_TEST_DISALLOW", "-fplugin=lint.so")
+	if err := checkCompilerFlags("TEST", []string{"-fplugin=faster.so"}); err != nil {
+		t.Fatalf("unexpected error for -fplugin=faster.so: %v", err)
+	}
+	if err := checkCompilerFlags("TEST", []string{"-fplugin=lint.so"}); err == nil {
+		t.Fatalf("missing error for -fplugin=lint.so: %v", err)
+	}
+}

cgo/testdata/basic.go

@@ -0,0 +1,3 @@
+package main
+
+import "C"

cgo/testdata/basic.out.go

@@ -0,0 +1,26 @@
+package main
+
+import "unsafe"
+
+var _ unsafe.Pointer
+
+type C.int16_t = int16
+type C.int32_t = int32
+type C.int64_t = int64
+type C.int8_t = int8
+type C.uint16_t = uint16
+type C.uint32_t = uint32
+type C.uint64_t = uint64
+type C.uint8_t = uint8
+type C.uintptr_t = uintptr
+type C.char uint8
+type C.int int32
+type C.long int32
+type C.longlong int64
+type C.schar int8
+type C.short int16
+type C.uchar uint8
+type C.uint uint32
+type C.ulong uint32
+type C.ulonglong uint64
+type C.ushort uint16

cgo/testdata/const.go

@@ -0,0 +1,12 @@
+package main
+
+/*
+#define foo 3
+#define bar foo
+*/
+import "C"
+
+const (
+	Foo = C.foo
+	Bar = C.bar
+)

cgo/testdata/const.out.go

@@ -0,0 +1,29 @@
+package main
+
+import "unsafe"
+
+var _ unsafe.Pointer
+
+const C.bar = C.foo
+const C.foo = 3
+
+type C.int16_t = int16
+type C.int32_t = int32
+type C.int64_t = int64
+type C.int8_t = int8
+type C.uint16_t = uint16
+type C.uint32_t = uint32
+type C.uint64_t = uint64
+type C.uint8_t = uint8
+type C.uintptr_t = uintptr
+type C.char uint8
+type C.int int32
+type C.long int32
+type C.longlong int64
+type C.schar int8
+type C.short int16
+type C.uchar uint8
+type C.uint uint32
+type C.ulong uint32
+type C.ulonglong uint64
+type C.ushort uint16

cgo/testdata/errors.go

@@ -0,0 +1,33 @@
+package main
+
+/*
+#warning some warning
+
+typedef struct {
+	int x;
+	int y;
+} point_t;
+
+typedef someType noType; // undefined type
+
+#define SOME_CONST_1 5) // invalid const syntax
+#define SOME_CONST_2 6) // const not used (so no error)
+#define SOME_CONST_3 1234 // const too large for byte
+*/
+import "C"
+
+// Make sure that errors for the following lines won't change with future
+// additions to the CGo preamble.
+//line errors.go:100
+var (
+	// constant too large
+	_ C.uint8_t = 2 << 10
+
+	// z member does not exist
+	_ C.point_t = C.point_t{z: 3}
+
+	// constant has syntax error
+	_ = C.SOME_CONST_1
+
+	_ byte = C.SOME_CONST_3
+)

cgo/testdata/errors.out.go

@@ -0,0 +1,43 @@
+// CGo errors:
+//     testdata/errors.go:4:2: warning: some warning
+//     testdata/errors.go:11:9: error: unknown type name 'someType'
+//     testdata/errors.go:13:23: unexpected token )
+
+// Type checking errors after CGo processing:
+//     testdata/errors.go:102: 2 << 10 (untyped int constant 2048) overflows uint8
+//     testdata/errors.go:105: unknown field z in struct literal
+//     testdata/errors.go:108: undeclared name: C.SOME_CONST_1
+//     testdata/errors.go:110: C.SOME_CONST_3 (untyped int constant 1234) overflows byte
+
+package main
+
+import "unsafe"
+
+var _ unsafe.Pointer
+
+const C.SOME_CONST_3 = 1234
+
+type C.int16_t = int16
+type C.int32_t = int32
+type C.int64_t = int64
+type C.int8_t = int8
+type C.uint16_t = uint16
+type C.uint32_t = uint32
+type C.uint64_t = uint64
+type C.uint8_t = uint8
+type C.uintptr_t = uintptr
+type C.char uint8
+type C.int int32
+type C.long int32
+type C.longlong int64
+type C.schar int8
+type C.short int16
+type C.uchar uint8
+type C.uint uint32
+type C.ulong uint32
+type C.ulonglong uint64
+type C.ushort uint16
+type C.point_t = struct {
+	x C.int
+	y C.int
+}

cgo/testdata/flags.go

@@ -0,0 +1,26 @@
+package main
+
+/*
+// this name doesn't exist
+#cgo  NOFLAGS: -foo
+
+// unknown flag
+#cgo CFLAGS: -fdoes-not-exist -DNOTDEFINED
+
+#cgo CFLAGS: -DFOO
+
+#if defined(FOO)
+#define BAR 3
+#else
+#define BAR 5
+#endif
+
+#if defined(NOTDEFINED)
+#warning flag must not be defined
+#endif
+*/
+import "C"
+
+var (
+	_ = C.BAR
+)

cgo/testdata/flags.out.go

@@ -0,0 +1,32 @@
+// CGo errors:
+//     testdata/flags.go:5:7: invalid #cgo line: NOFLAGS
+//     testdata/flags.go:8:13: invalid flag: -fdoes-not-exist
+
+package main
+
+import "unsafe"
+
+var _ unsafe.Pointer
+
+const C.BAR = 3
+
+type C.int16_t = int16
+type C.int32_t = int32
+type C.int64_t = int64
+type C.int8_t = int8
+type C.uint16_t = uint16
+type C.uint32_t = uint32
+type C.uint64_t = uint64
+type C.uint8_t = uint8
+type C.uintptr_t = uintptr
+type C.char uint8
+type C.int int32
+type C.long int32
+type C.longlong int64
+type C.schar int8
+type C.short int16
+type C.uchar uint8
+type C.uint uint32
+type C.ulong uint32
+type C.ulonglong uint64
+type C.ushort uint16

cgo/testdata/types.go

@@ -0,0 +1,165 @@
+package main
+
+/*
+// Simple typedef.
+typedef int myint;
+
+// Structs, with or without name.
+typedef struct {
+	int x;
+	int y;
+} point2d_t;
+typedef struct point3d {
+	int x;
+	int y;
+	int z;
+} point3d_t;
+
+// Structs with reserved field names.
+struct type1 {
+	// All these fields should be renamed.
+	int type;
+	int _type;
+	int __type;
+};
+struct type2 {
+	// This field should not be renamed.
+	int _type;
+};
+
+// Unions.
+typedef union {
+	// Union should be treated as a struct.
+	int i;
+} union1_t;
+typedef union {
+	// Union must contain a single field and have special getters/setters.
+	int    i;
+	double d;
+	short  s;
+} union3_t;
+typedef union union2d {
+	int i;
+	double d[2];
+} union2d_t;
+typedef union {
+	unsigned char arr[10];
+} unionarray_t;
+
+// Nested structs and unions.
+typedef struct {
+	point2d_t begin;
+	point2d_t end;
+	int       tag;
+	union {
+		point2d_t area;
+		point3d_t solid;
+	} coord;
+} struct_nested_t;
+typedef union {
+	point3d_t    point;
+	unionarray_t array;
+	union3_t     thing;
+} union_nested_t;
+
+// Enums. These define constant numbers. All these constants must be given the
+// correct number.
+typedef enum option {
+	optionA,
+	optionB,
+	optionC = -5,
+	optionD,
+	optionE = 10,
+	optionF,
+	optionG,
+} option_t;
+enum unused {
+	unused1 = 5,
+};
+
+// Anonymous enum.
+typedef enum {
+	option2A = 20,
+} option2_t;
+
+// Various types that are usually translated directly to Go types, but storing
+// them in a struct reveals them.
+typedef struct {
+	float  f;
+	double d;
+	int    *ptr;
+} types_t;
+
+// Arrays.
+typedef int myIntArray[10];
+
+// Bitfields.
+typedef struct {
+	unsigned char start;
+	unsigned char a : 5;
+	unsigned char b : 1;
+	unsigned char c : 2;
+	unsigned char :0; // new field
+	unsigned char d : 6;
+	unsigned char e : 3;
+	// Note that C++ allows bitfields bigger than the underlying type.
+} bitfield_t;
+*/
+import "C"
+
+var (
+	// Simple typedefs.
+	_ C.myint
+
+	// Structs.
+	_ C.point2d_t
+	_ C.point3d_t
+	_ C.struct_point3d
+
+	// Structs with reserved field names.
+	_ C.struct_type1
+	_ C.struct_type2
+
+	// Unions.
+	_ C.union1_t
+	_ C.union3_t
+	_ C.union2d_t
+	_ C.unionarray_t
+
+	// Nested structs and unions.
+	_ C.struct_nested_t
+	_ C.union_nested_t
+
+	// Enums (anonymous and named).
+	_ C.option_t
+	_ C.enum_option
+	_ C.option2_t
+
+	// Various types.
+	_ C.types_t
+
+	// Arrays.
+	_ C.myIntArray
+)
+
+// Test bitfield accesses.
+func accessBitfields() {
+	var x C.bitfield_t
+	x.start = 3
+	x.set_bitfield_a(4)
+	x.set_bitfield_b(1)
+	x.set_bitfield_c(2)
+	x.d = 10
+	x.e = 5
+	var _ C.uchar = x.bitfield_a()
+}
+
+// Test union accesses.
+func accessUnion() {
+	var union1 C.union1_t
+	union1.i = 5
+
+	var union2d C.union2d_t
+	var _ *C.int = union2d.unionfield_i()
+	var _ *[2]float64 = union2d.unionfield_d()
+}

cgo/testdata/types.out.go

@@ -0,0 +1,119 @@
+package main
+
+import "unsafe"
+
+var _ unsafe.Pointer
+
+const C.option2A = 20
+const C.optionA = 0
+const C.optionB = 1
+const C.optionC = -5
+const C.optionD = -4
+const C.optionE = 10
+const C.optionF = 11
+const C.optionG = 12
+const C.unused1 = 5
+
+type C.int16_t = int16
+type C.int32_t = int32
+type C.int64_t = int64
+type C.int8_t = int8
+type C.uint16_t = uint16
+type C.uint32_t = uint32
+type C.uint64_t = uint64
+type C.uint8_t = uint8
+type C.uintptr_t = uintptr
+type C.char uint8
+type C.int int32
+type C.long int32
+type C.longlong int64
+type C.schar int8
+type C.short int16
+type C.uchar uint8
+type C.uint uint32
+type C.ulong uint32
+type C.ulonglong uint64
+type C.ushort uint16
+type C.bitfield_t = C.struct_4
+type C.myIntArray = [10]C.int
+type C.myint = C.int
+type C.option2_t = C.uint
+type C.option_t = C.enum_option
+type C.point2d_t = struct {
+	x C.int
+	y C.int
+}
+type C.point3d_t = C.struct_point3d
+type C.struct_nested_t = struct {
+	begin C.point2d_t
+	end   C.point2d_t
+	tag   C.int
+
+	coord C.union_2
+}
+type C.types_t = struct {
+	f   float32
+	d   float64
+	ptr *C.int
+}
+type C.union1_t = struct{ i C.int }
+type C.union2d_t = C.union_union2d
+type C.union3_t = C.union_1
+type C.union_nested_t = C.union_3
+type C.unionarray_t = struct{ arr [10]C.uchar }
+
+func (s *C.struct_4) bitfield_a() C.uchar          { return s.__bitfield_1 & 0x1f }
+func (s *C.struct_4) set_bitfield_a(value C.uchar) { s.__bitfield_1 = s.__bitfield_1&^0x1f | value&0x1f<<0 }
+func (s *C.struct_4) bitfield_b() C.uchar {
+	return s.__bitfield_1 >> 5 & 0x1
+}
+func (s *C.struct_4) set_bitfield_b(value C.uchar) { s.__bitfield_1 = s.__bitfield_1&^0x20 | value&0x1<<5 }
+func (s *C.struct_4) bitfield_c() C.uchar {
+	return s.__bitfield_1 >> 6
+}
+func (s *C.struct_4) set_bitfield_c(value C.uchar,
+
+) { s.__bitfield_1 = s.__bitfield_1&0x3f | value<<6 }
+
+type C.struct_4 struct {
+	start        C.uchar
+	__bitfield_1 C.uchar
+
+	d C.uchar
+	e C.uchar
+}
+type C.struct_point3d struct {
+	x C.int
+	y C.int
+	z C.int
+}
+type C.struct_type1 struct {
+	_type   C.int
+	__type  C.int
+	___type C.int
+}
+type C.struct_type2 struct{ _type C.int }
+
+func (union *C.union_1) unionfield_i() *C.int   { return (*C.int)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_1) unionfield_d() *float64 { return (*float64)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_1) unionfield_s() *C.short { return (*C.short)(unsafe.Pointer(&union.$union)) }
+
+type C.union_1 struct{ $union uint64 }
+
+func (union *C.union_2) unionfield_area() *C.point2d_t  { return (*C.point2d_t)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_2) unionfield_solid() *C.point3d_t { return (*C.point3d_t)(unsafe.Pointer(&union.$union)) }
+
+type C.union_2 struct{ $union [3]uint32 }
+
+func (union *C.union_3) unionfield_point() *C.point3d_t    { return (*C.point3d_t)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_3) unionfield_array() *C.unionarray_t { return (*C.unionarray_t)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_3) unionfield_thing() *C.union3_t     { return (*C.union3_t)(unsafe.Pointer(&union.$union)) }
+
+type C.union_3 struct{ $union [2]uint64 }
+
+func (union *C.union_union2d) unionfield_i() *C.int      { return (*C.int)(unsafe.Pointer(&union.$union)) }
+func (union *C.union_union2d) unionfield_d() *[2]float64 { return (*[2]float64)(unsafe.Pointer(&union.$union)) }
+
+type C.union_union2d struct{ $union [2]uint64 }
+type C.enum_option C.int
+type C.enum_unused C.uint

commands.go

@@ -1,33 +0,0 @@
-package main
-
-import (
-	"errors"
-	"os"
-	"os/exec"
-	"strings"
-)
-
-// Commands used by the compilation process might have different file names
-// across operating systems and distributions.
-var commands = map[string][]string{
-	"clang":   {"clang-8"},
-	"ld.lld":  {"ld.lld-8", "ld.lld"},
-	"wasm-ld": {"wasm-ld-8", "wasm-ld"},
-}
-
-func execCommand(cmdNames []string, args ...string) error {
-	for _, cmdName := range cmdNames {
-		cmd := exec.Command(cmdName, args...)
-		cmd.Stdout = os.Stdout
-		cmd.Stderr = os.Stderr
-		err := cmd.Run()
-		if err != nil {
-			if err, ok := err.(*exec.Error); ok && err.Err == exec.ErrNotFound {
-				// this command was not found, try the next
-				continue
-			}
-		}
-		return nil
-	}
-	return errors.New("none of these commands were found in your $PATH: " + strings.Join(cmdNames, " "))
-}

compileopts/config.go

@@ -0,0 +1,221 @@
+// Package compileopts contains the configuration for a single to-be-built
+// binary.
+package compileopts
+
+import (
+	"errors"
+	"fmt"
+	"regexp"
+	"strconv"
+	"strings"
+
+	"github.com/tinygo-org/tinygo/goenv"
+)
+
+// Config keeps all configuration affecting the build in a single struct.
+type Config struct {
+	Options        *Options
+	Target         *TargetSpec
+	GoMinorVersion int
+	ClangHeaders   string // Clang built-in header include path
+	TestConfig     TestConfig
+}
+
+// Triple returns the LLVM target triple, like armv6m-none-eabi.
+func (c *Config) Triple() string {
+	return c.Target.Triple
+}
+
+// CPU returns the LLVM CPU name, like atmega328p or arm7tdmi. It may return an
+// empty string if the CPU name is not known.
+func (c *Config) CPU() string {
+	return c.Target.CPU
+}
+
+// Features returns a list of features this CPU supports. For example, for a
+// RISC-V processor, that could be ["+a", "+c", "+m"]. For many targets, an
+// empty list will be returned.
+func (c *Config) Features() []string {
+	return c.Target.Features
+}
+
+// GOOS returns the GOOS of the target. This might not always be the actual OS:
+// for example, bare-metal targets will usually pretend to be linux to get the
+// standard library to compile.
+func (c *Config) GOOS() string {
+	return c.Target.GOOS
+}
+
+// GOARCH returns the GOARCH of the target. This might not always be the actual
+// archtecture: for example, the AVR target is not supported by the Go standard
+// library so such targets will usually pretend to be linux/arm.
+func (c *Config) GOARCH() string {
+	return c.Target.GOARCH
+}
+
+// BuildTags returns the complete list of build tags used during this build.
+func (c *Config) BuildTags() []string {
+	tags := append(c.Target.BuildTags, []string{"tinygo", "gc." + c.GC(), "scheduler." + c.Scheduler()}...)
+	for i := 1; i <= c.GoMinorVersion; i++ {
+		tags = append(tags, fmt.Sprintf("go1.%d", i))
+	}
+	if extraTags := strings.Fields(c.Options.Tags); len(extraTags) != 0 {
+		tags = append(tags, extraTags...)
+	}
+	return tags
+}
+
+// GC returns the garbage collection strategy in use on this platform. Valid
+// values are "none", "leaking", and "conservative".
+func (c *Config) GC() string {
+	if c.Options.GC != "" {
+		return c.Options.GC
+	}
+	if c.Target.GC != "" {
+		return c.Target.GC
+	}
+	return "conservative"
+}
+
+// NeedsStackObjects returns true if the compiler should insert stack objects
+// that can be traced by the garbage collector.
+func (c *Config) NeedsStackObjects() bool {
+	if c.GC() != "conservative" {
+		return false
+	}
+	for _, tag := range c.BuildTags() {
+		if tag == "baremetal" {
+			return false
+		}
+	}
+
+	return true
+}
+
+// Scheduler returns the scheduler implementation. Valid values are "coroutines"
+// and "tasks".
+func (c *Config) Scheduler() string {
+	if c.Options.Scheduler != "" {
+		return c.Options.Scheduler
+	}
+	if c.Target.Scheduler != "" {
+		return c.Target.Scheduler
+	}
+	// Fall back to coroutines, which are supported everywhere.
+	return "coroutines"
+}
+
+// PanicStrategy returns the panic strategy selected for this target. Valid
+// values are "print" (print the panic value, then exit) or "trap" (issue a trap
+// instruction).
+func (c *Config) PanicStrategy() string {
+	return c.Options.PanicStrategy
+}
+
+// CFlags returns the flags to pass to the C compiler. This is necessary for CGo
+// preprocessing.
+func (c *Config) CFlags() []string {
+	cflags := append([]string{}, c.Options.CFlags...)
+	for _, flag := range c.Target.CFlags {
+		cflags = append(cflags, strings.Replace(flag, "{root}", goenv.Get("TINYGOROOT"), -1))
+	}
+	return cflags
+}
+
+// LDFlags returns the flags to pass to the linker. A few more flags are needed
+// (like the one for the compiler runtime), but this represents the majority of
+// the flags.
+func (c *Config) LDFlags() []string {
+	root := goenv.Get("TINYGOROOT")
+	// Merge and adjust LDFlags.
+	ldflags := append([]string{}, c.Options.LDFlags...)
+	for _, flag := range c.Target.LDFlags {
+		ldflags = append(ldflags, strings.Replace(flag, "{root}", root, -1))
+	}
+	ldflags = append(ldflags, "-L", root)
+	if c.Target.GOARCH == "wasm" {
+		// Round heap size to next multiple of 65536 (the WebAssembly page
+		// size).
+		heapSize := (c.Options.HeapSize + (65536 - 1)) &^ (65536 - 1)
+		ldflags = append(ldflags, "--initial-memory="+strconv.FormatInt(heapSize, 10))
+	}
+	if c.Target.LinkerScript != "" {
+		ldflags = append(ldflags, "-T", c.Target.LinkerScript)
+	}
+	return ldflags
+}
+
+// ExtraFiles returns the list of extra files to be built and linked with the
+// executable. This can include extra C and assembly files.
+func (c *Config) ExtraFiles() []string {
+	return c.Target.ExtraFiles
+}
+
+// DumpSSA returns whether to dump Go SSA while compiling (-dumpssa flag). Only
+// enable this for debugging.
+func (c *Config) DumpSSA() bool {
+	return c.Options.DumpSSA
+}
+
+// VerifyIR returns whether to run extra checks on the IR. This is normally
+// disabled but enabled during testing.
+func (c *Config) VerifyIR() bool {
+	return c.Options.VerifyIR
+}
+
+// Debug returns whether to add debug symbols to the IR, for debugging with GDB
+// and similar.
+func (c *Config) Debug() bool {
+	return c.Options.Debug
+}
+
+// Programmer returns the flash method and OpenOCD interface name given a
+// particular configuration. It may either be all configured in the target JSON
+// file or be modified using the -programmmer command-line option.
+func (c *Config) Programmer() (method, openocdInterface string) {
+	switch c.Options.Programmer {
+	case "":
+		// No configuration supplied.
+		return c.Target.FlashMethod, c.Target.OpenOCDInterface
+	case "openocd", "msd", "command":
+		// The -programmer flag only specifies the flash method.
+		return c.Options.Programmer, c.Target.OpenOCDInterface
+	default:
+		// The -programmer flag specifies something else, assume it specifies
+		// the OpenOCD interface name.
+		return "openocd", c.Options.Programmer
+	}
+}
+
+// OpenOCDConfiguration returns a list of command line arguments to OpenOCD.
+// This list of command-line arguments is based on the various OpenOCD-related
+// flags in the target specification.
+func (c *Config) OpenOCDConfiguration() (args []string, err error) {
+	_, openocdInterface := c.Programmer()
+	if openocdInterface == "" {
+		return nil, errors.New("OpenOCD programmer not set")
+	}
+	if !regexp.MustCompile("^[\\p{L}0-9_-]+$").MatchString(openocdInterface) {
+		return nil, fmt.Errorf("OpenOCD programmer has an invalid name: %#v", openocdInterface)
+	}
+	if c.Target.OpenOCDTarget == "" {
+		return nil, errors.New("OpenOCD chip not set")
+	}
+	if !regexp.MustCompile("^[\\p{L}0-9_-]+$").MatchString(c.Target.OpenOCDTarget) {
+		return nil, fmt.Errorf("OpenOCD target has an invalid name: %#v", c.Target.OpenOCDTarget)
+	}
+	if c.Target.OpenOCDTransport != "" && c.Target.OpenOCDTransport != "swd" {
+		return nil, fmt.Errorf("unknown OpenOCD transport: %#v", c.Target.OpenOCDTransport)
+	}
+	args = []string{"-f", "interface/" + openocdInterface + ".cfg"}
+	if c.Target.OpenOCDTransport != "" {
+		args = append(args, "-c", "transport select "+c.Target.OpenOCDTransport)
+	}
+	args = append(args, "-f", "target/"+c.Target.OpenOCDTarget+".cfg")
+	return args, nil
+}
+
+type TestConfig struct {
+	CompileTestBinary bool
+	// TODO: Filter the test functions to run, include verbose flag, etc
+}

compileopts/options.go

@@ -0,0 +1,23 @@
+package compileopts
+
+// Options contains extra options to give to the compiler. These options are
+// usually passed from the command line.
+type Options struct {
+	Target        string
+	Opt           string
+	GC            string
+	PanicStrategy string
+	Scheduler     string
+	PrintIR       bool
+	DumpSSA       bool
+	VerifyIR      bool
+	Debug         bool
+	PrintSizes    string
+	CFlags        []string
+	LDFlags       []string
+	Tags          string
+	WasmAbi       string
+	HeapSize      int64
+	TestConfig    TestConfig
+	Programmer    string
+}

compileopts/target.go

@@ -1,16 +1,17 @@
-package main
+package compileopts
+
+// This file loads a target specification from a JSON file.
 
 import (
 	"encoding/json"
 	"errors"
-	"fmt"
 	"io"
 	"os"
-	"os/exec"
-	"os/user"
 	"path/filepath"
 	"runtime"
 	"strings"
+
+	"github.com/tinygo-org/tinygo/goenv"
 )
 
 // Target specification for a given target. Used for bare metal targets.
@@ -19,24 +20,32 @@ import (
 // https://doc.rust-lang.org/nightly/nightly-rustc/rustc_target/spec/struct.TargetOptions.html
 // https://github.com/shepmaster/rust-arduino-blink-led-no-core-with-cargo/blob/master/blink/arduino.json
 type TargetSpec struct {
-	Inherits   []string `json:"inherits"`
-	Triple     string   `json:"llvm-target"`
-	CPU        string   `json:"cpu"`
-	GOOS       string   `json:"goos"`
-	GOARCH     string   `json:"goarch"`
-	BuildTags  []string `json:"build-tags"`
-	GC         string   `json:"gc"`
-	Compiler   string   `json:"compiler"`
-	Linker     string   `json:"linker"`
-	RTLib      string   `json:"rtlib"` // compiler runtime library (libgcc, compiler-rt)
-	CFlags     []string `json:"cflags"`
-	LDFlags    []string `json:"ldflags"`
-	ExtraFiles []string `json:"extra-files"`
-	Emulator   []string `json:"emulator"`
-	Flasher    string   `json:"flash"`
-	OCDDaemon  []string `json:"ocd-daemon"`
-	GDB        string   `json:"gdb"`
-	GDBCmds    []string `json:"gdb-initial-cmds"`
+	Inherits         []string `json:"inherits"`
+	Triple           string   `json:"llvm-target"`
+	CPU              string   `json:"cpu"`
+	Features         []string `json:"features"`
+	GOOS             string   `json:"goos"`
+	GOARCH           string   `json:"goarch"`
+	BuildTags        []string `json:"build-tags"`
+	GC               string   `json:"gc"`
+	Scheduler        string   `json:"scheduler"`
+	Compiler         string   `json:"compiler"`
+	Linker           string   `json:"linker"`
+	RTLib            string   `json:"rtlib"` // compiler runtime library (libgcc, compiler-rt)
+	CFlags           []string `json:"cflags"`
+	LDFlags          []string `json:"ldflags"`
+	LinkerScript     string   `json:"linkerscript"`
+	ExtraFiles       []string `json:"extra-files"`
+	Emulator         []string `json:"emulator"`
+	FlashCommand     string   `json:"flash-command"`
+	GDB              string   `json:"gdb"`
+	PortReset        string   `json:"flash-1200-bps-reset"`
+	FlashMethod      string   `json:"flash-method"`
+	FlashVolume      string   `json:"msd-volume-name"`
+	FlashFilename    string   `json:"msd-firmware-name"`
+	OpenOCDInterface string   `json:"openocd-interface"`
+	OpenOCDTarget    string   `json:"openocd-target"`
+	OpenOCDTransport string   `json:"openocd-transport"`
 }
 
 // copyProperties copies all properties that are set in spec2 into itself.
@@ -50,6 +59,7 @@ func (spec *TargetSpec) copyProperties(spec2 *TargetSpec) {
 	if spec2.CPU != "" {
 		spec.CPU = spec2.CPU
 	}
+	spec.Features = append(spec.Features, spec2.Features...)
 	if spec2.GOOS != "" {
 		spec.GOOS = spec2.GOOS
 	}
@@ -60,6 +70,9 @@ func (spec *TargetSpec) copyProperties(spec2 *TargetSpec) {
 	if spec2.GC != "" {
 		spec.GC = spec2.GC
 	}
+	if spec2.Scheduler != "" {
+		spec.Scheduler = spec2.Scheduler
+	}
 	if spec2.Compiler != "" {
 		spec.Compiler = spec2.Compiler
 	}
@@ -71,21 +84,39 @@ func (spec *TargetSpec) copyProperties(spec2 *TargetSpec) {
 	}
 	spec.CFlags = append(spec.CFlags, spec2.CFlags...)
 	spec.LDFlags = append(spec.LDFlags, spec2.LDFlags...)
+	if spec2.LinkerScript != "" {
+		spec.LinkerScript = spec2.LinkerScript
+	}
 	spec.ExtraFiles = append(spec.ExtraFiles, spec2.ExtraFiles...)
 	if len(spec2.Emulator) != 0 {
 		spec.Emulator = spec2.Emulator
 	}
-	if spec2.Flasher != "" {
-		spec.Flasher = spec2.Flasher
-	}
-	if len(spec2.OCDDaemon) != 0 {
-		spec.OCDDaemon = spec2.OCDDaemon
+	if spec2.FlashCommand != "" {
+		spec.FlashCommand = spec2.FlashCommand
 	}
 	if spec2.GDB != "" {
 		spec.GDB = spec2.GDB
 	}
-	if len(spec2.GDBCmds) != 0 {
-		spec.GDBCmds = spec2.GDBCmds
+	if spec2.PortReset != "" {
+		spec.PortReset = spec2.PortReset
+	}
+	if spec2.FlashMethod != "" {
+		spec.FlashMethod = spec2.FlashMethod
+	}
+	if spec2.FlashVolume != "" {
+		spec.FlashVolume = spec2.FlashVolume
+	}
+	if spec2.FlashFilename != "" {
+		spec.FlashFilename = spec2.FlashFilename
+	}
+	if spec2.OpenOCDInterface != "" {
+		spec.OpenOCDInterface = spec2.OpenOCDInterface
+	}
+	if spec2.OpenOCDTarget != "" {
+		spec.OpenOCDTarget = spec2.OpenOCDTarget
+	}
+	if spec2.OpenOCDTransport != "" {
+		spec.OpenOCDTransport = spec2.OpenOCDTransport
 	}
 }
 
@@ -100,10 +131,18 @@ func (spec *TargetSpec) load(r io.Reader) error {
 	return nil
 }
 
-// loadFromName loads the given target from the targets/ directory inside the
-// compiler sources.
-func (spec *TargetSpec) loadFromName(name string) error {
-	path := filepath.Join(sourceDir(), "targets", strings.ToLower(name)+".json")
+// loadFromGivenStr loads the TargetSpec from the given string that could be:
+// - targets/ directory inside the compiler sources
+// - a relative or absolute path to custom (project specific) target specification .json file;
+//   the Inherits[] could contain the files from target folder (ex. stm32f4disco)
+//   as well as path to custom files (ex. myAwesomeProject.json)
+func (spec *TargetSpec) loadFromGivenStr(str string) error {
+	path := ""
+	if strings.HasSuffix(str, ".json") {
+		path, _ = filepath.Abs(str)
+	} else {
+		path = filepath.Join(goenv.Get("TINYGOROOT"), "targets", strings.ToLower(str)+".json")
+	}
 	fp, err := os.Open(path)
 	if err != nil {
 		return err
@@ -118,7 +157,7 @@ func (spec *TargetSpec) resolveInherits() error {
 	newSpec := &TargetSpec{}
 	for _, name := range spec.Inherits {
 		subtarget := &TargetSpec{}
-		err := subtarget.loadFromName(name)
+		err := subtarget.loadFromGivenStr(name)
 		if err != nil {
 			return err
 		}
@@ -141,14 +180,8 @@ func LoadTarget(target string) (*TargetSpec, error) {
 	if target == "" {
 		// Configure based on GOOS/GOARCH environment variables (falling back to
 		// runtime.GOOS/runtime.GOARCH), and generate a LLVM target based on it.
-		goos := os.Getenv("GOOS")
-		if goos == "" {
-			goos = runtime.GOOS
-		}
-		goarch := os.Getenv("GOARCH")
-		if goarch == "" {
-			goarch = runtime.GOARCH
-		}
+		goos := goenv.Get("GOOS")
+		goarch := goenv.Get("GOARCH")
 		llvmos := goos
 		llvmarch := map[string]string{
 			"386":   "i386",
@@ -168,7 +201,7 @@ func LoadTarget(target string) (*TargetSpec, error) {
 	// See whether there is a target specification for this target (e.g.
 	// Arduino).
 	spec := &TargetSpec{}
-	err := spec.loadFromName(target)
+	err := spec.loadFromGivenStr(target)
 	if err == nil {
 		// Successfully loaded this target from a built-in .json file. Make sure
 		// it includes all parents as specified in the "inherits" key.
@@ -185,7 +218,7 @@ func LoadTarget(target string) (*TargetSpec, error) {
 		// Load target from given triple, ignore GOOS/GOARCH environment
 		// variables.
 		tripleSplit := strings.Split(target, "-")
-		if len(tripleSplit) == 1 {
+		if len(tripleSplit) < 3 {
 			return nil, errors.New("expected a full LLVM target or a custom target in -target flag")
 		}
 		goos := tripleSplit[2]
@@ -208,14 +241,15 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 	// No target spec available. Use the default one, useful on most systems
 	// with a regular OS.
 	spec := TargetSpec{
-		Triple:    triple,
-		GOOS:      goos,
-		GOARCH:    goarch,
-		BuildTags: []string{goos, goarch},
-		Compiler:  "clang",
-		Linker:    "cc",
-		GDB:       "gdb",
-		GDBCmds:   []string{"run"},
+		Triple:      triple,
+		GOOS:        goos,
+		GOARCH:      goarch,
+		BuildTags:   []string{goos, goarch},
+		Compiler:    "clang",
+		Linker:      "cc",
+		GDB:         "gdb",
+		PortReset:   "false",
+		FlashMethod: "native",
 	}
 	if goos == "darwin" {
 		spec.LDFlags = append(spec.LDFlags, "-Wl,-dead_strip")
@@ -241,130 +275,3 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 	}
 	return &spec, nil
 }
-
-// Return the TINYGOROOT, or exit with an error.
-func sourceDir() string {
-	// Use $TINYGOROOT as root, if available.
-	root := os.Getenv("TINYGOROOT")
-	if root != "" {
-		if !isSourceDir(root) {
-			fmt.Fprintln(os.Stderr, "error: $TINYGOROOT was not set to the correct root")
-			os.Exit(1)
-		}
-		return root
-	}
-
-	// Find root from executable path.
-	path, err := os.Executable()
-	if err != nil {
-		// Very unlikely. Bail out if it happens.
-		panic("could not get executable path: " + err.Error())
-	}
-	root = filepath.Dir(filepath.Dir(path))
-	if isSourceDir(root) {
-		return root
-	}
-
-	// Fallback: use the original directory from where it was built
-	// https://stackoverflow.com/a/32163888/559350
-	_, path, _, _ = runtime.Caller(0)
-	root = filepath.Dir(path)
-	if isSourceDir(root) {
-		return root
-	}
-
-	fmt.Fprintln(os.Stderr, "error: could not autodetect root directory, set the TINYGOROOT environment variable to override")
-	os.Exit(1)
-	panic("unreachable")
-}
-
-// isSourceDir returns true if the directory looks like a TinyGo source directory.
-func isSourceDir(root string) bool {
-	_, err := os.Stat(filepath.Join(root, "src/runtime/internal/sys/zversion.go"))
-	if err != nil {
-		return false
-	}
-	_, err = os.Stat(filepath.Join(root, "src/device/arm/arm.go"))
-	return err == nil
-}
-
-func getGopath() string {
-	gopath := os.Getenv("GOPATH")
-	if gopath != "" {
-		return gopath
-	}
-
-	// fallback
-	home := getHomeDir()
-	return filepath.Join(home, "go")
-}
-
-func getHomeDir() string {
-	u, err := user.Current()
-	if err != nil {
-		panic("cannot get current user: " + err.Error())
-	}
-	if u.HomeDir == "" {
-		// This is very unlikely, so panic here.
-		// Not the nicest solution, however.
-		panic("could not find home directory")
-	}
-	return u.HomeDir
-}
-
-// getGoroot returns an appropriate GOROOT from various sources. If it can't be
-// found, it returns an empty string.
-func getGoroot() string {
-	goroot := os.Getenv("GOROOT")
-	if goroot != "" {
-		// An explicitly set GOROOT always has preference.
-		return goroot
-	}
-
-	// Check for the location of the 'go' binary and base GOROOT on that.
-	binpath, err := exec.LookPath("go")
-	if err == nil {
-		binpath, err = filepath.EvalSymlinks(binpath)
-		if err == nil {
-			goroot := filepath.Dir(filepath.Dir(binpath))
-			if isGoroot(goroot) {
-				return goroot
-			}
-		}
-	}
-
-	// Check what GOROOT was at compile time.
-	if isGoroot(runtime.GOROOT()) {
-		return runtime.GOROOT()
-	}
-
-	// Check for some standard locations, as a last resort.
-	var candidates []string
-	switch runtime.GOOS {
-	case "linux":
-		candidates = []string{
-			"/usr/local/go", // manually installed
-			"/usr/lib/go",   // from the distribution
-		}
-	case "darwin":
-		candidates = []string{
-			"/usr/local/go",             // manually installed
-			"/usr/local/opt/go/libexec", // from Homebrew
-		}
-	}
-
-	for _, candidate := range candidates {
-		if isGoroot(candidate) {
-			return candidate
-		}
-	}
-
-	// Can't find GOROOT...
-	return ""
-}
-
-// isGoroot checks whether the given path looks like a GOROOT.
-func isGoroot(goroot string) bool {
-	_, err := os.Stat(filepath.Join(goroot, "src", "runtime", "internal", "sys", "zversion.go"))
-	return err == nil
-}

compileopts/target_test.go

@@ -0,0 +1,19 @@
+package compileopts
+
+import "testing"
+
+func TestLoadTarget(t *testing.T) {
+	_, err := LoadTarget("arduino")
+	if err != nil {
+		t.Error("LoadTarget test failed:", err)
+	}
+
+	_, err = LoadTarget("notexist")
+	if err == nil {
+		t.Error("LoadTarget should have failed with non existing target")
+	}
+
+	if err.Error() != "expected a full LLVM target or a custom target in -target flag" {
+		t.Error("LoadTarget failed for wrong reason:", err)
+	}
+}

compiler/asserts.go

@@ -22,7 +22,7 @@ func (c *Compiler) emitLookupBoundsCheck(frame *Frame, arrayLen, index llvm.Valu
 	if index.Type().IntTypeWidth() < arrayLen.Type().IntTypeWidth() {
 		// Sometimes, the index can be e.g. an uint8 or int8, and we have to
 		// correctly extend that type.
-		if indexType.(*types.Basic).Info()&types.IsUnsigned == 0 {
+		if indexType.Underlying().(*types.Basic).Info()&types.IsUnsigned == 0 {
 			index = c.builder.CreateZExt(index, arrayLen.Type(), "")
 		} else {
 			index = c.builder.CreateSExt(index, arrayLen.Type(), "")
@@ -42,7 +42,7 @@ func (c *Compiler) emitLookupBoundsCheck(frame *Frame, arrayLen, index llvm.Valu
 
 	// Fail: this is a nil pointer, exit with a panic.
 	c.builder.SetInsertPointAtEnd(faultBlock)
-	c.createRuntimeCall("lookuppanic", nil, "")
+	c.createRuntimeCall("lookupPanic", nil, "")
 	c.builder.CreateUnreachable()
 
 	// Ok: this is a valid pointer.
@@ -56,7 +56,7 @@ func (c *Compiler) emitLookupBoundsCheck(frame *Frame, arrayLen, index llvm.Valu
 // normal meaning) and for creating a new slice, where 'capacity' means the
 // biggest possible slice capacity, 'low' means len and 'high' means cap. The
 // logic is the same in both cases.
-func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high llvm.Value, lowType, highType *types.Basic) {
+func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high, max llvm.Value, lowType, highType, maxType *types.Basic) {
 	if frame.fn.IsNoBounds() {
 		// The //go:nobounds pragma was added to the function to avoid bounds
 		// checking.
@@ -71,6 +71,9 @@ func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high llvm.V
 	if high.Type().IntTypeWidth() > capacityType.IntTypeWidth() {
 		capacityType = high.Type()
 	}
+	if max.Type().IntTypeWidth() > capacityType.IntTypeWidth() {
+		capacityType = max.Type()
+	}
 	if capacityType != capacity.Type() {
 		capacity = c.builder.CreateZExt(capacity, capacityType, "")
 	}
@@ -90,6 +93,13 @@ func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high llvm.V
 			high = c.builder.CreateSExt(high, capacityType, "")
 		}
 	}
+	if max.Type().IntTypeWidth() < capacityType.IntTypeWidth() {
+		if maxType.Info()&types.IsUnsigned != 0 {
+			max = c.builder.CreateZExt(max, capacityType, "")
+		} else {
+			max = c.builder.CreateSExt(max, capacityType, "")
+		}
+	}
 
 	faultBlock := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "slice.outofbounds")
 	nextBlock := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "slice.next")
@@ -97,13 +107,15 @@ func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high llvm.V
 
 	// Now do the bounds check: low > high || high > capacity
 	outOfBounds1 := c.builder.CreateICmp(llvm.IntUGT, low, high, "slice.lowhigh")
-	outOfBounds2 := c.builder.CreateICmp(llvm.IntUGT, high, capacity, "slice.highcap")
-	outOfBounds := c.builder.CreateOr(outOfBounds1, outOfBounds2, "slice.outofbounds")
+	outOfBounds2 := c.builder.CreateICmp(llvm.IntUGT, high, max, "slice.highmax")
+	outOfBounds3 := c.builder.CreateICmp(llvm.IntUGT, max, capacity, "slice.maxcap")
+	outOfBounds := c.builder.CreateOr(outOfBounds1, outOfBounds2, "slice.lowmax")
+	outOfBounds = c.builder.CreateOr(outOfBounds, outOfBounds3, "slice.lowcap")
 	c.builder.CreateCondBr(outOfBounds, faultBlock, nextBlock)
 
 	// Fail: this is a nil pointer, exit with a panic.
 	c.builder.SetInsertPointAtEnd(faultBlock)
-	c.createRuntimeCall("slicepanic", nil, "")
+	c.createRuntimeCall("slicePanic", nil, "")
 	c.builder.CreateUnreachable()
 
 	// Ok: this is a valid pointer.
@@ -114,6 +126,11 @@ func (c *Compiler) emitSliceBoundsCheck(frame *Frame, capacity, low, high llvm.V
 // has no effect in well-behaved programs, but makes sure no uncaught nil
 // pointer dereferences exist in valid Go code.
 func (c *Compiler) emitNilCheck(frame *Frame, ptr llvm.Value, blockPrefix string) {
+	// Check whether we need to emit this check at all.
+	if !ptr.IsAGlobalValue().IsNil() {
+		return
+	}
+
 	// Check whether this is a nil pointer.
 	faultBlock := c.ctx.AddBasicBlock(frame.fn.LLVMFn, blockPrefix+".nil")
 	nextBlock := c.ctx.AddBasicBlock(frame.fn.LLVMFn, blockPrefix+".next")
@@ -141,7 +158,7 @@ func (c *Compiler) emitNilCheck(frame *Frame, ptr llvm.Value, blockPrefix string
 
 	// Fail: this is a nil pointer, exit with a panic.
 	c.builder.SetInsertPointAtEnd(faultBlock)
-	c.createRuntimeCall("nilpanic", nil, "")
+	c.createRuntimeCall("nilPanic", nil, "")
 	c.builder.CreateUnreachable()
 
 	// Ok: this is a valid pointer.

compiler/calls.go

@@ -1,6 +1,7 @@
 package compiler
 
 import (
+	"fmt"
 	"golang.org/x/tools/go/ssa"
 	"tinygo.org/x/go-llvm"
 )
@@ -20,6 +21,9 @@ func (c *Compiler) createRuntimeCall(fnName string, args []llvm.Value, name stri
 		panic("trying to call runtime." + fnName)
 	}
 	fn := c.ir.GetFunction(member.(*ssa.Function))
+	if fn.LLVMFn.IsNil() {
+		panic(fmt.Errorf("function %s does not appear in LLVM IR", fnName))
+	}
 	if !fn.IsExported() {
 		args = append(args, llvm.Undef(c.i8ptrType))            // unused context parameter
 		args = append(args, llvm.ConstPointerNull(c.i8ptrType)) // coroutine handle
@@ -38,7 +42,7 @@ func (c *Compiler) createCall(fn llvm.Value, args []llvm.Value, name string) llv
 }
 
 // Expand an argument type to a list that can be used in a function call
-// paramter list.
+// parameter list.
 func (c *Compiler) expandFormalParamType(t llvm.Type) []llvm.Type {
 	switch t.TypeKind() {
 	case llvm.StructTypeKind:
@@ -163,7 +167,7 @@ func (c *Compiler) collapseFormalParamInternal(t llvm.Type, fields []llvm.Value)
 	switch t.TypeKind() {
 	case llvm.StructTypeKind:
 		if len(c.flattenAggregateType(t)) <= MaxFieldsPerParam {
-			value := c.getZeroValue(t)
+			value := llvm.ConstNull(t)
 			for i, subtyp := range t.StructElementTypes() {
 				structField, remaining := c.collapseFormalParamInternal(subtyp, fields)
 				fields = remaining

compiler/channel.go

@@ -6,67 +6,53 @@ package compiler
 import (
 	"go/types"
 
+	"github.com/tinygo-org/tinygo/compiler/llvmutil"
 	"golang.org/x/tools/go/ssa"
 	"tinygo.org/x/go-llvm"
 )
 
-// emitMakeChan returns a new channel value for the given channel type.
-func (c *Compiler) emitMakeChan(expr *ssa.MakeChan) (llvm.Value, error) {
-	chanType := c.mod.GetTypeByName("runtime.channel")
-	size := c.targetData.TypeAllocSize(chanType)
-	sizeValue := llvm.ConstInt(c.uintptrType, size, false)
-	ptr := c.createRuntimeCall("alloc", []llvm.Value{sizeValue}, "chan.alloc")
-	ptr = c.builder.CreateBitCast(ptr, llvm.PointerType(chanType, 0), "chan")
-	return ptr, nil
+func (c *Compiler) emitMakeChan(frame *Frame, expr *ssa.MakeChan) llvm.Value {
+	elementSize := c.targetData.TypeAllocSize(c.getLLVMType(expr.Type().(*types.Chan).Elem()))
+	elementSizeValue := llvm.ConstInt(c.uintptrType, elementSize, false)
+	bufSize := c.getValue(frame, expr.Size)
+	return c.createRuntimeCall("chanMake", []llvm.Value{elementSizeValue, bufSize}, "")
 }
 
 // emitChanSend emits a pseudo chan send operation. It is lowered to the actual
 // channel send operation during goroutine lowering.
 func (c *Compiler) emitChanSend(frame *Frame, instr *ssa.Send) {
-	valueType := c.getLLVMType(instr.X.Type())
 	ch := c.getValue(frame, instr.Chan)
 	chanValue := c.getValue(frame, instr.X)
-	valueSize := llvm.ConstInt(c.uintptrType, c.targetData.TypeAllocSize(chanValue.Type()), false)
-	coroutine := c.createRuntimeCall("getCoroutine", nil, "")
 
 	// store value-to-send
-	c.builder.SetInsertPointBefore(coroutine.InstructionParent().Parent().EntryBasicBlock().FirstInstruction())
-	valueAlloca := c.builder.CreateAlloca(valueType, "chan.value")
-	c.builder.SetInsertPointBefore(coroutine)
-	c.builder.SetInsertPointAtEnd(coroutine.InstructionParent())
+	valueType := c.getLLVMType(instr.X.Type())
+	valueAlloca, valueAllocaCast, valueAllocaSize := c.createTemporaryAlloca(valueType, "chan.value")
 	c.builder.CreateStore(chanValue, valueAlloca)
-	valueAllocaCast := c.builder.CreateBitCast(valueAlloca, c.i8ptrType, "chan.value.i8ptr")
 
 	// Do the send.
-	c.createRuntimeCall("chanSend", []llvm.Value{coroutine, ch, valueAllocaCast, valueSize}, "")
+	c.createRuntimeCall("chanSend", []llvm.Value{ch, valueAllocaCast}, "")
 
-	// Make sure CoroSplit includes the alloca in the coroutine frame.
-	// This is a bit dirty, but it works (at least in LLVM 8).
-	valueSizeI64 := llvm.ConstInt(c.ctx.Int64Type(), c.targetData.TypeAllocSize(chanValue.Type()), false)
-	c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{valueSizeI64, valueAllocaCast}, "")
+	// End the lifetime of the alloca.
+	// This also works around a bug in CoroSplit, at least in LLVM 8:
+	// https://bugs.llvm.org/show_bug.cgi?id=41742
+	c.emitLifetimeEnd(valueAllocaCast, valueAllocaSize)
 }
 
 // emitChanRecv emits a pseudo chan receive operation. It is lowered to the
 // actual channel receive operation during goroutine lowering.
 func (c *Compiler) emitChanRecv(frame *Frame, unop *ssa.UnOp) llvm.Value {
 	valueType := c.getLLVMType(unop.X.Type().(*types.Chan).Elem())
-	valueSize := llvm.ConstInt(c.uintptrType, c.targetData.TypeAllocSize(valueType), false)
 	ch := c.getValue(frame, unop.X)
-	coroutine := c.createRuntimeCall("getCoroutine", nil, "")
 
 	// Allocate memory to receive into.
-	c.builder.SetInsertPointBefore(coroutine.InstructionParent().Parent().EntryBasicBlock().FirstInstruction())
-	valueAlloca := c.builder.CreateAlloca(valueType, "chan.value")
-	c.builder.SetInsertPointBefore(coroutine)
-	c.builder.SetInsertPointAtEnd(coroutine.InstructionParent())
-	valueAllocaCast := c.builder.CreateBitCast(valueAlloca, c.i8ptrType, "chan.value.i8ptr")
+	valueAlloca, valueAllocaCast, valueAllocaSize := c.createTemporaryAlloca(valueType, "chan.value")
 
 	// Do the receive.
-	c.createRuntimeCall("chanRecv", []llvm.Value{coroutine, ch, valueAllocaCast, valueSize}, "")
+	commaOk := c.createRuntimeCall("chanRecv", []llvm.Value{ch, valueAllocaCast}, "")
 	received := c.builder.CreateLoad(valueAlloca, "chan.received")
+	c.emitLifetimeEnd(valueAllocaCast, valueAllocaSize)
+
 	if unop.CommaOk {
-		commaOk := c.createRuntimeCall("getTaskPromiseData", []llvm.Value{coroutine}, "chan.commaOk.wide")
-		commaOk = c.builder.CreateTrunc(commaOk, c.ctx.Int1Type(), "chan.commaOk")
 		tuple := llvm.Undef(c.ctx.StructType([]llvm.Type{valueType, c.ctx.Int1Type()}, false))
 		tuple = c.builder.CreateInsertValue(tuple, received, 0, "")
 		tuple = c.builder.CreateInsertValue(tuple, commaOk, 1, "")
@@ -78,8 +64,174 @@ func (c *Compiler) emitChanRecv(frame *Frame, unop *ssa.UnOp) llvm.Value {
 
 // emitChanClose closes the given channel.
 func (c *Compiler) emitChanClose(frame *Frame, param ssa.Value) {
-	valueType := c.getLLVMType(param.Type().(*types.Chan).Elem())
-	valueSize := llvm.ConstInt(c.uintptrType, c.targetData.TypeAllocSize(valueType), false)
 	ch := c.getValue(frame, param)
-	c.createRuntimeCall("chanClose", []llvm.Value{ch, valueSize}, "")
+	c.createRuntimeCall("chanClose", []llvm.Value{ch}, "")
+}
+
+// emitSelect emits all IR necessary for a select statements. That's a
+// non-trivial amount of code because select is very complex to implement.
+func (c *Compiler) emitSelect(frame *Frame, expr *ssa.Select) llvm.Value {
+	if len(expr.States) == 0 {
+		// Shortcuts for some simple selects.
+		llvmType := c.getLLVMType(expr.Type())
+		if expr.Blocking {
+			// Blocks forever:
+			//     select {}
+			c.createRuntimeCall("deadlock", nil, "")
+			return llvm.Undef(llvmType)
+		} else {
+			// No-op:
+			//     select {
+			//     default:
+			//     }
+			retval := llvm.Undef(llvmType)
+			retval = c.builder.CreateInsertValue(retval, llvm.ConstInt(c.intType, 0xffffffffffffffff, true), 0, "")
+			return retval // {-1, false}
+		}
+	}
+
+	// This code create a (stack-allocated) slice containing all the select
+	// cases and then calls runtime.chanSelect to perform the actual select
+	// statement.
+	// Simple selects (blocking and with just one case) are already transformed
+	// into regular chan operations during SSA construction so we don't have to
+	// optimize such small selects.
+
+	// Go through all the cases. Create the selectStates slice and and
+	// determine the receive buffer size and alignment.
+	recvbufSize := uint64(0)
+	recvbufAlign := 0
+	hasReceives := false
+	var selectStates []llvm.Value
+	chanSelectStateType := c.getLLVMRuntimeType("chanSelectState")
+	for _, state := range expr.States {
+		ch := c.getValue(frame, state.Chan)
+		selectState := llvm.ConstNull(chanSelectStateType)
+		selectState = c.builder.CreateInsertValue(selectState, ch, 0, "")
+		switch state.Dir {
+		case types.RecvOnly:
+			// Make sure the receive buffer is big enough and has the correct alignment.
+			llvmType := c.getLLVMType(state.Chan.Type().(*types.Chan).Elem())
+			if size := c.targetData.TypeAllocSize(llvmType); size > recvbufSize {
+				recvbufSize = size
+			}
+			if align := c.targetData.ABITypeAlignment(llvmType); align > recvbufAlign {
+				recvbufAlign = align
+			}
+			hasReceives = true
+		case types.SendOnly:
+			// Store this value in an alloca and put a pointer to this alloca
+			// in the send state.
+			sendValue := c.getValue(frame, state.Send)
+			alloca := llvmutil.CreateEntryBlockAlloca(c.builder, sendValue.Type(), "select.send.value")
+			c.builder.CreateStore(sendValue, alloca)
+			ptr := c.builder.CreateBitCast(alloca, c.i8ptrType, "")
+			selectState = c.builder.CreateInsertValue(selectState, ptr, 1, "")
+		default:
+			panic("unreachable")
+		}
+		selectStates = append(selectStates, selectState)
+	}
+
+	// Create a receive buffer, where the received value will be stored.
+	recvbuf := llvm.Undef(c.i8ptrType)
+	if hasReceives {
+		allocaType := llvm.ArrayType(c.ctx.Int8Type(), int(recvbufSize))
+		recvbufAlloca, _, _ := c.createTemporaryAlloca(allocaType, "select.recvbuf.alloca")
+		recvbufAlloca.SetAlignment(recvbufAlign)
+		recvbuf = c.builder.CreateGEP(recvbufAlloca, []llvm.Value{
+			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+		}, "select.recvbuf")
+	}
+
+	// Create the states slice (allocated on the stack).
+	statesAllocaType := llvm.ArrayType(chanSelectStateType, len(selectStates))
+	statesAlloca, statesI8, statesSize := c.createTemporaryAlloca(statesAllocaType, "select.states.alloca")
+	for i, state := range selectStates {
+		// Set each slice element to the appropriate channel.
+		gep := c.builder.CreateGEP(statesAlloca, []llvm.Value{
+			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+			llvm.ConstInt(c.ctx.Int32Type(), uint64(i), false),
+		}, "")
+		c.builder.CreateStore(state, gep)
+	}
+	statesPtr := c.builder.CreateGEP(statesAlloca, []llvm.Value{
+		llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+		llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+	}, "select.states")
+	statesLen := llvm.ConstInt(c.uintptrType, uint64(len(selectStates)), false)
+
+	// Do the select in the runtime.
+	var results llvm.Value
+	if expr.Blocking {
+		// Stack-allocate operation structures.
+		// If these were simply created as a slice, they would heap-allocate.
+		chBlockAllocaType := llvm.ArrayType(c.getLLVMRuntimeType("channelBlockedList"), len(selectStates))
+		chBlockAlloca, chBlockAllocaPtr, chBlockSize := c.createTemporaryAlloca(chBlockAllocaType, "select.block.alloca")
+		chBlockLen := llvm.ConstInt(c.uintptrType, uint64(len(selectStates)), false)
+		chBlockPtr := c.builder.CreateGEP(chBlockAlloca, []llvm.Value{
+			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+		}, "select.block")
+
+		results = c.createRuntimeCall("chanSelect", []llvm.Value{
+			recvbuf,
+			statesPtr, statesLen, statesLen, // []chanSelectState
+			chBlockPtr, chBlockLen, chBlockLen, // []channelBlockList
+		}, "select.result")
+
+		// Terminate the lifetime of the operation structures.
+		c.emitLifetimeEnd(chBlockAllocaPtr, chBlockSize)
+	} else {
+		results = c.createRuntimeCall("tryChanSelect", []llvm.Value{
+			recvbuf,
+			statesPtr, statesLen, statesLen, // []chanSelectState
+		}, "select.result")
+	}
+
+	// Terminate the lifetime of the states alloca.
+	c.emitLifetimeEnd(statesI8, statesSize)
+
+	// The result value does not include all the possible received values,
+	// because we can't load them in advance. Instead, the *ssa.Extract
+	// instruction will treat a *ssa.Select specially and load it there inline.
+	// Store the receive alloca in a sidetable until we hit this extract
+	// instruction.
+	if frame.selectRecvBuf == nil {
+		frame.selectRecvBuf = make(map[*ssa.Select]llvm.Value)
+	}
+	frame.selectRecvBuf[expr] = recvbuf
+
+	return results
+}
+
+// getChanSelectResult returns the special values from a *ssa.Extract expression
+// when extracting a value from a select statement (*ssa.Select). Because
+// *ssa.Select cannot load all values in advance, it does this later in the
+// *ssa.Extract expression.
+func (c *Compiler) getChanSelectResult(frame *Frame, expr *ssa.Extract) llvm.Value {
+	if expr.Index == 0 {
+		// index
+		value := c.getValue(frame, expr.Tuple)
+		index := c.builder.CreateExtractValue(value, expr.Index, "")
+		if index.Type().IntTypeWidth() < c.intType.IntTypeWidth() {
+			index = c.builder.CreateSExt(index, c.intType, "")
+		}
+		return index
+	} else if expr.Index == 1 {
+		// comma-ok
+		value := c.getValue(frame, expr.Tuple)
+		return c.builder.CreateExtractValue(value, expr.Index, "")
+	} else {
+		// Select statements are (index, ok, ...) where ... is a number of
+		// received values, depending on how many receive statements there
+		// are. They are all combined into one alloca (because only one
+		// receive can proceed at a time) so we'll get that alloca, bitcast
+		// it to the correct type, and dereference it.
+		recvbuf := frame.selectRecvBuf[expr.Tuple.(*ssa.Select)]
+		typ := llvm.PointerType(c.getLLVMType(expr.Type()), 0)
+		ptr := c.builder.CreateBitCast(recvbuf, typ, "")
+		return c.builder.CreateLoad(ptr, "")
+	}
 }

compiler/check.go

@@ -0,0 +1,172 @@
+package compiler
+
+// This file implements a set of sanity checks for the IR that is generated.
+// It can catch some mistakes that LLVM's verifier cannot.
+
+import (
+	"errors"
+	"fmt"
+
+	"tinygo.org/x/go-llvm"
+)
+
+func (c *Compiler) checkType(t llvm.Type, checked map[llvm.Type]struct{}, specials map[llvm.TypeKind]llvm.Type) error {
+	// prevent infinite recursion for self-referential types
+	if _, ok := checked[t]; ok {
+		return nil
+	}
+	checked[t] = struct{}{}
+
+	// check for any context mismatches
+	switch {
+	case t.Context() == c.ctx:
+		// this is correct
+	case t.Context() == llvm.GlobalContext():
+		// somewhere we accidentally used the global context instead of a real context
+		return fmt.Errorf("type %q uses global context", t.String())
+	default:
+		// we used some other context by accident
+		return fmt.Errorf("type %q uses context %v instead of the main context %v", t.Context(), c.ctx)
+	}
+
+	// if this is a composite type, check the components of the type
+	switch t.TypeKind() {
+	case llvm.VoidTypeKind, llvm.LabelTypeKind, llvm.TokenTypeKind, llvm.MetadataTypeKind:
+		// there should only be one of any of these
+		if s, ok := specials[t.TypeKind()]; !ok {
+			specials[t.TypeKind()] = t
+		} else if s != t {
+			return fmt.Errorf("duplicate special type %q: %v and %v", t.TypeKind().String(), t, s)
+		}
+	case llvm.FloatTypeKind, llvm.DoubleTypeKind, llvm.X86_FP80TypeKind, llvm.FP128TypeKind, llvm.PPC_FP128TypeKind:
+		// floating point numbers are primitives - nothing to recurse
+	case llvm.IntegerTypeKind:
+		// integers are primitives - nothing to recurse
+	case llvm.FunctionTypeKind:
+		// check arguments and return(s)
+		for i, v := range t.ParamTypes() {
+			if err := c.checkType(v, checked, specials); err != nil {
+				return fmt.Errorf("failed to verify argument %d of type %s: %s", i, t.String(), err.Error())
+			}
+		}
+		if err := c.checkType(t.ReturnType(), checked, specials); err != nil {
+			return fmt.Errorf("failed to verify return type of type %s: %s", t.String(), err.Error())
+		}
+	case llvm.StructTypeKind:
+		// check all elements
+		for i, v := range t.StructElementTypes() {
+			if err := c.checkType(v, checked, specials); err != nil {
+				return fmt.Errorf("failed to verify type of field %d of struct type %s: %s", i, t.String(), err.Error())
+			}
+		}
+	case llvm.ArrayTypeKind:
+		// check element type
+		if err := c.checkType(t.ElementType(), checked, specials); err != nil {
+			return fmt.Errorf("failed to verify element type of array type %s: %s", t.String(), err.Error())
+		}
+	case llvm.PointerTypeKind:
+		// check underlying type
+		if err := c.checkType(t.ElementType(), checked, specials); err != nil {
+			return fmt.Errorf("failed to verify underlying type of pointer type %s: %s", t.String(), err.Error())
+		}
+	case llvm.VectorTypeKind:
+		// check element type
+		if err := c.checkType(t.ElementType(), checked, specials); err != nil {
+			return fmt.Errorf("failed to verify element type of vector type %s: %s", t.String(), err.Error())
+		}
+	default:
+		return fmt.Errorf("unrecognized kind %q of type %s", t.TypeKind(), t.String())
+	}
+
+	return nil
+}
+
+func (c *Compiler) checkValue(v llvm.Value, types map[llvm.Type]struct{}, specials map[llvm.TypeKind]llvm.Type) error {
+	// check type
+	if err := c.checkType(v.Type(), types, specials); err != nil {
+		return fmt.Errorf("failed to verify type of value: %s", err.Error())
+	}
+
+	// check if this is an undefined void
+	if v.IsUndef() && v.Type().TypeKind() == llvm.VoidTypeKind {
+		return errors.New("encountered undefined void value")
+	}
+
+	return nil
+}
+
+func (c *Compiler) checkInstruction(inst llvm.Value, types map[llvm.Type]struct{}, specials map[llvm.TypeKind]llvm.Type) error {
+	// check value properties
+	if err := c.checkValue(inst, types, specials); err != nil {
+		return errorAt(inst, err.Error())
+	}
+
+	// check operands
+	for i := 0; i < inst.OperandsCount(); i++ {
+		if err := c.checkValue(inst.Operand(i), types, specials); err != nil {
+			return errorAt(inst, fmt.Sprintf("failed to validate operand %d of instruction %q: %s", i, inst.Name(), err.Error()))
+		}
+	}
+
+	return nil
+}
+
+func (c *Compiler) checkBasicBlock(bb llvm.BasicBlock, types map[llvm.Type]struct{}, specials map[llvm.TypeKind]llvm.Type) []error {
+	// check basic block value and type
+	var errs []error
+	if err := c.checkValue(bb.AsValue(), types, specials); err != nil {
+		errs = append(errs, errorAt(bb.Parent(), fmt.Sprintf("failed to validate value of basic block %s: %v", bb.AsValue().Name(), err)))
+	}
+
+	// check instructions
+	for inst := bb.FirstInstruction(); !inst.IsNil(); inst = llvm.NextInstruction(inst) {
+		if err := c.checkInstruction(inst, types, specials); err != nil {
+			errs = append(errs, err)
+		}
+	}
+
+	return errs
+}
+
+func (c *Compiler) checkFunction(fn llvm.Value, types map[llvm.Type]struct{}, specials map[llvm.TypeKind]llvm.Type) []error {
+	// check function value and type
+	var errs []error
+	if err := c.checkValue(fn, types, specials); err != nil {
+		errs = append(errs, fmt.Errorf("failed to validate value of function %s: %s", fn.Name(), err.Error()))
+	}
+
+	// check basic blocks
+	for bb := fn.FirstBasicBlock(); !bb.IsNil(); bb = llvm.NextBasicBlock(bb) {
+		errs = append(errs, c.checkBasicBlock(bb, types, specials)...)
+	}
+
+	return errs
+}
+
+func (c *Compiler) checkModule() []error {
+	// check for any context mismatches
+	var errs []error
+	switch {
+	case c.mod.Context() == c.ctx:
+		// this is correct
+	case c.mod.Context() == llvm.GlobalContext():
+		// somewhere we accidentally used the global context instead of a real context
+		errs = append(errs, errors.New("module uses global context"))
+	default:
+		// we used some other context by accident
+		errs = append(errs, fmt.Errorf("module uses context %v instead of the main context %v", c.mod.Context(), c.ctx))
+	}
+
+	types := map[llvm.Type]struct{}{}
+	specials := map[llvm.TypeKind]llvm.Type{}
+	for fn := c.mod.FirstFunction(); !fn.IsNil(); fn = llvm.NextFunction(fn) {
+		errs = append(errs, c.checkFunction(fn, types, specials)...)
+	}
+	for g := c.mod.FirstGlobal(); !g.IsNil(); g = llvm.NextGlobal(g) {
+		if err := c.checkValue(g, types, specials); err != nil {
+			errs = append(errs, fmt.Errorf("failed to verify global %s of module: %s", g.Name(), err.Error()))
+		}
+	}
+
+	return errs
+}

compiler/defer.go

@@ -14,6 +14,7 @@ package compiler
 //     frames.
 
 import (
+	"github.com/tinygo-org/tinygo/compiler/llvmutil"
 	"github.com/tinygo-org/tinygo/ir"
 	"golang.org/x/tools/go/ssa"
 	"tinygo.org/x/go-llvm"
@@ -29,11 +30,45 @@ func (c *Compiler) deferInitFunc(frame *Frame) {
 	frame.deferClosureFuncs = make(map[*ir.Function]int)
 
 	// Create defer list pointer.
-	deferType := llvm.PointerType(c.mod.GetTypeByName("runtime._defer"), 0)
+	deferType := llvm.PointerType(c.getLLVMRuntimeType("_defer"), 0)
 	frame.deferPtr = c.builder.CreateAlloca(deferType, "deferPtr")
 	c.builder.CreateStore(llvm.ConstPointerNull(deferType), frame.deferPtr)
 }
 
+// isInLoop checks if there is a path from a basic block to itself.
+func isInLoop(start *ssa.BasicBlock) bool {
+	// Use a breadth-first search to scan backwards through the block graph.
+	queue := []*ssa.BasicBlock{start}
+	checked := map[*ssa.BasicBlock]struct{}{}
+
+	for len(queue) > 0 {
+		// pop a block off of the queue
+		block := queue[len(queue)-1]
+		queue = queue[:len(queue)-1]
+
+		// Search through predecessors.
+		// Searching backwards means that this is pretty fast when the block is close to the start of the function.
+		// Defers are often placed near the start of the function.
+		for _, pred := range block.Preds {
+			if pred == start {
+				// cycle found
+				return true
+			}
+
+			if _, ok := checked[pred]; ok {
+				// block already checked
+				continue
+			}
+
+			// add to queue and checked map
+			queue = append(queue, pred)
+			checked[pred] = struct{}{}
+		}
+	}
+
+	return false
+}
+
 // emitDefer emits a single defer instruction, to be run when this function
 // returns.
 func (c *Compiler) emitDefer(frame *Frame, instr *ssa.Defer) {
@@ -122,13 +157,26 @@ func (c *Compiler) emitDefer(frame *Frame, instr *ssa.Defer) {
 
 	// Make a struct out of the collected values to put in the defer frame.
 	deferFrameType := c.ctx.StructType(valueTypes, false)
-	deferFrame := c.getZeroValue(deferFrameType)
+	deferFrame := llvm.ConstNull(deferFrameType)
 	for i, value := range values {
 		deferFrame = c.builder.CreateInsertValue(deferFrame, value, i, "")
 	}
 
-	// Put this struct in an alloca.
-	alloca := c.builder.CreateAlloca(deferFrameType, "defer.alloca")
+	// Put this struct in an allocation.
+	var alloca llvm.Value
+	if !isInLoop(instr.Block()) {
+		// This can safely use a stack allocation.
+		alloca = llvmutil.CreateEntryBlockAlloca(c.builder, deferFrameType, "defer.alloca")
+	} else {
+		// This may be hit a variable number of times, so use a heap allocation.
+		size := c.targetData.TypeAllocSize(deferFrameType)
+		sizeValue := llvm.ConstInt(c.uintptrType, size, false)
+		allocCall := c.createRuntimeCall("alloc", []llvm.Value{sizeValue}, "defer.alloc.call")
+		alloca = c.builder.CreateBitCast(allocCall, llvm.PointerType(deferFrameType, 0), "defer.alloc")
+	}
+	if c.NeedsStackObjects() {
+		c.trackPointer(alloca)
+	}
 	c.builder.CreateStore(deferFrame, alloca)
 
 	// Push it on top of the linked list by replacing deferPtr.
@@ -154,10 +202,10 @@ func (c *Compiler) emitRunDefers(frame *Frame) {
 	//     }
 
 	// Create loop.
-	loophead := llvm.AddBasicBlock(frame.fn.LLVMFn, "rundefers.loophead")
-	loop := llvm.AddBasicBlock(frame.fn.LLVMFn, "rundefers.loop")
-	unreachable := llvm.AddBasicBlock(frame.fn.LLVMFn, "rundefers.default")
-	end := llvm.AddBasicBlock(frame.fn.LLVMFn, "rundefers.end")
+	loophead := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "rundefers.loophead")
+	loop := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "rundefers.loop")
+	unreachable := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "rundefers.default")
+	end := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "rundefers.end")
 	c.builder.CreateBr(loophead)
 
 	// Create loop head:
@@ -189,7 +237,7 @@ func (c *Compiler) emitRunDefers(frame *Frame) {
 		// Create switch case, for example:
 		//     case 0:
 		//         // run first deferred call
-		block := llvm.AddBasicBlock(frame.fn.LLVMFn, "rundefers.callback")
+		block := c.ctx.AddBasicBlock(frame.fn.LLVMFn, "rundefers.callback")
 		sw.AddCase(llvm.ConstInt(c.uintptrType, uint64(i), false), block)
 		c.builder.SetInsertPointAtEnd(block)
 		switch callback := callback.(type) {
@@ -200,7 +248,7 @@ func (c *Compiler) emitRunDefers(frame *Frame) {
 			}
 
 			// Get the real defer struct type and cast to it.
-			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.mod.GetTypeByName("runtime._defer"), 0), c.i8ptrType}
+			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.getLLVMRuntimeType("_defer"), 0), c.i8ptrType}
 			for _, arg := range callback.Args {
 				valueTypes = append(valueTypes, c.getLLVMType(arg.Type()))
 			}
@@ -231,7 +279,7 @@ func (c *Compiler) emitRunDefers(frame *Frame) {
 			// Direct call.
 
 			// Get the real defer struct type and cast to it.
-			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.mod.GetTypeByName("runtime._defer"), 0)}
+			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.getLLVMRuntimeType("_defer"), 0)}
 			for _, param := range callback.Params {
 				valueTypes = append(valueTypes, c.getLLVMType(param.Type()))
 			}
@@ -260,7 +308,7 @@ func (c *Compiler) emitRunDefers(frame *Frame) {
 		case *ssa.MakeClosure:
 			// Get the real defer struct type and cast to it.
 			fn := c.ir.GetFunction(callback.Fn.(*ssa.Function))
-			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.mod.GetTypeByName("runtime._defer"), 0)}
+			valueTypes := []llvm.Type{c.uintptrType, llvm.PointerType(c.getLLVMRuntimeType("_defer"), 0)}
 			params := fn.Signature.Params()
 			for i := 0; i < params.Len(); i++ {
 				valueTypes = append(valueTypes, c.getLLVMType(params.At(i).Type()))

compiler/errors.go

@@ -1,8 +1,12 @@
 package compiler
 
 import (
+	"go/scanner"
 	"go/token"
 	"go/types"
+	"path/filepath"
+
+	"tinygo.org/x/go-llvm"
 )
 
 func (c *Compiler) makeError(pos token.Pos, msg string) types.Error {
@@ -16,3 +20,42 @@ func (c *Compiler) makeError(pos token.Pos, msg string) types.Error {
 func (c *Compiler) addError(pos token.Pos, msg string) {
 	c.diagnostics = append(c.diagnostics, c.makeError(pos, msg))
 }
+
+// errorAt returns an error value at the location of the instruction.
+// The location information may not be complete as it depends on debug
+// information in the IR.
+func errorAt(inst llvm.Value, msg string) scanner.Error {
+	return scanner.Error{
+		Pos: getPosition(inst),
+		Msg: msg,
+	}
+}
+
+// getPosition returns the position information for the given value, as far as
+// it is available.
+func getPosition(val llvm.Value) token.Position {
+	if !val.IsAInstruction().IsNil() {
+		loc := val.InstructionDebugLoc()
+		if loc.IsNil() {
+			return token.Position{}
+		}
+		file := loc.LocationScope().ScopeFile()
+		return token.Position{
+			Filename: filepath.Join(file.FileDirectory(), file.FileFilename()),
+			Line:     int(loc.LocationLine()),
+			Column:   int(loc.LocationColumn()),
+		}
+	} else if !val.IsAFunction().IsNil() {
+		loc := val.Subprogram()
+		if loc.IsNil() {
+			return token.Position{}
+		}
+		file := loc.ScopeFile()
+		return token.Position{
+			Filename: filepath.Join(file.FileDirectory(), file.FileFilename()),
+			Line:     int(loc.SubprogramLine()),
+		}
+	} else {
+		return token.Position{}
+	}
+}

compiler/func-lowering.go

@@ -1,269 +0,0 @@
-package compiler
-
-// This file lowers func values into their final form. This is necessary for
-// funcValueSwitch, which needs full program analysis.
-
-import (
-	"sort"
-	"strconv"
-
-	"tinygo.org/x/go-llvm"
-)
-
-// funcSignatureInfo keeps information about a single signature and its uses.
-type funcSignatureInfo struct {
-	sig                     llvm.Value   // *uint8 to identify the signature
-	funcValueWithSignatures []llvm.Value // slice of runtime.funcValueWithSignature
-}
-
-// funcWithUses keeps information about a single function used as func value and
-// the assigned function ID. More commonly used functions are assigned a lower
-// ID.
-type funcWithUses struct {
-	funcPtr  llvm.Value
-	useCount int // how often this function is used in a func value
-	id       int // assigned ID
-}
-
-// Slice to sort functions by their use counts, or else their name if they're
-// used equally often.
-type funcWithUsesList []*funcWithUses
-
-func (l funcWithUsesList) Len() int { return len(l) }
-func (l funcWithUsesList) Less(i, j int) bool {
-	if l[i].useCount != l[j].useCount {
-		// return the reverse: we want the highest use counts sorted first
-		return l[i].useCount > l[j].useCount
-	}
-	iName := l[i].funcPtr.Name()
-	jName := l[j].funcPtr.Name()
-	return iName < jName
-}
-func (l funcWithUsesList) Swap(i, j int) {
-	l[i], l[j] = l[j], l[i]
-}
-
-// LowerFuncValue lowers the runtime.funcValueWithSignature type and
-// runtime.getFuncPtr function to their final form.
-func (c *Compiler) LowerFuncValues() {
-	if c.funcImplementation() != funcValueSwitch {
-		return
-	}
-
-	// Find all func values used in the program with their signatures.
-	funcValueWithSignaturePtr := llvm.PointerType(c.mod.GetTypeByName("runtime.funcValueWithSignature"), 0)
-	signatures := map[string]*funcSignatureInfo{}
-	for global := c.mod.FirstGlobal(); !global.IsNil(); global = llvm.NextGlobal(global) {
-		if global.Type() != funcValueWithSignaturePtr {
-			continue
-		}
-		sig := llvm.ConstExtractValue(global.Initializer(), []uint32{1})
-		name := sig.Name()
-		if info, ok := signatures[name]; ok {
-			info.funcValueWithSignatures = append(info.funcValueWithSignatures, global)
-		} else {
-			signatures[name] = &funcSignatureInfo{
-				sig:                     sig,
-				funcValueWithSignatures: []llvm.Value{global},
-			}
-		}
-	}
-
-	// Sort the signatures, for deterministic execution.
-	names := make([]string, 0, len(signatures))
-	for name := range signatures {
-		names = append(names, name)
-	}
-	sort.Strings(names)
-
-	for _, name := range names {
-		info := signatures[name]
-		functions := make(funcWithUsesList, len(info.funcValueWithSignatures))
-		for i, use := range info.funcValueWithSignatures {
-			var useCount int
-			for _, use2 := range getUses(use) {
-				useCount += len(getUses(use2))
-			}
-			functions[i] = &funcWithUses{
-				funcPtr:  llvm.ConstExtractValue(use.Initializer(), []uint32{0}).Operand(0),
-				useCount: useCount,
-			}
-		}
-		sort.Sort(functions)
-
-		for i, fn := range functions {
-			fn.id = i + 1
-			for _, ptrtoint := range getUses(fn.funcPtr) {
-				if ptrtoint.IsAConstantExpr().IsNil() || ptrtoint.Opcode() != llvm.PtrToInt {
-					continue
-				}
-				for _, funcValueWithSignatureConstant := range getUses(ptrtoint) {
-					for _, funcValueWithSignatureGlobal := range getUses(funcValueWithSignatureConstant) {
-						for _, use := range getUses(funcValueWithSignatureGlobal) {
-							if ptrtoint.IsAConstantExpr().IsNil() || ptrtoint.Opcode() != llvm.PtrToInt {
-								panic("expected const ptrtoint")
-							}
-							use.ReplaceAllUsesWith(llvm.ConstInt(c.uintptrType, uint64(fn.id), false))
-						}
-					}
-				}
-			}
-		}
-
-		for _, getFuncPtrCall := range getUses(info.sig) {
-			if getFuncPtrCall.IsACallInst().IsNil() {
-				continue
-			}
-			if getFuncPtrCall.CalledValue().Name() != "runtime.getFuncPtr" {
-				panic("expected all call uses to be runtime.getFuncPtr")
-			}
-			funcID := getFuncPtrCall.Operand(1)
-			switch len(functions) {
-			case 0:
-				// There are no functions used in a func value that implement
-				// this signature. The only possible value is a nil value.
-				for _, inttoptr := range getUses(getFuncPtrCall) {
-					if inttoptr.IsAIntToPtrInst().IsNil() {
-						panic("expected inttoptr")
-					}
-					nilptr := llvm.ConstPointerNull(inttoptr.Type())
-					inttoptr.ReplaceAllUsesWith(nilptr)
-					inttoptr.EraseFromParentAsInstruction()
-				}
-				getFuncPtrCall.EraseFromParentAsInstruction()
-			case 1:
-				// There is exactly one function with this signature that is
-				// used in a func value. The func value itself can be either nil
-				// or this one function.
-				c.builder.SetInsertPointBefore(getFuncPtrCall)
-				zero := llvm.ConstInt(c.uintptrType, 0, false)
-				isnil := c.builder.CreateICmp(llvm.IntEQ, funcID, zero, "")
-				funcPtrNil := llvm.ConstPointerNull(functions[0].funcPtr.Type())
-				funcPtr := c.builder.CreateSelect(isnil, funcPtrNil, functions[0].funcPtr, "")
-				for _, inttoptr := range getUses(getFuncPtrCall) {
-					if inttoptr.IsAIntToPtrInst().IsNil() {
-						panic("expected inttoptr")
-					}
-					inttoptr.ReplaceAllUsesWith(funcPtr)
-					inttoptr.EraseFromParentAsInstruction()
-				}
-				getFuncPtrCall.EraseFromParentAsInstruction()
-			default:
-				// There are multiple functions used in a func value that
-				// implement this signature.
-				// What we'll do is transform the following:
-				//     rawPtr := runtime.getFuncPtr(fn)
-				//     if func.rawPtr == nil {
-				//         runtime.nilpanic()
-				//     }
-				//     result := func.rawPtr(...args, func.context)
-				// into this:
-				//     if false {
-				//         runtime.nilpanic()
-				//     }
-				//     var result // Phi
-				//     switch fn.id {
-				//     case 0:
-				//         runtime.nilpanic()
-				//     case 1:
-				//         result = call first implementation...
-				//     case 2:
-				//         result = call second implementation...
-				//     default:
-				//         unreachable
-				//     }
-
-				// Remove some casts, checks, and the old call which we're going
-				// to replace.
-				var funcCall llvm.Value
-				for _, inttoptr := range getUses(getFuncPtrCall) {
-					if inttoptr.IsAIntToPtrInst().IsNil() {
-						panic("expected inttoptr")
-					}
-					for _, ptrUse := range getUses(inttoptr) {
-						if !ptrUse.IsABitCastInst().IsNil() {
-							for _, bitcastUse := range getUses(ptrUse) {
-								if bitcastUse.IsACallInst().IsNil() || bitcastUse.CalledValue().Name() != "runtime.isnil" {
-									panic("expected a call to runtime.isnil")
-								}
-								bitcastUse.ReplaceAllUsesWith(llvm.ConstInt(c.ctx.Int1Type(), 0, false))
-								bitcastUse.EraseFromParentAsInstruction()
-							}
-							ptrUse.EraseFromParentAsInstruction()
-						} else if !ptrUse.IsACallInst().IsNil() && ptrUse.CalledValue() == inttoptr {
-							if !funcCall.IsNil() {
-								panic("multiple calls on a single runtime.getFuncPtr")
-							}
-							funcCall = ptrUse
-						} else {
-							panic("unexpected getFuncPtrCall")
-						}
-					}
-				}
-				if funcCall.IsNil() {
-					panic("expected exactly one call use of a runtime.getFuncPtr")
-				}
-
-				// The block that cannot be reached with correct funcValues (to
-				// help the optimizer).
-				c.builder.SetInsertPointBefore(funcCall)
-				defaultBlock := llvm.AddBasicBlock(funcCall.InstructionParent().Parent(), "func.default")
-				c.builder.SetInsertPointAtEnd(defaultBlock)
-				c.builder.CreateUnreachable()
-
-				// Create the switch.
-				c.builder.SetInsertPointBefore(funcCall)
-				sw := c.builder.CreateSwitch(funcID, defaultBlock, len(functions)+1)
-
-				// Split right after the switch. We will need to insert a few
-				// basic blocks in this gap.
-				nextBlock := c.splitBasicBlock(sw, llvm.NextBasicBlock(sw.InstructionParent()), "func.next")
-
-				// The 0 case, which is actually a nil check.
-				nilBlock := llvm.InsertBasicBlock(nextBlock, "func.nil")
-				c.builder.SetInsertPointAtEnd(nilBlock)
-				c.createRuntimeCall("nilpanic", nil, "")
-				c.builder.CreateUnreachable()
-				sw.AddCase(llvm.ConstInt(c.uintptrType, 0, false), nilBlock)
-
-				// Gather the list of parameters for every call we're going to
-				// make.
-				callParams := make([]llvm.Value, funcCall.OperandsCount()-1)
-				for i := range callParams {
-					callParams[i] = funcCall.Operand(i)
-				}
-
-				// If the call produces a value, we need to get it using a PHI
-				// node.
-				phiBlocks := make([]llvm.BasicBlock, len(functions))
-				phiValues := make([]llvm.Value, len(functions))
-				for i, fn := range functions {
-					// Insert a switch case.
-					bb := llvm.InsertBasicBlock(nextBlock, "func.call"+strconv.Itoa(fn.id))
-					c.builder.SetInsertPointAtEnd(bb)
-					result := c.builder.CreateCall(fn.funcPtr, callParams, "")
-					c.builder.CreateBr(nextBlock)
-					sw.AddCase(llvm.ConstInt(c.uintptrType, uint64(fn.id), false), bb)
-					phiBlocks[i] = bb
-					phiValues[i] = result
-				}
-				// Create the PHI node so that the call result flows into the
-				// next block (after the split). This is only necessary when the
-				// call produced a value.
-				if funcCall.Type().TypeKind() != llvm.VoidTypeKind {
-					c.builder.SetInsertPointBefore(nextBlock.FirstInstruction())
-					phi := c.builder.CreatePHI(funcCall.Type(), "")
-					phi.AddIncoming(phiValues, phiBlocks)
-					funcCall.ReplaceAllUsesWith(phi)
-				}
-
-				// Finally, remove the old instructions.
-				funcCall.EraseFromParentAsInstruction()
-				for _, inttoptr := range getUses(getFuncPtrCall) {
-					inttoptr.EraseFromParentAsInstruction()
-				}
-				getFuncPtrCall.EraseFromParentAsInstruction()
-			}
-		}
-	}
-}

compiler/func.go

@@ -32,10 +32,15 @@ const (
 // funcImplementation picks an appropriate func value implementation for the
 // target.
 func (c *Compiler) funcImplementation() funcValueImplementation {
-	if c.GOARCH == "wasm" {
+	// Always pick the switch implementation, as it allows the use of blocking
+	// inside a function that is used as a func value.
+	switch c.Scheduler() {
+	case "coroutines":
 		return funcValueSwitch
-	} else {
+	case "tasks":
 		return funcValueDoubleword
+	default:
+		panic("unknown scheduler type")
 	}
 }
 
@@ -48,11 +53,11 @@ func (c *Compiler) createFuncValue(funcPtr, context llvm.Value, sig *types.Signa
 		// Closure is: {context, function pointer}
 		funcValueScalar = funcPtr
 	case funcValueSwitch:
-		sigGlobal := c.getFuncSignature(sig)
+		sigGlobal := c.getTypeCode(sig)
 		funcValueWithSignatureGlobalName := funcPtr.Name() + "$withSignature"
 		funcValueWithSignatureGlobal := c.mod.NamedGlobal(funcValueWithSignatureGlobalName)
 		if funcValueWithSignatureGlobal.IsNil() {
-			funcValueWithSignatureType := c.mod.GetTypeByName("runtime.funcValueWithSignature")
+			funcValueWithSignatureType := c.getLLVMRuntimeType("funcValueWithSignature")
 			funcValueWithSignature := llvm.ConstNamedStruct(funcValueWithSignatureType, []llvm.Value{
 				llvm.ConstPtrToInt(funcPtr, c.uintptrType),
 				sigGlobal,
@@ -73,22 +78,6 @@ func (c *Compiler) createFuncValue(funcPtr, context llvm.Value, sig *types.Signa
 	return funcValue
 }
 
-// getFuncSignature returns a global for identification of a particular function
-// signature. It is used in runtime.funcValueWithSignature and in calls to
-// getFuncPtr.
-func (c *Compiler) getFuncSignature(sig *types.Signature) llvm.Value {
-	typeCodeName := getTypeCodeName(sig)
-	sigGlobalName := "reflect/types.type:" + typeCodeName
-	sigGlobal := c.mod.NamedGlobal(sigGlobalName)
-	if sigGlobal.IsNil() {
-		sigGlobal = llvm.AddGlobal(c.mod, c.ctx.Int8Type(), sigGlobalName)
-		sigGlobal.SetInitializer(llvm.Undef(c.ctx.Int8Type()))
-		sigGlobal.SetGlobalConstant(true)
-		sigGlobal.SetLinkage(llvm.InternalLinkage)
-	}
-	return sigGlobal
-}
-
 // extractFuncScalar returns some scalar that can be used in comparisons. It is
 // a cheap operation.
 func (c *Compiler) extractFuncScalar(funcValue llvm.Value) llvm.Value {
@@ -103,14 +92,14 @@ func (c *Compiler) extractFuncContext(funcValue llvm.Value) llvm.Value {
 
 // decodeFuncValue extracts the context and the function pointer from this func
 // value. This may be an expensive operation.
-func (c *Compiler) decodeFuncValue(funcValue llvm.Value, sig *types.Signature) (funcPtr, context llvm.Value, err error) {
+func (c *Compiler) decodeFuncValue(funcValue llvm.Value, sig *types.Signature) (funcPtr, context llvm.Value) {
 	context = c.builder.CreateExtractValue(funcValue, 0, "")
 	switch c.funcImplementation() {
 	case funcValueDoubleword:
 		funcPtr = c.builder.CreateExtractValue(funcValue, 1, "")
 	case funcValueSwitch:
 		llvmSig := c.getRawFuncType(sig)
-		sigGlobal := c.getFuncSignature(sig)
+		sigGlobal := c.getTypeCode(sig)
 		funcPtr = c.createRuntimeCall("getFuncPtr", []llvm.Value{funcValue, sigGlobal}, "")
 		funcPtr = c.builder.CreateIntToPtr(funcPtr, llvmSig, "")
 	default:
@@ -126,7 +115,7 @@ func (c *Compiler) getFuncType(typ *types.Signature) llvm.Type {
 		rawPtr := c.getRawFuncType(typ)
 		return c.ctx.StructType([]llvm.Type{c.i8ptrType, rawPtr}, false)
 	case funcValueSwitch:
-		return c.mod.GetTypeByName("runtime.funcValue")
+		return c.getLLVMRuntimeType("funcValue")
 	default:
 		panic("unimplemented func value variant")
 	}

compiler/gc-precise.go

@@ -1,104 +0,0 @@
-package compiler
-
-import (
-	"math/big"
-
-	"tinygo.org/x/go-llvm"
-)
-
-func (c *Compiler) addGlobalsBitmap() {
-	if c.mod.NamedGlobal("runtime.trackedGlobalsStart").IsNil() {
-		return // nothing to do: no GC in use
-	}
-
-	var trackedGlobals []llvm.Value
-	var trackedGlobalTypes []llvm.Type
-	for global := c.mod.FirstGlobal(); !global.IsNil(); global = llvm.NextGlobal(global) {
-		if global.IsDeclaration() {
-			continue
-		}
-		typ := global.Type().ElementType()
-		ptrs := c.getPointerBitmap(typ, global.Name())
-		if ptrs.BitLen() == 0 {
-			continue
-		}
-		trackedGlobals = append(trackedGlobals, global)
-		trackedGlobalTypes = append(trackedGlobalTypes, typ)
-	}
-
-	//
-	globalsBundleType := c.ctx.StructType(trackedGlobalTypes, false)
-	globalsBundle := llvm.AddGlobal(c.mod, globalsBundleType, "tinygo.trackedGlobals")
-	globalsBundle.SetLinkage(llvm.InternalLinkage)
-	globalsBundle.SetUnnamedAddr(true)
-	initializer := llvm.Undef(globalsBundleType)
-	for i, global := range trackedGlobals {
-		initializer = llvm.ConstInsertValue(initializer, global.Initializer(), []uint32{uint32(i)})
-		gep := llvm.ConstGEP(globalsBundle, []llvm.Value{
-			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-			llvm.ConstInt(c.ctx.Int32Type(), uint64(i), false),
-		})
-		global.ReplaceAllUsesWith(gep)
-		global.EraseFromParentAsGlobal()
-	}
-	globalsBundle.SetInitializer(initializer)
-
-	trackedGlobalsStart := llvm.ConstPtrToInt(globalsBundle, c.uintptrType)
-	c.mod.NamedGlobal("runtime.trackedGlobalsStart").SetInitializer(trackedGlobalsStart)
-
-	alignment := c.targetData.PrefTypeAlignment(c.i8ptrType)
-	trackedGlobalsLength := llvm.ConstInt(c.uintptrType, c.targetData.TypeAllocSize(globalsBundleType)/uint64(alignment), false)
-	c.mod.NamedGlobal("runtime.trackedGlobalsLength").SetInitializer(trackedGlobalsLength)
-
-	bitmapBytes := c.getPointerBitmap(globalsBundleType, "globals bundle").Bytes()
-	bitmapValues := make([]llvm.Value, len(bitmapBytes))
-	for i, b := range bitmapBytes {
-		bitmapValues[len(bitmapBytes)-i-1] = llvm.ConstInt(c.ctx.Int8Type(), uint64(b), false)
-	}
-	bitmapArray := llvm.ConstArray(llvm.ArrayType(c.ctx.Int8Type(), len(bitmapBytes)), bitmapValues)
-	bitmapNew := llvm.AddGlobal(c.mod, bitmapArray.Type(), "runtime.trackedGlobalsBitmap.tmp")
-	bitmapOld := c.mod.NamedGlobal("runtime.trackedGlobalsBitmap")
-	bitmapOld.ReplaceAllUsesWith(bitmapNew)
-	bitmapNew.SetInitializer(bitmapArray)
-	bitmapNew.SetName("runtime.trackedGlobalsBitmap")
-}
-
-func (c *Compiler) getPointerBitmap(typ llvm.Type, name string) *big.Int {
-	alignment := c.targetData.PrefTypeAlignment(c.i8ptrType)
-	switch typ.TypeKind() {
-	case llvm.IntegerTypeKind, llvm.FloatTypeKind, llvm.DoubleTypeKind:
-		return big.NewInt(0)
-	case llvm.PointerTypeKind:
-		return big.NewInt(1)
-	case llvm.StructTypeKind:
-		ptrs := big.NewInt(0)
-		for i, subtyp := range typ.StructElementTypes() {
-			subptrs := c.getPointerBitmap(subtyp, name)
-			if subptrs.BitLen() == 0 {
-				continue
-			}
-			offset := c.targetData.ElementOffset(typ, i)
-			if offset%uint64(alignment) != 0 {
-				panic("precise GC: global contains unaligned pointer: " + name)
-			}
-			subptrs.Lsh(subptrs, uint(offset)/uint(alignment))
-			ptrs.Or(ptrs, subptrs)
-		}
-		return ptrs
-	case llvm.ArrayTypeKind:
-		subtyp := typ.ElementType()
-		subptrs := c.getPointerBitmap(subtyp, name)
-		ptrs := big.NewInt(0)
-		if subptrs.BitLen() == 0 {
-			return ptrs
-		}
-		elementSize := c.targetData.TypeAllocSize(subtyp)
-		for i := 0; i < typ.ArrayLength(); i++ {
-			ptrs.Lsh(ptrs, uint(elementSize)/uint(alignment))
-			ptrs.Or(ptrs, subptrs)
-		}
-		return ptrs
-	default:
-		panic("unknown type kind of global: " + name)
-	}
-}

compiler/gc.go

@@ -0,0 +1,106 @@
+package compiler
+
+// This file provides IR transformations necessary for precise and portable
+// garbage collectors.
+
+import (
+	"go/token"
+
+	"golang.org/x/tools/go/ssa"
+	"tinygo.org/x/go-llvm"
+)
+
+// trackExpr inserts pointer tracking intrinsics for the GC if the expression is
+// one of the expressions that need this.
+func (c *Compiler) trackExpr(frame *Frame, expr ssa.Value, value llvm.Value) {
+	// There are uses of this expression, Make sure the pointers
+	// are tracked during GC.
+	switch expr := expr.(type) {
+	case *ssa.Alloc, *ssa.MakeChan, *ssa.MakeMap:
+		// These values are always of pointer type in IR.
+		c.trackPointer(value)
+	case *ssa.Call, *ssa.Convert, *ssa.MakeClosure, *ssa.MakeInterface, *ssa.MakeSlice, *ssa.Next:
+		if !value.IsNil() {
+			c.trackValue(value)
+		}
+	case *ssa.Select:
+		if alloca, ok := frame.selectRecvBuf[expr]; ok {
+			if alloca.IsAUndefValue().IsNil() {
+				c.trackPointer(alloca)
+			}
+		}
+	case *ssa.UnOp:
+		switch expr.Op {
+		case token.MUL:
+			// Pointer dereference.
+			c.trackValue(value)
+		case token.ARROW:
+			// Channel receive operator.
+			// It's not necessary to look at commaOk here, because in that
+			// case it's just an aggregate and trackValue will extract the
+			// pointer in there (if there is one).
+			c.trackValue(value)
+		}
+	}
+}
+
+// trackValue locates pointers in a value (possibly an aggregate) and tracks the
+// individual pointers
+func (c *Compiler) trackValue(value llvm.Value) {
+	typ := value.Type()
+	switch typ.TypeKind() {
+	case llvm.PointerTypeKind:
+		c.trackPointer(value)
+	case llvm.StructTypeKind:
+		if !typeHasPointers(typ) {
+			return
+		}
+		numElements := typ.StructElementTypesCount()
+		for i := 0; i < numElements; i++ {
+			subValue := c.builder.CreateExtractValue(value, i, "")
+			c.trackValue(subValue)
+		}
+	case llvm.ArrayTypeKind:
+		if !typeHasPointers(typ) {
+			return
+		}
+		numElements := typ.ArrayLength()
+		for i := 0; i < numElements; i++ {
+			subValue := c.builder.CreateExtractValue(value, i, "")
+			c.trackValue(subValue)
+		}
+	}
+}
+
+// trackPointer creates a call to runtime.trackPointer, bitcasting the poitner
+// first if needed. The input value must be of LLVM pointer type.
+func (c *Compiler) trackPointer(value llvm.Value) {
+	if value.Type() != c.i8ptrType {
+		value = c.builder.CreateBitCast(value, c.i8ptrType, "")
+	}
+	c.createRuntimeCall("trackPointer", []llvm.Value{value}, "")
+}
+
+// typeHasPointers returns whether this type is a pointer or contains pointers.
+// If the type is an aggregate type, it will check whether there is a pointer
+// inside.
+func typeHasPointers(t llvm.Type) bool {
+	switch t.TypeKind() {
+	case llvm.PointerTypeKind:
+		return true
+	case llvm.StructTypeKind:
+		for _, subType := range t.StructElementTypes() {
+			if typeHasPointers(subType) {
+				return true
+			}
+		}
+		return false
+	case llvm.ArrayTypeKind:
+		if typeHasPointers(t.ElementType()) {
+			return true
+		}
+		return false
+	default:
+		return false
+	}
+}

compiler/goroutine.go

@@ -0,0 +1,138 @@
+package compiler
+
+// This file implements the 'go' keyword to start a new goroutine. See
+// goroutine-lowering.go for more details.
+
+import "tinygo.org/x/go-llvm"
+
+// emitStartGoroutine starts a new goroutine with the provided function pointer
+// and parameters.
+// In general, you should pass all regular parameters plus the context parameter.
+// There is one exception: the task-based scheduler needs to have the function
+// pointer passed in as a parameter too in addition to the context.
+//
+// Because a go statement doesn't return anything, return undef.
+func (c *Compiler) emitStartGoroutine(funcPtr llvm.Value, params []llvm.Value) llvm.Value {
+	switch c.Scheduler() {
+	case "tasks":
+		paramBundle := c.emitPointerPack(params)
+		paramBundle = c.builder.CreatePtrToInt(paramBundle, c.uintptrType, "")
+
+		calleeValue := c.createGoroutineStartWrapper(funcPtr)
+		c.createRuntimeCall("startGoroutine", []llvm.Value{calleeValue, paramBundle}, "")
+	case "coroutines":
+		// We roundtrip through runtime.makeGoroutine as a signal (to find these
+		// calls) and to break any optimizations LLVM will try to do: they are
+		// invalid if we called this as a regular function to be updated later.
+		calleeValue := c.builder.CreatePtrToInt(funcPtr, c.uintptrType, "")
+		calleeValue = c.createRuntimeCall("makeGoroutine", []llvm.Value{calleeValue}, "")
+		calleeValue = c.builder.CreateIntToPtr(calleeValue, funcPtr.Type(), "")
+		c.createCall(calleeValue, append(params, llvm.ConstPointerNull(c.i8ptrType)), "")
+	default:
+		panic("unreachable")
+	}
+	return llvm.Undef(funcPtr.Type().ElementType().ReturnType())
+}
+
+// createGoroutineStartWrapper creates a wrapper for the task-based
+// implementation of goroutines. For example, to call a function like this:
+//
+//     func add(x, y int) int { ... }
+//
+// It creates a wrapper like this:
+//
+//     func add$gowrapper(ptr *unsafe.Pointer) {
+//         args := (*struct{
+//             x, y int
+//         })(ptr)
+//         add(args.x, args.y)
+//     }
+//
+// This is useful because the task-based goroutine start implementation only
+// allows a single (pointer) argument to the newly started goroutine. Also, it
+// ignores the return value because newly started goroutines do not have a
+// return value.
+func (c *Compiler) createGoroutineStartWrapper(fn llvm.Value) llvm.Value {
+	var wrapper llvm.Value
+
+	if !fn.IsAFunction().IsNil() {
+		// See whether this wrapper has already been created. If so, return it.
+		name := fn.Name()
+		wrapper = c.mod.NamedFunction(name + "$gowrapper")
+		if !wrapper.IsNil() {
+			return c.builder.CreatePtrToInt(wrapper, c.uintptrType, "")
+		}
+
+		// Save the current position in the IR builder.
+		currentBlock := c.builder.GetInsertBlock()
+		defer c.builder.SetInsertPointAtEnd(currentBlock)
+
+		// Create the wrapper.
+		wrapperType := llvm.FunctionType(c.ctx.VoidType(), []llvm.Type{c.i8ptrType}, false)
+		wrapper = llvm.AddFunction(c.mod, name+"$gowrapper", wrapperType)
+		wrapper.SetLinkage(llvm.PrivateLinkage)
+		wrapper.SetUnnamedAddr(true)
+		entry := c.ctx.AddBasicBlock(wrapper, "entry")
+		c.builder.SetInsertPointAtEnd(entry)
+
+		// Create the list of params for the call.
+		paramTypes := fn.Type().ElementType().ParamTypes()
+		params := c.emitPointerUnpack(wrapper.Param(0), paramTypes[:len(paramTypes)-1])
+		params = append(params, llvm.Undef(c.i8ptrType))
+
+		// Create the call.
+		c.builder.CreateCall(fn, params, "")
+
+	} else {
+		// For a function pointer like this:
+		//
+		//     var funcPtr func(x, y int) int
+		//
+		// A wrapper like the following is created:
+		//
+		//     func .gowrapper(ptr *unsafe.Pointer) {
+		//         args := (*struct{
+		//             x, y int
+		//             fn   func(x, y int) int
+		//         })(ptr)
+		//         args.fn(x, y)
+		//     }
+		//
+		// With a bit of luck, identical wrapper functions like these can be
+		// merged into one.
+
+		// Save the current position in the IR builder.
+		currentBlock := c.builder.GetInsertBlock()
+		defer c.builder.SetInsertPointAtEnd(currentBlock)
+
+		// Create the wrapper.
+		wrapperType := llvm.FunctionType(c.ctx.VoidType(), []llvm.Type{c.i8ptrType}, false)
+		wrapper = llvm.AddFunction(c.mod, ".gowrapper", wrapperType)
+		wrapper.SetLinkage(llvm.InternalLinkage)
+		wrapper.SetUnnamedAddr(true)
+		entry := c.ctx.AddBasicBlock(wrapper, "entry")
+		c.builder.SetInsertPointAtEnd(entry)
+
+		// Get the list of parameters, with the extra parameters at the end.
+		paramTypes := fn.Type().ElementType().ParamTypes()
+		paramTypes[len(paramTypes)-1] = fn.Type() // the last element is the function pointer
+		params := c.emitPointerUnpack(wrapper.Param(0), paramTypes)
+
+		// Get the function pointer.
+		fnPtr := params[len(params)-1]
+
+		// Ignore the last param, which isn't used anymore.
+		// TODO: avoid this extra "parent handle" parameter in most functions.
+		params[len(params)-1] = llvm.Undef(c.i8ptrType)
+
+		// Create the call.
+		c.builder.CreateCall(fnPtr, params, "")
+	}
+
+	// Finish the function. Every basic block must end in a terminator, and
+	// because goroutines never return a value we can simply return void.
+	c.builder.CreateRetVoid()
+
+	// Return a ptrtoint of the wrapper, not the function itself.
+	return c.builder.CreatePtrToInt(wrapper, c.uintptrType, "")
+}

compiler/inlineasm.go

@@ -18,10 +18,19 @@ import (
 //     func ReadRegister(name string) uintptr
 //
 // The register name must be a constant, for example "sp".
-func (c *Compiler) emitReadRegister(args []ssa.Value) (llvm.Value, error) {
+func (c *Compiler) emitReadRegister(name string, args []ssa.Value) (llvm.Value, error) {
 	fnType := llvm.FunctionType(c.uintptrType, []llvm.Type{}, false)
 	regname := constant.StringVal(args[0].(*ssa.Const).Value)
-	target := llvm.InlineAsm(fnType, "mov $0, "+regname, "=r", false, false, 0)
+	var asm string
+	switch name {
+	case "device/arm.ReadRegister":
+		asm = "mov $0, " + regname
+	case "device/riscv.ReadRegister":
+		asm = "mv $0, " + regname
+	default:
+		panic("unknown architecture")
+	}
+	target := llvm.InlineAsm(fnType, asm, "=r", false, false, 0)
 	return c.builder.CreateCall(target, nil, ""), nil
 }
 

compiler/interface.go

@@ -28,14 +28,14 @@ func (c *Compiler) parseMakeInterface(val llvm.Value, typ types.Type, pos token.
 	itfMethodSetGlobal := c.getTypeMethodSet(typ)
 	itfConcreteTypeGlobal := c.mod.NamedGlobal("typeInInterface:" + itfTypeCodeGlobal.Name())
 	if itfConcreteTypeGlobal.IsNil() {
-		typeInInterface := c.mod.GetTypeByName("runtime.typeInInterface")
+		typeInInterface := c.getLLVMRuntimeType("typeInInterface")
 		itfConcreteTypeGlobal = llvm.AddGlobal(c.mod, typeInInterface, "typeInInterface:"+itfTypeCodeGlobal.Name())
 		itfConcreteTypeGlobal.SetInitializer(llvm.ConstNamedStruct(typeInInterface, []llvm.Value{itfTypeCodeGlobal, itfMethodSetGlobal}))
 		itfConcreteTypeGlobal.SetGlobalConstant(true)
 		itfConcreteTypeGlobal.SetLinkage(llvm.PrivateLinkage)
 	}
 	itfTypeCode := c.builder.CreatePtrToInt(itfConcreteTypeGlobal, c.uintptrType, "")
-	itf := llvm.Undef(c.mod.GetTypeByName("runtime._interface"))
+	itf := llvm.Undef(c.getLLVMRuntimeType("_interface"))
 	itf = c.builder.CreateInsertValue(itf, itfTypeCode, 0, "")
 	itf = c.builder.CreateInsertValue(itf, itfValue, 1, "")
 	return itf
@@ -45,27 +45,100 @@ func (c *Compiler) parseMakeInterface(val llvm.Value, typ types.Type, pos token.
 // It returns a pointer to an external global which should be replaced with the
 // real type in the interface lowering pass.
 func (c *Compiler) getTypeCode(typ types.Type) llvm.Value {
-	globalName := "type:" + getTypeCodeName(typ)
+	globalName := "reflect/types.type:" + getTypeCodeName(typ)
 	global := c.mod.NamedGlobal(globalName)
 	if global.IsNil() {
-		global = llvm.AddGlobal(c.mod, c.mod.GetTypeByName("runtime.typecodeID"), globalName)
+		// Create a new typecode global.
+		global = llvm.AddGlobal(c.mod, c.getLLVMRuntimeType("typecodeID"), globalName)
+		// Some type classes contain more information for underlying types or
+		// element types. Store it directly in the typecode global to make
+		// reflect lowering simpler.
+		var references llvm.Value
+		var length int64
+		switch typ := typ.(type) {
+		case *types.Named:
+			references = c.getTypeCode(typ.Underlying())
+		case *types.Chan:
+			references = c.getTypeCode(typ.Elem())
+		case *types.Pointer:
+			references = c.getTypeCode(typ.Elem())
+		case *types.Slice:
+			references = c.getTypeCode(typ.Elem())
+		case *types.Array:
+			references = c.getTypeCode(typ.Elem())
+			length = typ.Len()
+		case *types.Struct:
+			// Take a pointer to the typecodeID of the first field (if it exists).
+			structGlobal := c.makeStructTypeFields(typ)
+			references = llvm.ConstBitCast(structGlobal, global.Type())
+		}
+		if !references.IsNil() {
+			// Set the 'references' field of the runtime.typecodeID struct.
+			globalValue := llvm.ConstNull(global.Type().ElementType())
+			globalValue = llvm.ConstInsertValue(globalValue, references, []uint32{0})
+			if length != 0 {
+				lengthValue := llvm.ConstInt(c.uintptrType, uint64(length), false)
+				globalValue = llvm.ConstInsertValue(globalValue, lengthValue, []uint32{1})
+			}
+			global.SetInitializer(globalValue)
+			global.SetLinkage(llvm.PrivateLinkage)
+		}
 		global.SetGlobalConstant(true)
 	}
 	return global
 }
 
+// makeStructTypeFields creates a new global that stores all type information
+// related to this struct type, and returns the resulting global. This global is
+// actually an array of all the fields in the structs.
+func (c *Compiler) makeStructTypeFields(typ *types.Struct) llvm.Value {
+	// The global is an array of runtime.structField structs.
+	runtimeStructField := c.getLLVMRuntimeType("structField")
+	structGlobalType := llvm.ArrayType(runtimeStructField, typ.NumFields())
+	structGlobal := llvm.AddGlobal(c.mod, structGlobalType, "reflect/types.structFields")
+	structGlobalValue := llvm.ConstNull(structGlobalType)
+	for i := 0; i < typ.NumFields(); i++ {
+		fieldGlobalValue := llvm.ConstNull(runtimeStructField)
+		fieldGlobalValue = llvm.ConstInsertValue(fieldGlobalValue, c.getTypeCode(typ.Field(i).Type()), []uint32{0})
+		fieldName := c.makeGlobalArray([]byte(typ.Field(i).Name()), "reflect/types.structFieldName", c.ctx.Int8Type())
+		fieldName.SetLinkage(llvm.PrivateLinkage)
+		fieldName.SetUnnamedAddr(true)
+		fieldName = llvm.ConstGEP(fieldName, []llvm.Value{
+			llvm.ConstInt(llvm.Int32Type(), 0, false),
+			llvm.ConstInt(llvm.Int32Type(), 0, false),
+		})
+		fieldGlobalValue = llvm.ConstInsertValue(fieldGlobalValue, fieldName, []uint32{1})
+		if typ.Tag(i) != "" {
+			fieldTag := c.makeGlobalArray([]byte(typ.Tag(i)), "reflect/types.structFieldTag", c.ctx.Int8Type())
+			fieldTag.SetLinkage(llvm.PrivateLinkage)
+			fieldTag.SetUnnamedAddr(true)
+			fieldTag = llvm.ConstGEP(fieldTag, []llvm.Value{
+				llvm.ConstInt(llvm.Int32Type(), 0, false),
+				llvm.ConstInt(llvm.Int32Type(), 0, false),
+			})
+			fieldGlobalValue = llvm.ConstInsertValue(fieldGlobalValue, fieldTag, []uint32{2})
+		}
+		if typ.Field(i).Embedded() {
+			fieldEmbedded := llvm.ConstInt(c.ctx.Int1Type(), 1, false)
+			fieldGlobalValue = llvm.ConstInsertValue(fieldGlobalValue, fieldEmbedded, []uint32{3})
+		}
+		structGlobalValue = llvm.ConstInsertValue(structGlobalValue, fieldGlobalValue, []uint32{uint32(i)})
+	}
+	structGlobal.SetInitializer(structGlobalValue)
+	structGlobal.SetUnnamedAddr(true)
+	structGlobal.SetLinkage(llvm.PrivateLinkage)
+	return structGlobal
+}
+
 // getTypeCodeName returns a name for this type that can be used in the
 // interface lowering pass to assign type codes as expected by the reflect
 // package. See getTypeCodeNum.
 func getTypeCodeName(t types.Type) string {
-	name := ""
-	if named, ok := t.(*types.Named); ok {
-		name = "~" + named.String() + ":"
-		t = t.Underlying()
-	}
 	switch t := t.(type) {
+	case *types.Named:
+		return "named:" + t.String()
 	case *types.Array:
-		return "array:" + name + strconv.FormatInt(t.Len(), 10) + ":" + getTypeCodeName(t.Elem())
+		return "array:" + strconv.FormatInt(t.Len(), 10) + ":" + getTypeCodeName(t.Elem())
 	case *types.Basic:
 		var kind string
 		switch t.Kind() {
@@ -108,21 +181,21 @@ func getTypeCodeName(t types.Type) string {
 		default:
 			panic("unknown basic type: " + t.Name())
 		}
-		return "basic:" + name + kind
+		return "basic:" + kind
 	case *types.Chan:
-		return "chan:" + name + getTypeCodeName(t.Elem())
+		return "chan:" + getTypeCodeName(t.Elem())
 	case *types.Interface:
 		methods := make([]string, t.NumMethods())
 		for i := 0; i < t.NumMethods(); i++ {
 			methods[i] = getTypeCodeName(t.Method(i).Type())
 		}
-		return "interface:" + name + "{" + strings.Join(methods, ",") + "}"
+		return "interface:" + "{" + strings.Join(methods, ",") + "}"
 	case *types.Map:
 		keyType := getTypeCodeName(t.Key())
 		elemType := getTypeCodeName(t.Elem())
-		return "map:" + name + "{" + keyType + "," + elemType + "}"
+		return "map:" + "{" + keyType + "," + elemType + "}"
 	case *types.Pointer:
-		return "pointer:" + name + getTypeCodeName(t.Elem())
+		return "pointer:" + getTypeCodeName(t.Elem())
 	case *types.Signature:
 		params := make([]string, t.Params().Len())
 		for i := 0; i < t.Params().Len(); i++ {
@@ -132,19 +205,22 @@ func getTypeCodeName(t types.Type) string {
 		for i := 0; i < t.Results().Len(); i++ {
 			results[i] = getTypeCodeName(t.Results().At(i).Type())
 		}
-		return "func:" + name + "{" + strings.Join(params, ",") + "}{" + strings.Join(results, ",") + "}"
+		return "func:" + "{" + strings.Join(params, ",") + "}{" + strings.Join(results, ",") + "}"
 	case *types.Slice:
-		return "slice:" + name + getTypeCodeName(t.Elem())
+		return "slice:" + getTypeCodeName(t.Elem())
 	case *types.Struct:
 		elems := make([]string, t.NumFields())
-		if t.NumFields() > 2 && t.Field(0).Name() == "C union" {
-			// TODO: report this as a normal error instead of panicking.
-			panic("cgo unions are not allowed in interfaces")
-		}
 		for i := 0; i < t.NumFields(); i++ {
-			elems[i] = getTypeCodeName(t.Field(i).Type())
+			embedded := ""
+			if t.Field(i).Embedded() {
+				embedded = "#"
+			}
+			elems[i] = embedded + t.Field(i).Name() + ":" + getTypeCodeName(t.Field(i).Type())
+			if t.Tag(i) != "" {
+				elems[i] += "`" + t.Tag(i) + "`"
+			}
 		}
-		return "struct:" + name + "{" + strings.Join(elems, ",") + "}"
+		return "struct:" + "{" + strings.Join(elems, ",") + "}"
 	default:
 		panic("unknown type: " + t.String())
 	}
@@ -163,11 +239,11 @@ func (c *Compiler) getTypeMethodSet(typ types.Type) llvm.Value {
 	ms := c.ir.Program.MethodSets.MethodSet(typ)
 	if ms.Len() == 0 {
 		// no methods, so can leave that one out
-		return llvm.ConstPointerNull(llvm.PointerType(c.mod.GetTypeByName("runtime.interfaceMethodInfo"), 0))
+		return llvm.ConstPointerNull(llvm.PointerType(c.getLLVMRuntimeType("interfaceMethodInfo"), 0))
 	}
 
 	methods := make([]llvm.Value, ms.Len())
-	interfaceMethodInfoType := c.mod.GetTypeByName("runtime.interfaceMethodInfo")
+	interfaceMethodInfoType := c.getLLVMRuntimeType("interfaceMethodInfo")
 	for i := 0; i < ms.Len(); i++ {
 		method := ms.At(i)
 		signatureGlobal := c.getMethodSignature(method.Obj().(*types.Func))
@@ -203,14 +279,14 @@ func (c *Compiler) getInterfaceMethodSet(typ *types.Named) llvm.Value {
 		return llvm.ConstGEP(global, []llvm.Value{zero, zero})
 	}
 
-	// Every method is a *i16 reference indicating the signature of this method.
+	// Every method is a *i8 reference indicating the signature of this method.
 	methods := make([]llvm.Value, typ.Underlying().(*types.Interface).NumMethods())
 	for i := range methods {
 		method := typ.Underlying().(*types.Interface).Method(i)
 		methods[i] = c.getMethodSignature(method)
 	}
 
-	value := llvm.ConstArray(methods[0].Type(), methods)
+	value := llvm.ConstArray(c.i8ptrType, methods)
 	global = llvm.AddGlobal(c.mod, value.Type(), typ.String()+"$interface")
 	global.SetInitializer(value)
 	global.SetGlobalConstant(true)
@@ -219,7 +295,7 @@ func (c *Compiler) getInterfaceMethodSet(typ *types.Named) llvm.Value {
 }
 
 // getMethodSignature returns a global variable which is a reference to an
-// external *i16 indicating the indicating the signature of this method. It is
+// external *i8 indicating the indicating the signature of this method. It is
 // used during the interface lowering pass.
 func (c *Compiler) getMethodSignature(method *types.Func) llvm.Value {
 	signature := ir.MethodSignature(method)
@@ -300,7 +376,7 @@ func (c *Compiler) parseTypeAssert(frame *Frame, expr *ssa.TypeAssert) llvm.Valu
 	// Continue after the if statement.
 	c.builder.SetInsertPointAtEnd(nextBlock)
 	phi := c.builder.CreatePHI(assertedType, "typeassert.value")
-	phi.AddIncoming([]llvm.Value{c.getZeroValue(assertedType), valueOk}, []llvm.BasicBlock{prevBlock, okBlock})
+	phi.AddIncoming([]llvm.Value{llvm.ConstNull(assertedType), valueOk}, []llvm.BasicBlock{prevBlock, okBlock})
 
 	if expr.CommaOk {
 		tuple := c.ctx.ConstStruct([]llvm.Value{llvm.Undef(assertedType), llvm.Undef(c.ctx.Int1Type())}, false) // create empty tuple
@@ -385,6 +461,9 @@ func (c *Compiler) getInterfaceInvokeWrapper(f *ir.Function) llvm.Value {
 	paramTypes := append([]llvm.Type{c.i8ptrType}, fnType.ParamTypes()[len(expandedReceiverType):]...)
 	wrapFnType := llvm.FunctionType(fnType.ReturnType(), paramTypes, false)
 	wrapper = llvm.AddFunction(c.mod, wrapperName, wrapFnType)
+	if f.LLVMFn.LastParam().Name() == "parentHandle" {
+		wrapper.LastParam().SetName("parentHandle")
+	}
 	c.interfaceInvokeWrappers = append(c.interfaceInvokeWrappers, interfaceInvokeWrapper{
 		fn:           f,
 		wrapper:      wrapper,
@@ -403,7 +482,7 @@ func (c *Compiler) createInterfaceInvokeWrapper(state interfaceInvokeWrapper) {
 	wrapper.SetUnnamedAddr(true)
 
 	// add debug info if needed
-	if c.Debug {
+	if c.Debug() {
 		pos := c.ir.Program.Fset.Position(fn.Pos())
 		difunc := c.attachDebugInfoRaw(fn, wrapper, "$invoke", pos.Filename, pos.Line)
 		c.builder.SetCurrentDebugLocation(uint(pos.Line), uint(pos.Column), difunc, llvm.Metadata{})

compiler/interrupt.go

@@ -0,0 +1,91 @@
+package compiler
+
+import (
+	"strconv"
+	"strings"
+
+	"golang.org/x/tools/go/ssa"
+	"tinygo.org/x/go-llvm"
+)
+
+// emitInterruptGlobal creates a new runtime/interrupt.Interrupt struct that
+// will be lowered to a real interrupt during interrupt lowering.
+//
+// This two-stage approach allows unused interrupts to be optimized away if
+// necessary.
+func (c *Compiler) emitInterruptGlobal(frame *Frame, instr *ssa.CallCommon) (llvm.Value, error) {
+	// Get the interrupt number, which must be a compile-time constant.
+	id, ok := instr.Args[0].(*ssa.Const)
+	if !ok {
+		return llvm.Value{}, c.makeError(instr.Pos(), "interrupt ID is not a constant")
+	}
+
+	// Get the func value, which also must be a compile time constant.
+	// Note that bound functions are allowed if the function has a pointer
+	// receiver and is a global. This is rather strict but still allows for
+	// idiomatic Go code.
+	funcValue := c.getValue(frame, instr.Args[1])
+	if funcValue.IsAConstant().IsNil() {
+		// Try to determine the cause of the non-constantness for a nice error
+		// message.
+		switch instr.Args[1].(type) {
+		case *ssa.MakeClosure:
+			// This may also be a bound method.
+			return llvm.Value{}, c.makeError(instr.Pos(), "closures are not supported in interrupt.New")
+		}
+		// Fall back to a generic error.
+		return llvm.Value{}, c.makeError(instr.Pos(), "interrupt function must be constant")
+	}
+
+	// Create a new global of type runtime/interrupt.handle. Globals of this
+	// type are lowered in the interrupt lowering pass.
+	globalType := c.ir.Program.ImportedPackage("runtime/interrupt").Type("handle").Type()
+	globalLLVMType := c.getLLVMType(globalType)
+	globalName := "runtime/interrupt.$interrupt" + strconv.FormatInt(id.Int64(), 10)
+	if global := c.mod.NamedGlobal(globalName); !global.IsNil() {
+		return llvm.Value{}, c.makeError(instr.Pos(), "interrupt redeclared in this program")
+	}
+	global := llvm.AddGlobal(c.mod, globalLLVMType, globalName)
+	global.SetLinkage(llvm.PrivateLinkage)
+	global.SetGlobalConstant(true)
+	initializer := llvm.ConstNull(globalLLVMType)
+	initializer = llvm.ConstInsertValue(initializer, funcValue, []uint32{0})
+	initializer = llvm.ConstInsertValue(initializer, llvm.ConstInt(c.intType, uint64(id.Int64()), true), []uint32{1, 0})
+	global.SetInitializer(initializer)
+
+	// Add debug info to the interrupt global.
+	if c.Debug() {
+		pos := c.ir.Program.Fset.Position(instr.Pos())
+		diglobal := c.dibuilder.CreateGlobalVariableExpression(c.difiles[pos.Filename], llvm.DIGlobalVariableExpression{
+			Name:        "interrupt" + strconv.FormatInt(id.Int64(), 10),
+			LinkageName: globalName,
+			File:        c.getDIFile(pos.Filename),
+			Line:        pos.Line,
+			Type:        c.getDIType(globalType),
+			Expr:        c.dibuilder.CreateExpression(nil),
+			LocalToUnit: false,
+		})
+		global.AddMetadata(0, diglobal)
+	}
+
+	// Create the runtime/interrupt.Interrupt type. It is a struct with a single
+	// member of type int.
+	num := llvm.ConstPtrToInt(global, c.intType)
+	interrupt := llvm.ConstNamedStruct(c.mod.GetTypeByName("runtime/interrupt.Interrupt"), []llvm.Value{num})
+
+	// Add dummy "use" call for AVR, because interrupts may be used even though
+	// they are never referenced again. This is unlike Cortex-M or the RISC-V
+	// PLIC where each interrupt must be enabled using the interrupt number, and
+	// thus keeps the Interrupt object alive.
+	// This call is removed during interrupt lowering.
+	if strings.HasPrefix(c.Triple(), "avr") {
+		useFn := c.mod.NamedFunction("runtime/interrupt.use")
+		if useFn.IsNil() {
+			useFnType := llvm.FunctionType(c.ctx.VoidType(), []llvm.Type{interrupt.Type()}, false)
+			useFn = llvm.AddFunction(c.mod, "runtime/interrupt.use", useFnType)
+		}
+		c.builder.CreateCall(useFn, []llvm.Value{interrupt}, "")
+	}
+
+	return interrupt, nil
+}

compiler/llvm.go

@@ -1,6 +1,7 @@
 package compiler
 
 import (
+	"github.com/tinygo-org/tinygo/compiler/llvmutil"
 	"tinygo.org/x/go-llvm"
 )
 
@@ -22,115 +23,47 @@ func getUses(value llvm.Value) []llvm.Value {
 	return uses
 }
 
-// createEntryBlockAlloca creates a new alloca in the entry block, even though
-// the IR builder is located elsewhere. It assumes that the insert point is
-// after the last instruction in the current block. Also, it adds lifetime
-// information to the IR signalling that the alloca won't be used before this
-// point.
+// createTemporaryAlloca creates a new alloca in the entry block and adds
+// lifetime start infromation in the IR signalling that the alloca won't be used
+// before this point.
 //
 // This is useful for creating temporary allocas for intrinsics. Don't forget to
-// end the lifetime after you're done with it.
-func (c *Compiler) createEntryBlockAlloca(t llvm.Type, name string) (alloca, bitcast, size llvm.Value) {
-	currentBlock := c.builder.GetInsertBlock()
-	c.builder.SetInsertPointBefore(currentBlock.Parent().EntryBasicBlock().FirstInstruction())
-	alloca = c.builder.CreateAlloca(t, name)
-	c.builder.SetInsertPointAtEnd(currentBlock)
-	bitcast = c.builder.CreateBitCast(alloca, c.i8ptrType, name+".bitcast")
-	size = llvm.ConstInt(c.ctx.Int64Type(), c.targetData.TypeAllocSize(t), false)
-	c.builder.CreateCall(c.getLifetimeStartFunc(), []llvm.Value{size, bitcast}, "")
-	return
+// end the lifetime using emitLifetimeEnd after you're done with it.
+func (c *Compiler) createTemporaryAlloca(t llvm.Type, name string) (alloca, bitcast, size llvm.Value) {
+	return llvmutil.CreateTemporaryAlloca(c.builder, c.mod, t, name)
 }
 
-// getLifetimeStartFunc returns the llvm.lifetime.start intrinsic and creates it
-// first if it doesn't exist yet.
-func (c *Compiler) getLifetimeStartFunc() llvm.Value {
-	fn := c.mod.NamedFunction("llvm.lifetime.start.p0i8")
-	if fn.IsNil() {
-		fnType := llvm.FunctionType(c.ctx.VoidType(), []llvm.Type{c.ctx.Int64Type(), c.i8ptrType}, false)
-		fn = llvm.AddFunction(c.mod, "llvm.lifetime.start.p0i8", fnType)
-	}
-	return fn
+// emitLifetimeEnd signals the end of an (alloca) lifetime by calling the
+// llvm.lifetime.end intrinsic. It is commonly used together with
+// createTemporaryAlloca.
+func (c *Compiler) emitLifetimeEnd(ptr, size llvm.Value) {
+	llvmutil.EmitLifetimeEnd(c.builder, c.mod, ptr, size)
 }
 
-// getLifetimeEndFunc returns the llvm.lifetime.end intrinsic and creates it
-// first if it doesn't exist yet.
-func (c *Compiler) getLifetimeEndFunc() llvm.Value {
-	fn := c.mod.NamedFunction("llvm.lifetime.end.p0i8")
-	if fn.IsNil() {
-		fnType := llvm.FunctionType(c.ctx.VoidType(), []llvm.Type{c.ctx.Int64Type(), c.i8ptrType}, false)
-		fn = llvm.AddFunction(c.mod, "llvm.lifetime.end.p0i8", fnType)
-	}
-	return fn
+// emitPointerPack packs the list of values into a single pointer value using
+// bitcasts, or else allocates a value on the heap if it cannot be packed in the
+// pointer value directly. It returns the pointer with the packed data.
+func (c *Compiler) emitPointerPack(values []llvm.Value) llvm.Value {
+	return llvmutil.EmitPointerPack(c.builder, c.mod, c.Config, values)
 }
 
-// splitBasicBlock splits a LLVM basic block into two parts. All instructions
-// after afterInst are moved into a new basic block (created right after the
-// current one) with the given name.
-func (c *Compiler) splitBasicBlock(afterInst llvm.Value, insertAfter llvm.BasicBlock, name string) llvm.BasicBlock {
-	oldBlock := afterInst.InstructionParent()
-	newBlock := c.ctx.InsertBasicBlock(insertAfter, name)
-	var nextInstructions []llvm.Value // values to move
-
-	// Collect to-be-moved instructions.
-	inst := afterInst
-	for {
-		inst = llvm.NextInstruction(inst)
-		if inst.IsNil() {
-			break
-		}
-		nextInstructions = append(nextInstructions, inst)
-	}
-
-	// Move instructions.
-	c.builder.SetInsertPointAtEnd(newBlock)
-	for _, inst := range nextInstructions {
-		inst.RemoveFromParentAsInstruction()
-		c.builder.Insert(inst)
-	}
-
-	// Find PHI nodes to update.
-	var phiNodes []llvm.Value // PHI nodes to update
-	for bb := insertAfter.Parent().FirstBasicBlock(); !bb.IsNil(); bb = llvm.NextBasicBlock(bb) {
-		for inst := bb.FirstInstruction(); !inst.IsNil(); inst = llvm.NextInstruction(inst) {
-			if inst.IsAPHINode().IsNil() {
-				continue
-			}
-			needsUpdate := false
-			incomingCount := inst.IncomingCount()
-			for i := 0; i < incomingCount; i++ {
-				if inst.IncomingBlock(i) == oldBlock {
-					needsUpdate = true
-					break
-				}
-			}
-			if !needsUpdate {
-				// PHI node has no incoming edge from the old block.
-				continue
-			}
-			phiNodes = append(phiNodes, inst)
-		}
-	}
-
-	// Update PHI nodes.
-	for _, phi := range phiNodes {
-		c.builder.SetInsertPointBefore(phi)
-		newPhi := c.builder.CreatePHI(phi.Type(), "")
-		incomingCount := phi.IncomingCount()
-		incomingVals := make([]llvm.Value, incomingCount)
-		incomingBlocks := make([]llvm.BasicBlock, incomingCount)
-		for i := 0; i < incomingCount; i++ {
-			value := phi.IncomingValue(i)
-			block := phi.IncomingBlock(i)
-			if block == oldBlock {
-				block = newBlock
-			}
-			incomingVals[i] = value
-			incomingBlocks[i] = block
-		}
-		newPhi.AddIncoming(incomingVals, incomingBlocks)
-		phi.ReplaceAllUsesWith(newPhi)
-		phi.EraseFromParentAsInstruction()
-	}
-
-	return newBlock
+// emitPointerUnpack extracts a list of values packed using emitPointerPack.
+func (c *Compiler) emitPointerUnpack(ptr llvm.Value, valueTypes []llvm.Type) []llvm.Value {
+	return llvmutil.EmitPointerUnpack(c.builder, c.mod, ptr, valueTypes)
+}
+
+// makeGlobalArray creates a new LLVM global with the given name and integers as
+// contents, and returns the global.
+// Note that it is left with the default linkage etc., you should set
+// linkage/constant/etc properties yourself.
+func (c *Compiler) makeGlobalArray(buf []byte, name string, elementType llvm.Type) llvm.Value {
+	globalType := llvm.ArrayType(elementType, len(buf))
+	global := llvm.AddGlobal(c.mod, globalType, name)
+	value := llvm.Undef(globalType)
+	for i := 0; i < len(buf); i++ {
+		ch := uint64(buf[i])
+		value = llvm.ConstInsertValue(value, llvm.ConstInt(elementType, ch, false), []uint32{uint32(i)})
+	}
+	global.SetInitializer(value)
+	return global
 }

compiler/llvmutil/llvm.go

@@ -0,0 +1,168 @@
+// Package llvmutil contains utility functions used across multiple compiler
+// packages. For example, they may be used by both the compiler pacakge and
+// transformation packages.
+//
+// Normally, utility packages are avoided. However, in this case, the utility
+// functions are non-trivial and hard to get right. Copying them to multiple
+// places would be a big risk if only one of them is updated.
+package llvmutil
+
+import "tinygo.org/x/go-llvm"
+
+// CreateEntryBlockAlloca creates a new alloca in the entry block, even though
+// the IR builder is located elsewhere. It assumes that the insert point is
+// at the end of the current block.
+func CreateEntryBlockAlloca(builder llvm.Builder, t llvm.Type, name string) llvm.Value {
+	currentBlock := builder.GetInsertBlock()
+	entryBlock := currentBlock.Parent().EntryBasicBlock()
+	if entryBlock.FirstInstruction().IsNil() {
+		builder.SetInsertPointAtEnd(entryBlock)
+	} else {
+		builder.SetInsertPointBefore(entryBlock.FirstInstruction())
+	}
+	alloca := builder.CreateAlloca(t, name)
+	builder.SetInsertPointAtEnd(currentBlock)
+	return alloca
+}
+
+// CreateTemporaryAlloca creates a new alloca in the entry block and adds
+// lifetime start infromation in the IR signalling that the alloca won't be used
+// before this point.
+//
+// This is useful for creating temporary allocas for intrinsics. Don't forget to
+// end the lifetime using emitLifetimeEnd after you're done with it.
+func CreateTemporaryAlloca(builder llvm.Builder, mod llvm.Module, t llvm.Type, name string) (alloca, bitcast, size llvm.Value) {
+	ctx := t.Context()
+	targetData := llvm.NewTargetData(mod.DataLayout())
+	i8ptrType := llvm.PointerType(ctx.Int8Type(), 0)
+	alloca = CreateEntryBlockAlloca(builder, t, name)
+	bitcast = builder.CreateBitCast(alloca, i8ptrType, name+".bitcast")
+	size = llvm.ConstInt(ctx.Int64Type(), targetData.TypeAllocSize(t), false)
+	builder.CreateCall(getLifetimeStartFunc(mod), []llvm.Value{size, bitcast}, "")
+	return
+}
+
+// CreateInstructionAlloca creates an alloca in the entry block, and places lifetime control intrinsics around the instruction
+func CreateInstructionAlloca(builder llvm.Builder, mod llvm.Module, t llvm.Type, inst llvm.Value, name string) llvm.Value {
+	ctx := mod.Context()
+	targetData := llvm.NewTargetData(mod.DataLayout())
+	i8ptrType := llvm.PointerType(ctx.Int8Type(), 0)
+
+	alloca := CreateEntryBlockAlloca(builder, t, name)
+	builder.SetInsertPointBefore(inst)
+	bitcast := builder.CreateBitCast(alloca, i8ptrType, name+".bitcast")
+	size := llvm.ConstInt(ctx.Int64Type(), targetData.TypeAllocSize(t), false)
+	builder.CreateCall(getLifetimeStartFunc(mod), []llvm.Value{size, bitcast}, "")
+	if next := llvm.NextInstruction(inst); !next.IsNil() {
+		builder.SetInsertPointBefore(next)
+	} else {
+		builder.SetInsertPointAtEnd(inst.InstructionParent())
+	}
+	builder.CreateCall(getLifetimeEndFunc(mod), []llvm.Value{size, bitcast}, "")
+	return alloca
+}
+
+// EmitLifetimeEnd signals the end of an (alloca) lifetime by calling the
+// llvm.lifetime.end intrinsic. It is commonly used together with
+// createTemporaryAlloca.
+func EmitLifetimeEnd(builder llvm.Builder, mod llvm.Module, ptr, size llvm.Value) {
+	builder.CreateCall(getLifetimeEndFunc(mod), []llvm.Value{size, ptr}, "")
+}
+
+// getLifetimeStartFunc returns the llvm.lifetime.start intrinsic and creates it
+// first if it doesn't exist yet.
+func getLifetimeStartFunc(mod llvm.Module) llvm.Value {
+	fn := mod.NamedFunction("llvm.lifetime.start.p0i8")
+	ctx := mod.Context()
+	i8ptrType := llvm.PointerType(ctx.Int8Type(), 0)
+	if fn.IsNil() {
+		fnType := llvm.FunctionType(ctx.VoidType(), []llvm.Type{ctx.Int64Type(), i8ptrType}, false)
+		fn = llvm.AddFunction(mod, "llvm.lifetime.start.p0i8", fnType)
+	}
+	return fn
+}
+
+// getLifetimeEndFunc returns the llvm.lifetime.end intrinsic and creates it
+// first if it doesn't exist yet.
+func getLifetimeEndFunc(mod llvm.Module) llvm.Value {
+	fn := mod.NamedFunction("llvm.lifetime.end.p0i8")
+	ctx := mod.Context()
+	i8ptrType := llvm.PointerType(ctx.Int8Type(), 0)
+	if fn.IsNil() {
+		fnType := llvm.FunctionType(ctx.VoidType(), []llvm.Type{ctx.Int64Type(), i8ptrType}, false)
+		fn = llvm.AddFunction(mod, "llvm.lifetime.end.p0i8", fnType)
+	}
+	return fn
+}
+
+// SplitBasicBlock splits a LLVM basic block into two parts. All instructions
+// after afterInst are moved into a new basic block (created right after the
+// current one) with the given name.
+func SplitBasicBlock(builder llvm.Builder, afterInst llvm.Value, insertAfter llvm.BasicBlock, name string) llvm.BasicBlock {
+	oldBlock := afterInst.InstructionParent()
+	newBlock := afterInst.Type().Context().InsertBasicBlock(insertAfter, name)
+	var nextInstructions []llvm.Value // values to move
+
+	// Collect to-be-moved instructions.
+	inst := afterInst
+	for {
+		inst = llvm.NextInstruction(inst)
+		if inst.IsNil() {
+			break
+		}
+		nextInstructions = append(nextInstructions, inst)
+	}
+
+	// Move instructions.
+	builder.SetInsertPointAtEnd(newBlock)
+	for _, inst := range nextInstructions {
+		inst.RemoveFromParentAsInstruction()
+		builder.Insert(inst)
+	}
+
+	// Find PHI nodes to update.
+	var phiNodes []llvm.Value // PHI nodes to update
+	for bb := insertAfter.Parent().FirstBasicBlock(); !bb.IsNil(); bb = llvm.NextBasicBlock(bb) {
+		for inst := bb.FirstInstruction(); !inst.IsNil(); inst = llvm.NextInstruction(inst) {
+			if inst.IsAPHINode().IsNil() {
+				continue
+			}
+			needsUpdate := false
+			incomingCount := inst.IncomingCount()
+			for i := 0; i < incomingCount; i++ {
+				if inst.IncomingBlock(i) == oldBlock {
+					needsUpdate = true
+					break
+				}
+			}
+			if !needsUpdate {
+				// PHI node has no incoming edge from the old block.
+				continue
+			}
+			phiNodes = append(phiNodes, inst)
+		}
+	}
+
+	// Update PHI nodes.
+	for _, phi := range phiNodes {
+		builder.SetInsertPointBefore(phi)
+		newPhi := builder.CreatePHI(phi.Type(), "")
+		incomingCount := phi.IncomingCount()
+		incomingVals := make([]llvm.Value, incomingCount)
+		incomingBlocks := make([]llvm.BasicBlock, incomingCount)
+		for i := 0; i < incomingCount; i++ {
+			value := phi.IncomingValue(i)
+			block := phi.IncomingBlock(i)
+			if block == oldBlock {
+				block = newBlock
+			}
+			incomingVals[i] = value
+			incomingBlocks[i] = block
+		}
+		newPhi.AddIncoming(incomingVals, incomingBlocks)
+		phi.ReplaceAllUsesWith(newPhi)
+		phi.EraseFromParentAsInstruction()
+	}
+
+	return newBlock
+}

compiler/llvmutil/wordpack.go

@@ -0,0 +1,141 @@
+package llvmutil
+
+// This file contains utility functions to pack and unpack sets of values. It
+// can take in a list of values and tries to store it efficiently in the pointer
+// itself if possible and legal.
+
+import (
+	"github.com/tinygo-org/tinygo/compileopts"
+	"tinygo.org/x/go-llvm"
+)
+
+// EmitPointerPack packs the list of values into a single pointer value using
+// bitcasts, or else allocates a value on the heap if it cannot be packed in the
+// pointer value directly. It returns the pointer with the packed data.
+func EmitPointerPack(builder llvm.Builder, mod llvm.Module, config *compileopts.Config, values []llvm.Value) llvm.Value {
+	ctx := mod.Context()
+	targetData := llvm.NewTargetData(mod.DataLayout())
+	i8ptrType := llvm.PointerType(mod.Context().Int8Type(), 0)
+	uintptrType := ctx.IntType(llvm.NewTargetData(mod.DataLayout()).PointerSize() * 8)
+
+	valueTypes := make([]llvm.Type, len(values))
+	for i, value := range values {
+		valueTypes[i] = value.Type()
+	}
+	packedType := ctx.StructType(valueTypes, false)
+
+	// Allocate memory for the packed data.
+	var packedAlloc, packedHeapAlloc llvm.Value
+	size := targetData.TypeAllocSize(packedType)
+	if size == 0 {
+		return llvm.ConstPointerNull(i8ptrType)
+	} else if len(values) == 1 && values[0].Type().TypeKind() == llvm.PointerTypeKind {
+		return builder.CreateBitCast(values[0], i8ptrType, "pack.ptr")
+	} else if size <= targetData.TypeAllocSize(i8ptrType) {
+		// Packed data fits in a pointer, so store it directly inside the
+		// pointer.
+		if len(values) == 1 && values[0].Type().TypeKind() == llvm.IntegerTypeKind {
+			// Try to keep this cast in SSA form.
+			return builder.CreateIntToPtr(values[0], i8ptrType, "pack.int")
+		}
+		// Because packedType is a struct and we have to cast it to a *i8, store
+		// it in an alloca first for bitcasting (store+bitcast+load).
+		packedAlloc, _, _ = CreateTemporaryAlloca(builder, mod, packedType, "")
+	} else {
+		// Packed data is bigger than a pointer, so allocate it on the heap.
+		sizeValue := llvm.ConstInt(uintptrType, size, false)
+		alloc := mod.NamedFunction("runtime.alloc")
+		packedHeapAlloc = builder.CreateCall(alloc, []llvm.Value{
+			sizeValue,
+			llvm.Undef(i8ptrType),            // unused context parameter
+			llvm.ConstPointerNull(i8ptrType), // coroutine handle
+		}, "")
+		if config.NeedsStackObjects() {
+			trackPointer := mod.NamedFunction("runtime.trackPointer")
+			builder.CreateCall(trackPointer, []llvm.Value{
+				packedHeapAlloc,
+				llvm.Undef(i8ptrType),            // unused context parameter
+				llvm.ConstPointerNull(i8ptrType), // coroutine handle
+			}, "")
+		}
+		packedAlloc = builder.CreateBitCast(packedHeapAlloc, llvm.PointerType(packedType, 0), "")
+	}
+	// Store all values in the alloca or heap pointer.
+	for i, value := range values {
+		indices := []llvm.Value{
+			llvm.ConstInt(ctx.Int32Type(), 0, false),
+			llvm.ConstInt(ctx.Int32Type(), uint64(i), false),
+		}
+		gep := builder.CreateInBoundsGEP(packedAlloc, indices, "")
+		builder.CreateStore(value, gep)
+	}
+
+	if packedHeapAlloc.IsNil() {
+		// Load value (as *i8) from the alloca.
+		packedAlloc = builder.CreateBitCast(packedAlloc, llvm.PointerType(i8ptrType, 0), "")
+		result := builder.CreateLoad(packedAlloc, "")
+		packedPtr := builder.CreateBitCast(packedAlloc, i8ptrType, "")
+		packedSize := llvm.ConstInt(ctx.Int64Type(), targetData.TypeAllocSize(packedAlloc.Type()), false)
+		EmitLifetimeEnd(builder, mod, packedPtr, packedSize)
+		return result
+	} else {
+		// Get the original heap allocation pointer, which already is an *i8.
+		return packedHeapAlloc
+	}
+}
+
+// EmitPointerUnpack extracts a list of values packed using EmitPointerPack.
+func EmitPointerUnpack(builder llvm.Builder, mod llvm.Module, ptr llvm.Value, valueTypes []llvm.Type) []llvm.Value {
+	ctx := mod.Context()
+	targetData := llvm.NewTargetData(mod.DataLayout())
+	i8ptrType := llvm.PointerType(mod.Context().Int8Type(), 0)
+	uintptrType := ctx.IntType(llvm.NewTargetData(mod.DataLayout()).PointerSize() * 8)
+
+	packedType := ctx.StructType(valueTypes, false)
+
+	// Get a correctly-typed pointer to the packed data.
+	var packedAlloc, packedRawAlloc llvm.Value
+	size := targetData.TypeAllocSize(packedType)
+	if size == 0 {
+		// No data to unpack.
+	} else if len(valueTypes) == 1 && valueTypes[0].TypeKind() == llvm.PointerTypeKind {
+		// A single pointer is always stored directly.
+		return []llvm.Value{builder.CreateBitCast(ptr, valueTypes[0], "unpack.ptr")}
+	} else if size <= targetData.TypeAllocSize(i8ptrType) {
+		// Packed data stored directly in pointer.
+		if len(valueTypes) == 1 && valueTypes[0].TypeKind() == llvm.IntegerTypeKind {
+			// Keep this cast in SSA form.
+			return []llvm.Value{builder.CreatePtrToInt(ptr, valueTypes[0], "unpack.int")}
+		}
+		// Fallback: load it using an alloca.
+		packedRawAlloc, _, _ = CreateTemporaryAlloca(builder, mod, llvm.PointerType(i8ptrType, 0), "unpack.raw.alloc")
+		packedRawValue := builder.CreateBitCast(ptr, llvm.PointerType(i8ptrType, 0), "unpack.raw.value")
+		builder.CreateStore(packedRawValue, packedRawAlloc)
+		packedAlloc = builder.CreateBitCast(packedRawAlloc, llvm.PointerType(packedType, 0), "unpack.alloc")
+	} else {
+		// Packed data stored on the heap. Bitcast the passed-in pointer to the
+		// correct pointer type.
+		packedAlloc = builder.CreateBitCast(ptr, llvm.PointerType(packedType, 0), "unpack.raw.ptr")
+	}
+	// Load each value from the packed data.
+	values := make([]llvm.Value, len(valueTypes))
+	for i, valueType := range valueTypes {
+		if targetData.TypeAllocSize(valueType) == 0 {
+			// This value has length zero, so there's nothing to load.
+			values[i] = llvm.ConstNull(valueType)
+			continue
+		}
+		indices := []llvm.Value{
+			llvm.ConstInt(ctx.Int32Type(), 0, false),
+			llvm.ConstInt(ctx.Int32Type(), uint64(i), false),
+		}
+		gep := builder.CreateInBoundsGEP(packedAlloc, indices, "")
+		values[i] = builder.CreateLoad(gep, "")
+	}
+	if !packedRawAlloc.IsNil() {
+		allocPtr := builder.CreateBitCast(packedRawAlloc, i8ptrType, "")
+		allocSize := llvm.ConstInt(ctx.Int64Type(), targetData.TypeAllocSize(uintptrType), false)
+		EmitLifetimeEnd(builder, mod, allocPtr, allocSize)
+	}
+	return values
+}

compiler/map.go

@@ -15,7 +15,7 @@ func (c *Compiler) emitMapLookup(keyType, valueType types.Type, m, key llvm.Valu
 	// Allocate the memory for the resulting type. Do not zero this memory: it
 	// will be zeroed by the hashmap get implementation if the key is not
 	// present in the map.
-	mapValueAlloca, mapValuePtr, mapValueSize := c.createEntryBlockAlloca(llvmValueType, "hashmap.value")
+	mapValueAlloca, mapValuePtr, mapValueSize := c.createTemporaryAlloca(llvmValueType, "hashmap.value")
 
 	// Do the lookup. How it is done depends on the key type.
 	var commaOkValue llvm.Value
@@ -27,21 +27,21 @@ func (c *Compiler) emitMapLookup(keyType, valueType types.Type, m, key llvm.Valu
 		// key can be compared with runtime.memequal
 		// Store the key in an alloca, in the entry block to avoid dynamic stack
 		// growth.
-		mapKeyAlloca, mapKeyPtr, mapKeySize := c.createEntryBlockAlloca(key.Type(), "hashmap.key")
+		mapKeyAlloca, mapKeyPtr, mapKeySize := c.createTemporaryAlloca(key.Type(), "hashmap.key")
 		c.builder.CreateStore(key, mapKeyAlloca)
 		// Fetch the value from the hashmap.
 		params := []llvm.Value{m, mapKeyPtr, mapValuePtr}
 		commaOkValue = c.createRuntimeCall("hashmapBinaryGet", params, "")
-		c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{mapKeySize, mapKeyPtr}, "")
+		c.emitLifetimeEnd(mapKeyPtr, mapKeySize)
 	} else {
 		// Not trivially comparable using memcmp.
-		return llvm.Value{}, c.makeError(pos, "only strings, bools, ints or structs of bools/ints are supported as map keys, but got: "+keyType.String())
+		return llvm.Value{}, c.makeError(pos, "only strings, bools, ints, pointers or structs of bools/ints are supported as map keys, but got: "+keyType.String())
 	}
 
 	// Load the resulting value from the hashmap. The value is set to the zero
 	// value if the key doesn't exist in the hashmap.
 	mapValue := c.builder.CreateLoad(mapValueAlloca, "")
-	c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{mapValueSize, mapValuePtr}, "")
+	c.emitLifetimeEnd(mapValuePtr, mapValueSize)
 
 	if commaOk {
 		tuple := llvm.Undef(c.ctx.StructType([]llvm.Type{llvmValueType, c.ctx.Int1Type()}, false))
@@ -54,7 +54,7 @@ func (c *Compiler) emitMapLookup(keyType, valueType types.Type, m, key llvm.Valu
 }
 
 func (c *Compiler) emitMapUpdate(keyType types.Type, m, key, value llvm.Value, pos token.Pos) {
-	valueAlloca, valuePtr, valueSize := c.createEntryBlockAlloca(value.Type(), "hashmap.value")
+	valueAlloca, valuePtr, valueSize := c.createTemporaryAlloca(value.Type(), "hashmap.value")
 	c.builder.CreateStore(value, valueAlloca)
 	keyType = keyType.Underlying()
 	if t, ok := keyType.(*types.Basic); ok && t.Info()&types.IsString != 0 {
@@ -63,15 +63,15 @@ func (c *Compiler) emitMapUpdate(keyType types.Type, m, key, value llvm.Value, p
 		c.createRuntimeCall("hashmapStringSet", params, "")
 	} else if hashmapIsBinaryKey(keyType) {
 		// key can be compared with runtime.memequal
-		keyAlloca, keyPtr, keySize := c.createEntryBlockAlloca(key.Type(), "hashmap.key")
+		keyAlloca, keyPtr, keySize := c.createTemporaryAlloca(key.Type(), "hashmap.key")
 		c.builder.CreateStore(key, keyAlloca)
 		params := []llvm.Value{m, keyPtr, valuePtr}
 		c.createRuntimeCall("hashmapBinarySet", params, "")
-		c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{keySize, keyPtr}, "")
+		c.emitLifetimeEnd(keyPtr, keySize)
 	} else {
-		c.addError(pos, "only strings, bools, ints or structs of bools/ints are supported as map keys, but got: "+keyType.String())
+		c.addError(pos, "only strings, bools, ints, pointers or structs of bools/ints are supported as map keys, but got: "+keyType.String())
 	}
-	c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{valueSize, valuePtr}, "")
+	c.emitLifetimeEnd(valuePtr, valueSize)
 }
 
 func (c *Compiler) emitMapDelete(keyType types.Type, m, key llvm.Value, pos token.Pos) error {
@@ -82,14 +82,14 @@ func (c *Compiler) emitMapDelete(keyType types.Type, m, key llvm.Value, pos toke
 		c.createRuntimeCall("hashmapStringDelete", params, "")
 		return nil
 	} else if hashmapIsBinaryKey(keyType) {
-		keyAlloca, keyPtr, keySize := c.createEntryBlockAlloca(key.Type(), "hashmap.key")
+		keyAlloca, keyPtr, keySize := c.createTemporaryAlloca(key.Type(), "hashmap.key")
 		c.builder.CreateStore(key, keyAlloca)
 		params := []llvm.Value{m, keyPtr}
 		c.createRuntimeCall("hashmapBinaryDelete", params, "")
-		c.builder.CreateCall(c.getLifetimeEndFunc(), []llvm.Value{keySize, keyPtr}, "")
+		c.emitLifetimeEnd(keyPtr, keySize)
 		return nil
 	} else {
-		return c.makeError(pos, "only strings, bools, ints or structs of bools/ints are supported as map keys, but got: "+keyType.String())
+		return c.makeError(pos, "only strings, bools, ints, pointers or structs of bools/ints are supported as map keys, but got: "+keyType.String())
 	}
 }
 
@@ -121,6 +121,8 @@ func hashmapIsBinaryKey(keyType types.Type) bool {
 	switch keyType := keyType.(type) {
 	case *types.Basic:
 		return keyType.Info()&(types.IsBoolean|types.IsInteger) != 0
+	case *types.Pointer:
+		return true
 	case *types.Struct:
 		for i := 0; i < keyType.NumFields(); i++ {
 			fieldType := keyType.Field(i).Type().Underlying()

compiler/optimizer.go

@@ -3,12 +3,13 @@ package compiler
 import (
 	"errors"
 
+	"github.com/tinygo-org/tinygo/transform"
 	"tinygo.org/x/go-llvm"
 )
 
 // Run the LLVM optimizer over the module.
 // The inliner can be disabled (if necessary) by passing 0 to the inlinerThreshold.
-func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) error {
+func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) []error {
 	builder := llvm.NewPassManagerBuilder()
 	defer builder.Dispose()
 	builder.SetOptLevel(optLevel)
@@ -18,8 +19,16 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 	}
 	builder.AddCoroutinePassesToExtensionPoints()
 
-	if c.PanicStrategy == "trap" {
-		c.replacePanicsWithTrap() // -panic=trap
+	if c.PanicStrategy() == "trap" {
+		transform.ReplacePanicsWithTrap(c.mod) // -panic=trap
+	}
+
+	// run a check of all of our code
+	if c.VerifyIR() {
+		errs := c.checkModule()
+		if errs != nil {
+			return errs
+		}
 	}
 
 	// Run function passes for each function.
@@ -36,19 +45,27 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 		// Run some preparatory passes for the Go optimizer.
 		goPasses := llvm.NewPassManager()
 		defer goPasses.Dispose()
-		goPasses.AddGlobalOptimizerPass()
 		goPasses.AddGlobalDCEPass()
+		goPasses.AddGlobalOptimizerPass()
 		goPasses.AddConstantPropagationPass()
 		goPasses.AddAggressiveDCEPass()
 		goPasses.AddFunctionAttrsPass()
 		goPasses.Run(c.mod)
 
 		// Run Go-specific optimization passes.
-		c.OptimizeMaps()
-		c.OptimizeStringToBytes()
-		c.OptimizeAllocs()
-		c.LowerInterfaces()
-		c.LowerFuncValues()
+		transform.OptimizeMaps(c.mod)
+		transform.OptimizeStringToBytes(c.mod)
+		transform.OptimizeAllocs(c.mod)
+		transform.LowerInterfaces(c.mod)
+
+		errs := transform.LowerInterruptRegistrations(c.mod)
+		if len(errs) > 0 {
+			return errs
+		}
+
+		if c.funcImplementation() == funcValueSwitch {
+			transform.LowerFuncValues(c.mod)
+		}
 
 		// After interfaces are lowered, there are many more opportunities for
 		// interprocedural optimizations. To get them to work, function
@@ -56,8 +73,8 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 		goPasses.Run(c.mod)
 
 		// Run TinyGo-specific interprocedural optimizations.
-		c.OptimizeAllocs()
-		c.OptimizeStringToBytes()
+		transform.OptimizeAllocs(c.mod)
+		transform.OptimizeStringToBytes(c.mod)
 
 		// Lower runtime.isnil calls to regular nil comparisons.
 		isnil := c.mod.NamedFunction("runtime.isnil")
@@ -77,19 +94,30 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 
 		err := c.LowerGoroutines()
 		if err != nil {
-			return err
+			return []error{err}
 		}
 	} else {
 		// Must be run at any optimization level.
-		c.LowerInterfaces()
-		c.LowerFuncValues()
+		transform.LowerInterfaces(c.mod)
+		if c.funcImplementation() == funcValueSwitch {
+			transform.LowerFuncValues(c.mod)
+		}
 		err := c.LowerGoroutines()
 		if err != nil {
-			return err
+			return []error{err}
+		}
+		errs := transform.LowerInterruptRegistrations(c.mod)
+		if len(errs) > 0 {
+			return errs
+		}
+	}
+	if c.VerifyIR() {
+		if errs := c.checkModule(); errs != nil {
+			return errs
 		}
 	}
 	if err := c.Verify(); err != nil {
-		return errors.New("optimizations caused a verification failure")
+		return []error{errors.New("optimizations caused a verification failure")}
 	}
 
 	if sizeLevel >= 2 {
@@ -102,6 +130,15 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 		}
 	}
 
+	// After TinyGo-specific transforms have finished, undo exporting these functions.
+	for _, name := range c.getFunctionsUsedInTransforms() {
+		fn := c.mod.NamedFunction(name)
+		if fn.IsNil() {
+			continue
+		}
+		fn.SetLinkage(llvm.InternalLinkage)
+	}
+
 	// Run function passes again, because without it, llvm.coro.size.i32()
 	// doesn't get lowered.
 	for fn := c.mod.FirstFunction(); !fn.IsNil(); fn = llvm.NextFunction(fn) {
@@ -115,268 +152,13 @@ func (c *Compiler) Optimize(optLevel, sizeLevel int, inlinerThreshold uint) erro
 	builder.Populate(modPasses)
 	modPasses.Run(c.mod)
 
-	if c.gcIsPrecise() {
-		c.addGlobalsBitmap()
+	hasGCPass := transform.AddGlobalsBitmap(c.mod)
+	hasGCPass = transform.MakeGCStackSlots(c.mod) || hasGCPass
+	if hasGCPass {
 		if err := c.Verify(); err != nil {
-			return errors.New("GC pass caused a verification failure")
+			return []error{errors.New("GC pass caused a verification failure")}
 		}
 	}
 
 	return nil
 }
-
-// Replace panic calls with calls to llvm.trap, to reduce code size. This is the
-// -panic=trap intrinsic.
-func (c *Compiler) replacePanicsWithTrap() {
-	trap := c.mod.NamedFunction("llvm.trap")
-	for _, name := range []string{"runtime._panic", "runtime.runtimePanic"} {
-		fn := c.mod.NamedFunction(name)
-		if fn.IsNil() {
-			continue
-		}
-		for _, use := range getUses(fn) {
-			if use.IsACallInst().IsNil() || use.CalledValue() != fn {
-				panic("expected use of a panic function to be a call")
-			}
-			c.builder.SetInsertPointBefore(use)
-			c.builder.CreateCall(trap, nil, "")
-		}
-	}
-}
-
-// Eliminate created but not used maps.
-//
-// In the future, this should statically allocate created but never modified
-// maps. This has not yet been implemented, however.
-func (c *Compiler) OptimizeMaps() {
-	hashmapMake := c.mod.NamedFunction("runtime.hashmapMake")
-	if hashmapMake.IsNil() {
-		// nothing to optimize
-		return
-	}
-
-	hashmapBinarySet := c.mod.NamedFunction("runtime.hashmapBinarySet")
-	hashmapStringSet := c.mod.NamedFunction("runtime.hashmapStringSet")
-
-	for _, makeInst := range getUses(hashmapMake) {
-		updateInsts := []llvm.Value{}
-		unknownUses := false // are there any uses other than setting a value?
-
-		for _, use := range getUses(makeInst) {
-			if use := use.IsACallInst(); !use.IsNil() {
-				switch use.CalledValue() {
-				case hashmapBinarySet, hashmapStringSet:
-					updateInsts = append(updateInsts, use)
-				default:
-					unknownUses = true
-				}
-			} else {
-				unknownUses = true
-			}
-		}
-
-		if !unknownUses {
-			// This map can be entirely removed, as it is only created but never
-			// used.
-			for _, inst := range updateInsts {
-				inst.EraseFromParentAsInstruction()
-			}
-			makeInst.EraseFromParentAsInstruction()
-		}
-	}
-}
-
-// Transform runtime.stringToBytes(...) calls into const []byte slices whenever
-// possible. This optimizes the following pattern:
-//     w.Write([]byte("foo"))
-// where Write does not store to the slice.
-func (c *Compiler) OptimizeStringToBytes() {
-	stringToBytes := c.mod.NamedFunction("runtime.stringToBytes")
-	if stringToBytes.IsNil() {
-		// nothing to optimize
-		return
-	}
-
-	for _, call := range getUses(stringToBytes) {
-		strptr := call.Operand(0)
-		strlen := call.Operand(1)
-
-		// strptr is always constant because strings are always constant.
-
-		convertedAllUses := true
-		for _, use := range getUses(call) {
-			nilValue := llvm.Value{}
-			if use.IsAExtractValueInst() == nilValue {
-				convertedAllUses = false
-				continue
-			}
-			switch use.Type().TypeKind() {
-			case llvm.IntegerTypeKind:
-				// A length (len or cap). Propagate the length value.
-				use.ReplaceAllUsesWith(strlen)
-				use.EraseFromParentAsInstruction()
-			case llvm.PointerTypeKind:
-				// The string pointer itself.
-				if !c.isReadOnly(use) {
-					convertedAllUses = false
-					continue
-				}
-				use.ReplaceAllUsesWith(strptr)
-				use.EraseFromParentAsInstruction()
-			default:
-				// should not happen
-				panic("unknown return type of runtime.stringToBytes: " + use.Type().String())
-			}
-		}
-		if convertedAllUses {
-			// Call to runtime.stringToBytes can be eliminated: both the input
-			// and the output is constant.
-			call.EraseFromParentAsInstruction()
-		}
-	}
-}
-
-// Basic escape analysis: translate runtime.alloc calls into alloca
-// instructions.
-func (c *Compiler) OptimizeAllocs() {
-	allocator := c.mod.NamedFunction("runtime.alloc")
-	if allocator.IsNil() {
-		// nothing to optimize
-		return
-	}
-
-	heapallocs := getUses(allocator)
-	for _, heapalloc := range heapallocs {
-		nilValue := llvm.Value{}
-		if heapalloc.Operand(0).IsAConstant() == nilValue {
-			// Do not allocate variable length arrays on the stack.
-			continue
-		}
-		size := heapalloc.Operand(0).ZExtValue()
-		if size > 256 {
-			// The maximum value for a stack allocation.
-			// TODO: tune this, this is just a random value.
-			continue
-		}
-
-		// In general the pattern is:
-		//     %0 = call i8* @runtime.alloc(i32 %size)
-		//     %1 = bitcast i8* %0 to type*
-		//     (use %1 only)
-		// But the bitcast might sometimes be dropped when allocating an *i8.
-		// The 'bitcast' variable below is thus usually a bitcast of the
-		// heapalloc but not always.
-		bitcast := heapalloc // instruction that creates the value
-		if uses := getUses(heapalloc); len(uses) == 1 && uses[0].IsABitCastInst() != nilValue {
-			// getting only bitcast use
-			bitcast = uses[0]
-		}
-		if !c.doesEscape(bitcast) {
-			// Insert alloca in the entry block. Do it here so that mem2reg can
-			// promote it to a SSA value.
-			fn := bitcast.InstructionParent().Parent()
-			c.builder.SetInsertPointBefore(fn.EntryBasicBlock().FirstInstruction())
-			alignment := c.targetData.ABITypeAlignment(c.i8ptrType)
-			sizeInWords := (size + uint64(alignment) - 1) / uint64(alignment)
-			allocaType := llvm.ArrayType(c.ctx.IntType(alignment*8), int(sizeInWords))
-			alloca := c.builder.CreateAlloca(allocaType, "stackalloc.alloca")
-			zero := c.getZeroValue(alloca.Type().ElementType())
-			c.builder.CreateStore(zero, alloca)
-			stackalloc := c.builder.CreateBitCast(alloca, bitcast.Type(), "stackalloc")
-			bitcast.ReplaceAllUsesWith(stackalloc)
-			if heapalloc != bitcast {
-				bitcast.EraseFromParentAsInstruction()
-			}
-			heapalloc.EraseFromParentAsInstruction()
-		}
-	}
-}
-
-// Very basic escape analysis.
-func (c *Compiler) doesEscape(value llvm.Value) bool {
-	uses := getUses(value)
-	for _, use := range uses {
-		nilValue := llvm.Value{}
-		if use.IsAGetElementPtrInst() != nilValue {
-			if c.doesEscape(use) {
-				return true
-			}
-		} else if use.IsABitCastInst() != nilValue {
-			// A bitcast escapes if the casted-to value escapes.
-			if c.doesEscape(use) {
-				return true
-			}
-		} else if use.IsALoadInst() != nilValue {
-			// Load does not escape.
-		} else if use.IsAStoreInst() != nilValue {
-			// Store only escapes when the value is stored to, not when the
-			// value is stored into another value.
-			if use.Operand(0) == value {
-				return true
-			}
-		} else if use.IsACallInst() != nilValue {
-			if !c.hasFlag(use, value, "nocapture") {
-				return true
-			}
-		} else if use.IsAICmpInst() != nilValue {
-			// Comparing pointers don't let the pointer escape.
-			// This is often a compiler-inserted nil check.
-		} else {
-			// Unknown instruction, might escape.
-			return true
-		}
-	}
-
-	// does not escape
-	return false
-}
-
-// Check whether the given value (which is of pointer type) is never stored to.
-func (c *Compiler) isReadOnly(value llvm.Value) bool {
-	uses := getUses(value)
-	for _, use := range uses {
-		nilValue := llvm.Value{}
-		if use.IsAGetElementPtrInst() != nilValue {
-			if !c.isReadOnly(use) {
-				return false
-			}
-		} else if use.IsACallInst() != nilValue {
-			if !c.hasFlag(use, value, "readonly") {
-				return false
-			}
-		} else {
-			// Unknown instruction, might not be readonly.
-			return false
-		}
-	}
-	return true
-}
-
-// Check whether all uses of this param as parameter to the call have the given
-// flag. In most cases, there will only be one use but a function could take the
-// same parameter twice, in which case both must have the flag.
-// A flag can be any enum flag, like "readonly".
-func (c *Compiler) hasFlag(call, param llvm.Value, kind string) bool {
-	fn := call.CalledValue()
-	nilValue := llvm.Value{}
-	if fn.IsAFunction() == nilValue {
-		// This is not a function but something else, like a function pointer.
-		return false
-	}
-	kindID := llvm.AttributeKindID(kind)
-	for i := 0; i < fn.ParamsCount(); i++ {
-		if call.Operand(i) != param {
-			// This is not the parameter we're checking.
-			continue
-		}
-		index := i + 1 // param attributes start at 1
-		attr := fn.GetEnumAttributeAtIndex(index, kindID)
-		nilAttribute := llvm.Attribute{}
-		if attr == nilAttribute {
-			// At least one parameter doesn't have the flag (there may be
-			// multiple).
-			return false
-		}
-	}
-	return true
-}

compiler/reflect.go

@@ -1,156 +0,0 @@
-package compiler
-
-import (
-	"math/big"
-	"strings"
-)
-
-var basicTypes = map[string]int64{
-	"bool":       1,
-	"int":        2,
-	"int8":       3,
-	"int16":      4,
-	"int32":      5,
-	"int64":      6,
-	"uint":       7,
-	"uint8":      8,
-	"uint16":     9,
-	"uint32":     10,
-	"uint64":     11,
-	"uintptr":    12,
-	"float32":    13,
-	"float64":    14,
-	"complex64":  15,
-	"complex128": 16,
-	"string":     17,
-	"unsafeptr":  18,
-}
-
-func (c *Compiler) assignTypeCodes(typeSlice typeInfoSlice) {
-	fn := c.mod.NamedFunction("reflect.ValueOf")
-	if fn.IsNil() {
-		// reflect.ValueOf is never used, so we can use the most efficient
-		// encoding possible.
-		for i, t := range typeSlice {
-			t.num = uint64(i + 1)
-		}
-		return
-	}
-
-	// Assign typecodes the way the reflect package expects.
-	fallbackIndex := 1
-	namedTypes := make(map[string]int)
-	for _, t := range typeSlice {
-		if t.name[:5] != "type:" {
-			panic("expected type name to start with 'type:'")
-		}
-		num := c.getTypeCodeNum(t.name[5:], &fallbackIndex, namedTypes)
-		if num.BitLen() > c.uintptrType.IntTypeWidth() || !num.IsUint64() {
-			// TODO: support this in some way, using a side table for example.
-			// That's less efficient but better than not working at all.
-			// Particularly important on systems with 16-bit pointers (e.g.
-			// AVR).
-			panic("compiler: could not store type code number inside interface type code")
-		}
-		t.num = num.Uint64()
-	}
-}
-
-// getTypeCodeNum returns the typecode for a given type as expected by the
-// reflect package. Also see getTypeCodeName, which serializes types to a string
-// based on a types.Type value for this function.
-func (c *Compiler) getTypeCodeNum(id string, fallbackIndex *int, namedTypes map[string]int) *big.Int {
-	// Note: see src/reflect/type.go for bit allocations.
-	// A type can be named or unnamed. Example of both:
-	//     basic:~foo:uint64
-	//     basic:uint64
-	// Extract the class (basic, slice, pointer, etc.), the name, and the
-	// contents of this type ID string. Allocate bits based on that, as
-	// src/runtime/types.go expects.
-	class := id[:strings.IndexByte(id, ':')]
-	value := id[len(class)+1:]
-	name := ""
-	if value[0] == '~' {
-		name = value[1:strings.IndexByte(value, ':')]
-		value = value[len(name)+2:]
-	}
-	if class == "basic" {
-		// Basic types follow the following bit pattern:
-		//    ...xxxxx0
-		// where xxxxx is allocated for the 18 possible basic types and all the
-		// upper bits are used to indicate the named type.
-		num, ok := basicTypes[value]
-		if !ok {
-			panic("invalid basic type: " + id)
-		}
-		if name != "" {
-			// This type is named, set the upper bits to the name ID.
-			num |= int64(getNamedTypeNum(namedTypes, name)) << 5
-		}
-		return big.NewInt(num << 1)
-	} else {
-		// Complex types use the following bit pattern:
-		//    ...nxxx1
-		// where xxx indicates the complex type (any non-basic type). The upper
-		// bits contain whatever the type contains. Types that wrap a single
-		// other type (channel, interface, pointer, slice) just contain the bits
-		// of the wrapped type. Other types (like struct) have a different
-		// method of encoding the contents of the type.
-		var num *big.Int
-		var classNumber int64
-		switch class {
-		case "chan":
-			num = c.getTypeCodeNum(value, fallbackIndex, namedTypes)
-			classNumber = 0
-		case "interface":
-			num = big.NewInt(int64(*fallbackIndex))
-			*fallbackIndex++
-			classNumber = 1
-		case "pointer":
-			num = c.getTypeCodeNum(value, fallbackIndex, namedTypes)
-			classNumber = 2
-		case "slice":
-			num = c.getTypeCodeNum(value, fallbackIndex, namedTypes)
-			classNumber = 3
-		case "array":
-			num = big.NewInt(int64(*fallbackIndex))
-			*fallbackIndex++
-			classNumber = 4
-		case "func":
-			num = big.NewInt(int64(*fallbackIndex))
-			*fallbackIndex++
-			classNumber = 5
-		case "map":
-			num = big.NewInt(int64(*fallbackIndex))
-			*fallbackIndex++
-			classNumber = 6
-		case "struct":
-			num = big.NewInt(int64(*fallbackIndex))
-			*fallbackIndex++
-			classNumber = 7
-		default:
-			panic("unknown type kind: " + id)
-		}
-		if name == "" {
-			num.Lsh(num, 5).Or(num, big.NewInt((classNumber<<1)+1))
-		} else {
-			// TODO: store num in a sidetable
-			num = big.NewInt(int64(getNamedTypeNum(namedTypes, name))<<1 | 1)
-			num.Lsh(num, 4).Or(num, big.NewInt((classNumber<<1)+1))
-		}
-		return num
-	}
-}
-
-// getNamedTypeNum returns an appropriate (unique) number for the given named
-// type. If the name already has a number that number is returned, else a new
-// number is returned. The number is always non-zero.
-func getNamedTypeNum(namedTypes map[string]int, name string) int {
-	if num, ok := namedTypes[name]; ok {
-		return num
-	} else {
-		num = len(namedTypes) + 1
-		namedTypes[name] = num
-		return num
-	}
-}

compiler/sizes.go

@@ -63,12 +63,6 @@ func (s *StdSizes) Alignof(T types.Type) int64 {
 
 func (s *StdSizes) Offsetsof(fields []*types.Var) []int64 {
 	offsets := make([]int64, len(fields))
-	if len(fields) > 1 && fields[0].Name() == "C union" {
-		// This struct contains the magic "C union" field which indicates that
-		// this is actually a union from CGo.
-		// All fields in the union start at 0 so return that.
-		return offsets // all fields are still set to 0
-	}
 	var o int64
 	for i, f := range fields {
 		a := s.Alignof(f.Type())
@@ -116,6 +110,9 @@ func (s *StdSizes) Sizeof(T types.Type) int64 {
 		if k == types.UnsafePointer {
 			return s.PtrSize
 		}
+		if k == types.Invalid {
+			return 0 // only relevant when there is a type error somewhere
+		}
 		panic("unknown basic type: " + t.String())
 	case *types.Array:
 		n := t.Len()
@@ -143,29 +140,12 @@ func (s *StdSizes) Sizeof(T types.Type) int64 {
 				maxAlign = al
 			}
 		}
-		if fields[0].Name() == "C union" {
-			// Magic field that indicates this is a CGo union and not a struct.
-			// The size is the biggest element, aligned to the element with the
-			// biggest alignment. This is not necessarily the same, for example
-			// in the following union:
-			//     union { int32_t l; int16_t s[3] }
-			maxSize := int64(0)
-			for _, field := range fields[1:] {
-				si := s.Sizeof(field.Type())
-				if si > maxSize {
-					maxSize = si
-				}
-			}
-			return align(maxSize, maxAlign)
-		} else {
-			// This is a regular struct.
-			// Pick the size that fits this struct and add some alignment. Some
-			// structs have some extra padding at the end which should also be
-			// taken care of:
-			//     struct { int32 n; byte b }
-			offsets := s.Offsetsof(fields)
-			return align(offsets[n-1]+s.Sizeof(fields[n-1].Type()), maxAlign)
-		}
+		// Pick the size that fits this struct and add some alignment. Some
+		// structs have some extra padding at the end which should also be taken
+		// care of:
+		//     struct { int32 n; byte b }
+		offsets := s.Offsetsof(fields)
+		return align(offsets[n-1]+s.Sizeof(fields[n-1].Type()), maxAlign)
 	case *types.Interface:
 		return s.PtrSize * 2
 	case *types.Pointer:

compiler/symbol.go

@@ -0,0 +1,147 @@
+package compiler
+
+// This file manages symbols, that is, functions and globals. It reads their
+// pragmas, determines the link name, etc.
+
+import (
+	"go/ast"
+	"go/token"
+	"go/types"
+	"strconv"
+	"strings"
+
+	"github.com/tinygo-org/tinygo/loader"
+	"golang.org/x/tools/go/ssa"
+	"tinygo.org/x/go-llvm"
+)
+
+// globalInfo contains some information about a specific global. By default,
+// linkName is equal to .RelString(nil) on a global and extern is false, but for
+// some symbols this is different (due to //go:extern for example).
+type globalInfo struct {
+	linkName string // go:extern
+	extern   bool   // go:extern
+	align    int    // go:align
+}
+
+// loadASTComments loads comments on globals from the AST, for use later in the
+// program. In particular, they are required for //go:extern pragmas on globals.
+func (c *Compiler) loadASTComments(lprogram *loader.Program) {
+	c.astComments = map[string]*ast.CommentGroup{}
+	for _, pkgInfo := range lprogram.Sorted() {
+		for _, file := range pkgInfo.Files {
+			for _, decl := range file.Decls {
+				switch decl := decl.(type) {
+				case *ast.GenDecl:
+					switch decl.Tok {
+					case token.VAR:
+						if len(decl.Specs) != 1 {
+							continue
+						}
+						for _, spec := range decl.Specs {
+							switch spec := spec.(type) {
+							case *ast.ValueSpec: // decl.Tok == token.VAR
+								for _, name := range spec.Names {
+									id := pkgInfo.Pkg.Path() + "." + name.Name
+									c.astComments[id] = decl.Doc
+								}
+							}
+						}
+					}
+				}
+			}
+		}
+	}
+}
+
+// getGlobal returns a LLVM IR global value for a Go SSA global. It is added to
+// the LLVM IR if it has not been added already.
+func (c *Compiler) getGlobal(g *ssa.Global) llvm.Value {
+	info := c.getGlobalInfo(g)
+	llvmGlobal := c.mod.NamedGlobal(info.linkName)
+	if llvmGlobal.IsNil() {
+		typ := g.Type().(*types.Pointer).Elem()
+		llvmType := c.getLLVMType(typ)
+		llvmGlobal = llvm.AddGlobal(c.mod, llvmType, info.linkName)
+		if !info.extern {
+			llvmGlobal.SetInitializer(llvm.ConstNull(llvmType))
+			llvmGlobal.SetLinkage(llvm.InternalLinkage)
+		}
+
+		// Set alignment from the //go:align comment.
+		var alignInBits uint32
+		if info.align < 0 || info.align&(info.align-1) != 0 {
+			// Check for power-of-two (or 0).
+			// See: https://stackoverflow.com/a/108360
+			c.addError(g.Pos(), "global variable alignment must be a positive power of two")
+		} else {
+			// Set the alignment only when it is a power of two.
+			alignInBits = uint32(info.align) ^ uint32(info.align-1)
+			if info.align > c.targetData.ABITypeAlignment(llvmType) {
+				llvmGlobal.SetAlignment(info.align)
+			}
+		}
+
+		if c.Debug() {
+			// Add debug info.
+			// TODO: this should be done for every global in the program, not just
+			// the ones that are referenced from some code.
+			pos := c.ir.Program.Fset.Position(g.Pos())
+			diglobal := c.dibuilder.CreateGlobalVariableExpression(c.difiles[pos.Filename], llvm.DIGlobalVariableExpression{
+				Name:        g.RelString(nil),
+				LinkageName: info.linkName,
+				File:        c.getDIFile(pos.Filename),
+				Line:        pos.Line,
+				Type:        c.getDIType(typ),
+				LocalToUnit: false,
+				Expr:        c.dibuilder.CreateExpression(nil),
+				AlignInBits: alignInBits,
+			})
+			llvmGlobal.AddMetadata(0, diglobal)
+		}
+	}
+	return llvmGlobal
+}
+
+// getGlobalInfo returns some information about a specific global.
+func (c *Compiler) getGlobalInfo(g *ssa.Global) globalInfo {
+	info := globalInfo{}
+	if strings.HasPrefix(g.Name(), "C.") {
+		// Created by CGo: such a name cannot be created by regular C code.
+		info.linkName = g.Name()[2:]
+		info.extern = true
+	} else {
+		// Pick the default linkName.
+		info.linkName = g.RelString(nil)
+		// Check for //go: pragmas, which may change the link name (among
+		// others).
+		doc := c.astComments[info.linkName]
+		if doc != nil {
+			info.parsePragmas(doc)
+		}
+	}
+	return info
+}
+
+// Parse //go: pragma comments from the source. In particular, it parses the
+// //go:extern pragma on globals.
+func (info *globalInfo) parsePragmas(doc *ast.CommentGroup) {
+	for _, comment := range doc.List {
+		if !strings.HasPrefix(comment.Text, "//go:") {
+			continue
+		}
+		parts := strings.Fields(comment.Text)
+		switch parts[0] {
+		case "//go:extern":
+			info.extern = true
+			if len(parts) == 2 {
+				info.linkName = parts[1]
+			}
+		case "//go:align":
+			align, err := strconv.Atoi(parts[1])
+			if err == nil {
+				info.align = align
+			}
+		}
+	}
+}

compiler/syscall.go

@@ -4,7 +4,6 @@ package compiler
 // compiler builtins.
 
 import (
-	"go/constant"
 	"strconv"
 
 	"golang.org/x/tools/go/ssa"
@@ -14,11 +13,11 @@ import (
 // emitSyscall emits an inline system call instruction, depending on the target
 // OS/arch.
 func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value, error) {
-	num, _ := constant.Uint64Val(call.Args[0].(*ssa.Const).Value)
+	num := c.getValue(frame, call.Args[0])
 	var syscallResult llvm.Value
 	switch {
-	case c.GOARCH == "amd64":
-		if c.GOOS == "darwin" {
+	case c.GOARCH() == "amd64":
+		if c.GOOS() == "darwin" {
 			// Darwin adds this magic number to system call numbers:
 			//
 			// > Syscall classes for 64-bit system call entry.
@@ -29,12 +28,12 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 			// > All system classes enter the kernel via the syscall instruction.
 			//
 			// Source: https://opensource.apple.com/source/xnu/xnu-792.13.8/osfmk/mach/i386/syscall_sw.h
-			num += 0x2000000
+			num = c.builder.CreateOr(num, llvm.ConstInt(c.uintptrType, 0x2000000, false), "")
 		}
 		// Sources:
 		//   https://stackoverflow.com/a/2538212
 		//   https://en.wikibooks.org/wiki/X86_Assembly/Interfacing_with_Linux#syscall
-		args := []llvm.Value{llvm.ConstInt(c.uintptrType, num, false)}
+		args := []llvm.Value{num}
 		argTypes := []llvm.Type{c.uintptrType}
 		// Constraints will look something like:
 		//   "={rax},0,{rdi},{rsi},{rdx},{r10},{r8},{r9},~{rcx},~{r11}"
@@ -59,7 +58,33 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 		fnType := llvm.FunctionType(c.uintptrType, argTypes, false)
 		target := llvm.InlineAsm(fnType, "syscall", constraints, true, false, llvm.InlineAsmDialectIntel)
 		syscallResult = c.builder.CreateCall(target, args, "")
-	case c.GOARCH == "arm" && c.GOOS == "linux":
+	case c.GOARCH() == "386" && c.GOOS() == "linux":
+		// Sources:
+		//   syscall(2) man page
+		//   https://stackoverflow.com/a/2538212
+		//   https://en.wikibooks.org/wiki/X86_Assembly/Interfacing_with_Linux#int_0x80
+		args := []llvm.Value{num}
+		argTypes := []llvm.Type{c.uintptrType}
+		// Constraints will look something like:
+		//   "={eax},0,{ebx},{ecx},{edx},{esi},{edi},{ebp}"
+		constraints := "={eax},0"
+		for i, arg := range call.Args[1:] {
+			constraints += "," + [...]string{
+				"{ebx}",
+				"{ecx}",
+				"{edx}",
+				"{esi}",
+				"{edi}",
+				"{ebp}",
+			}[i]
+			llvmValue := c.getValue(frame, arg)
+			args = append(args, llvmValue)
+			argTypes = append(argTypes, llvmValue.Type())
+		}
+		fnType := llvm.FunctionType(c.uintptrType, argTypes, false)
+		target := llvm.InlineAsm(fnType, "int 0x80", constraints, true, false, llvm.InlineAsmDialectIntel)
+		syscallResult = c.builder.CreateCall(target, args, "")
+	case c.GOARCH() == "arm" && c.GOOS() == "linux":
 		// Implement the EABI system call convention for Linux.
 		// Source: syscall(2) man page.
 		args := []llvm.Value{}
@@ -81,7 +106,7 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 			args = append(args, llvmValue)
 			argTypes = append(argTypes, llvmValue.Type())
 		}
-		args = append(args, llvm.ConstInt(c.uintptrType, num, false))
+		args = append(args, num)
 		argTypes = append(argTypes, c.uintptrType)
 		constraints += ",{r7}" // syscall number
 		for i := len(call.Args) - 1; i < 4; i++ {
@@ -91,7 +116,7 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 		fnType := llvm.FunctionType(c.uintptrType, argTypes, false)
 		target := llvm.InlineAsm(fnType, "svc #0", constraints, true, false, 0)
 		syscallResult = c.builder.CreateCall(target, args, "")
-	case c.GOARCH == "arm64" && c.GOOS == "linux":
+	case c.GOARCH() == "arm64" && c.GOOS() == "linux":
 		// Source: syscall(2) man page.
 		args := []llvm.Value{}
 		argTypes := []llvm.Type{}
@@ -111,7 +136,7 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 			args = append(args, llvmValue)
 			argTypes = append(argTypes, llvmValue.Type())
 		}
-		args = append(args, llvm.ConstInt(c.uintptrType, num, false))
+		args = append(args, num)
 		argTypes = append(argTypes, c.uintptrType)
 		constraints += ",{x8}" // syscall number
 		for i := len(call.Args) - 1; i < 8; i++ {
@@ -124,10 +149,10 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 		target := llvm.InlineAsm(fnType, "svc #0", constraints, true, false, 0)
 		syscallResult = c.builder.CreateCall(target, args, "")
 	default:
-		return llvm.Value{}, c.makeError(call.Pos(), "unknown GOOS/GOARCH for syscall: "+c.GOOS+"/"+c.GOARCH)
+		return llvm.Value{}, c.makeError(call.Pos(), "unknown GOOS/GOARCH for syscall: "+c.GOOS()+"/"+c.GOARCH())
 	}
-	switch c.GOOS {
-	case "linux":
+	switch c.GOOS() {
+	case "linux", "freebsd":
 		// Return values: r0, r1 uintptr, err Errno
 		// Pseudocode:
 		//     var err uintptr
@@ -140,7 +165,7 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 		inrange2 := c.builder.CreateICmp(llvm.IntSGT, syscallResult, llvm.ConstInt(c.uintptrType, 0xfffffffffffff000, true), "") // -4096
 		hasError := c.builder.CreateAnd(inrange1, inrange2, "")
 		errResult := c.builder.CreateSelect(hasError, c.builder.CreateSub(zero, syscallResult, ""), zero, "syscallError")
-		retval := llvm.Undef(llvm.StructType([]llvm.Type{c.uintptrType, c.uintptrType, c.uintptrType}, false))
+		retval := llvm.Undef(c.ctx.StructType([]llvm.Type{c.uintptrType, c.uintptrType, c.uintptrType}, false))
 		retval = c.builder.CreateInsertValue(retval, syscallResult, 0, "")
 		retval = c.builder.CreateInsertValue(retval, zero, 1, "")
 		retval = c.builder.CreateInsertValue(retval, errResult, 2, "")
@@ -156,12 +181,12 @@ func (c *Compiler) emitSyscall(frame *Frame, call *ssa.CallCommon) (llvm.Value,
 		zero := llvm.ConstInt(c.uintptrType, 0, false)
 		hasError := c.builder.CreateICmp(llvm.IntNE, syscallResult, llvm.ConstInt(c.uintptrType, 0, false), "")
 		errResult := c.builder.CreateSelect(hasError, syscallResult, zero, "syscallError")
-		retval := llvm.Undef(llvm.StructType([]llvm.Type{c.uintptrType, c.uintptrType, c.uintptrType}, false))
+		retval := llvm.Undef(c.ctx.StructType([]llvm.Type{c.uintptrType, c.uintptrType, c.uintptrType}, false))
 		retval = c.builder.CreateInsertValue(retval, syscallResult, 0, "")
 		retval = c.builder.CreateInsertValue(retval, zero, 1, "")
 		retval = c.builder.CreateInsertValue(retval, errResult, 2, "")
 		return retval, nil
 	default:
-		return llvm.Value{}, c.makeError(call.Pos(), "unknown GOOS/GOARCH for syscall: "+c.GOOS+"/"+c.GOARCH)
+		return llvm.Value{}, c.makeError(call.Pos(), "unknown GOOS/GOARCH for syscall: "+c.GOOS()+"/"+c.GOARCH())
 	}
 }

compiler/wordpack.go

@@ -1,108 +0,0 @@
-package compiler
-
-// This file contains utility functions to pack and unpack sets of values. It
-// can take in a list of values and tries to store it efficiently in the pointer
-// itself if possible and legal.
-
-import (
-	"tinygo.org/x/go-llvm"
-)
-
-// emitPointerPack packs the list of values into a single pointer value using
-// bitcasts, or else allocates a value on the heap if it cannot be packed in the
-// pointer value directly. It returns the pointer with the packed data.
-func (c *Compiler) emitPointerPack(values []llvm.Value) llvm.Value {
-	valueTypes := make([]llvm.Type, len(values))
-	for i, value := range values {
-		valueTypes[i] = value.Type()
-	}
-	packedType := c.ctx.StructType(valueTypes, false)
-
-	// Allocate memory for the packed data.
-	var packedAlloc, packedHeapAlloc llvm.Value
-	size := c.targetData.TypeAllocSize(packedType)
-	if size == 0 {
-		return llvm.ConstPointerNull(c.i8ptrType)
-	} else if len(values) == 1 && values[0].Type().TypeKind() == llvm.PointerTypeKind {
-		return c.builder.CreateBitCast(values[0], c.i8ptrType, "pack.ptr")
-	} else if size <= c.targetData.TypeAllocSize(c.i8ptrType) && !c.gcIsPrecise() {
-		// Packed data fits in a pointer, so store it directly inside the
-		// pointer.
-		if len(values) == 1 && values[0].Type().TypeKind() == llvm.IntegerTypeKind {
-			// Try to keep this cast in SSA form.
-			return c.builder.CreateIntToPtr(values[0], c.i8ptrType, "pack.int")
-		}
-		// Because packedType is a struct and we have to cast it to a *i8, store
-		// it in an alloca first for bitcasting (store+bitcast+load).
-		packedAlloc = c.builder.CreateAlloca(packedType, "")
-	} else {
-		// Packed data is bigger than a pointer, so allocate it on the heap.
-		sizeValue := llvm.ConstInt(c.uintptrType, size, false)
-		packedHeapAlloc = c.createRuntimeCall("alloc", []llvm.Value{sizeValue}, "")
-		packedAlloc = c.builder.CreateBitCast(packedHeapAlloc, llvm.PointerType(packedType, 0), "")
-	}
-	// Store all values in the alloca or heap pointer.
-	for i, value := range values {
-		indices := []llvm.Value{
-			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-			llvm.ConstInt(c.ctx.Int32Type(), uint64(i), false),
-		}
-		gep := c.builder.CreateInBoundsGEP(packedAlloc, indices, "")
-		c.builder.CreateStore(value, gep)
-	}
-
-	if packedHeapAlloc.IsNil() {
-		// Load value (as *i8) from the alloca.
-		packedAlloc = c.builder.CreateBitCast(packedAlloc, llvm.PointerType(c.i8ptrType, 0), "")
-		return c.builder.CreateLoad(packedAlloc, "")
-	} else {
-		// Get the original heap allocation pointer, which already is an *i8.
-		return c.builder.CreateBitCast(packedAlloc, c.i8ptrType, "")
-	}
-}
-
-// emitPointerUnpack extracts a list of values packed using emitPointerPack.
-func (c *Compiler) emitPointerUnpack(ptr llvm.Value, valueTypes []llvm.Type) []llvm.Value {
-	packedType := c.ctx.StructType(valueTypes, false)
-
-	// Get a correctly-typed pointer to the packed data.
-	var packedAlloc llvm.Value
-	size := c.targetData.TypeAllocSize(packedType)
-	if size == 0 {
-		// No data to unpack.
-	} else if len(valueTypes) == 1 && valueTypes[0].TypeKind() == llvm.PointerTypeKind {
-		// A single pointer is always stored directly.
-		return []llvm.Value{c.builder.CreateBitCast(ptr, valueTypes[0], "unpack.ptr")}
-	} else if size <= c.targetData.TypeAllocSize(c.i8ptrType) && !c.gcIsPrecise() {
-		// Packed data stored directly in pointer.
-		if len(valueTypes) == 1 && valueTypes[0].TypeKind() == llvm.IntegerTypeKind {
-			// Keep this cast in SSA form.
-			return []llvm.Value{c.builder.CreatePtrToInt(ptr, valueTypes[0], "unpack.int")}
-		}
-		// Fallback: load it using an alloca.
-		packedRawAlloc := c.builder.CreateAlloca(llvm.PointerType(c.i8ptrType, 0), "unpack.raw.alloc")
-		packedRawValue := c.builder.CreateBitCast(ptr, llvm.PointerType(c.i8ptrType, 0), "unpack.raw.value")
-		c.builder.CreateStore(packedRawValue, packedRawAlloc)
-		packedAlloc = c.builder.CreateBitCast(packedRawAlloc, llvm.PointerType(packedType, 0), "unpack.alloc")
-	} else {
-		// Packed data stored on the heap. Bitcast the passed-in pointer to the
-		// correct pointer type.
-		packedAlloc = c.builder.CreateBitCast(ptr, llvm.PointerType(packedType, 0), "unpack.raw.ptr")
-	}
-	// Load each value from the packed data.
-	values := make([]llvm.Value, len(valueTypes))
-	for i, valueType := range valueTypes {
-		if c.targetData.TypeAllocSize(valueType) == 0 {
-			// This value has length zero, so there's nothing to load.
-			values[i] = c.getZeroValue(valueType)
-			continue
-		}
-		indices := []llvm.Value{
-			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-			llvm.ConstInt(c.ctx.Int32Type(), uint64(i), false),
-		}
-		gep := c.builder.CreateInBoundsGEP(packedAlloc, indices, "")
-		values[i] = c.builder.CreateLoad(gep, "")
-	}
-	return values
-}

go.mod

@@ -0,0 +1,14 @@
+module github.com/tinygo-org/tinygo
+
+go 1.11
+
+require (
+	github.com/blakesmith/ar v0.0.0-20150311145944-8bd4349a67f2
+	github.com/creack/goselect v0.1.0 // indirect
+	github.com/google/shlex v0.0.0-20181106134648-c34317bd91bf
+	github.com/marcinbor85/gohex v0.0.0-20180128172054-7a43cd876e46
+	go.bug.st/serial.v1 v0.0.0-20180827123349-5f7892a7bb45
+	golang.org/x/sys v0.0.0-20191010194322-b09406accb47 // indirect
+	golang.org/x/tools v0.0.0-20190227180812-8dcc6e70cdef
+	tinygo.org/x/go-llvm v0.0.0-20200104190746-1ff21df33566
+)

go.sum

@@ -0,0 +1,36 @@
+github.com/blakesmith/ar v0.0.0-20150311145944-8bd4349a67f2 h1:oMCHnXa6CCCafdPDbMh/lWRhRByN0VFLvv+g+ayx1SI=
+github.com/blakesmith/ar v0.0.0-20150311145944-8bd4349a67f2/go.mod h1:PkYb9DJNAwrSvRx5DYA+gUcOIgTGVMNkfSCbZM8cWpI=
+github.com/creack/goselect v0.1.0 h1:4QiXIhcpSQF50XGaBsFzesjwX/1qOY5bOveQPmN9CXY=
+github.com/creack/goselect v0.1.0/go.mod h1:gHrIcH/9UZDn2qgeTUeW5K9eZsVYCH6/60J/FHysWyE=
+github.com/golang/protobuf v1.2.0/go.mod h1:6lQm79b+lXiMfvg/cZm0SGofjICqVBUtrP5yJMmIC1U=
+github.com/google/shlex v0.0.0-20181106134648-c34317bd91bf h1:7+FW5aGwISbqUtkfmIpZJGRgNFg2ioYPvFaUxdqpDsg=
+github.com/google/shlex v0.0.0-20181106134648-c34317bd91bf/go.mod h1:RpwtwJQFrIEPstU94h88MWPXP2ektJZ8cZ0YntAmXiE=
+github.com/marcinbor85/gohex v0.0.0-20180128172054-7a43cd876e46 h1:wXG2bA8fO7Vv7lLk2PihFMTqmbT173Tje39oKzQ50Mo=
+github.com/marcinbor85/gohex v0.0.0-20180128172054-7a43cd876e46/go.mod h1:Pb6XcsXyropB9LNHhnqaknG/vEwYztLkQzVCHv8sQ3M=
+go.bug.st/serial.v1 v0.0.0-20180827123349-5f7892a7bb45 h1:mACY1anK6HNCZtm/DK2Rf2ZPHggVqeB0+7rY9Gl6wyI=
+go.bug.st/serial.v1 v0.0.0-20180827123349-5f7892a7bb45/go.mod h1:dRSl/CVCTf56CkXgJMDOdSwNfo2g1orOGE/gBGdvjZw=
+golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
+golang.org/x/net v0.0.0-20190213061140-3a22650c66bd/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
+golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
+golang.org/x/sys v0.0.0-20191010194322-b09406accb47 h1:/XfQ9z7ib8eEJX2hdgFTZJ/ntt0swNk5oYBziWeTCvY=
+golang.org/x/sys v0.0.0-20191010194322-b09406accb47/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
+golang.org/x/tools v0.0.0-20190227180812-8dcc6e70cdef h1:ymc9FeDom3RIEA3coKokSllBB1hRcMT0tZ1W3Jf9Ids=
+golang.org/x/tools v0.0.0-20190227180812-8dcc6e70cdef/go.mod h1:9Yl7xja0Znq3iFh3HoIrodX9oNMXvdceNzlUR8zjMvY=
+google.golang.org/appengine v1.4.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4=
+tinygo.org/x/go-llvm v0.0.0-20190224120431-7707ae5d1261 h1:rJS2Hga39YAnm7DE4qrPm6Dr/67EOojL0XPzvbEeBiw=
+tinygo.org/x/go-llvm v0.0.0-20190224120431-7707ae5d1261/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20190818154551-95bc4ffe1add h1:dFjMH1sLhYADg8UQm7DB56B7e+TfvAmWmEZLhyv3r/w=
+tinygo.org/x/go-llvm v0.0.0-20190818154551-95bc4ffe1add/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20191103182207-90b6e4bdc0b9 h1:d6rAX39a3C0pKrY5HcojEGyN8w9ocU0v7X28lC/TRKU=
+tinygo.org/x/go-llvm v0.0.0-20191103182207-90b6e4bdc0b9/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20191103200204-37e93e3f04e2 h1:Q5Hv3e5cLMGkiYwYgZL1Zrv6nb/EY+DJpRWrdO6ws6o=
+tinygo.org/x/go-llvm v0.0.0-20191103200204-37e93e3f04e2/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20191113125529-bad6d01809e8 h1:9Bfvso+tTVQg16UzOA614NaYA4x8vsRBNtd3eBrXwp0=
+tinygo.org/x/go-llvm v0.0.0-20191113125529-bad6d01809e8/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20191124211856-b2db3df3f257 h1:o8VDylrMN7gWemBMu8rEyuogKPhcLTdx5KrUAp9macc=
+tinygo.org/x/go-llvm v0.0.0-20191124211856-b2db3df3f257/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20191215173731-ad71f3d24aae h1:s8J5EyxCkHxXB08UI3gk9W9IS/ekizRvSX+PfZxnAB0=
+tinygo.org/x/go-llvm v0.0.0-20191215173731-ad71f3d24aae/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=
+tinygo.org/x/go-llvm v0.0.0-20200104190746-1ff21df33566 h1:a4y30bTf7U0zDA75v2PTL+XQ2OzJetj19gK8XwQpUNY=
+tinygo.org/x/go-llvm v0.0.0-20200104190746-1ff21df33566/go.mod h1:fv1F0BSNpxMfCL0zF3M4OPFbgYHnhtB6ST0HvUtu/LE=

goenv/goenv.go

@@ -0,0 +1,189 @@
+// Package goenv returns environment variables that are used in various parts of
+// the compiler. You can query it manually with the `tinygo env` subcommand.
+package goenv
+
+import (
+	"fmt"
+	"os"
+	"os/exec"
+	"os/user"
+	"path/filepath"
+	"runtime"
+)
+
+// Keys is a slice of all available environment variable keys.
+var Keys = []string{
+	"GOOS",
+	"GOARCH",
+	"GOROOT",
+	"GOPATH",
+	"GOCACHE",
+	"TINYGOROOT",
+}
+
+// TINYGOROOT is the path to the final location for checking tinygo files. If
+// unset (by a -X ldflag), then sourceDir() will fallback to the original build
+// directory.
+var TINYGOROOT string
+
+// Get returns a single environment variable, possibly calculating it on-demand.
+// The empty string is returned for unknown environment variables.
+func Get(name string) string {
+	switch name {
+	case "GOOS":
+		if dir := os.Getenv("GOOS"); dir != "" {
+			return dir
+		}
+		return runtime.GOOS
+	case "GOARCH":
+		if dir := os.Getenv("GOARCH"); dir != "" {
+			return dir
+		}
+		return runtime.GOARCH
+	case "GOROOT":
+		return getGoroot()
+	case "GOPATH":
+		if dir := os.Getenv("GOPATH"); dir != "" {
+			return dir
+		}
+
+		// fallback
+		home := getHomeDir()
+		return filepath.Join(home, "go")
+	case "GOCACHE":
+		// Get the cache directory, usually ~/.cache/tinygo
+		dir, err := os.UserCacheDir()
+		if err != nil {
+			panic("could not find cache dir: " + err.Error())
+		}
+		return filepath.Join(dir, "tinygo")
+	case "TINYGOROOT":
+		return sourceDir()
+	default:
+		return ""
+	}
+}
+
+// Return the TINYGOROOT, or exit with an error.
+func sourceDir() string {
+	// Use $TINYGOROOT as root, if available.
+	root := os.Getenv("TINYGOROOT")
+	if root != "" {
+		if !isSourceDir(root) {
+			fmt.Fprintln(os.Stderr, "error: $TINYGOROOT was not set to the correct root")
+			os.Exit(1)
+		}
+		return root
+	}
+
+	if TINYGOROOT != "" {
+		if !isSourceDir(TINYGOROOT) {
+			fmt.Fprintln(os.Stderr, "error: TINYGOROOT was not set to the correct root")
+			os.Exit(1)
+		}
+		return TINYGOROOT
+	}
+
+	// Find root from executable path.
+	path, err := os.Executable()
+	if err != nil {
+		// Very unlikely. Bail out if it happens.
+		panic("could not get executable path: " + err.Error())
+	}
+	root = filepath.Dir(filepath.Dir(path))
+	if isSourceDir(root) {
+		return root
+	}
+
+	// Fallback: use the original directory from where it was built
+	// https://stackoverflow.com/a/32163888/559350
+	_, path, _, _ = runtime.Caller(0)
+	root = filepath.Dir(filepath.Dir(path))
+	if isSourceDir(root) {
+		return root
+	}
+
+	fmt.Fprintln(os.Stderr, "error: could not autodetect root directory, set the TINYGOROOT environment variable to override")
+	os.Exit(1)
+	panic("unreachable")
+}
+
+// isSourceDir returns true if the directory looks like a TinyGo source directory.
+func isSourceDir(root string) bool {
+	_, err := os.Stat(filepath.Join(root, "src/runtime/internal/sys/zversion.go"))
+	if err != nil {
+		return false
+	}
+	_, err = os.Stat(filepath.Join(root, "src/device/arm/arm.go"))
+	return err == nil
+}
+
+func getHomeDir() string {
+	u, err := user.Current()
+	if err != nil {
+		panic("cannot get current user: " + err.Error())
+	}
+	if u.HomeDir == "" {
+		// This is very unlikely, so panic here.
+		// Not the nicest solution, however.
+		panic("could not find home directory")
+	}
+	return u.HomeDir
+}
+
+// getGoroot returns an appropriate GOROOT from various sources. If it can't be
+// found, it returns an empty string.
+func getGoroot() string {
+	goroot := os.Getenv("GOROOT")
+	if goroot != "" {
+		// An explicitly set GOROOT always has preference.
+		return goroot
+	}
+
+	// Check for the location of the 'go' binary and base GOROOT on that.
+	binpath, err := exec.LookPath("go")
+	if err == nil {
+		binpath, err = filepath.EvalSymlinks(binpath)
+		if err == nil {
+			goroot := filepath.Dir(filepath.Dir(binpath))
+			if isGoroot(goroot) {
+				return goroot
+			}
+		}
+	}
+
+	// Check what GOROOT was at compile time.
+	if isGoroot(runtime.GOROOT()) {
+		return runtime.GOROOT()
+	}
+
+	// Check for some standard locations, as a last resort.
+	var candidates []string
+	switch runtime.GOOS {
+	case "linux":
+		candidates = []string{
+			"/usr/local/go", // manually installed
+			"/usr/lib/go",   // from the distribution
+		}
+	case "darwin":
+		candidates = []string{
+			"/usr/local/go",             // manually installed
+			"/usr/local/opt/go/libexec", // from Homebrew
+		}
+	}
+
+	for _, candidate := range candidates {
+		if isGoroot(candidate) {
+			return candidate
+		}
+	}
+
+	// Can't find GOROOT...
+	return ""
+}
+
+// isGoroot checks whether the given path looks like a GOROOT.
+func isGoroot(goroot string) bool {
+	_, err := os.Stat(filepath.Join(goroot, "src", "runtime", "internal", "sys", "zversion.go"))
+	return err == nil
+}

interp/errors.go

@@ -3,15 +3,92 @@ package interp
 // This file provides useful types for errors encountered during IR evaluation.
 
 import (
+	"errors"
+	"go/scanner"
+	"go/token"
+	"path/filepath"
+
 	"tinygo.org/x/go-llvm"
 )
 
+// errUnreachable is returned when an unreachable instruction is executed. This
+// error should not be visible outside of the interp package.
+var errUnreachable = errors.New("interp: unreachable executed")
+
+// Unsupported is the specific error that is returned when an unsupported
+// instruction is hit while trying to interpret all initializers.
 type Unsupported struct {
-	Inst llvm.Value
+	ImportPath string
+	Inst       llvm.Value
+	Pos        token.Position
 }
 
 func (e Unsupported) Error() string {
 	// TODO: how to return the actual instruction string?
 	// It looks like LLVM provides no function for that...
-	return "interp: unsupported instruction"
+	return scanner.Error{
+		Pos: e.Pos,
+		Msg: "interp: unsupported instruction",
+	}.Error()
+}
+
+// unsupportedInstructionError returns a new "unsupported instruction" error for
+// the given instruction. It includes source location information, when
+// available.
+func (e *evalPackage) unsupportedInstructionError(inst llvm.Value) *Unsupported {
+	return &Unsupported{
+		ImportPath: e.packagePath,
+		Inst:       inst,
+		Pos:        getPosition(inst),
+	}
+}
+
+// Error encapsulates compile-time interpretation errors with an associated
+// import path. The errors may not have a precise location attached.
+type Error struct {
+	ImportPath string
+	Errs       []scanner.Error
+}
+
+// Error returns the string of the first error in the list of errors.
+func (e Error) Error() string {
+	return e.Errs[0].Error()
+}
+
+// errorAt returns an error value for the currently interpreted package at the
+// location of the instruction. The location information may not be complete as
+// it depends on debug information in the IR.
+func (e *evalPackage) errorAt(inst llvm.Value, msg string) Error {
+	return Error{
+		ImportPath: e.packagePath,
+		Errs:       []scanner.Error{errorAt(inst, msg)},
+	}
+}
+
+// errorAt returns an error value at the location of the instruction.
+// The location information may not be complete as it depends on debug
+// information in the IR.
+func errorAt(inst llvm.Value, msg string) scanner.Error {
+	return scanner.Error{
+		Pos: getPosition(inst),
+		Msg: msg,
+	}
+}
+
+// getPosition returns the position information for the given instruction, as
+// far as it is available.
+func getPosition(inst llvm.Value) token.Position {
+	if inst.IsAInstruction().IsNil() {
+		return token.Position{}
+	}
+	loc := inst.InstructionDebugLoc()
+	if loc.IsNil() {
+		return token.Position{}
+	}
+	file := loc.LocationScope().ScopeFile()
+	return token.Position{
+		Filename: filepath.Join(file.FileDirectory(), file.FileFilename()),
+		Line:     int(loc.LocationLine()),
+		Column:   int(loc.LocationColumn()),
+	}
 }

interp/frame.go

@@ -4,21 +4,17 @@ package interp
 // functions.
 
 import (
-	"errors"
 	"strings"
 
 	"tinygo.org/x/go-llvm"
 )
 
 type frame struct {
-	*Eval
-	fn      llvm.Value
-	pkgName string
-	locals  map[llvm.Value]Value
+	*evalPackage
+	fn     llvm.Value
+	locals map[llvm.Value]Value
 }
 
-var ErrUnreachable = errors.New("interp: unreachable executed")
-
 // evalBasicBlock evaluates a single basic block, returning the return value (if
 // ending with a ret instruction), a list of outgoing basic blocks (if not
 // ending with a ret instruction), or an error on failure.
@@ -79,14 +75,14 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				fr.locals[inst] = &LocalValue{fr.Eval, fr.builder.CreateXor(lhs, rhs, "")}
 
 			default:
-				return nil, nil, &Unsupported{inst}
+				return nil, nil, fr.unsupportedInstructionError(inst)
 			}
 
 		// Memory operators
 		case !inst.IsAAllocaInst().IsNil():
 			allocType := inst.Type().ElementType()
-			alloca := llvm.AddGlobal(fr.Mod, allocType, fr.pkgName+"$alloca")
-			alloca.SetInitializer(getZeroValue(allocType))
+			alloca := llvm.AddGlobal(fr.Mod, allocType, fr.packagePath+"$alloca")
+			alloca.SetInitializer(llvm.ConstNull(allocType))
 			alloca.SetLinkage(llvm.InternalLinkage)
 			fr.locals[inst] = &LocalValue{
 				Underlying: alloca,
@@ -101,7 +97,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				value = operand.Load()
 			}
 			if value.Type() != inst.Type() {
-				panic("interp: load: type does not match")
+				return nil, nil, fr.errorAt(inst, "interp: load: type does not match")
 			}
 			fr.locals[inst] = fr.getValue(value)
 		case !inst.IsAStoreInst().IsNil():
@@ -125,15 +121,13 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 					// Not a constant operation.
 					// This should be detected by the scanner, but isn't at the
 					// moment.
-					panic("todo: non-const gep")
+					return nil, nil, fr.errorAt(inst, "todo: non-const gep")
 				}
 				indices[i] = uint32(operand.Value().ZExtValue())
 			}
 			result := value.GetElementPtr(indices)
 			if result.Type() != inst.Type() {
-				println(" expected:", inst.Type().String())
-				println(" actual:  ", result.Type().String())
-				panic("interp: gep: type does not match")
+				return nil, nil, fr.errorAt(inst, "interp: gep: type does not match")
 			}
 			fr.locals[inst] = result
 
@@ -176,6 +170,20 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 					continue // special case: bitcast of alloc
 				}
 			}
+			if _, ok := fr.getLocal(operand).(*MapValue); ok {
+				// Special case for runtime.trackPointer calls.
+				// Note: this might not be entirely sound in some rare cases
+				// where the map is stored in a dirty global.
+				uses := getUses(inst)
+				if len(uses) == 1 {
+					use := uses[0]
+					if !use.IsACallInst().IsNil() && !use.CalledValue().IsAFunction().IsNil() && use.CalledValue().Name() == "runtime.trackPointer" {
+						continue
+					}
+				}
+				// It is not possible in Go to bitcast a map value to a pointer.
+				return nil, nil, fr.errorAt(inst, "unimplemented: bitcast of map")
+			}
 			value := fr.getLocal(operand).(*LocalValue)
 			fr.locals[inst] = &LocalValue{fr.Eval, fr.builder.CreateBitCast(value.Value(), inst.Type(), "")}
 
@@ -184,6 +192,33 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 			lhs := fr.getLocal(inst.Operand(0)).(*LocalValue).Underlying
 			rhs := fr.getLocal(inst.Operand(1)).(*LocalValue).Underlying
 			predicate := inst.IntPredicate()
+			if predicate == llvm.IntEQ {
+				var lhsZero, rhsZero bool
+				var ok1, ok2 bool
+				if lhs.Type().TypeKind() == llvm.PointerTypeKind {
+					// Unfortunately, the const propagation in the IR builder
+					// doesn't handle pointer compares of inttoptr values. So we
+					// implement it manually here.
+					lhsZero, ok1 = isPointerNil(lhs)
+					rhsZero, ok2 = isPointerNil(rhs)
+				}
+				if lhs.Type().TypeKind() == llvm.IntegerTypeKind {
+					lhsZero, ok1 = isZero(lhs)
+					rhsZero, ok2 = isZero(rhs)
+				}
+				if ok1 && ok2 {
+					if lhsZero && rhsZero {
+						// Both are zero, so this icmp is always evaluated to true.
+						fr.locals[inst] = &LocalValue{fr.Eval, llvm.ConstInt(fr.Mod.Context().Int1Type(), 1, false)}
+						continue
+					}
+					if lhsZero != rhsZero {
+						// Only one of them is zero, so this comparison must return false.
+						fr.locals[inst] = &LocalValue{fr.Eval, llvm.ConstInt(fr.Mod.Context().Int1Type(), 0, false)}
+						continue
+					}
+				}
+			}
 			fr.locals[inst] = &LocalValue{fr.Eval, fr.builder.CreateICmp(predicate, lhs, rhs, "")}
 		case !inst.IsAFCmpInst().IsNil():
 			lhs := fr.getLocal(inst.Operand(0)).(*LocalValue).Underlying
@@ -214,13 +249,13 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				if size != typeSize {
 					// allocate an array
 					if size%typeSize != 0 {
-						return nil, nil, &Unsupported{inst}
+						return nil, nil, fr.unsupportedInstructionError(inst)
 					}
 					elementCount = int(size / typeSize)
 					allocType = llvm.ArrayType(allocType, elementCount)
 				}
-				alloc := llvm.AddGlobal(fr.Mod, allocType, fr.pkgName+"$alloc")
-				alloc.SetInitializer(getZeroValue(allocType))
+				alloc := llvm.AddGlobal(fr.Mod, allocType, fr.packagePath+"$alloc")
+				alloc.SetInitializer(llvm.ConstNull(allocType))
 				alloc.SetLinkage(llvm.InternalLinkage)
 				result := &LocalValue{
 					Underlying: alloc,
@@ -237,24 +272,55 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				valueSize := inst.Operand(1).ZExtValue()
 				fr.locals[inst] = &MapValue{
 					Eval:      fr.Eval,
-					PkgName:   fr.pkgName,
+					PkgName:   fr.packagePath,
 					KeySize:   int(keySize),
 					ValueSize: int(valueSize),
 				}
 			case callee.Name() == "runtime.hashmapStringSet":
 				// set a string key in the map
-				m := fr.getLocal(inst.Operand(0)).(*MapValue)
-				// "key" is a Go string value, which in the TinyGo calling convention is split up
-				// into separate pointer and length parameters.
 				keyBuf := fr.getLocal(inst.Operand(1)).(*LocalValue)
 				keyLen := fr.getLocal(inst.Operand(2)).(*LocalValue)
 				valPtr := fr.getLocal(inst.Operand(3)).(*LocalValue)
+				m, ok := fr.getLocal(inst.Operand(0)).(*MapValue)
+				if !ok || !keyBuf.IsConstant() || !keyLen.IsConstant() || !valPtr.IsConstant() {
+					// The mapassign operation could not be done at compile
+					// time. Do it at runtime instead.
+					m := fr.getLocal(inst.Operand(0)).Value()
+					fr.markDirty(m)
+					llvmParams := []llvm.Value{
+						m,                                    // *runtime.hashmap
+						fr.getLocal(inst.Operand(1)).Value(), // key.ptr
+						fr.getLocal(inst.Operand(2)).Value(), // key.len
+						fr.getLocal(inst.Operand(3)).Value(), // value (unsafe.Pointer)
+						fr.getLocal(inst.Operand(4)).Value(), // context
+						fr.getLocal(inst.Operand(5)).Value(), // parentHandle
+					}
+					fr.builder.CreateCall(callee, llvmParams, "")
+					continue
+				}
+				// "key" is a Go string value, which in the TinyGo calling convention is split up
+				// into separate pointer and length parameters.
 				m.PutString(keyBuf, keyLen, valPtr)
 			case callee.Name() == "runtime.hashmapBinarySet":
 				// set a binary (int etc.) key in the map
-				m := fr.getLocal(inst.Operand(0)).(*MapValue)
 				keyBuf := fr.getLocal(inst.Operand(1)).(*LocalValue)
 				valPtr := fr.getLocal(inst.Operand(2)).(*LocalValue)
+				m, ok := fr.getLocal(inst.Operand(0)).(*MapValue)
+				if !ok || !keyBuf.IsConstant() || !valPtr.IsConstant() {
+					// The mapassign operation could not be done at compile
+					// time. Do it at runtime instead.
+					m := fr.getLocal(inst.Operand(0)).Value()
+					fr.markDirty(m)
+					llvmParams := []llvm.Value{
+						m,                                    // *runtime.hashmap
+						fr.getLocal(inst.Operand(1)).Value(), // key
+						fr.getLocal(inst.Operand(2)).Value(), // value
+						fr.getLocal(inst.Operand(3)).Value(), // context
+						fr.getLocal(inst.Operand(4)).Value(), // parentHandle
+					}
+					fr.builder.CreateCall(callee, llvmParams, "")
+					continue
+				}
 				m.PutBinary(keyBuf, valPtr)
 			case callee.Name() == "runtime.stringConcat":
 				// adding two strings together
@@ -271,7 +337,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				}
 				globalType := llvm.ArrayType(fr.Mod.Context().Int8Type(), len(result))
 				globalValue := llvm.ConstArray(fr.Mod.Context().Int8Type(), vals)
-				global := llvm.AddGlobal(fr.Mod, globalType, fr.pkgName+"$stringconcat")
+				global := llvm.AddGlobal(fr.Mod, globalType, fr.packagePath+"$stringconcat")
 				global.SetInitializer(globalValue)
 				global.SetLinkage(llvm.InternalLinkage)
 				global.SetGlobalConstant(true)
@@ -279,10 +345,53 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				stringType := fr.Mod.GetTypeByName("runtime._string")
 				retPtr := llvm.ConstGEP(global, getLLVMIndices(fr.Mod.Context().Int32Type(), []uint32{0, 0}))
 				retLen := llvm.ConstInt(stringType.StructElementTypes()[1], uint64(len(result)), false)
-				ret := getZeroValue(stringType)
+				ret := llvm.ConstNull(stringType)
 				ret = llvm.ConstInsertValue(ret, retPtr, []uint32{0})
 				ret = llvm.ConstInsertValue(ret, retLen, []uint32{1})
 				fr.locals[inst] = &LocalValue{fr.Eval, ret}
+			case callee.Name() == "runtime.sliceCopy":
+				elementSize := fr.getLocal(inst.Operand(4)).(*LocalValue).Value().ZExtValue()
+				dstArray := fr.getLocal(inst.Operand(0)).(*LocalValue).stripPointerCasts()
+				srcArray := fr.getLocal(inst.Operand(1)).(*LocalValue).stripPointerCasts()
+				dstLen := fr.getLocal(inst.Operand(2)).(*LocalValue)
+				srcLen := fr.getLocal(inst.Operand(3)).(*LocalValue)
+				if elementSize != 1 && dstArray.Type().ElementType().TypeKind() == llvm.ArrayTypeKind && srcArray.Type().ElementType().TypeKind() == llvm.ArrayTypeKind {
+					// Slice data pointers are created by adding a global array
+					// and getting the address of the first element using a GEP.
+					// However, before the compiler can pass it to
+					// runtime.sliceCopy, it has to perform a bitcast to a *i8,
+					// to make it a unsafe.Pointer. Now, when the IR builder
+					// sees a bitcast of a GEP with zero indices, it will make
+					// a bitcast of the original array instead of the GEP,
+					// which breaks our assumptions.
+					// Re-add this GEP, in the hope that it it is then of the correct type...
+					dstArray = dstArray.GetElementPtr([]uint32{0, 0}).(*LocalValue)
+					srcArray = srcArray.GetElementPtr([]uint32{0, 0}).(*LocalValue)
+				}
+				if fr.Eval.TargetData.TypeAllocSize(dstArray.Type().ElementType()) != elementSize {
+					return nil, nil, fr.errorAt(inst, "interp: slice dst element size does not match pointer type")
+				}
+				if fr.Eval.TargetData.TypeAllocSize(srcArray.Type().ElementType()) != elementSize {
+					return nil, nil, fr.errorAt(inst, "interp: slice src element size does not match pointer type")
+				}
+				if dstArray.Type() != srcArray.Type() {
+					return nil, nil, fr.errorAt(inst, "interp: slice element types don't match")
+				}
+				length := dstLen.Value().SExtValue()
+				if srcLength := srcLen.Value().SExtValue(); srcLength < length {
+					length = srcLength
+				}
+				if length < 0 {
+					return nil, nil, fr.errorAt(inst, "interp: trying to copy a slice with negative length?")
+				}
+				for i := int64(0); i < length; i++ {
+					// *dst = *src
+					dstArray.Store(srcArray.Load())
+					// dst++
+					dstArray = dstArray.GetElementPtr([]uint32{1}).(*LocalValue)
+					// src++
+					srcArray = srcArray.GetElementPtr([]uint32{1}).(*LocalValue)
+				}
 			case callee.Name() == "runtime.stringToBytes":
 				// convert a string to a []byte
 				bufPtr := fr.getLocal(inst.Operand(0))
@@ -294,7 +403,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				}
 				globalType := llvm.ArrayType(fr.Mod.Context().Int8Type(), len(result))
 				globalValue := llvm.ConstArray(fr.Mod.Context().Int8Type(), vals)
-				global := llvm.AddGlobal(fr.Mod, globalType, fr.pkgName+"$bytes")
+				global := llvm.AddGlobal(fr.Mod, globalType, fr.packagePath+"$bytes")
 				global.SetInitializer(globalValue)
 				global.SetLinkage(llvm.InternalLinkage)
 				global.SetGlobalConstant(true)
@@ -302,7 +411,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				sliceType := inst.Type()
 				retPtr := llvm.ConstGEP(global, getLLVMIndices(fr.Mod.Context().Int32Type(), []uint32{0, 0}))
 				retLen := llvm.ConstInt(sliceType.StructElementTypes()[1], uint64(len(result)), false)
-				ret := getZeroValue(sliceType)
+				ret := llvm.ConstNull(sliceType)
 				ret = llvm.ConstInsertValue(ret, retPtr, []uint32{0}) // ptr
 				ret = llvm.ConstInsertValue(ret, retLen, []uint32{1}) // len
 				ret = llvm.ConstInsertValue(ret, retLen, []uint32{2}) // cap
@@ -311,16 +420,16 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				typecode := fr.getLocal(inst.Operand(0)).(*LocalValue).Underlying
 				interfaceMethodSet := fr.getLocal(inst.Operand(1)).(*LocalValue).Underlying
 				if typecode.IsAConstantExpr().IsNil() || typecode.Opcode() != llvm.PtrToInt {
-					panic("interp: expected typecode to be a ptrtoint")
+					return nil, nil, fr.errorAt(inst, "interp: expected typecode to be a ptrtoint")
 				}
 				typecode = typecode.Operand(0)
 				if interfaceMethodSet.IsAConstantExpr().IsNil() || interfaceMethodSet.Opcode() != llvm.GetElementPtr {
-					panic("interp: expected method set in runtime.interfaceImplements to be a constant gep")
+					return nil, nil, fr.errorAt(inst, "interp: expected method set in runtime.interfaceImplements to be a constant gep")
 				}
 				interfaceMethodSet = interfaceMethodSet.Operand(0).Initializer()
 				methodSet := llvm.ConstExtractValue(typecode.Initializer(), []uint32{1})
 				if methodSet.IsAConstantExpr().IsNil() || methodSet.Opcode() != llvm.GetElementPtr {
-					panic("interp: expected method set to be a constant gep")
+					return nil, nil, fr.errorAt(inst, "interp: expected method set to be a constant gep")
 				}
 				methodSet = methodSet.Operand(0).Initializer()
 
@@ -349,6 +458,10 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				fr.locals[inst] = &LocalValue{fr.Eval, llvm.ConstInt(fr.Mod.Context().Int64Type(), 0, false)}
 			case callee.Name() == "llvm.dbg.value":
 				// do nothing
+			case strings.HasPrefix(callee.Name(), "llvm.lifetime."):
+				// do nothing
+			case callee.Name() == "runtime.trackPointer":
+				// do nothing
 			case strings.HasPrefix(callee.Name(), "runtime.print") || callee.Name() == "runtime._panic":
 				// This are all print instructions, which necessarily have side
 				// effects but no results.
@@ -388,7 +501,10 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 					params = append(params, local)
 				}
 				var ret Value
-				scanResult := fr.Eval.hasSideEffects(callee)
+				scanResult, err := fr.hasSideEffects(callee)
+				if err != nil {
+					return nil, nil, err
+				}
 				if scanResult.severity == sideEffectLimited || dirtyParams && scanResult.severity != sideEffectAll {
 					// Side effect is bounded. This means the operation invokes
 					// side effects (like calling an external function) but it
@@ -411,7 +527,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 					//     compile time.
 					//   * Unbounded: cannot call at runtime so we'll try to
 					//     interpret anyway and hope for the best.
-					ret, err = fr.function(callee, params, fr.pkgName, indent+"    ")
+					ret, err = fr.function(callee, params, indent+"    ")
 					if err != nil {
 						return nil, nil, err
 					}
@@ -421,7 +537,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				}
 			default:
 				// function pointers, etc.
-				return nil, nil, &Unsupported{inst}
+				return nil, nil, fr.unsupportedInstructionError(inst)
 			}
 		case !inst.IsAExtractValueInst().IsNil():
 			agg := fr.getLocal(inst.Operand(0)).(*LocalValue) // must be constant
@@ -431,7 +547,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				fr.locals[inst] = fr.getValue(newValue)
 			} else {
 				if len(indices) != 1 {
-					return nil, nil, errors.New("cannot handle extractvalue with not exactly 1 index")
+					return nil, nil, fr.errorAt(inst, "interp: cannot handle extractvalue with not exactly 1 index")
 				}
 				fr.locals[inst] = &LocalValue{fr.Eval, fr.builder.CreateExtractValue(agg.Underlying, int(indices[0]), inst.Name())}
 			}
@@ -444,7 +560,7 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 				fr.locals[inst] = &LocalValue{fr.Eval, newValue}
 			} else {
 				if len(indices) != 1 {
-					return nil, nil, errors.New("cannot handle insertvalue with not exactly 1 index")
+					return nil, nil, fr.errorAt(inst, "interp: cannot handle insertvalue with not exactly 1 index")
 				}
 				fr.locals[inst] = &LocalValue{fr.Eval, fr.builder.CreateInsertValue(agg.Underlying, val.Value(), int(indices[0]), inst.Name())}
 			}
@@ -457,21 +573,25 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 			// conditional branch (if/then/else)
 			cond := fr.getLocal(inst.Operand(0)).Value()
 			if cond.Type() != fr.Mod.Context().Int1Type() {
-				panic("expected an i1 in a branch instruction")
+				return nil, nil, fr.errorAt(inst, "expected an i1 in a branch instruction")
 			}
 			thenBB := inst.Operand(1)
 			elseBB := inst.Operand(2)
-			if !cond.IsConstant() {
-				return nil, nil, errors.New("interp: branch on a non-constant")
-			} else {
-				switch cond.ZExtValue() {
-				case 0: // false
-					return nil, []llvm.Value{thenBB}, nil // then
-				case 1: // true
-					return nil, []llvm.Value{elseBB}, nil // else
-				default:
-					panic("branch was not true or false")
-				}
+			if !cond.IsAInstruction().IsNil() {
+				return nil, nil, fr.errorAt(inst, "interp: branch on a non-constant")
+			}
+			if !cond.IsAConstantExpr().IsNil() {
+				// This may happen when the instruction builder could not
+				// const-fold some instructions.
+				return nil, nil, fr.errorAt(inst, "interp: branch on a non-const-propagated constant expression")
+			}
+			switch cond {
+			case llvm.ConstInt(fr.Mod.Context().Int1Type(), 0, false): // false
+				return nil, []llvm.Value{thenBB}, nil // then
+			case llvm.ConstInt(fr.Mod.Context().Int1Type(), 1, false): // true
+				return nil, []llvm.Value{elseBB}, nil // else
+			default:
+				return nil, nil, fr.errorAt(inst, "branch was not true or false")
 			}
 		case !inst.IsABranchInst().IsNil() && inst.OperandsCount() == 1:
 			// unconditional branch (goto)
@@ -479,10 +599,11 @@ func (fr *frame) evalBasicBlock(bb, incoming llvm.BasicBlock, indent string) (re
 		case !inst.IsAUnreachableInst().IsNil():
 			// Unreachable was reached (e.g. after a call to panic()).
 			// Report this as an error, as it is not supposed to happen.
-			return nil, nil, ErrUnreachable
+			// This is a sentinel error value.
+			return nil, nil, errUnreachable
 
 		default:
-			return nil, nil, &Unsupported{inst}
+			return nil, nil, fr.unsupportedInstructionError(inst)
 		}
 	}
 
@@ -496,6 +617,7 @@ func (fr *frame) getLocal(v llvm.Value) Value {
 	} else if value := fr.getValue(v); value != nil {
 		return value
 	} else {
+		// This should not happen under normal circumstances.
 		panic("cannot find value")
 	}
 }

interp/interp.go

@@ -7,7 +7,6 @@ package interp
 // methods.
 
 import (
-	"errors"
 	"strings"
 
 	"tinygo.org/x/go-llvm"
@@ -22,9 +21,17 @@ type Eval struct {
 	sideEffectFuncs map[llvm.Value]*sideEffectResult // cache of side effect scan results
 }
 
+// evalPackage encapsulates the Eval type for just a single package. The Eval
+// type keeps state across the whole program, the evalPackage type keeps extra
+// state for the currently interpreted package.
+type evalPackage struct {
+	*Eval
+	packagePath string
+}
+
 // Run evaluates the function with the given name and then eliminates all
 // callers.
-func Run(mod llvm.Module, targetData llvm.TargetData, debug bool) error {
+func Run(mod llvm.Module, debug bool) error {
 	if debug {
 		println("\ncompile-time evaluation:")
 	}
@@ -32,7 +39,7 @@ func Run(mod llvm.Module, targetData llvm.TargetData, debug bool) error {
 	name := "runtime.initAll"
 	e := &Eval{
 		Mod:          mod,
-		TargetData:   targetData,
+		TargetData:   llvm.NewTargetData(mod.DataLayout()),
 		Debug:        debug,
 		dirtyGlobals: map[llvm.Value]struct{}{},
 	}
@@ -56,7 +63,7 @@ func Run(mod llvm.Module, targetData llvm.TargetData, debug bool) error {
 			break // ret void
 		}
 		if inst.IsACallInst().IsNil() || inst.CalledValue().IsAFunction().IsNil() {
-			return errors.New("expected all instructions in " + name + " to be direct calls")
+			return errorAt(inst, "interp: expected all instructions in "+name+" to be direct calls")
 		}
 		initCalls = append(initCalls, inst)
 	}
@@ -66,13 +73,17 @@ func Run(mod llvm.Module, targetData llvm.TargetData, debug bool) error {
 	for _, call := range initCalls {
 		initName := call.CalledValue().Name()
 		if !strings.HasSuffix(initName, ".init") {
-			return errors.New("expected all instructions in " + name + " to be *.init() calls")
+			return errorAt(call, "interp: expected all instructions in "+name+" to be *.init() calls")
 		}
 		pkgName := initName[:len(initName)-5]
 		fn := call.CalledValue()
 		call.EraseFromParentAsInstruction()
-		_, err := e.Function(fn, []Value{&LocalValue{e, undefPtr}, &LocalValue{e, undefPtr}}, pkgName)
-		if err == ErrUnreachable {
+		evalPkg := evalPackage{
+			Eval:        e,
+			packagePath: pkgName,
+		}
+		_, err := evalPkg.function(fn, []Value{&LocalValue{e, undefPtr}, &LocalValue{e, undefPtr}}, "")
+		if err == errUnreachable {
 			break
 		}
 		if err != nil {
@@ -83,16 +94,14 @@ func Run(mod llvm.Module, targetData llvm.TargetData, debug bool) error {
 	return nil
 }
 
-func (e *Eval) Function(fn llvm.Value, params []Value, pkgName string) (Value, error) {
-	return e.function(fn, params, pkgName, "")
-}
-
-func (e *Eval) function(fn llvm.Value, params []Value, pkgName, indent string) (Value, error) {
+// function interprets the given function. The params are the function params
+// and the indent is the string indentation to use when dumping all interpreted
+// instructions.
+func (e *evalPackage) function(fn llvm.Value, params []Value, indent string) (Value, error) {
 	fr := frame{
-		Eval:    e,
-		fn:      fn,
-		pkgName: pkgName,
-		locals:  make(map[llvm.Value]Value),
+		evalPackage: e,
+		fn:          fn,
+		locals:      make(map[llvm.Value]Value),
 	}
 	for i, param := range fn.Params() {
 		fr.locals[param] = params[i]

interp/interp_test.go

@@ -0,0 +1,102 @@
+package interp
+
+import (
+	"io/ioutil"
+	"os"
+	"strings"
+	"testing"
+
+	"tinygo.org/x/go-llvm"
+)
+
+func TestInterp(t *testing.T) {
+	for _, name := range []string{
+		"basic",
+		"slice-copy",
+		"consteval",
+		"map",
+	} {
+		name := name // make tc local to this closure
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+			runTest(t, "testdata/"+name)
+		})
+	}
+}
+
+func runTest(t *testing.T, pathPrefix string) {
+	// Read the input IR.
+	ctx := llvm.NewContext()
+	buf, err := llvm.NewMemoryBufferFromFile(pathPrefix + ".ll")
+	os.Stat(pathPrefix + ".ll") // make sure this file is tracked by `go test` caching
+	if err != nil {
+		t.Fatalf("could not read file %s: %v", pathPrefix+".ll", err)
+	}
+	mod, err := ctx.ParseIR(buf)
+	if err != nil {
+		t.Fatalf("could not load module:\n%v", err)
+	}
+
+	// Perform the transform.
+	err = Run(mod, false)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Run some cleanup passes to get easy-to-read outputs.
+	pm := llvm.NewPassManager()
+	defer pm.Dispose()
+	pm.AddGlobalOptimizerPass()
+	pm.AddDeadStoreEliminationPass()
+	pm.Run(mod)
+
+	// Read the expected output IR.
+	out, err := ioutil.ReadFile(pathPrefix + ".out.ll")
+	if err != nil {
+		t.Fatalf("could not read output file %s: %v", pathPrefix+".out.ll", err)
+	}
+
+	// See whether the transform output matches with the expected output IR.
+	expected := string(out)
+	actual := mod.String()
+	if !fuzzyEqualIR(expected, actual) {
+		t.Logf("output does not match expected output:\n%s", actual)
+		t.Fail()
+	}
+}
+
+// fuzzyEqualIR returns true if the two LLVM IR strings passed in are roughly
+// equal. That means, only relevant lines are compared (excluding comments
+// etc.).
+func fuzzyEqualIR(s1, s2 string) bool {
+	lines1 := filterIrrelevantIRLines(strings.Split(s1, "\n"))
+	lines2 := filterIrrelevantIRLines(strings.Split(s2, "\n"))
+	if len(lines1) != len(lines2) {
+		return false
+	}
+	for i, line := range lines1 {
+		if line != lines2[i] {
+			return false
+		}
+	}
+
+	return true
+}
+
+// filterIrrelevantIRLines removes lines from the input slice of strings that
+// are not relevant in comparing IR. For example, empty lines and comments are
+// stripped out.
+func filterIrrelevantIRLines(lines []string) []string {
+	var out []string
+	for _, line := range lines {
+		line = strings.TrimSpace(line) // drop '\r' on Windows
+		if line == "" || line[0] == ';' {
+			continue
+		}
+		if strings.HasPrefix(line, "source_filename = ") {
+			continue
+		}
+		out = append(out, line)
+	}
+	return out
+}

interp/scan.go

@@ -1,11 +1,28 @@
 package interp
 
 import (
+	"strings"
+
 	"tinygo.org/x/go-llvm"
 )
 
 type sideEffectSeverity int
 
+func (severity sideEffectSeverity) String() string {
+	switch severity {
+	case sideEffectInProgress:
+		return "in progress"
+	case sideEffectNone:
+		return "none"
+	case sideEffectLimited:
+		return "limited"
+	case sideEffectAll:
+		return "all"
+	default:
+		return "unknown"
+	}
+}
+
 const (
 	sideEffectInProgress sideEffectSeverity = iota // computing side effects is in progress (for recursive functions)
 	sideEffectNone                                 // no side effects at all (pure)
@@ -23,26 +40,36 @@ type sideEffectResult struct {
 // hasSideEffects scans this function and all descendants, recursively. It
 // returns whether this function has side effects and if it does, which globals
 // it mentions anywhere in this function or any called functions.
-func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
-	switch fn.Name() {
-	case "runtime.alloc":
+func (e *evalPackage) hasSideEffects(fn llvm.Value) (*sideEffectResult, error) {
+	name := fn.Name()
+	switch {
+	case name == "runtime.alloc":
 		// Cannot be scanned but can be interpreted.
-		return &sideEffectResult{severity: sideEffectNone}
-	case "runtime.nanotime":
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case name == "runtime.nanotime":
 		// Fixed value at compile time.
-		return &sideEffectResult{severity: sideEffectNone}
-	case "runtime._panic":
-		return &sideEffectResult{severity: sideEffectLimited}
-	case "runtime.interfaceImplements":
-		return &sideEffectResult{severity: sideEffectNone}
-	case "llvm.dbg.value":
-		return &sideEffectResult{severity: sideEffectNone}
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case name == "runtime._panic":
+		return &sideEffectResult{severity: sideEffectLimited}, nil
+	case name == "runtime.interfaceImplements":
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case name == "runtime.sliceCopy":
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case name == "runtime.trackPointer":
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case name == "llvm.dbg.value":
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	case strings.HasPrefix(name, "llvm.lifetime."):
+		return &sideEffectResult{severity: sideEffectNone}, nil
+	}
+	if fn.IsDeclaration() {
+		return &sideEffectResult{severity: sideEffectLimited}, nil
 	}
 	if e.sideEffectFuncs == nil {
 		e.sideEffectFuncs = make(map[llvm.Value]*sideEffectResult)
 	}
 	if se, ok := e.sideEffectFuncs[fn]; ok {
-		return se
+		return se, nil
 	}
 	result := &sideEffectResult{
 		severity:        sideEffectInProgress,
@@ -53,6 +80,7 @@ func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
 	for bb := fn.EntryBasicBlock(); !bb.IsNil(); bb = llvm.NextBasicBlock(bb) {
 		for inst := bb.FirstInstruction(); !inst.IsNil(); inst = llvm.NextInstruction(inst) {
 			if inst.IsAInstruction().IsNil() {
+				// Should not happen in valid IR.
 				panic("not an instruction")
 			}
 
@@ -69,7 +97,7 @@ func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
 			switch inst.InstructionOpcode() {
 			case llvm.IndirectBr, llvm.Invoke:
 				// Not emitted by the compiler.
-				panic("unknown instructions")
+				return nil, e.errorAt(inst, "unknown instructions")
 			case llvm.Call:
 				child := inst.CalledValue()
 				if !child.IsAInlineAsm().IsNil() {
@@ -95,7 +123,10 @@ func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
 					}
 					continue
 				}
-				childSideEffects := e.hasSideEffects(child)
+				childSideEffects, err := e.hasSideEffects(child)
+				if err != nil {
+					return nil, err
+				}
 				switch childSideEffects.severity {
 				case sideEffectInProgress, sideEffectNone:
 					// no side effects or recursive function - continue scanning
@@ -109,7 +140,16 @@ func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
 				default:
 					panic("unreachable")
 				}
-			case llvm.Load, llvm.Store:
+			case llvm.Load:
+				if inst.IsVolatile() {
+					result.updateSeverity(sideEffectLimited)
+				}
+				if _, ok := e.dirtyGlobals[inst.Operand(0)]; ok {
+					if e.hasLocalSideEffects(dirtyLocals, inst) {
+						result.updateSeverity(sideEffectLimited)
+					}
+				}
+			case llvm.Store:
 				if inst.IsVolatile() {
 					result.updateSeverity(sideEffectLimited)
 				}
@@ -128,7 +168,7 @@ func (e *Eval) hasSideEffects(fn llvm.Value) *sideEffectResult {
 		// No side effect was reported for this function.
 		result.severity = sideEffectNone
 	}
-	return result
+	return result, nil
 }
 
 // hasLocalSideEffects checks whether the given instruction flows into a branch
@@ -143,6 +183,7 @@ func (e *Eval) hasLocalSideEffects(dirtyLocals map[llvm.Value]struct{}, inst llv
 	for use := inst.FirstUse(); !use.IsNil(); use = use.NextUse() {
 		user := use.User()
 		if user.IsAInstruction().IsNil() {
+			// Should not happen in valid IR.
 			panic("user not an instruction")
 		}
 		switch user.InstructionOpcode() {
@@ -162,10 +203,13 @@ func (e *Eval) hasLocalSideEffects(dirtyLocals map[llvm.Value]struct{}, inst llv
 				// Already handled in (*Eval).hasSideEffects.
 				continue
 			}
-			// But a store might also store to an alloca, in which case all uses
-			// of the alloca (possibly indirect through a GEP, bitcast, etc.)
-			// must be marked dirty.
-			panic("todo: store")
+			// This store might affect all kinds of values. While it is
+			// certainly possible to traverse through all of them, the easiest
+			// option right now is to just assume the worst and say that this
+			// function has side effects.
+			// TODO: traverse through all stores and mark all relevant allocas /
+			// globals dirty.
+			return true
 		default:
 			// All instructions that take 0 or more operands (1 or more if it
 			// was a use) and produce a result.

interp/scan_test.go

@@ -0,0 +1,93 @@
+package interp
+
+import (
+	"os"
+	"sort"
+	"testing"
+
+	"tinygo.org/x/go-llvm"
+)
+
+var scanTestTable = []struct {
+	name            string
+	severity        sideEffectSeverity
+	mentionsGlobals []string
+}{
+	{"returnsConst", sideEffectNone, nil},
+	{"returnsArg", sideEffectNone, nil},
+	{"externalCallOnly", sideEffectNone, nil},
+	{"externalCallAndReturn", sideEffectLimited, nil},
+	{"externalCallBranch", sideEffectLimited, nil},
+	{"readCleanGlobal", sideEffectNone, []string{"cleanGlobalInt"}},
+	{"readDirtyGlobal", sideEffectLimited, []string{"dirtyGlobalInt"}},
+	{"callFunctionPointer", sideEffectAll, []string{"functionPointer"}},
+	{"getDirtyPointer", sideEffectLimited, nil},
+	{"storeToPointer", sideEffectLimited, nil},
+}
+
+func TestScan(t *testing.T) {
+	t.Parallel()
+
+	// Read the input IR.
+	path := "testdata/scan.ll"
+	ctx := llvm.NewContext()
+	buf, err := llvm.NewMemoryBufferFromFile(path)
+	os.Stat(path) // make sure this file is tracked by `go test` caching
+	if err != nil {
+		t.Fatalf("could not read file %s: %v", path, err)
+	}
+	mod, err := ctx.ParseIR(buf)
+	if err != nil {
+		t.Fatalf("could not load module:\n%v", err)
+	}
+
+	// Check all to-be-tested functions.
+	for _, tc := range scanTestTable {
+		// Create an eval object, for testing.
+		e := &Eval{
+			Mod:          mod,
+			TargetData:   llvm.NewTargetData(mod.DataLayout()),
+			dirtyGlobals: map[llvm.Value]struct{}{},
+		}
+
+		// Mark some globals dirty, for testing.
+		e.markDirty(mod.NamedGlobal("dirtyGlobalInt"))
+
+		// Scan for side effects.
+		fn := mod.NamedFunction(tc.name)
+		if fn.IsNil() {
+			t.Errorf("scan test: could not find tested function %s in the IR", tc.name)
+			continue
+		}
+		evalPkg := &evalPackage{e, "testdata"}
+		result, err := evalPkg.hasSideEffects(fn)
+		if err != nil {
+			t.Errorf("scan test: failed to scan %s for side effects: %v", fn.Name(), err)
+		}
+
+		// Check whether the result is what we expect.
+		if result.severity != tc.severity {
+			t.Errorf("scan test: function %s should have severity %s but it has %s", tc.name, tc.severity, result.severity)
+		}
+
+		// Check whether the mentioned globals match with what we'd expect.
+		mentionsGlobalNames := make([]string, 0, len(result.mentionsGlobals))
+		for global := range result.mentionsGlobals {
+			mentionsGlobalNames = append(mentionsGlobalNames, global.Name())
+		}
+		sort.Strings(mentionsGlobalNames)
+		globalsMismatch := false
+		if len(result.mentionsGlobals) != len(tc.mentionsGlobals) {
+			globalsMismatch = true
+		} else {
+			for i, globalName := range mentionsGlobalNames {
+				if tc.mentionsGlobals[i] != globalName {
+					globalsMismatch = true
+				}
+			}
+		}
+		if globalsMismatch {
+			t.Errorf("scan test: expected %s to mention globals %v, but it mentions globals %v", tc.name, tc.mentionsGlobals, mentionsGlobalNames)
+		}
+	}
+}

interp/testdata/basic.ll

@@ -0,0 +1,60 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--linux"
+
+@main.v1 = internal global i64 0
+@main.nonConst1 = global [4 x i64] zeroinitializer
+@main.nonConst2 = global i64 0
+
+declare void @runtime.printint64(i64) unnamed_addr
+
+declare void @runtime.printnl() unnamed_addr
+
+define void @runtime.initAll() unnamed_addr {
+entry:
+  call void @runtime.init()
+  call void @main.init()
+  ret void
+}
+
+define void @main() unnamed_addr {
+entry:
+  %0 = load i64, i64* @main.v1
+  call void @runtime.printint64(i64 %0)
+  call void @runtime.printnl()
+  ret void
+}
+
+define internal void @runtime.init() unnamed_addr {
+entry:
+  ret void
+}
+
+define internal void @main.init() unnamed_addr {
+entry:
+  store i64 3, i64* @main.v1
+  call void @"main.init#1"()
+
+  ; test the following pattern:
+  ;     func someValue() int // extern function
+  ;     var nonConst1 = [4]int{someValue(), 0, 0, 0}
+  %value1 = call i64 @someValue()
+  %gep1 = getelementptr [4 x i64], [4 x i64]* @main.nonConst1, i32 0, i32 0
+  store i64 %value1, i64* %gep1
+
+  ; Test that the global really is marked dirty:
+  ;     var nonConst2 = nonConst1[0]
+  %gep2 = getelementptr [4 x i64], [4 x i64]* @main.nonConst1, i32 0, i32 0
+  %value2 = load i64, i64* %gep2
+  store i64 %value2, i64* @main.nonConst2
+
+  ret void
+}
+
+define internal void @"main.init#1"() unnamed_addr {
+entry:
+  call void @runtime.printint64(i64 5)
+  call void @runtime.printnl()
+  ret void
+}
+
+declare i64 @someValue()

interp/testdata/basic.out.ll

@@ -0,0 +1,29 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--linux"
+
+@main.nonConst1 = local_unnamed_addr global [4 x i64] zeroinitializer
+@main.nonConst2 = local_unnamed_addr global i64 0
+
+declare void @runtime.printint64(i64) unnamed_addr
+
+declare void @runtime.printnl() unnamed_addr
+
+define void @runtime.initAll() unnamed_addr {
+entry:
+  call void @runtime.printint64(i64 5)
+  call void @runtime.printnl()
+  %value1 = call i64 @someValue()
+  store i64 %value1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @main.nonConst1, i32 0, i32 0)
+  %value2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @main.nonConst1, i32 0, i32 0)
+  store i64 %value2, i64* @main.nonConst2
+  ret void
+}
+
+define void @main() unnamed_addr {
+entry:
+  call void @runtime.printint64(i64 3)
+  call void @runtime.printnl()
+  ret void
+}
+
+declare i64 @someValue() local_unnamed_addr

interp/testdata/consteval.ll

@@ -0,0 +1,42 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--linux"
+
+@intToPtrResult = global i8 0
+@ptrToIntResult = global i8 0
+
+define void @runtime.initAll() {
+  call void @main.init()
+  ret void
+}
+
+define internal void @main.init() {
+  call void @testIntToPtr()
+  call void @testPtrToInt()
+  ret void
+}
+
+define internal void @testIntToPtr() {
+  %nil = icmp eq i8* inttoptr (i64 1024 to i8*), null
+  br i1 %nil, label %a, label %b
+a:
+  ; should not be reached
+  store i8 1, i8* @intToPtrResult
+  ret void
+b:
+  ; should be reached
+  store i8 2, i8* @intToPtrResult
+  ret void
+}
+
+define internal void @testPtrToInt() {
+  %zero = icmp eq i64 ptrtoint (i8* @ptrToIntResult to i64), 0
+  br i1 %zero, label %a, label %b
+a:
+  ; should not be reached
+  store i8 1, i8* @ptrToIntResult
+  ret void
+b:
+  ; should be reached
+  store i8 2, i8* @ptrToIntResult
+  ret void
+}

interp/testdata/consteval.out.ll

@@ -0,0 +1,9 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--linux"
+
+@intToPtrResult = local_unnamed_addr global i8 2
+@ptrToIntResult = local_unnamed_addr global i8 2
+
+define void @runtime.initAll() local_unnamed_addr {
+  ret void
+}

interp/testdata/map.ll

@@ -0,0 +1,76 @@
+target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+target triple = "armv6m-none-eabi"
+
+%runtime._string = type { i8*, i32 }
+%runtime.hashmap = type { %runtime.hashmap*, i8*, i32, i8, i8, i8 }
+
+@main.m = global %runtime.hashmap* null
+@main.binaryMap = global %runtime.hashmap* null
+@main.stringMap = global %runtime.hashmap* null
+@main.init.string = internal unnamed_addr constant [7 x i8] c"CONNECT"
+
+declare %runtime.hashmap* @runtime.hashmapMake(i8, i8, i32, i8* %context, i8* %parentHandle)
+declare void @runtime.hashmapBinarySet(%runtime.hashmap*, i8*, i8*, i8* %context, i8* %parentHandle)
+declare void @runtime.hashmapStringSet(%runtime.hashmap*, i8*, i32, i8*, i8* %context, i8* %parentHandle)
+declare void @llvm.lifetime.end.p0i8(i64, i8*)
+declare void @llvm.lifetime.start.p0i8(i64, i8*)
+
+define void @runtime.initAll() unnamed_addr {
+entry:
+  call void @main.init(i8* undef, i8* null)
+  ret void
+}
+
+define internal void @main.init(i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+; Test that hashmap optimizations generally work (even with lifetimes).
+  %hashmap.key = alloca i8
+  %hashmap.value = alloca %runtime._string
+  %0 = call %runtime.hashmap* @runtime.hashmapMake(i8 1, i8 8, i32 1, i8* undef, i8* null)
+  %hashmap.value.bitcast = bitcast %runtime._string* %hashmap.value to i8*
+  call void @llvm.lifetime.start.p0i8(i64 8, i8* %hashmap.value.bitcast)
+  store %runtime._string { i8* getelementptr inbounds ([7 x i8], [7 x i8]* @main.init.string, i32 0, i32 0), i32 7 }, %runtime._string* %hashmap.value
+  call void @llvm.lifetime.start.p0i8(i64 1, i8* %hashmap.key)
+  store i8 1, i8* %hashmap.key
+  call void @runtime.hashmapBinarySet(%runtime.hashmap* %0, i8* %hashmap.key, i8* %hashmap.value.bitcast, i8* undef, i8* null)
+  call void @llvm.lifetime.end.p0i8(i64 1, i8* %hashmap.key)
+  call void @llvm.lifetime.end.p0i8(i64 8, i8* %hashmap.value.bitcast)
+  store %runtime.hashmap* %0, %runtime.hashmap** @main.m
+
+  ; Other tests, that can be done in a separate function.
+  call void @main.testNonConstantBinarySet()
+  call void @main.testNonConstantStringSet()
+  ret void
+}
+
+; Test that a map loaded from a global can still be used for mapassign
+; operations (with binary keys).
+define internal void @main.testNonConstantBinarySet() {
+  %hashmap.key = alloca i8
+  %hashmap.value = alloca i8
+  ; Create hashmap from global. This breaks the normal hashmapBinarySet
+  ; optimization, to test the fallback.
+  %map.new = call %runtime.hashmap* @runtime.hashmapMake(i8 1, i8 1, i32 1, i8* undef, i8* null)
+  store %runtime.hashmap* %map.new, %runtime.hashmap** @main.binaryMap
+  %map = load %runtime.hashmap*, %runtime.hashmap** @main.binaryMap
+  ; Do the binary set to the newly loaded map.
+  store i8 1, i8* %hashmap.key
+  store i8 2, i8* %hashmap.value
+  call void @runtime.hashmapBinarySet(%runtime.hashmap* %map, i8* %hashmap.key, i8* %hashmap.value, i8* undef, i8* null)
+  ret void
+}
+
+; Test that a map loaded from a global can still be used for mapassign
+; operations (with string keys).
+define internal void @main.testNonConstantStringSet() {
+  %hashmap.value = alloca i8
+  ; Create hashmap from global. This breaks the normal hashmapStringSet
+  ; optimization, to test the fallback.
+  %map.new = call %runtime.hashmap* @runtime.hashmapMake(i8 8, i8 1, i32 1, i8* undef, i8* null)
+  store %runtime.hashmap* %map.new, %runtime.hashmap** @main.stringMap
+  %map = load %runtime.hashmap*, %runtime.hashmap** @main.stringMap
+  ; Do the string set to the newly loaded map.
+  store i8 2, i8* %hashmap.value
+  call void @runtime.hashmapStringSet(%runtime.hashmap* %map, i8* getelementptr inbounds ([7 x i8], [7 x i8]* @main.init.string, i32 0, i32 0), i32 7, i8* %hashmap.value, i8* undef, i8* null)
+  ret void
+}

interp/testdata/map.out.ll

@@ -0,0 +1,28 @@
+target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+target triple = "armv6m-none-eabi"
+
+%runtime.hashmap = type { %runtime.hashmap*, i8*, i32, i8, i8, i8 }
+%runtime._string = type { i8*, i32 }
+
+@main.m = local_unnamed_addr global %runtime.hashmap* @"main$map"
+@main.binaryMap = local_unnamed_addr global %runtime.hashmap* @"main$map.4"
+@main.stringMap = local_unnamed_addr global %runtime.hashmap* @"main$map.6"
+@main.init.string = internal unnamed_addr constant [7 x i8] c"CONNECT"
+@"main$mapbucket" = internal unnamed_addr global { [8 x i8], i8*, [8 x i8], [8 x %runtime._string] } { [8 x i8] c"\04\00\00\00\00\00\00\00", i8* null, [8 x i8] c"\01\00\00\00\00\00\00\00", [8 x %runtime._string] [%runtime._string { i8* getelementptr inbounds ([7 x i8], [7 x i8]* @main.init.string, i32 0, i32 0), i32 7 }, %runtime._string zeroinitializer, %runtime._string zeroinitializer, %runtime._string zeroinitializer, %runtime._string zeroinitializer, %runtime._string zeroinitializer, %runtime._string zeroinitializer, %runtime._string zeroinitializer] }
+@"main$map" = internal unnamed_addr global %runtime.hashmap { %runtime.hashmap* null, i8* getelementptr inbounds ({ [8 x i8], i8*, [8 x i8], [8 x %runtime._string] }, { [8 x i8], i8*, [8 x i8], [8 x %runtime._string] }* @"main$mapbucket", i32 0, i32 0, i32 0), i32 1, i8 1, i8 8, i8 0 }
+@"main$alloca.2" = internal global i8 1
+@"main$alloca.3" = internal global i8 2
+@"main$map.4" = internal unnamed_addr global %runtime.hashmap { %runtime.hashmap* null, i8* null, i32 0, i8 1, i8 1, i8 0 }
+@"main$alloca.5" = internal global i8 2
+@"main$map.6" = internal unnamed_addr global %runtime.hashmap { %runtime.hashmap* null, i8* null, i32 0, i8 8, i8 1, i8 0 }
+
+declare void @runtime.hashmapBinarySet(%runtime.hashmap*, i8*, i8*, i8*, i8*) local_unnamed_addr
+
+declare void @runtime.hashmapStringSet(%runtime.hashmap*, i8*, i32, i8*, i8*, i8*) local_unnamed_addr
+
+define void @runtime.initAll() unnamed_addr {
+entry:
+  call void @runtime.hashmapBinarySet(%runtime.hashmap* @"main$map.4", i8* @"main$alloca.2", i8* @"main$alloca.3", i8* undef, i8* null)
+  call void @runtime.hashmapStringSet(%runtime.hashmap* @"main$map.6", i8* getelementptr inbounds ([7 x i8], [7 x i8]* @main.init.string, i32 0, i32 0), i32 7, i8* @"main$alloca.5", i8* undef, i8* null)
+  ret void
+}

interp/testdata/scan.ll

@@ -0,0 +1,61 @@
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64--linux"
+
+define i64 @returnsConst() {
+  ret i64 0
+}
+
+define i64 @returnsArg(i64 %arg) {
+  ret i64 %arg
+}
+
+declare i64 @externalCall()
+
+define i64 @externalCallOnly() {
+  %result = call i64 @externalCall()
+  ret i64 0
+}
+
+define i64 @externalCallAndReturn() {
+  %result = call i64 @externalCall()
+  ret i64 %result
+}
+
+define i64 @externalCallBranch() {
+  %result = call i64 @externalCall()
+  %zero = icmp eq i64 %result, 0
+  br i1 %zero, label %if.then, label %if.done
+
+if.then:
+  ret i64 2
+
+if.done:
+  ret i64 4
+}
+
+@cleanGlobalInt = global i64 5
+define i64 @readCleanGlobal() {
+  %global = load i64, i64* @cleanGlobalInt
+  ret i64 %global
+}
+
+@dirtyGlobalInt = global i64 5
+define i64 @readDirtyGlobal() {
+  %global = load i64, i64* @dirtyGlobalInt
+  ret i64 %global
+}
+
+declare i64* @getDirtyPointer()
+
+define void @storeToPointer() {
+  %ptr = call i64* @getDirtyPointer()
+  store i64 3, i64* %ptr
+  ret void
+}
+
+@functionPointer = global i64()* null
+define i64 @callFunctionPointer() {
+  %fp = load i64()*, i64()** @functionPointer
+  %result = call i64 %fp()
+  ret i64 %result
+}