Bug with pthread_getspecific() and signals

[Archie Cobbs]

Daniel Eischen wrote:
> I don't think libc_r differentiates between synchronous
> and asynchronous signals.  It'll look for threads in
> sigwait(), then sigsuspend(), then any thread with the
> signal unmasked.

Well that would certainly explain my problem. By the way, this application
is a Java virtual machine, trying to use signals to detect null pointer
dereferences (and divide by zero). How is one supposed to do this if
synchronous signals are not delivered to the thread that generated them?
It then becomes impossible to detect in which thread the signal occurred.

Is this a bug in libc_r? Does POSIX not specify this "same thread" behavior
for synchronous signals?

If it does, can you show me in which FreeBSD releases the bug is fixed?

If not, how do other applications implement this??

> Try -stable or -current with libpthread and libthr.

Unfortunately I don't have easy access to either such machine.
I've been told that the same problem occurs with libthr though.

In any case, if the thread library is not always sending synchronous
signals to the generating thread then no further research is necessary.

Thanks,
-Archie

__________________________________________________________________________
Archie Cobbs      *        CTO, Awarix        *      http://www.awarix.com