On Nov 24, 2013, at 7:18 PM, Jeff Trawick <traw...@gmail.com> wrote:
> On Sat, Nov 23, 2013 at 5:39 PM, Yann Ylavic <ylavic....@gmail.com> wrote:
> Couldn't ap_queue_info_try_get_idler() and the event_pre_config() check use :
>
>
> prev_idlers = apr_atomic_add32((apr_uint32_t *)&(queue_info->idlers), -1);
>
> like ap_queue_info_wait_for_idler() does ?
>
> I think that's correct...
>
> What the code really wants is new_idlers, so edit that slightly to
>
> new_idlers = apr_atomic_add32((apr_uint32_t *)&(queue_info->idlers), -1) - 1;
> if (new_idlers <= 0) {
> ...
> return APR_EAGAIN;
> }
It just wants to see if there are any idle ones... If there wasn't
before the dec, then we need to return APR_EAGAIN. Recall that
a 0 means "no idlers" and neg means 'this many blocking', and
the logic is the same in both cases, minus the "wake up" scenario.
