gback at cs.utah.edu
Wed Oct 21 11:16:45 PDT 1998
> I think in retrospect that we should put suspend/restore back in - in the name of
> compatibility. However I'm loathed to put the ThreadDeath exception back because
> it was truly aweful and difficult to port to native thread systems.
I did manage to port it to our own pthreads implementation. Then again,
its our own implementation (where we can ensure that things like
pthread_cancel and pthread_testcancel and pthread_setcanceltype are
implemented and work --- unlike 9 out of 10 pthreads implementations
out there.) I agree that it's likely different for other native thread
A configuration option (--with-threaddeath) might be an option, too.
Personally I think that Thread.destroy() in combination with a privileged
operation akin to pthread_canceldisable() (to temporarily hold off cancellation)
is the way to go, but the Java API doesn't allow for that.
More information about the kaffe