mod_security-2.9.0-1.fc21.20150410.rh.x86_64
mod_security disabled and the crashes are gone
well, the exactly same build with identical options on F20 with GCC 4.8
is running here in production perfectly - somehow GCC 4.9 seems to be
involved here
i will track that down to GCC optimize flags and report it to
mod_security upstream - strange that the crashes only happen in context
of /plugins/jqueryui/js/i18n/jquery.ui.datepicker-de.js *and* modsec
Am 27.04.2015 um 13:08 schrieb Reindl Harald:
Am 27.04.2015 um 13:01 schrieb Reindl Harald:
> Am 27.04.2015 um 12:49 schrieb drago01:
>> On Mon, Apr 27, 2015 at 11:12 AM, Reindl Harald
>> <h.reindl(a)thelounge.net> wrote:
>>> am i the only one facing reproduceable crashes of apache worker
>>> process each
>>> time a roundcube 1.0.5 page is loaded? all relevant packages (apache,
>>> php..)
>>> here are built from identical srpms andthe only difference is the
>>> Fedora 21
>>> userland and a newer GCC
>>>
>>> that affects the same way PHP 5.5 as PHP 5.6
>>
>> You need a backtrace (better with debug symbols) to track it down ..
>> strace output doesn't really tell much here
>
> not that easy with non-debug builds and in context of a preforking
> apache with mod_php hecne the question if anybody can see that issue too
>
> it's not really obvious because there is nohting missing in the browser
> and so it seems to happen very late at the request, you can see only the
> loglines in the apache errorlog
>
> [Mon Apr 27 12:51:33.076997 2015] [core:notice] [pid 85570] AH00052:
> child pid 85572 exit signal Segmentation fault (11)
uhm - it's not PHP - it's httpd crashing by request the js file
/plugins/jqueryui/js/i18n/jquery.ui.datepicker-de.js
need to investigate that deeper if it's the gcc optimize options in
general and what part of httpd can crash by deliver a static js file
(mod_deflate and so on are active)