From: Thomas Schwinge <tschwinge@gnu.org>
Date: Sat, 28 Jul 2007 18:25:42 GMT
Subject: Re: [PATCH] Don't rely on unspecified behavior
Message-ID: <20070728182542.GA22651@fencepost.gnu.org>
In-Reply-To: <20070728173948.GD23337@cip.informatik.uni-erlangen.de>

Hello!

On Sat, Jul 28, 2007 at 07:39:48PM +0200, Thomas Glanzmann wrote:
> > Calling access(p, m) with p == NULL is not specified, so don't do
> > that.  On GNU/Hurd systems doing so will result in an SIGSEGV.
> 
> a friend of mine choked on this one when tried git for the second time
> (the first time "git-repack -a -d -f" screwed his repository after the
> initial checkout. This is fixed for a long time). Lucky me that he had
> his libusbdriver in LD_PRELOAD which could not handle the NULL argument.
> And I always thought libc would make the check before it does the system
> call or does GNU/hurts not use the gnu libc?

GNU/Hurd systems do (obviously ;-) use the GNU libc.  The glibc
maintainer Roland McGrath explicitly told me that ``access (NULL, m)''
shall not be caught as it is not specified and thus must not be invoked
like this.


I noticed that the patch I sent was prepared for an old version of the
file.  I'll send an updated patch that applies to the current revision.


Regards,
 Thomas