possible open() problem again

[Adam Jackson]

On Thu, 2007-08-23 at 19:05 -0500, Josh Boyer wrote:
> On Thu, 23 Aug 2007 17:53:03 -0500
> Eric Sandeen <sandeen at redhat.com> wrote:
> > Alan Cox wrote:
> > > On Thu, Aug 23, 2007 at 05:05:53PM -0400, Jakub Jelinek wrote:
> > >> Well, ISO C doesn't cover fcntl.h, if it would, it would certainly not
> > >> be a perfectly legal C code, as per ISO C99, 7.1.3/1.
> > >> As fcntl.h is a POSIX header, you need to consider that standard
> > >> and that standard says that this code is not perfectly legal.
> > > 
> > > Yes.  And the ANSI C spec also makes structure assignment internally
> > > self-inconsistent, forbids using pointers out of the allocated range even
> > > for comparison and so on...
> > > 
> > > That doesn't create an excuse for doing these things, and doing them breaks
> > > lots of code and annoys people. You'll figure this out when you try and
> > > push it in a product people pay for, and also that just about everything
> > > the idiot macro hack does can be done by a 1 line of perl search of the
> > > source trees.
> > 
> > Hm, in fact, *if* this is only for the F8 devel phase (is it?) then
> > maybe an rpm macro in the prep phase to do that perl search would be
> > less painful...
> 
> I think this will show up in upstream glibc 2.7 when it is released.

That's sort of a tautology.  glibc releases when Fedora releases, at
this point.

- ajax