> Because we have to connect() to the server enough times to wake up
> each thread (ap_daemons_limit * ap_threads_per_child for the threaded
> MPM), we will end up writing that many '!' characters anyway so there
> is no harm in more than one thread in a process consuming a char from
> the pod.  There are plenty to go around.
> 
> The old pod support unique to threaded only wrote to the pod once per
> process so it had to ensure that no more than one thread per process
> would consume a character.

I still don't understand why the parent doesn't just close the pod.
Anything attempting to read from it will then get an EOF, which happens
to be more reliable than reading a special character N times.  Furthermore,
you can then add the pod to the select/poll list so that all of the children
wake up when it closes (except for the ones working or in accept).

Am I missing something?

....Roy

Reply via email to