On Wed, Dec 20, 2017 at 6:03 PM, Eric Covener <cove...@gmail.com> wrote:
> On Wed, Dec 20, 2017 at 11:28 AM, Yann Ylavic <ylavic....@gmail.com> wrote:
>> On Wed, Dec 20, 2017 at 4:51 PM, Eric Covener <cove...@gmail.com> wrote:
>>> is the "rc" check just covering further bases? Should we trace or set
>>> a note here?
>> For me "rc" would catch at least the DECLINED case (no hook at all),
>> though it may not possibly happen with default builtin http module
>> (does --with-http=shared work?).
>> I also think that it's easier for a module which asks for the MPM to
>> finish the connection, to simply return an error in a
>> process_connection hook than something like OK + CONN_STATE_LINGER (+-
>> c->aborted = 1), .
> Maybe the header could be clarified one way or another.
Done (hopefully) in r1818951, does it work for you?
>> I'd be fine with a c->notes, but wouldn't that be useful for a
>> pre_close_connection hook only?
>> Or do you mean an ap_log_cerror() at some specific level (which I'd be
>> fine with too)?
> Yes, I guess notes are not so useful for such an early failure. Maybe
> ap_log_cerror() at some traceN level would be good.
Also in r1818951.