[mythtv] Frontend dying on aspect ratio change

[Ed Wildgoose]

Could someone please at least download the file and let me know if they 
can get a good backtrace from it?

I updated my threads library to NPTL and rebuild a few things, but I'm 
still having problems getting a bt.  At least now it does segfault every 
time though...

I think this is also related to the problem reported on 17/3/03 by Jon 
Whitear, titled "Mythfrontend dies" - the link below does at least have 
a file which reliably reproduces the problem

Thanks

Ed W


> The following snippet from one of my recordings *usually* kills the 
> frontend at the point the aspect ratio changes from 16:9 to 4:3 (just 
> as Friends starts)
>
> http://www.wildgooses.com/downloads/duff.nuv   (around 7Mb and FWIW: 
> it came from a dvb-t device)
>
> I'm having difficulty debugging it though.  It doesn't *always* die  
> for starters (but > 50% of the time it does).  Also, gdb is acting 
> rather wierd for me, "thread apply all bt full" is just doing nothing 
> (although "bt full" still seems to do something) - I have tried 
> updating to gdb 6 and a newer 2.6.5 kernel, but no change..?  I also 
> can't see how to get gdb to stop just before the app dies, and so I 
> can't see how to find out which line it is actually excuting on?  
> (It's not dying like a segfault would) Finally, whats with all the 
> Sig32 signals in the most recent cvs version? Is "handle SIG32 nostop" 
> the way everyone else supresses them?
>
> As near as I can tell though, it's probably related to the "unknown 
> conversion" error messages coming back (which seem to be in the OSD 
> code), and the "GetNextFreeFrame() served a busy frame" messages.  
> However, these don't reliably turn up every time.  Could it in fact be 
> a chunk of code which isn't handling SIG32 signals though (hence the 
> reason it doesn't happen everytime, bit of a race condition)?
>
> Any help appreciated
>
> Ed W