Bug 334572

Summary:	patch to glibc-2.5-35.5@686 makes apache2&php-cgi (with use of gettext) fail
Product:	[openSUSE] openSUSE 10.2	Reporter:	H. Hansen <cyberbeat>
Component:	Maintenance	Assignee:	Petr Baudis <pbaudis>
Status:	RESOLVED DUPLICATE	QA Contact:	E-mail List <qa-bugs>
Severity:	Major
Priority:	P5 - None
Version:	Final
Target Milestone:	---
Hardware:	i686
OS:	openSUSE 10.2
Whiteboard:
Found By:	Customer	Services Priority:
Business Priority:		Blocker:	---
Marketing QA Status:	---	IT Deployment:	---

Description H. Hansen 2007-10-17 14:09:35 UTC

I run a webserver. After applying the glibc-patch, there are apache errors "premature end of script headers" (on random php-scripts, which use gettext, not every script-run fails). after each failure, a php5-instance remains/hangs, so after a few hours, there are many frozen php5-cgi instances.

In the patch description I found the text "Also includes race condition fix for gettext", that is what I guess is the reason.

Comment 1 Petr Baudis 2007-10-17 15:06:29 UTC

Does this problem persist after restarting Apache?

Comment 2 H. Hansen 2007-10-17 17:24:22 UTC

yes, the problem persisted after restarting apache. stopping apache killed the php-instances, but after a while the same errors occured again.

I verified, that the reason for the problem was the patch: after downgrad glibc to original the problem disappeared.

Comment 3 H. Hansen 2007-10-17 17:26:34 UTC

i forgot to mark the checkbox for providing "needinfo"

Comment 4 Petr Baudis 2007-10-17 23:14:16 UTC

Can you find out where are instances hanging, please? Either try to strace them or attach gdb to them and use the bt command (or both :).

Comment 5 H. Hansen 2007-10-18 13:22:24 UTC

I don't know how to do such things. Also I would not apply that patch again on my productive server.

Comment 6 Petr Baudis 2007-10-18 13:29:36 UTC


*** This bug has been marked as a duplicate of bug 334777 ***

Comment 7 H. Hansen 2007-10-19 12:54:42 UTC

I have another question:

I downgraded glibc to original version to avoid this bug and marked the package as "locked" in yast.

Yast-automatic-online-update ignores the locked state and applies this patch again. Is the only solution to this, to completely turn of automatic-online-update? Or is that a bug, too?

Comment 8 Petr Baudis 2007-10-19 14:58:34 UTC

I'm sorry, I have no idea about this.