Re: [pulseaudio-discuss] Please help me to understand this

2011-04-13 Thread Daniel Mack
On Wed, Apr 13, 2011 at 6:03 PM, Claude Frantz
claudefrantzx...@lavabit.com wrote:
 Hello !

 While running a program, it entered a fatal error state. See about the
 backtrace when running under gdb.

 I'm not sure if this is an error in pulseaudio and if this bug is known.

AFAIK, this bug is not known, and I haven't seen any such effects yet.
Which program is that? Can you reproduce the bug?

To me, it smells like a memory corruption outside the PulseAudio code,
but that's just guessing. Could you try running the binary in
valgrind? Maybe that shows anything.

Daniel
___
pulseaudio-discuss mailing list
pulseaudio-discuss@mail.0pointer.de
https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss


Re: [pulseaudio-discuss] Please help me to understand this

2011-04-13 Thread Claude Frantz

On 04/13/2011 06:32 PM, Daniel Mack wrote:


AFAIK, this bug is not known, and I haven't seen any such effects yet.
Which program is that?


The program is wspr. It is started whith python and it includes some 
modules in FORTRAN and C.



Can you reproduce the bug?


Yes !


To me, it smells like a memory corruption outside the PulseAudio code,
but that's just guessing. Could you try running the binary in
valgrind? Maybe that shows anything.


I have tried it. You will find attached some extract from the output.

If you want, I can send you the full output.

Thanks for your help, Daniel !

Claude


==4461== HEAP SUMMARY:
==4461== in use at exit: 1,487 bytes in 47 blocks
==4461==   total heap usage: 77 allocs, 30 frees, 10,050 bytes allocated
==4461==
==4461== Searching for pointers to 47 not-freed blocks
==4461== Checked 312,572 bytes
==4461==
==4461== LEAK SUMMARY:
==4461==definitely lost: 0 bytes in 0 blocks
==4461==indirectly lost: 0 bytes in 0 blocks
==4461==  possibly lost: 0 bytes in 0 blocks
==4461==still reachable: 1,487 bytes in 47 blocks
==4461== suppressed: 0 bytes in 0 blocks

==4461== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 21 from 10)
--4461--
--4461-- used_suppression: 21 dl-hack3-cond-1
==4461==
==4461== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 21 from 10)

==4450== Thread 3:
==4450== Invalid write of size 2
==4450==at 0x5691BE: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x56BDE7: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x57536A: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x56AF49: Pa_WriteStream (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x884EA1C: soundout_ (sound.c:134)
==4450==by 0x885691D: tx_ (tx.f90:198)
==4450==by 0x458E98: start_thread (in /lib/libpthread-2.13.so)
==4450==by 0x397D2D: clone (in /lib/libc-2.13.so)
==4450==  Address 0x5f3d712 is 0 bytes after a block of size 8,194 alloc'd
==4450==at 0x400522F: calloc (vg_replace_malloc.c:418)
==4450==by 0x570569: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x575161: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x57533F: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x56AF49: Pa_WriteStream (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x884EA1C: soundout_ (sound.c:134)
==4450==by 0x885691D: tx_ (tx.f90:198)
==4450==by 0x458E98: start_thread (in /lib/libpthread-2.13.so)
==4450==by 0x397D2D: clone (in /lib/libc-2.13.so)
==4450==
==4450== Invalid read of size 1
==4450==at 0x4007CE8: memcpy (mc_replace_strmem.c:602)
==4450==by 0x6BE9328: pa_stream_write (string3.h:52)
==4450==by 0x699BC64: ??? (in 
/usr/lib/alsa-lib/libasound_module_pcm_pulse.so)

==4450==by 0x6B1EB91: ??? (in /lib/libasound.so.2.0.0)
==4450==by 0x6ADB2E7: ??? (in /lib/libasound.so.2.0.0)
==4450==by 0x6B1EF96: ??? (in /lib/libasound.so.2.0.0)
==4450==by 0x6AD3AF4: snd_pcm_writei (in /lib/libasound.so.2.0.0)
==4450==by 0x57043D: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x5709C7: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x57537C: ??? (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x56AF49: Pa_WriteStream (in /usr/lib/libportaudio.so.2.0.0)
==4450==by 0x884EA1C: soundout_ (sound.c:134)
==4450==  Address 0x5f4570f is 1 bytes before a block of size 48 free'd
==4450==at 0x4005B0A: free (vg_replace_malloc.c:325)
==4450==by 0x11F2AD: TclpFree (in /usr/lib/libtcl8.5.so)
==4450==by 0x129E0D: Tcl_Free (in /usr/lib/libtcl8.5.so)
==4450==by 0x1E11DD: TclBN_mp_clear (in /usr/lib/libtcl8.5.so)
==4450==by 0x1E1516: TclBN_mp_div (in /usr/lib/libtcl8.5.so)
==4450==by 0x1C1D2D: ??? (in /usr/lib/libtcl8.5.so)
==4450==by 0x1CE72A: Tcl_PrintDouble (in /usr/lib/libtcl8.5.so)
==4450==by 0x1A7182: ??? (in /usr/lib/libtcl8.5.so)
==4450==by 0x1A7E1E: Tcl_GetString (in /usr/lib/libtcl8.5.so)
==4450==by 0x592CF80: ??? (in /usr/lib/libtk8.5.so)
==4450==by 0x59225B2: ??? (in /usr/lib/libtk8.5.so)
==4450==by 0x125764: ??? (in /usr/lib/libtcl8.5.so)
==4450==

___
pulseaudio-discuss mailing list
pulseaudio-discuss@mail.0pointer.de
https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss


Re: [pulseaudio-discuss] Please help me to understand this

2011-04-13 Thread Daniel Mack
On Wed, Apr 13, 2011 at 7:15 PM, Claude Frantz
claudefrantzx...@lavabit.com wrote:
 On 04/13/2011 06:32 PM, Daniel Mack wrote:

 AFAIK, this bug is not known, and I haven't seen any such effects yet.
 Which program is that?

 The program is wspr. It is started whith python and it includes some
 modules in FORTRAN and C.

 Can you reproduce the bug?

 Yes !

 To me, it smells like a memory corruption outside the PulseAudio code,
 but that's just guessing. Could you try running the binary in
 valgrind? Maybe that shows anything.

 I have tried it. You will find attached some extract from the output.

The invalid write sounds suspicious. I assume the test binaries from
the PulseAudio source tree work fine for you?

I think you need to put some more efford in debugging this. IOW, go
ahead and build a minimal PortAudio tool and see if it also shows the
bug. Or go the other way around and strip down the program you use and
see if you can get it to the point where it doesn't crash anymore.

Sorry I can't provide more information at this point, but please keep
us posted about anything you find.

Daniel
___
pulseaudio-discuss mailing list
pulseaudio-discuss@mail.0pointer.de
https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss


Re: [pulseaudio-discuss] Please help me to understand this

2011-04-13 Thread Claude Frantz

On 04/13/2011 07:40 PM, Daniel Mack wrote:


The invalid write sounds suspicious. I assume the test binaries from
the PulseAudio source tree work fine for you?


I'm using Fedora 14, all available upgrades installed. The version is 
reported as pulseaudio-0.9.21-7.fc14.i686



I think you need to put some more efford in debugging this. IOW, go
ahead and build a minimal PortAudio tool and see if it also shows the
bug.


My knowledge about pulseaudio is too limited.


Or go the other way around and strip down the program you use and
see if you can get it to the point where it doesn't crash anymore.


The same problem here. I do not really understand wspr's structure and 
operating mode.



Sorry I can't provide more information at this point,


Many thanks for your help, anyway.


but please keep
us posted about anything you find.


OK ! I will do it.

Claude

___
pulseaudio-discuss mailing list
pulseaudio-discuss@mail.0pointer.de
https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss