faad segfaults on amd64 due to incorrect pointer size

[Rocko]

When I run faad on amd64 with the command line "faad -b 2 -f 2 -q -w AAC2AC3Temp.aac" (-b 2 indicating 24 bit output), I get a segmentation fault in fwrite. The same command on the same data works fine in a 32 bit system.

The problem is due to the write_audio_24bit() function in frontend/audio.c, which uses a long* pointer (ie which is 8 bytes on amd64 but only 4 on x86):

static int write_audio_24bit(audio_file *aufile, void *sample_buffer,
                             unsigned int samples)
{
    int ret;
    unsigned int i;
    long *sample_buffer24 = (long*)sample_buffer;
    char *data = malloc(samples*aufile->bits_per_sample*sizeof(char)/8);

Changing the definition of sample_buffer24 to make it an int* fixes the problem.

The function write_audio_32bit needs a similar change and stops the segfault happening if you use "-b 3" for 32 bit samples. (There is also similar code in aacDECdrop/audio.c but I'm not sure where that is used.)

ProblemType: Bug
DistroRelease: Ubuntu 10.10
Package: faad 2.7-4
Uname: Linux 2.6.36-iwlwifi-unify-scan x86_64
NonfreeKernelModules: nvidia
Architecture: amd64
CheckboxSubmission: 1bd8e90541d49b96c13cbfcc9baf103b
CheckboxSystem: d00f84de8a555815fa1c4660280da308
Date: Sun Oct 24 15:21:17 2010
InstallationMedia: Ubuntu 10.04 "Lucid Lynx" - Alpha amd64 (20100224.1)
ProcEnviron:
 LANG=en_AU.utf8
 SHELL=/bin/bash
SourcePackage: faad2

[Rocko (rockorequin)]

This patch (against faad 2.7 from sourceforge) stops the segfault. I tested it works on a 32 bit platform as well (as it should since int pointers are 4 bytes long on both LP64 and ILP32 platforms).

[Launchpad Janitor (janitor)]

This bug was fixed in the package faad2 - 2.7-6ubuntu1

---------------
faad2 (2.7-6ubuntu1) natty; urgency=low

  [ Angel Abad <email address hidden> ]
  * Merge from debian unstable (LP: #680474). Remaining changes:
    - debian/patches/fix_ftbfs_with_gcc4.5.patch
      + Correctly declare lrintf in libfaad/common.h to avoid a conflict of
        declaration in mathcalls.h to fix FTBFS on i386

  [ Bhavani Shankar <email address hidden> ]
  * Remove useless chunk from the patch I created above to fix FTBFS
    which happened due to my mistake. Sigh!

faad2 (2.7-6) unstable; urgency=high

  [ Alessio Treglia ]
  * Fix segmentation fault in faad due to an incorrect pointer size
    (Closes: #603807, LP: #665802).
  * Add gbp config file.

  [ Andres Mejia ]
  * Revert changes in 2.7-5. Upload of 2.7-5 was unintentionally done.
  * Refresh patches.
 -- Angel Abad <email address hidden> Tue, 23 Nov 2010 13:48:21 +0100

[Tomas Herfert (herfik)]

Hello,
it seems the same bug is in Ubuntu 10.04 (LTS).
Would it be possible to fix it also there?
Or should I create a new bug?

Thanks

[Rolf Leggewie (r0lf)]

lucid has seen the end of its life and is no longer receiving any updates. Marking the lucid task for this ticket as "Won't Fix".