On 2012-08-10 at 04:18:35 [+0200], Matt Madia <mattmadia@xxxxxxxxx> wrote: > On 8/9/12, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > On 2012-08-09 at 00:22:58 [+0200], Matt Madia <mattmadia@xxxxxxxxx> wrote: > >> On 8/8/12, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > >> > Why not define a _haiku_build_size_t? > >> > >> I honesty didn't consider doing that. Should this go in > >> headers/build/config_target/target_types.h? > >> typedef _HAIKU_BUILD_FEATURE___SIZE_TYPE__ _haiku_build_size_t ; > > > > I believe stddef.h is supposed to provide the size_t type, so the > > corresponding target header would be a more suitable place. > > But isn't stddef.h one of the headers that won't be getting a target_* > counterpart? If that's the case (I don't recall which compiler headers are wrapped by OS headers in FreeBSD), then the whole point is moot. stddef.h defines size_t, so you can use that directly and don't need a _haiku_build_size_t. > >> > That aside, __SIZE_TYPE__ is defined directly by the compiler (e.g. try > >> > "echo __SIZE_TYPE__ | gcc -E -"), so there's no point in wrapping it > >> > either. > >> > >> Since you mentioned using _haiku_build_size_t, the visual uniformity > >> does appeal to me. > > > > I wasn't arguing against the typedef. I was just pointing out that > > _HAIKU_BUILD_FEATURE___SIZE_TYPE__ is superfluous, since __SIZE_TYPE__ is > > not defined in a header, but by the compiler itself, so you can just as > > well use it directly. I also don't think it is required by any standard, > > so > > > > it might be gcc specific. Consequently size_t would need to be imported > > like the fixed-width integer types. > > And that would be by copying PRINT_TYPE_INFO(size_t); and related code > into test_stddef_header.cpp? Please don't copy but rather move shared stuff to a header included by both. Other than that, yes, if it is actually needed. CU, Ingo