On Wed, Apr 16, 2014 at 4:37 PM, Klaus Schmidinger Klaus.Schmidinger@tvdr.de wrote:

On 14.04.2014 21:58, Ville Skyttä wrote:

...

Hello,

I'm seeing an EIT related crash of VDR 2.0.6 every now and then, no idea if there's a way to trigger it on demand somehow nor do I know when exactly it happens; I just sometimes find a core file for it. gdb backtrace attached, my VDR is quite a patched one (the one from Fedora) but the line numbers appear to make sense for vanilla 2.0.6 too.

The only explanation I can think of is that some other thread is accessing the global 'Schedules' list after releasing its SchedulesLock. Are there any EPG related plugins in use?

Yes, epgsearch 1.0.1beta5 and epgfixer 0.3.1. FWIW, none of the patches I have touch eit*.[hc] or epg.[hc].

On Thu, Apr 17, 2014 at 10:56 AM, Klaus Schmidinger Klaus.Schmidinger@tvdr.de wrote:

On 17.04.2014 09:04, Ville Skyttä wrote:

...

On Wed, Apr 16, 2014 at 4:37 PM, Klaus Schmidinger Klaus.Schmidinger@tvdr.de wrote:

...

On 14.04.2014 21:58, Ville Skyttä wrote:

...

Hello,

I'm seeing an EIT related crash of VDR 2.0.6 every now and then, no idea if there's a way to trigger it on demand somehow nor do I know when exactly it happens; I just sometimes find a core file for it. gdb backtrace attached, my VDR is quite a patched one (the one from Fedora) but the line numbers appear to make sense for vanilla 2.0.6 too.

The only explanation I can think of is that some other thread is accessing the global 'Schedules' list after releasing its SchedulesLock. Are there any EPG related plugins in use?

Yes, epgsearch 1.0.1beta5 and epgfixer 0.3.1. FWIW, none of the patches I have touch eit*.[hc] or epg.[hc].

Well, maybe epgsearch or epgfixer is messing something up? Does the problem happen if these plugins are not in use?

It happens so rarely that it's most likely very hard to tell. I'll give it a shot without epgfixer but epgsearch is such an essential feature for this box that I don't think I'll be trying without it any time soon. I skimmed through both epgfixer and epgsearch sources quickly but didn't notice an obvious unlocked schedules access; it's entirely possible that I missed something though.

On 04/17/2014 08:06 PM, Matti Lehtimäki wrote:

On 04/17/2014 01:01 PM, Ville Skyttä wrote:

...

On Thu, Apr 17, 2014 at 10:56 AM, Klaus Schmidinger Klaus.Schmidinger@tvdr.de wrote:

...

Well, maybe epgsearch or epgfixer is messing something up? Does the problem happen if these plugins are not in use?

It happens so rarely that it's most likely very hard to tell. I'll give it a shot without epgfixer but epgsearch is such an essential feature for this box that I don't think I'll be trying without it any time soon. I skimmed through both epgfixer and epgsearch sources quickly but didn't notice an obvious unlocked schedules access; it's entirely possible that I missed something though.

I will go through the source of epgfixer thoroughly to see if any unlocked access to schedules can be found.

I could not find any unlocked use of schedules in epgfixer, so it should not be the cause of problems. There is actually only one situation where schedules are modified and that happens only if epg data is cloned.