long term bug advice

[PanGalactic]

I am currently running:

pulseaudio-0.9.21-1.fc12.x86_64

[atariman]

ok, i just found a link to fix the libpulse (pulseaudio) library...via removing it and installing something that works...I just installed the esound library, tested it with efence and it passed. gonna run the program and hope for the best...wish me luck.

[PanGalactic]

Good luck, atariman! :-)

[atariman]

Im back, unfortunately
so here goes.
after replacing the sound library I ran the program under efence in hopes of catching the bug...well much to my surprise is just didn't crash, it ran for three days and didn't crash. So I stopped the program and reran it outside of efence (recompiled with efence off) and once again it crashed...same errors as before with the backtrace. This is frustrating, I don't know if you have anymore suggestions but im plum stumped!

[PanGalactic]

That's sucks. I was hoping for the best. After a week went by, I assumed you had nailed it.

My only suggestion at this point is to compile with mudflap support and try again. It may catch something that efence missed. If that doesn't catch it, try valgrind.


@jsmith: do you have any ideas?

[jsmith]

I'm running out of ideas too. I'm not familiar with mudflap, but I've used valgrind.

[atariman]

yeah, ive used valgrind and removed all of the leaks detected that can be( meaning that the same result is given with a skeleton program with the libraries opened but no logic ) ill have to try mudflap see what goes on...id hate to port this over to java so late in the development this is really starting to get annoying. thanks for the help though guys ill get back to you.

[jsmith]

If you are getting a SIGABRT, then just for grins add a try-catch block around all of main:

int main() {
    try {
        /* your main() here
    } catch( const std::exception& ex ) {
        std::cout << "Caught exception: " << ex.what() << std::endl;
    } catch( ... ) {
        std::cout << "Caught through ellipsis" << std::endl;
    }
} */

Uncaught exceptions cause aborts.

[atariman]

it just an ABORTED message, but i suppose that might be worth a shot.

[jsmith]

Yeah, if it just says

Aborted

this means that the process caught SIGABRT. The default exception handler, called when an
uncaught exception is thrown, calls abort(), which does a raise( SIGABRT );

Maybe....

[atariman]

ok, well if you look at the backtrace, raise() is in whats veiwable from the backtrace, inside of libCoregl or something like that. Im getting slightly different backtrace than before. I should post it for you, but it's basically just a call to opengl lib and then raise() and then __kernel_vsyscall ().