Re: Threads signalling issue [was RE: threads change]

On Fri, Jul 14, 2006 at 05:39:09PM -0700, Jan Dubois wrote:
> On Fri, 14 Jul 2006, Artur Bergman wrote:
> > It is threadsafe in the sense you won't crash, but PL_sig_pending
> > might be set to 0 by the running thread and thus not cause the
> > signal to be delivered until another signal is raised.
> 
> Not true, the dispatcher checks the signals *after* resetting the
> PL_sig_pending flag:

Does "after" hold on an SMP machine without any form of locking, where
processor A makes the changes in order, but processor B sees the later
change before the former? I thought that without mutexes there was no
reliable before or after for SMP systems where memory is written back in
any order.

> void
> Perl_despatch_signals(pTHX)
> {
>     dVAR;
>     int sig;
>     PL_sig_pending = 0;
>     for (sig = 1; sig < SIG_SIZE; sig++) {
>         if (PL_psig_pend[sig]) {
>             PERL_BLOCKSIG_ADD(set, sig);
>             PL_psig_pend[sig] = 0;
>             PERL_BLOCKSIG_BLOCK(set);
>             (*PL_sighandlerp)(sig);
>             PERL_BLOCKSIG_UNBLOCK(set);
>         }
>     }
> }

Nicholas Clark