On Wed, 21 Feb 2007, William James wrote:
On 2/21/07, James Carlson <[EMAIL PROTECTED]> wrote:
William James writes:
> On 2/21/07, Vincenzo Sciarra <[EMAIL PROTECTED]> wrote:
> > All threads receive the signal. Better, the main thread receive the
signal but it shares the signal with other threads.
> >
> >
> >
> > Dott. Vincenzo Sciarra
>
> So all threads call the signal callback at the same time? Can I use
> pthread_mutex_lock() in the signal handler to prevent a race
> condition?
There's no race. For async signals, system just picks one thread for
delivery.
This is bad. Is there no function to define that signals should be
send to all threads for async signals?
I must say you're asking strange questions.
What could you gain by dispatching a signal more than once ? Why would you
want a thundering herd of N threads attempting to handle a signal, only so
that in the end N-1 of them find in their signal handler that it's no
longer pending because whoever came first did away with it ?
You can have all async signals targeted at a specific thread by setting
every thread's signal mask to "ignore" except the one that you wish to do
the signal processing in; that remaining 'signal daemon' one can then
simply sigwait() for whatever happens.
As far as thread synchronization / inter-thread communication goes,
signals aren't exactly a great mechanisms. You'd rather use things like
condition variables (which can be 'signaled'), or one of the other
synchronization primitives designed for threads.
FrankH.
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
