[MPlayer-users] Re: BUG - Garbled output playing realaudio
Bryan Alton
balton at eircom.net
Sat Oct 15 01:07:33 CEST 2005
Attached is a new patch for demux_real.c. It needs testing by other users
before I'd be happy to propose as a patch as it is addressing error
conditions which are hard to simulate.
Unlike the previous patch, this patch insert packets whenever packet loss
has been detected. In the case of gaps longer than a block (typically more
than 16 packets or 1.8sec) the code will insert the minimum number of
packet to maintain audio.
If large gaps of 15-25 secs occur, then the cache is likely to be depleted
by the time packets are inserted and the audio stream will break up. In
these circumstance the patch will close Mplayer down with an EOF if the
cache stays at 0% for about 10 seconds. This will force users to restart
rather than wait and hope that the stream may build up the cache and
stabilise which can't happen with a real-time live stream.
Mplayer's behaviour is now comparable with RealPlayer and in the case of
Mplayer "Pause" of more than 30 secs. - I think it is more stable.
I have tested on Linux (SuSe 9.3) on 10 live BBC streams and one german
station. At the end of testing I had one incident where MPlayer
inexplicably closed down. Shortly before all BBC streams stopped for about
2 minutes, couldn't be accessed from 2 different machines so I think there
was either network or system problems which may have had subsequent side
effects. The testing of gaps around 4-5 seconds has not been thorough
because it is hard to recreate the problem.
The testing that needs to done:
1. Windows version
2. verify that patch works for gaps typically 3-5 seconds. Ensure sound is
OK and cache does not gradually deplete.
3. Check other RealAudio feeds are unaffected.
4. Check full operation when used with other systems such as Squeezebox.
Functions such as stopping / starting / changing stations and into/out of
standby.
5. Test other functions when used with other systems such as Pause, FFWD,
REW - although these shouldn't work, the test is needed to ensure that
Slimserver & Squeezebox are not left in a bad state and no zombie process
are left behind.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: demux_real.c.diff
Type: text/x-diff
Size: 9610 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-users/attachments/20051015/39f81a3a/attachment.diff>
More information about the MPlayer-users
mailing list