[vlc-devel] Tracking "unhandled vlc exception"

[David Glaude]

This message is "for the record" and to increase the audience since it has
been discussed a bit on IRC.

Any help or advice on how to track "unhandled vlc exception" under windows
is welcome.

Problem summary:
With "1.1.0-pre3-20100423-0202" started from msys under Win32 on a multicast
MPEG4 stream described by an SDP file we get random "unhandled vlc
exception".
This seems to be a recent problem as we tested for days with vlc 1.0.5
without problem and with various nightly build of 1.1.0.
We now would like to identify the root cause and/or the release where this
problem started.

I started VLC with:
$ "C:/Program Files/VideoLAN/VLC/VLC.exe" -vvv --verbose=2 --rtsp-caching=40
--clock-jitter=0 rtsp://136.173.110.16/Test &> logs5.txt

Here is an extract from the timestamped log:
09:23:22:960000000 [00474398] main libvlc debug: VLC media player -
1.1.0-pre3-20100423-0202 The Luggage
[...]
09:37:05:188000000 [00e5d5d8] main input debug: Decoder buffering done in 1
ms
09:37:05:220000000 [01655f68] main video output warning: late picture
skipped (20000 > -187)
09:42:59:961000000 unhandled vlc exception
09:42:59:992000000 QObject::killTimers: timers cannot be stopped from
another thread

So after less than 20 minutes of playing the stream VLC did crash on an
"unhandled vlc exception".
There is no event in the log just before that could help identify where the
problem is located.

Same problem a bit later, but this time I saved the "crashdump".
It is exactly the same multicast stream except that this time the SDP file
was "local" (by the way --rtsp-caching=xx is not used in this case).

I started VLC with:
$ "C:/Program Files/VideoLAN/VLC/VLC.exe" -vvv --verbose=2 --rtsp-caching=40
--clock-jitter=0 MCAST.sdp &> logs6.txt

Here is an extract from the timestamped log:
10:21:48:339000000 [0047a428] main libvlc debug: VLC media player -
1.1.0-pre3-20100423-0202 The Luggage
[...]
11:14:43:052000000 [01123e08] directx generic debug: DirectXUpdateRects
image_src_clipped coords: 0,0,352,288
11:19:35:607000000 [0233b258] main video output warning: late picture
skipped (8260 > -192)
11:55:59:707000000 unhandled vlc exception
11:55:59:816000000 QObject::killTimers: timers cannot be stopped from
another thread

So it crashed after a arround one hour and a half without any indication on
the root cause.

We have run various version of VLC on this kind of stream and it is the
first version where we detect this kind of behaviour.
It is either related to this particular version of VLC or due to the fact
that we run from within Msys (to gather the log).

Can you (or we) do something with the crashdump?

Except if we get any other advice, we will try to run the debug version of
vlc nightly build with gdb and multiply the number of VLC version to test.

Thanks

David Glaude
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20100425/cc7db934/attachment.html>