Re: AC_SYS_LARGEFILE and old OSX framework

[Diorcet Yann]

Le 30/10/2013 20:56, Paul Eggert a écrit :
> On 10/30/2013 12:46 PM, Diorcet Yann wrote:
>> Could be an explanation. but how do you explain the issue on that
>> https://gitorious.org/glib/android/source/003dbec6c84798cb5c6a8759258f2b7180e5fa10:glib/gkeyfile.c#L354-1871
>> S_ISREG is done just after fstat ... and there is #include "config.h".
> I don't know Darwin, but one possible explanation is that
> it's not just struct stat that's affected; it's also 'open'.
> At least, that's the case for the off_t stuff on many platforms,
> and maybe Apple's doing the same for the ino_t stuff.
>
> Anyway, I'd try building the entire application with
> -D_DARWIN_USE_64_BIT_INODE.  If that makes the problem go
> away we're probably on the right track.  And if not, perhaps
> it's not an Autoconf problem anyway.
Great,
I tried your idea ... same result.

So i take a look the documentation
https://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPages/man2/stat.2.html

which said that < 10.5 the flag _DARWIN_USE_64_BIT_INODE should do
nothing (keep 32)

I quote:
 "
  32-bit       32-bit inode values are enabled, and the legacy
structures involving the ino_t type
                        are in use.  The macro
_DARWIN_FEATURE_64_BIT_INODE is not defined.
 "

in the headers

/*
 * _DARWIN_FEATURE_64_BIT_INODE indicates that the ino_t type is 64-bit, and
 * structures modified for 64-bit inodes (like struct stat) will be used.
 */
#if __DARWIN_64_BIT_INO_T
#define _DARWIN_FEATURE_64_BIT_INODE            1
#endif


the __DARWIN_64_BIT_INO_T flag is controlled by _DARWIN_USE_64_BIT_INODE

Maybe there is an issue in the OSX SDK