[mythtv] Automatic backtrace on crash (crashhandler)?

[Brad Templeton]

On Fri, Feb 25, 2005 at 12:18:51PM -0500, Isaac Richards wrote:
> On Friday 25 February 2005 08:05 am, Thomas Börkel wrote:
> > HI!
> >
> > When mythbackend or mythfrontend crashes, one has to get the source,
> > compile with debug option, let it run in gdb and hope to be able to
> > reproduce the crash.
> >
> > When aMule crashes, it automatically spits out a backtrace into the
> > console.
> >
> > I wonder, if this could be implemented in myth, too? I guess, this will
> > cost some performance, but if it would be only a few percent, maybe it
> > would be worth it.
> 
> The main issue is that for normal operation, you really don't want to be 
> running in debug mode (all opts off, etc), but you really need to be in debug 
> mode to get a proper backtrace.  There's also the issue of if this would work 
> for all threads, etc..
> 
Optimizations sometimes cause backtrace errors, but most of the time
they don't, certainly not on the most basic questions like what routine
did a seg fault occur in.  (More often they get the line wrong within
a function.)   But in many bugs, just knowing the approximate area is
enough.  If it's not, you can just recompile without -O.

I was surprised to see how much performance difference there now is on
video playback between debug and release due to -O, and with that the case
you're not going to see users running debug by default in any event.
(Could, I suppose, have a debug mode that optimizes only the video
playback, OSD and other high-CPU code, which are all that really need it
anyway.)