>wrowe 02/03/19 22:14:19
>
> Modified: server/mpm/winnt mpm_winnt.c
> Log:
> Fix a few listener-related lifetime issues [they are created in the
> open logs phase, only once.]
>
> - if (!set_listeners_noninheritable(pconf)) {
> + if (!set_listeners_noninheritable(s->process->pool)) {
Ok guys, this really, really sucks.
Listeners must survive the server restart, to be closed down
in ap_open_listen after the new listener is created. Is noone
watching pool scope here?
I think once we all [me included] pay better attention to pool scope,
our Apache 2.0 will be the best server yet. But with the migration
to APR and tighter reigns on pool lifetimes, this is going to be
painful for a while.
The alternative is to allow one pconf to survive server restart until after
the -next- post-config hook is run, but that seems like overkill.
Bill
