[mythtv] [BT] current-cvs frontend crash on recorded-shows screen
Isaac Richards
ijr at po.cwru.edu
Fri Jan 16 11:31:56 EST 2004
On Friday 16 January 2004 03:40 am, Isaac Richards wrote:
> On Friday 16 January 2004 03:10 am, Jason Hoos wrote:
> > His backtrace looks very similar to one I was trying to figure out a week
> > or two ago, and in my case it was caused at the end of a recording. I
> > noticed it after having recorded a one-minute test clip for other
> > reasons; if I left that clip selected on the playback screen, the
> > frontend would crash 4 out of 5 times while trying to loop from the end
> > to the beginning of the clip.
> >
> > I wasn't able to figure out the cause of the problem, but from what I did
> > see, it seemed like some thread deleted a buffer out from underneath the
> > 'cmmx_argb323' function in playbackbox.cpp (called from updateVideo line
> > 546 or so). The interesting thing that I noticed was this - right before
> > the "convert" call, I put in a printf that displayed the current values
> > of "frame" and "frame->buf". Right before the segfault occurs, the value
> > that got printed out for frame->buf matched what the backtrace said the
> > second argument to mmx_argb323 was. However, doing a 'print frame->buf'
> > at the GDB prompt after the segfault showed that frame->buf had been
> > reset to NULL.
> >
> > I just reran that clip and double-checked all this. I've attached
> > another backtrace including the aforementioned 'print' if you want to
> > look at it.
>
> Hmm.. I think I see the problem; I'll get a fix into CVS tomorrow. Should
> be pretty easy to slap a mutex around the critical section here. Too late
> now, though. =)
And, now it should be fixed in CVS. I'd appreciate it if you could test, as I
wasn't able to get it to die with a 30 second recording playing in a loop for
15 minutes or so even before I added the extra lock around things.
Isaac
More information about the mythtv-dev
mailing list