|
From: | Brent Fulgham |
Subject: | Possible -D_XOPEN_SOURCE=500 Discrepency |
Date: | Wed, 25 Jul 2001 15:36:26 -0700 |
The Hurd header files do not seem to accurately duplicate the behavior
of Linux when the "-D_XOPEN_SOURCE=500" compiler flag is used.
You can observe this when compiling the patch program (patch version
2.5.4). If _XOPEN_SOURCE is not defined, the "d_fileno" member of the
dirent structure is aliased as "d_ino" and patch will compile. But if
_XOPEN_SOURCE is set to 500, the Hurd library headers do not alias
d_fileno as d_ino, and the compile fails.
Not having the XOPEN specifications available, I was wondering if any
of the GLIBC hackers know what the proper behavior is. Based on a
review of the Linux and Hurd sources, the difference seems to be that
under Linux the dirent structure has a member named "d_ino" and
unconditionally aliases d_fileno to be d_ino:
#define d_fileno d_ino /* Backwards compatibility */
While under the Hurd, the opposite is done (the structure member
is named "d_fileno", and a conditional alias of d_ino is created
when XOPEN_SOURCE is not defined.
Does anyone know what the proper behavior is? Does the XOPEN
specification require a "real" "d_ino" to be present in dirent?
Or, should setting XOPEN_SOURCE=500 simply not prevent the #define
of d_ino to be an alias for d_ino?
Thanks,
-Brent
[Prev in Thread] | Current Thread | [Next in Thread] |