John Levon wrote:
> On Wed, Mar 24, 2004 at 10:55:11AM +0000, Angus Leeming wrote:
>
>> 1 if (current_child == -1)
>> 2 return;
>>
>> // Block the SIGCHLD signal.
>> 3 sigprocmask(SIG_BLOCK, &newMask, &oldMask);
>>
>> // Wait for an existing signal to finish being processed.
>> 4 while (child_guard) {}
>
> Signals are asynchronous, but they're not parallel. Your code won't
> simultaneously be executing this while() loop *and* executing signal
> code, it'll be one or the other. So child_guard can't possibly be
> set here.
Great! However...
I thought that I read that the handler could be implemented as a
separate thread? If this is true, then SMP machines will execute it
in parallel, no?
Incidentally, shouldn't we handle WIFCONTINUED in
handleCompletedProcesses()?
} else if (WIFSTOPPED(store.status)) {
lyxerr << "LyX: Child (pid: " << store.pid
<< ") stopped on signal "
<< WSTOPSIG(store.status)
<< ". Waiting for child to finish." << endl;
} else if (WIFCONTINUED(store.status)) {
lyxerr << "LyX: Child (pid: " << store.pid
<< ") that was stopped has been restarted."
<< " Waiting for child to finish." << endl;
} else {
lyxerr << "LyX: Something rotten happened while "
"waiting for child " << store.pid << endl;
// Child died, so pretend it returned 1
child.setRetValue(1);
remove_it = true;
}
--
Angus
