[linux-dvb] AW: Re: popen for ac3 and thread (ringbuffer) problem

["Andreas Vitting" <Andreas@huji.de>]

Sorry, i forgot a little bit information...

This happens only during replaying something (with and w/o ac3). Even if the
"popen" function is executed/started. It is reproducible every time!! Notice
that it happens also with movies w/o ac3 data (only the pipe is opened, no
data is send). In this state, the ac3 program simply waits into the fread
function. If i execute the ac3 program paralell to vdr, no influence, no
errors. Also no errors if i wait for the end of the replayed stream do
nothing during the replay. Only with the OSD the crash occured (not at the
same time - but always after a few seconds).
I have this tested with #define DEBUG_OSD and the errors also gone away!!

Andreas

-----Ursprüngliche Nachricht-----
Von: linux-dvb-bounce@linuxtv.org
[mailto:linux-dvb-bounce@linuxtv.org]Im Auftrag von Carsten Koch
Gesendet: Mittwoch, 15. August 2001 12:23
An: Linux-Dvb
Betreff: [linux-dvb] Re: popen for ac3 and thread (ringbuffer) problem


Klaus Schmidinger wrote:
>
> Andreas Vitting wrote:
> >
> > Hi Klaus,
> > do you know about the problem with the popen function ? You opened it
into a
> > new Thread, so the thread is the main process for the ac3 decoder
(-shell).
> > All standard outputs going to the pipe, isn't it ??
> > Here is my problem:
> > if i open the osd menu (with and w/o ac3 data), vdr crashes and after a
> > while the dvb driver too (after 10-20 seconds). I have changed for
> > experimental purposes the popen command to the cDvbApi init routine
(near to
> > OstOpen) and know, you don't believe this,no crashes anymore.
> > This behavior is as soon as we start the vdr with the -a switch (it
doesn't
> > matter, if the recordings were with or w/o ac3 data)
>
> Can anybody else confirm this behaviour?

I can't.
I have vdr running with -a all the time and I am
of course using the OSD all the time. vdr does NOT
crash here 10-20 seconds after I bring up the OSD.

In fact, I am pleased with the recent improvements
in driver/firmware stability. I often see less than
one ARM reset per day now:

vdr:/var/log # fgrep ARM messages
Aug  4 13:14:00 vdr kernel: dvb: ARM RESET
Aug  7 18:03:54 vdr kernel: dvb: ARM RESET
Aug  7 18:47:37 vdr kernel: dvb: ARM RESET
Aug  7 20:11:50 vdr kernel: dvb: ARM RESET
Aug  8 18:12:41 vdr kernel: dvb: ARM RESET
Aug 11 01:23:56 vdr kernel: dvb: ARM RESET
Aug 14 14:11:22 vdr kernel: dvb: ARM RESET
Aug 15 04:16:46 vdr kernel: dvb: ARM RESET


> I like the idea that the AC3 decoder ist started only when an actual
> replay is performed, and not running all the time,

I agree.
I would even prefer to delay starting up the AC3 decoder
until the first actual AC3 data block is sent. That would
make the patch I posted in my AC3 howto superfluous and it
would further reduce resource usage.


Carsten.


--
Info:
To unsubscribe send a mail to listar@linuxtv.org with "unsubscribe
linux-dvb" as subject.



--
Info:
To unsubscribe send a mail to listar@linuxtv.org with "unsubscribe linux-dvb" as subject.