On 12.05.2017 18:00, u-j...@aetey.se wrote: > On Fri, May 12, 2017 at 05:19:15PM +0200, Kamil Rytarowski wrote: >> On 11.05.2017 21:53, avih wrote: >>> <dalias> by providing the intrinsics (__builtin_va*) >>> <Shiz> musl relies on __builtin_va_list and the likes >> >> NetBSD libc has hard requirement on __builtin_va_list (or some >> equivalent builtin). Without it, we cannot use TCC on NetBSD. > > Would you elaborate in which way the requirement is "hard"? >
We require C and C++ compilers that ship with popular extensions on par with GCC. Old va_list has been retired with the following commit: https://github.com/NetBSD/src/commit/9735e84a7821b8a9cc42139d798ee188c4dd9368 This restricts NetBSD to GCC, Clang, PCC and perhaps ICC. > An implementation of va_list is not itself a part of the calling > conventions, it should be possible to use whatever mechanics the compiler > chooses, builtins, macros or inlined/static functions. > > Is the limitation like "no support to avoid a certain definition of > va_list which is hidden inside the standard includes"? Then in theory > it could still be worked around, by using a modified copy of up to > all the system includes (which would be not pretty but possible). > We set minimum reasonable set of features mandatory in the toolchain. Certain things cannot be reliably done in a portable way in libc, like alloca(3). Old port of TCC used to... pretend to be lint(1). This verification tool has relaxed requirements in our headers. We won't reintroduce any fallbacks to libc other than __tcc_va_list or similar. > Regards, > Rune > > > _______________________________________________ > Tinycc-devel mailing list > Tinycc-devel@nongnu.org > https://lists.nongnu.org/mailman/listinfo/tinycc-devel >
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Tinycc-devel mailing list Tinycc-devel@nongnu.org https://lists.nongnu.org/mailman/listinfo/tinycc-devel