Page 2 of 2

Posted: Thu Jan 10, 2008 8:33 pm
by scott
Yeah, that doesnt hurt either.

Posted: Sun Jan 13, 2008 3:23 pm
by Highland
I've had some weird experiences myself with PEAR this weekend. Not from PEAR itself but from a bizarre problem with include_path.

I want to Plesk 8.3 a couple of weeks ago. That changed the global include_path and broke some stuff. No biggie, we just added some ini_set into our PHP code and all was well. Went to PHP 5.2.5 this week (Wed I think) and watched it closely and no problems arose.

Yesterday I get a call from my boss telling me our checkout is down. I pour through code until I find that PEAR is no longer included and the ini_set is being ignored. This is strange since we hadn't messed with any config settings or code along those lines. So I added php_value include_path /usr/share/pear and restarted Apache and all was well. Did a test run to make sure we were good and that was that. Or so I thought.

Later my manager informs me it went back down. I'm thinking in four letter words at this point and I went and looked. Sure enough, PEAR was the culprit. I did a phpinfo() run and found the path was set to .: for local and admin settings despite the fact that php_value was set. Restarted Apache and the local value stayed /usr/share/pear for all of a minute before bouncing back to .: I ran all over the place and tried enverything I could think of but no matter what I did it would bounce between the two values before winding up with the admin value. Finally I bit the bullet and edited php.ini and made it .:/usr/share/pear, restarted and no problems since.

I didn't have time yesterday but today I poked around some and found that there's a bug in 5.2.5 regarding php_value and ini_set not working properly. Describes perfectly what happened to me yesterday and even has a patch that came out yesterday. I suspect it would deal with most of the problems everyone else has been having in this thread.
http://bugs.php.net/bug.php?id=43677&edit=1

Is there any chance of a 5.2.5-3 with this patch in it soon?

Posted: Sun Jan 13, 2008 3:57 pm
by breun
Good find! Yes, this also explains the include_path problems I've seen with PHP 5.2.5.

Posted: Sun Jan 13, 2008 5:39 pm
by scott
Good find, I'll try and get that in today.

Posted: Mon Jan 14, 2008 8:41 am
by scott
It's syncing to atomic-testing right now.

Posted: Mon Jan 14, 2008 9:07 am
by breun
Tried those packages and they seem to fix that darn include_path problem!

+1 from me.

Posted: Mon Jan 14, 2008 2:40 pm
by Highland
Not sure if this is related or not but I had some problems with grsec and httpd afterwards. Had to roll back to 5.2.3 to get it to stop. Could be unrelated but thought I'd toss it out there
# service httpd restart
Stopping httpd: [ OK ]
Starting httpd: [ OK ]
[root@u15222042 ~]# grsec: signal 11 sent to /usr/sbin/httpd[httpd:8849] uid/euid:0/0 gid/egid:0/0, parent /sbin/init[init:1] uid/euid:0/0 gid/egid:0/0

# rpm -q kernel
kernel-2.6.9-42.EL
kernel-2.6.9-42.0.3.EL
kernel-2.6.17-1.art
kernel-2.6.19-7.art

Posted: Mon Jan 14, 2008 7:18 pm
by breun
What kernel are you running (output of uname -a)? Looks like you don't have the latest ASL kernel installed.

Posted: Mon Jan 14, 2008 8:11 pm
by scott
Thats just a logging message, its not actually causing it. Are you using zend optimizer or anything like that?

Posted: Tue Jan 15, 2008 9:57 am
by Highland
Hadn't installed Zend. Not a big deal at the moment (am migrating to a new server for unrelated reasons) but I wasn't sure if it was related.