On Friday, January 11, 2013 07:38:26 AM Colin Guthrie wrote: > 'Twas brillig, and Thomas Spuhler at 11/01/13 14:25 did gyre and gimble: > > On Friday, January 11, 2013 03:21:52 AM Colin Guthrie wrote: > >> 'Twas brillig, and Thomas Spuhler at 08/01/13 14:25 did gyre and gimble: > >>> On Tuesday, January 08, 2013 04:36:26 AM Oliver Burger wrote: > >>>> Am 08.01.2013 11:04, schrieb Frederik Himpe: > >>>>> On Sat, 05 Jan 2013 10:20:35 -0700, Thomas Spuhler wrote: > >>>>>> If nobody objects I am going to rebuild php treadsafe > >>>>>> (--enable-maintainer-zts) > >>>>>> it is needed for running apache with mpm-module-worker instead of > >>>>>> prefork and php apache module worker is needed by 389-ds. It is now > >>>>>> declared as stable. > >>>>>> > >>>>>> But apache mpm-prefork starts with php compiled treadsafe apache > >>>>>> mpm-worker doesn't start with php compiled none-treadsafe > >>>>> > >>>>> You can run PHP with apache2-mpm-worker with mod_fcgid and php5-cgi > >>>>> or php5-fpm. > >>>>> > >>>>> This is what Debian Wheezy's README file for php says: > >>>>> > >>>>> Using PHP 5 with threaded webservers (e.g. apache2-mpm-worker) > >>>>> --------------------------------------------------------------------- > >>>>> - > >>>>> > >>>>> After much back-and-forth with upstream (and even building our > >>>>> packages thread-safe for a while), we're currently admitting > >>>>> defeat on that front, and are NOT building any thread-safe > >>>>> versions of PHP 5 for any webservers. Our recommendation is that, > >>>>> if you need to use a threaded webserver, you should use php5-fpm > >>>>> and interface to your webserver with FastCGI. > >>>>> > >>>>> I think Mageia is hurting itself by enabling too experimental > >>>>> features. Another example is the uas kernel module, of which I saw a > >>>>> complaint recently in this mailing list. This module is known to be > >>>>> buggy, and disabled in Debian. > >>>> > >>>> +1 > >>>> > >>>> That's what Oden told me yesterday on IRC. > >>>> > >>>> So as current php maintainer I vote against enabling this. > >>>> > >>>> Oliver > >>> > >>> I guess I then will revert it to NTS. > >>> Fedora builds it with both options. But it may be too late now going > >>> that way. > >> > >> FWIW, over the last few days I've had to rebuild several PHP modules to > >> get my system working properly. > >> > >> A couple days ago I had to rebuild both php-memcached and php-redis. > >> > >> Today I've had to rebuild php-memcached, php-redis and php-timezonedb. > >> > >> I'm not against enabling features etc. to try them out but when doing so > >> care really should be taken to rebuild all affected PHP modules not just > >> the core ones. > >> > >> Thanks. > >> > >> Col > > > > Why did you have to rebuild them as a result of the new features? TS has > > been reverted to NTS so, there are no new features. > > For php-timezonedb it specifically mentioned that it was built against > the TS version where as the core PHP was NTS (TS vs NTS in the build ids > it was comparing). > > > For the other two, they complained about missing symbols when the TS > version appeared. I forget the symbol itself but it was mentioned > global_config. > > They had the same error today (presumably after reverting) so had to > rebuild again. > > I guess enabling TS vs. NTS changes things more subtly than it might > appear. > > Easy enough to reproduce - just install them and run "php -v" on the > command line to see the list of errors. > > Col
Thanks Col. I saw one was upgraded after php TS was built, but it's now solved. I guess you rebuilt some and umeabot took care of the rest. There are currently no errors using php -v When I built the TS version, I was aware something may happen and I may had to go back. The problem is now, 389-ds needs apache running with worker which in turn needs php TS Fedora builds both modules, TS and NTS and loads them depending on mpm-worker or not. -- Best regards Thomas Spuhler
signature.asc
Description: This is a digitally signed message part.
