GNU bug report logs - #9077
coreutils-8.12: fiemap.h uses "struct fiemap_extent fm_extents[0];"

Previous Next

Full log

Message #22 received at 9077-done <at> debbugs.gnu.org (full text, mbox):

From: Pádraig Brady <P <at> draigBrady.com>
To: Paul Eggert <eggert <at> cs.ucla.edu>
Cc: 9077-done <at> debbugs.gnu.org, Joachim Schmitz <jojo <at> schmitz-digital.de>
Subject: Re: bug#9077: coreutils-8.12: fiemap.h uses "struct fiemap_extent
	fm_extents[0]; "
Date: Sat, 16 Jul 2011 09:37:38 +0100

On 15/07/11 23:45, Paul Eggert wrote:
> On 07/15/11 02:49, Pádraig Brady wrote:
>> +#ifdef __linux__
>> +  struct fiemap_extent fm_extents[0];
>> +#endif
> 
> This assumes that all GNU/Linux compiles will
> use a compiler that allows size-zero arrays;
> also, it assumes that sizeof (struct fiemap)
> == offsetof (struct fiemap, fm_extents).
> Although both assumptions are probably
> portable in practice in this particular case,
> it's better to avoid them, if avoiding them is easy.
> This will help future-proof the code if we alter the
> data structures, as altering them might cause problems
> such as what I recently encountered in GNU Emacs's C code.
> 
> How about the following further patch?  It should do the trick.

I tested that and it didn't work.
Even though the *fiemap structs are the same in mem
with and without the change, the ioctl() is returning ENOTTY
when the array is of size 1.  I guess this is because
the array element is unpacked for the ioctl which
is then failing kernel checks.

cheers,
Pádraig.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.