Hi,
Michael Haubenwallner wrote:
> How do gnulib's file-functions respect the different sets of
> large-file-defines among several platforms in general?
The autoconf macro AC_SYS_LARGEFILE does it all. It defines _FILE_OFFSET_BITS,
_LARGE_FILES, or whatever system-dependent magic is needed to get
On Mon, 2008-09-29 at 10:01 +0200, Michael Haubenwallner wrote:
> On Sun, 2008-09-28 at 14:59 +0200, Bruno Haible wrote:
> > A followup to this override of open() and fopen():
> >
> >
> > 2008-09-28 Bruno Haible <[EMAIL PROTECTED]>
>
> > * lib/stdio.in.h (fopen, freopen): Undefine before
On Sun, 2008-09-28 at 14:59 +0200, Bruno Haible wrote:
> A followup to this override of open() and fopen():
>
>
> 2008-09-28 Bruno Haible <[EMAIL PROTECTED]>
> * lib/stdio.in.h (fopen, freopen): Undefine before redefining. Needed
> with AIX xlc.
Maybe I should just try more hard
A followup to this override of open() and fopen():
2008-09-28 Bruno Haible <[EMAIL PROTECTED]>
* doc/posix-functions/freopen.texi: Mention the trailing slash problem.
* lib/stdio.in.h (fopen, freopen): Undefine before redefining. Needed
with AIX xlc.
* lib/fcnt
Paolo Bonzini gnu.org> writes:
>
>
> > ! fd = open (filename, O_RDONLY);
> > ! if (fd < 0)
> > return NULL;
> >
> > ! if (fstat (fd, &statbuf) >= 0 && !S_ISDIR (statbuf.st_mode))
> > ! {
> > ! errno = ENOTDIR;
> > ! return NULL;
> > ! }
>
> You're leaking fd
> ! fd = open (filename, O_RDONLY);
> ! if (fd < 0)
> return NULL;
>
> ! if (fstat (fd, &statbuf) >= 0 && !S_ISDIR (statbuf.st_mode))
> ! {
> ! errno = ENOTDIR;
> ! return NULL;
> ! }
You're leaking fd here. Same for open().
Paolo
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Eric Blake on 9/24/2008 6:51 AM:
> Thanks. However, right now, if both fopen-safer and fopen are in use,
> then it looks like fopen_safer is used without also picking up rpl_fopen.
> I'm still looking into the best way to make these modu
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 9/24/2008 5:50 AM:
> The bug appears on the same platforms, namely HP-UX and Solaris 9. Therefore
> for the moment an extra autoconf check is not needed. But I agree that it
> should be worked around. I'm applying the patch
Eric Blake wrote:
> > * lib/fopen.c (rpl_fopen): Return NULL if the mode specifies write
> > access and the filename ends in a slash. Code copied from lib/open.c.
>
> Should we also check for the bug, present in at least Irix 5.3, where
> fopen("/dev/null/","r") succeeds instead of failing
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
According to Bruno Haible on 8/30/2008 6:39 AM:
> Hi,
>
> The documented open() bug on Solaris [7-9] and HP-UX 11.00 exists also as a
> bug in fopen(). I'm applying this to fix it:
> * lib/fopen.c (rpl_fopen): Return NULL if the mode specifies
Hi,
The documented open() bug on Solaris [7-9] and HP-UX 11.00 exists also as a
bug in fopen(). I'm applying this to fix it:
2008-08-30 Bruno Haible <[EMAIL PROTECTED]>
* m4/fopen.m4 (gl_FUNC_FOPEN): Test against bug with trailing slash.
Code copied from m4/open.m4.
*
11 matches
Mail list logo
