[vdr] Re: Autopid 1.1.29-2 Issue#2

[Andreas Schultz <aschultz@warp10.net>]

On Saturday 03 May 2003 13:12, Klaus Schmidinger wrote:

[...]

> Just wondering: this apparently works fine in the original VDR (at least
> a quick test here didn't cause any problems), so why would a fix need
> "some major changes to the schedules handling in cMenuSchedule"?

It's a deadlock on the schedulesMutex. cMenuSchedule locks that Mutex and only 
releases it once destroyed. For the original VDR that is completly harmless. 
The autopid patch needs to scan the PMT/SDT during channel switch. This 
scanning currently happens in cSIProcessor::Action() which also tries/needs 
to lock the schedulesMutex causing the deadlock.
Avoiding the lock in cMenuSchedules would certainly require changes that IMHO 
could be considerd major. Using two different looks would be the alternative, 
but this would probably have en even bigger impact. Currently, i try to 
minimize the impact of the patch on vdr and move stuff more in a direction 
that you might eventualy approve of. This might even mean that all of the 
scanning stuff will be removed from eit.c again and only a SectionsScanner 
interface will remain. This should solve the deadlock without any changes to 
cMenuSchedule eventualy.

Andreas


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.