On Sat, Nov 8, 2014 at 12:05 AM, Ozkan Sezer <[email protected]> wrote: > On 11/7/14, Dongsheng Song <[email protected]> wrote: >> On Fri, Nov 7, 2014 at 7:51 PM, Ozkan Sezer <[email protected]> wrote: >>> On 11/7/14, Ray Donnelly <[email protected]> wrote: >>>> On Fri, Nov 7, 2014 at 11:10 AM, Ozkan Sezer <[email protected]> wrote: >>>>> On 11/7/14, Ruben Van Boxem <[email protected]> wrote: >>>>>> 2014-11-07 9:25 GMT+01:00 Ozkan Sezer <[email protected]>: >>>>>> >>>>>>> On 11/7/14, Dongsheng Song <[email protected]> wrote: >>>>>>> > If we define _POSIX_, then getpid (process.h) was hidden. >>>>>>> > Is it correct ? >>>>>>> > >>>>>>> > PS: MSVC 2012 is the last compiler which use _POSIX_, MSVC 2013 do >>>>>>> > not >>>>>>> > use _POSIX_ anymore. >>>>>>> > MSVC 2012/2013 guard getpid with !__STDC__. >>>>>>> >>>>>>> I believe (but not necessarily correct about iıt) that MSVC's _POSIX >>>>>>> symbol is intended for diffrerent purposes, i.e. windows posix >>>>>>> subsystem, >>>>>>> and I believe that we are doing a wrong thing with having those _POSIX >>>>>>> ifdefs in our headers.. Someone correct me if I'm wrong. >>>>>>> >>>>>> >>>>>> I have no idea, but be aware at least one reference in GCC showed up: >>>>>> https://gcc.gnu.org/bugzilla/attachment.cgi?id=20034&action=edit >>>>>> >>>>>> But maybe that's there exactly because _POSIX is in the MinGW-w64 >>>>>> headers... >>>>> >>>>> I remember that they defined _POSIX only because mingw-w64 headers >>>>> required it for certain declarations >>>> >>>> Also, should we consider renaming _POSIX to _POSIX_SOURCE? >>>> >>> >>> What I am saying is that those two have different meanings. >>> If we ever get rid of _POSIX we will possibly need to remove >>> certain stuff too. Things that currently are guarded by _POSIX >>> but should actually be guarded by _POSIX_SOURCE are there too >>> and they are another part of the story. >>> >> >> During port libressl to Windows, I feel very painful on such macros. >> I want to use _POSIX_C_SOURCE to replace _POSIX, _POSIX_, >> _POSIX_SOURCE in all headers. >> The first is normalization in _mingw.h.in: >> >> --- a/mingw-w64-headers/crt/_mingw.h.in >> +++ b/mingw-w64-headers/crt/_mingw.h.in >> @@ -7,6 +7,26 @@ >> #ifndef _INC__MINGW_H >> #define _INC__MINGW_H >> >> +#if defined(_POSIX) || defined(_POSIX_) \ >> + || defined(_POSIX_SOURCE) || defined(_POSIX_C_SOURCE) \ >> + || defined(_REENTRANT) || defined(_THREAD_SAFE) || >> defined(_POSIX_THREAD_SAFE_FUNCTIONS) >> + >> +/* This has been superseded by _POSIX_C_SOURCE. */ >> +#ifndef _POSIX_SOURCE >> +#define _POSIX_SOURCE 1 >> +#endif >> + >> +/* Functions like wcscasecmp() and strnlen() were only added in >> POSIX.1-2008 */ >> +#ifndef _POSIX_C_SOURCE >> +#define _POSIX_C_SOURCE 200809L >> +#endif >> + >> +#if (defined(_REENTRANT) || defined(_THREAD_SAFE)) && >> !defined(_POSIX_THREAD_SAFE_FUNCTIONS) >> +#define _POSIX_THREAD_SAFE_FUNCTIONS 200809L >> +#endif >> + >> +#endif >> + >> >> comments ? > > I'd say that this is lazy and wrong. IMO, doing it right way should be > changing the correct (only the correct) _POSIX guards to _POSIX_SOURCE > instead, and hope that the project requesting such features are actually > defining _POSIX_SOURCE. (you should possily also cover _GNU_SOURCE, etc) > > -- > O.S. >
If you use _POSIX_SOURCE guard, no more _POSIX guard, you must make _POSIX_SOURCE defined automatically when _POSIX defined for compatibility. PS：_POSIX_SOURCE has been superseded by _POSIX_C_SOURCE. ------------------------------------------------------------------------------ _______________________________________________ Mingw-w64-public mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
