Code like this is not allowed by the upstream Go CGo implementation, but
was allowed by TinyGo:
var _ int8 = C.int8_t(5)
The reason it shouldn't be allowed is a little bit complicated. While
it is true that C.int8_t is always the same underlying data type as Go
int8 (signed 8-bit integer), the C type is actually a typedef of one of
the base C types (usually unsigned char or signed char) which in turn do
_not_ map cleanly to Go types: the 'char' type is ambiguous (it may be
either signed or unsigned depending on the ABI) and types like 'int'
vary in size by ABI as well.
To make code more portable, I think it's better to match the upstream
implementation.
This is a large refactor of the cgo package. It should fix a number of
smaller problems and be a bit more strict (like upstream CGo): it for
example requires every Go file in a package to include the header files
it needs instead of piggybacking on imports in earlier files.
The main benefit is that it should be a bit more maintainable and easier
to add new features in the future (like static functions).
This breaks the tinygo.org/x/bluetooth package, which should be updated
before this change lands.
Previously, libclang was run on each fragment (import "C") separately.
However, in regular Go it's possible for later fragments to refer to
types in earlier fragments so they must have been parsed as one.
This commit changes the behavior to run only one C parser invocation for
each Go file.
This commit adds tests for CGo preprocessing. There are various errors
that can be reported while preprocessing, and they should integrate well
with the compiler (including accurate source location tracking).
Also allow CGo preprocessing to continue after Clang encountered an
error, for a better view of what happened.
Ayke van Laethem