Re: com php-src: Fixed bug #61728 (php-fpm SIGSEGV running friendica on nginx): NEWS ext/session/tests/bug61728.phpt main/main.c

Sent from my iPhone 在 2012-4-18，3:22，Yasuo Ohgaki <yohgaki@ohgaki.net> 写道： > Hi, > > 2012/4/17 Johannes Schlüter <johannes@schlueters.de>: >> On Sun, 2012-04-15 at 13:00 +0800, Laruence wrote: >>> On Sun, Apr 15, 2012 at 12:36 PM, Stas Malyshev <smalyshev@sugarcrm.com> wrote: >>>> Hi! >>>> >>>>>> I'm not sure this is a good idea - output buffers can include output >>>>>> functions IIRC, but you have already shut down modules. I would be very, >>>>> Stas, they are request shutdown, not module shutdown :) >>>> >>>> Request shutdown IIRC destroys functions and module's per-request data >>>> structures, which user-supplied output handlers may need. Did you check >>>> that user-supplied output handlers - especially those that use >>>> extensions and per-request data structures - actually work after this >>>> change? >>> >>> Ah, thanks, I got your point. but output handler is called in step 3: >>> /* 3. Flush all output buffers */ >>> >>> and as I said, all tests passed. >>> >>> anyway, you are right, we should be careful, I will double check it. :) >> >> I can't find something that breaks, but we should be really really >> really really careful with this Johannes, I got you point, I checked again, the out buffer deactivate function now only release output handler stack, since mike rewritten the whole implement. So i think this fix seems a appropriate fix:) >> shutdown order. changes there caused >> trouble in the past and I rather have a error message from output >> buffers saying "output buffer can't be used in shutdown" than changing >> the order ... >> > > It would be nice to fix this without error if it works. > How about gave it a chance for next bug fix release? > (i.e. 5.4.2) > yes, except the crash, this fix also fixed a bug that anything output in request shutdown phase will not send to client. Thanks > Regards, > > -- > Yasuo Ohgaki > yohgaki@ohgaki.net