Re: Mystery threads eating CPU cycles

FROM : Douglas Davidson
DATE : Mon Nov 11 19:57:25 2002

On Monday, November 11, 2002, at 10:31  AM, Mike Laster wrote:

> I have a server process that after about a day of running starts
> eating about 50% of the CPU cycles, even though it should be
> completely idle (no clients talking to it).  If I run sample on it, I
> see a couple of threads running that appear to be the source of the
> problem.  However, these aren't threads that I
> have created.  From what I've seen, any thread started with NSThread
> has their stack start with pthread_body(), and these don't:
>
>    1000 Thread_1103
>      1000 NSExceptionHandlerUncaughtSignalHandler
>        1000 _sigtramp
>          1000 NSExceptionHandlerUncaughtSignalHandler
>            1000 NSExceptionHandlerUncaughtSignalHandler [STACK TOP]
>    1000 Thread_1203
>      1000 malloc_zone_from_ptr
>        1000 0x33d0
>          1000 _sigtramp
>            1000 NSExceptionHandlerUncaughtSignalHandler
>              1000 NSExceptionHandlerUncaughtSignalHandler [STACK TOP]
>
>
> I'm linked into ExceptionHandling.framework which is supposed to catch
> system signals and turn them into NSExceptions.  It almost looks like
> NSExceptionHandlerUncaughtSignalHandler is getting a signal while
> handling one, and going into an infinite loop.  Is this a bug in
> ExceptionHandling.framework?

Probably so, but the original signal that started this probably would
otherwise have been fatal.  It seems likely that the odd stack traces
you are getting from sample are confused because of the signals; these
probably are ordinary threads, just with incomplete backtraces.  Try
turning off ExceptionHandling signal handling, running in gdb, and
trying to locate the source of the original signal.

Douglas Davidson
_______________________________________________
cocoa-dev mailing list | <email_removed>
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/cocoa-dev
Do not post admin requests to the list. They will be ignored.