Re: CrashReporter, signals and exception ports
Re: CrashReporter, signals and exception ports
- Subject: Re: CrashReporter, signals and exception ports
- From: Jean-Francois Im <email@hidden>
- Date: Mon, 09 Jan 2006 15:38:08 -0500
Steve Checkoway wrote:
We do this just with a signal handler. We catch:
static int signals[] =
{
SIGALRM, SIGBUS, SIGFPE, SIGHUP, SIGILL, SIGINT, SIGABRT,
SIGQUIT, SIGSEGV, SIGTRAP, SIGTERM, SIGVTALRM, SIGXCPU, SIGXFSZ,
#if defined(HAVE_DECL_SIGPWR) && HAVE_DECL_SIGPWR
SIGPWR,
#endif
-1
};
And then we set up the signal handlers using:
struct sigaction sa;
sa.sa_flags = 0;
if( p != NULL )
sa.sa_flags |= SA_ONSTACK;
sa.sa_flags |= SA_NODEFER;
sa.sa_flags |= SA_SIGINFO;
sigemptyset(&sa.sa_mask);
/* Set up our signal handlers. */
sa.sa_sigaction = SigHandler;
for( int i = 0; signals[i] != -1; ++i )
sigaction( signals[i], &sa, NULL );
where p is the pointer to our signal stack set up with sigaltstack.
Doing this we can shut down the full screen graphics context and
perform any other clean up. We actually give a stack trace (among other
things) ourself and setting the handlers seems to prevent the Crash
Reporter from running.
[snip]
We didn't have to do anything with exception ports to get this to work.
Aha! Reading the CrashReporter documentation seems to hint that setting
a signal handler using signal(3) doesn't work.
Thanks for the pointer, I'll investigate it.
Jean-Francois Im
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden