Hi all once again,
thank you for your comments.
There was no OSS bug in my case!
The main problem was that our GUI app which is responsible for starting
and killing several codecs did not wait for the processes being
effectively killed.
So my codec fell asleep until /dev/dsp's buffer was played and died
honestly thereafter.
I was able to improve my codec's 'realtime reactions' by explicitly
setting the fragment size (thank you, Iain).
But as mentioned above this was not the main mistake.
By the way: Since none of our codecs catches any signals it doesn't make
any difference whether they are killed by SIGKILL or SIGTERM.
Jan
--
Jan Weil
Heinrich-Hertz-Institut f�r Nachrichtentechnik Berlin GmbH
Dept. Image Processing - Embedded Multimedia Group
Phone: +49 30 31002-876
