Re: [orca-list] got loss of speech after exiting firefox

[Jason White <jason jasonjgw net>]

Steve Holmes <steve holmes88 gmail com> wrote:
 
> Oh, I agree with you fully.  The hassle here is it doesn't happen
> every time so it is nearly impossible to set debug logging up to try
> and catch the hang whenever it decides to show up.  To file a bug, you
> need to have repeatable and reproducable events to track.  

I suppose you could turn on debug logging and leave it on by default until the
next hang/crash. You could have the debug file written to /tmp where
it will be deleted if ythe system is rebooted, and just periodically delete it
and restart Orca from time to time to save disk space.

gdb is really great for debugging C code; it can be attached to running
processes so that if something hangs, you can still debug it. However, I don't
know whether there is a good debugger for Python code that can attach to the
interpreter process and extract useful details. My understanding from a
discussion elsewhere is that the Java virtual machine has one, but that
doesn't help us here - it just exemplifies what I'm thinking of as a solution.

If there's a segmentation fault you can turn on core dumps. Usually, though,
Python code exits with a stack trace. If it's a hang rather than a crash, the
debug log might help.

There are also tracing mechanisms being developed for Linux that might, in
future, simplify this still further, but they don't support userspace tracing
yet - just kernel-level tracing.