iana added a comment. In D157793#4585268 <https://reviews.llvm.org/D157793#4585268>, @aaron.ballman wrote:
> In D157793#4583698 <https://reviews.llvm.org/D157793#4583698>, @iana wrote: > >> In D157793#4583663 <https://reviews.llvm.org/D157793#4583663>, @MaskRay >> wrote: >> >>>> Apple needs a __need_ macro for va_list. Add one, and also ones for >>>> va_start/va_arg/va_end, __va_copy, va_copy. >>> >>> Do you have a link to the specification or the source that this is needed? >> >> We need is so that our <sys/_types/_va_list.h> doesn't have to redeclare >> va_list and doesn't pull in all of stdarg.h either. As far as I know there's >> no specification for this (or stddef.h) being include-able in pieces. > > I'm not opposed, but this adds significant complexity to support a > questionable use case -- the C standard library headers are not intended to > be partially included, so if the plan is to use these `__need` macros in all > of the headers we provide, I think that should go through an RFC process to > be sure we want to maintain the added complexity. (Also, who is "we" in this > case?) We is Apple. We have a <sys/_types/_va_list.h> header that is generating redeclaration warnings in modules mode. I can't just include <stdarg.h> instead since it would change the semantics of <sys/_types/_va_list.h>. stdarg.h and stddef.h appear to be the only clang headers that Apple needs to use in pieces though, I don't think we should add this complexity to any other headers for consistency or "just because". Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D157793/new/ https://reviews.llvm.org/D157793 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
