MSYS2/Folly unistd conflict in Windows builds

[ikappaki]

Issue Description

On Windows, folly::portability::unistd conflicts with the POSIX APIs provided by MSYS2's Clang64 unistd.h. This manifests in jank's unity builds, where both are included in the same compilation unit, resulting in errors such as:

error: call to 'close' is ambiguous

Fix to follow.

Reproduction Steps

This is intermittent and has been observed in CI builds, for example https://github.com/ikappaki/jank-win/actions/runs/32768895617/job/97564805421:

ccache D:\a\_temp\msys64\clang64\bin\clang++.exe -D_CINDEX_LIB_ -D_FILE_OFFSET_BITS=64 -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -ID:/a/jank-win/jank-win/compiler+runtime/build -ID:/a/jank-win/jank-win/compiler+runtime -isystem D:/a/jank-win/jank-win/compiler+runtime/include/cpp -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/nanobench/include -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/folly -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/bdwgc/include -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/immer -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/zip/src -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/cpptrace/include -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/cppinterop/include -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/cppinterop/lib -isystem D:/a/jank-win/jank-win/compiler+runtime/third-party/boost-preprocessor/include -isystem D:/a/jank-win/jank-win/compi
In file included from CMakeFiles\jank_compiler_lib.dir\Unity\unity_0_cxx.cxx:10:
D:/a/jank-win/jank-win/compiler+runtime/src/cpp/jank/util/clang.cpp:34:5: error: call to 'close' is ambiguous
   34 |     close(fd);
      |     ^~~~~
D:/a/_temp/msys64/clang64/include/io.h:184:15: note: candidate function
  184 |   int __cdecl close(int _FileHandle) __MINGW_ATTRIB_DEPRECATED_MSVC2005;
      |               ^
..\third-party\folly/folly/portability/Unistd.h:79:5: note: candidate function
   79 | int close(int fh);
      |     ^
1 error generated.
ninja: build stopped: subcommand failed.

jank health check

─ system ───────────────────────────────────────────────────────────────────────────────────────────
─ ✅ operating system: windows
─ ✅ default triple: x86_64-w64-windows-gnu

─ jank install ─────────────────────────────────────────────────────────────────────────────────────
─ ✅ jank version: jank-0.1-alpha
─ ✅ jank cmake build type: Release
─ ✅ jank resource dir: ../lib/jank/0.1
─ ✅ jank resolved resource dir: 0x191f2da5b60 (ignored for dev build)
─ ✅ jank user cache dir: D:\removeme\msys64-jank21\home\chaos/.cache/jank/x86_64-w64-windows-gnu-22b51bbf32e189e324391d02448e61e80d50ba2664bd258bfa7e19fca4d61b9a (not found)

Stack trace

No response

OS Version Information

MINGW64_NT-10.0

I acknowledge that:

• I have searched the jank repository (both open and closed Discussions and Issues) and confirm this is not a duplicate of an existing issue or discussion.
• I have checked the "Preview" tab on all text fields to ensure that everything looks right, and have wrapped all configuration and code in code blocks with a group of three backticks (```) on separate lines.

[jeaye]

I have a fix for these! Just merged it, as part of #960

[ikappaki]

Thanks! Using ::close to prefer the MSYS2 runtime declaration over Folly's is a nice workaround. I'm going to address the issue at the source by not exposing Folly's unistd APIs in the global scope, it is not really needed.

[ikappaki]

There's another related issue: MSYS2 introduces an exception_info macro through a transitive include, which conflicts with jank's class of the same name. Although a guard already exists to undefine the macro in jank, it depends on the order of the transitive include, which is not guaranteed and can result in the following error:

D:/src/jank/compiler+runtime/include/cpp/jank/runtime/obj/exception_info.hpp:32:20: error: too many arguments provided to function-like macro invocation
   32 |     exception_info(jtl::immutable_string const &message, object_ref const data);
      |                    ^
D:/removeme/msys64-jank21/clang64/include/excpt.h:53:9: note: macro 'exception_info' defined here
   53 | #define exception_info() ((struct _EXCEPTION_POINTERS *)_exception_info())
      |         ^
In file included from D:/src/jank/compiler+runtime/src/cpp/jank/runtime/obj/exception_info.cpp:3:
D:/src/jank/compiler+runtime/include/cpp/jank/runtime/obj/exception_info.hpp:32:5: warning: declaration does not declare anything [-Wmissing-declarations]
   32 |     exception_info(jtl::immutable_string const &message, object_ref const data);
      |     ^~~~~~~~~~~~~~
D:/src/jank/compiler+runtime/src/cpp/jank/runtime/obj/exception_info.cpp:15:34: error: too many arguments provided to function-like macro invocation
   15 |   exception_info::exception_info(jtl::immutable_string const &message, object_ref const data)
      |                                  ^
D:/removeme/msys64-jank21/clang64/include/excpt.h:53:9: note: macro 'exception_info' defined here
   53 | #define exception_info() ((struct _EXCEPTION_POINTERS *)_exception_info())
      |         ^
D:/src/jank/compiler+runtime/src/cpp/jank/runtime/obj/exception_info.cpp:16:5: error: unexpected ':' in nested name specifier; did you mean '::'?
   16 |     : object{ obj_type, obj_behaviors }
      |     ^
      |     ::
D:/src/jank/compiler+runtime/src/cpp/jank/runtime/obj/exception_info.cpp:16:13: error: expected unqualified-id
   16 |     : object{ obj_type, obj_behaviors }
      |             ^

I'm going to address this as well.

[jeaye]

Yeah, this one is unfortunate. I think we just need to move the #undef to be after all #include uses in exception_info.hpp. But let me know what you find! I'd prefer to not have to rename our jank object because of Windows.