Re: [MinGW-w64]Build gdb/ctf.c failed

[Kai Tietz <ktietz70@googlemail.com>]

2013/3/25 Eli Zaretskii <eliz@gnu.org>:
>> Date: Mon, 25 Mar 2013 09:06:05 +0100
>> From: Kai Tietz <ktietz70@googlemail.com>
>> Cc: asmwarrior <asmwarrior@gmail.com>, tromey@redhat.com, yao@codesourcery.com,
>>       gdb-patches@sourceware.org
>>
>> 2013/3/25 Eli Zaretskii <eliz@gnu.org>:
>> >> Date: Mon, 25 Mar 2013 08:57:59 +0800
>> >> From: asmwarrior <asmwarrior@gmail.com>
>> >> CC: Eli Zaretskii <eliz@gnu.org>, Yao Qi <yao@codesourcery.com>,
>> >>  gdb-patches@sourceware.org
>> >>
>> >> I found that _mkdir was declared in the file: direct.h in MinGW-w64 SDKs:
>> >> _CRTIMP int __cdecl _mkdir(const char *_Path);
>> >
>> > Isn't _mkdir also declared in io.h?
>>
>> No, it isn't.
>
> Too bad.  Gratuitous differences between the different MinGW variants
> are likely to become maintenance headaches in the long run.  Like in
> this case.
>
>> It is a flaw to declare it there.  The unistd.h header is a POSIX
>> one.  _mkdir is for sure no POSIX variant, so its declaration
>> doesn't belong somewhere else.
>
> Posix header files can very well (and do) have non-Posix stuff, if
> that stuff is guarded by suitable preprocessor conditionals that make
> it disappear when compiled with the -std= compiler switch which
> requires Posix without extensions.  So I'm surprised this argument is
> being brought up here.

You are driveling ... use the right version of the API.  there is no
need to use MS private variant here at all.