Bug#687182: program seg faults on armhs

[Armin Wagner]

Package: mpg321
Version: 0.2.13

The program crashed with a seg fault on armhs.

$ mpg321 --verbose audiotest_scelsi.mp3
[...]
Playing MPEG stream from audiotest_scelsi.mp3 ...
MPEG 1.0, Layer: III, Freq: 44100, mode: Stereo, modext: 0, BPF : 4480
Channels: 2, copyright: No, original: Yes, CRC: No, emphasis: 0.
Bitrate: 224 Kbits/s, Extension value: 0
Audio: 1:1 conversion, rate: 44100, encoding: signed 16 bit, channels: 2
Segmentation fault4], Time: 00:00.02 [03:11.59],

I'm running Ubuntu 12.04 with a PIO_ONLY kernel on an ARM hard-float device (beabglebone) with an attached USB-soundcard (Z305). I'm aware that the problem might be the result of my non-standard configuration and not of mpg321 itself.

However, I have to note that mp3 playback generally works with mpg123. I tested mpg321 version 0.2.13 and version 0.3.2. Both are segfaulting.

[Nanakos Chrysostomos]

Hi,
can you please be more verbose?
Try to attach an strace log of the output please.

Regards,
Chris

Post by Armin Wagner
Package: mpg321
Version: 0.2.13
The program crashed with a seg fault on armhs.
$ mpg321 --verbose audiotest_scelsi.mp3
[...]
Playing MPEG stream from audiotest_scelsi.mp3 ...
MPEG 1.0, Layer: III, Freq: 44100, mode: Stereo, modext: 0, BPF : 4480
Channels: 2, copyright: No, original: Yes, CRC: No, emphasis: 0.
Bitrate: 224 Kbits/s, Extension value: 0
Audio: 1:1 conversion, rate: 44100, encoding: signed 16 bit, channels: 2
Segmentation fault4], Time: 00:00.02 [03:11.59],
I'm running Ubuntu 12.04 with a PIO_ONLY kernel on an ARM hard-float device (beabglebone) with an attached USB-soundcard (Z305). I'm aware that the problem might be the result of my non-standard configuration and not of mpg321 itself.
However, I have to note that mp3 playback generally works with mpg123. I tested mpg321 version 0.2.13 and version 0.3.2. Both are segfaulting.

[Nanakos Chrysostomos]

Hello,
of course it helps. The problem seems to be the libmad library.
Can you please try madplay and if the same thing happens then surely
the problem is libmad. I don't have an armhs machine to make some
tests.

Hello,
Thank you for the fast reply. I'm not very experienced in debugging on
$ trace -o trace.txt mpg321 audiotest_scelsi.mp3
$ gdb 'mpg321' core
[�]
Reading symbols from /usr/bin/mpg321...(no debugging symbols
found)...done.
[New LWP 1175]
[Thread debugging using libthread_db enabled]
Using host libthread_db library
"/lib/arm-linux-gnueabihf/libthread_db.so.1".
Core was generated by `mpg321 audiotest_scelsi.mp3'.
Program terminated with signal 11, Segmentation fault.
#0 0x4005a760 in mad_layer_III () from
/usr/lib/arm-linux-gnueabihf/libmad.so.0

disass 0x4005a760

$ sudo apt-cache show libmad0
Package: libmad0
Priority: optional
Section: libs
Installed-Size: 126
Original-Maintainer: Mad Maintainers
Architecture: armhf
Source: libmad
Version: 0.15.1b-7ubuntu1
Depends: libc6 (>= 2.4)
Pre-Depends: multiarch-support
Filename: pool/main/libm/libmad/libmad0_0.15.1b-7ubuntu1_armhf.deb
If the problem comes from libmad it would explain why mpg123 (mpglib) is
working okay.
https://bugs.launchpad.net/ubuntu/+source/libmad/+bug/989846
Regards,
Armin

Hi,
can you please be more verbose?
Try to attach an strace log of the output please.
Regards,
Chris

Post by Armin Wagner
Package: mpg321
Version: 0.2.13
The program crashed with a seg fault on armhs.
$ mpg321 --verbose audiotest_scelsi.mp3
[...]
Playing MPEG stream from audiotest_scelsi.mp3 ...
MPEG 1.0, Layer: III, Freq: 44100, mode: Stereo, modext: 0, BPF : 448

0

Post by Armin Wagner
Channels: 2, copyright: No, original: Yes, CRC: No, emphasis: 0.
Bitrate: 224 Kbits/s, Extension value: 0
Audio: 1:1 conversion, rate: 44100, encoding: signed 16 bit, channels: 2
Segmentation fault4], Time: 00:00.02 [03:11.59],
I'm running Ubuntu 12.04 with a PIO_ONLY kernel on an ARM hard-float
device (beabglebone) with an attached USB-soundcard (Z305). I'm aware
that the problem might be the result of my non-standard configuration
and not of mpg321 itself.
However, I have to note that mp3 playback generally works with mpg123.
I tested mpg321 version 0.2.13 and version 0.3.2. Both are segfaulting.

[наб]

Control: reassign -1 libmad0

After compiling libmad 0.15.1b myself everything works again.
The necessary steps are documented here: https://igw.tuwien.ac.at/ceat/node/2#comment-162

The site is dead so inlining
-- >8 --
compiling libmad and mpg321 (because of seg fault)

Unfortunately my favorite player (which I'm using as a workaround for the missing mp3 support in Py3k) keeps crashing with a seg fault (see debian bug #687182). Strace showed me that libmad might be the culprit here (compare: Ubuntu bug #989846 ) so I decided to deinstall all related packages and compile the 0.15.1b version myself. Some notes:

cc1: error: unrecognized command line option '-fforce-mem'

fforce-mem is a deprecated gcc optimization option (see: gcc 4.3 changes). I had to remove the corresponding line (optimize="$optimize -fforce-addr) in the configure file.

A few lines above the architecture gets selected. Here I inserted the following line:

armv7*) arch="-marm" ;;

If you don't do this you will get an "selected processor does not support Thumb mode" error (thanks to Nikolai for this hint) Afterwards I also used the configure --disable-aso parameter 
 just to be on the save side.

After successfully compiling libmad I compiled madplay. The "error: id3tag.h was not found" error made me install the libid3tag0-dev package.

And the mpg321 configuration stopped with the following message:

checking for snd_mixer_selem_get_playback_volume in -lasound... no
configure: error: "ALSA lib is required to compile mpg321."

Of course alsalib is installed, but the dev package is missing: sudo apt-get install libasound2-dev

Testing mpg321 brought up the following error: "ALSA lib confmisc.c:768:(parse_card) cannot find card '0'". I have to specify the audio device. aplay -l helps here:

**** List of PLAYBACK Hardware Devices ****
card 1: Z305 [ Logitech Z305], device 0: USB Audio [USB Audio]

mpg321 -o alsa -a hw:1,0 scelsi.mp3

It works. Finally.
-- >8 --

So this a libmad ARM baseline issue on the submitter's Beagleboard
(this is a Beagleboard forum).