[mythtv] [BT] mythbackend segfaults

[Bruce Markey]

Isaac Richards wrote:
> On Wednesday 18 February 2004 04:27 am, Bruce Markey wrote:
> 
>>Isaac Richards wrote:
>>
>>>On Wednesday 18 February 2004 02:23 am, Bruce Markey wrote:
>>>
>>>>After months of stability, I've had several backend SEGVs
>>>>during recording on both the master and slaves. This started
>>>>about 24 hours ago and may have started after the last ffmpeg
>>>>resync. Before writing a restart script, I did a debug build
>>>>and grabed this backtrace. This seems to happen at a frequency
>>>>of about once per 1-2 hours.
>>>
>>>Saw something similar before I checked it in (would die in a matter of
>>>seconds), but I I had fixed the cause of that...   Could you try
>>>reproducing with current CVS?
>>
>>I have it installed now and added a few extra things to record
>>over night.
> 
> 
> You catch my second commit?  I think that that's a better fix, though the 
> first one should still mostly work..

Yes, I caught the NVR patch too.

> And just some more notes on it:  should have only affect mpeg4 encodes, and I 
> think it'd probably happen after roughly 35-40 minutes of a single recording.

Confirmed fix. It's easy to verify this now because the symptom
is so obvious. Recordings would end with the last recordedmarkup
flush for 35:47. I had seen this length a couple times and thought
it was a weird coincidence, problem with a cron job or system
resource limit. When I saw a file this length after a restart at
an odd time, I knew it wasn't cron. After I had applied your fixes
I viewed the truncated files and saw that every recording over a
half hour reported a length within 2 seconds of 35:47.

This doesn't happen after your commits so I guess I don't need
to recompile for optimization flags or write a restart script.
Maybe next time ;-).

Thanks,

--  bjm