Bug 628073

Summary:	floating point exceptions in msgmerge
Product:	[Fedora] Fedora	Reporter:	Matthias Clasen <mclasen>
Component:	gettext	Assignee:	Jens Petersen <petersen>
Status:	CLOSED WORKSFORME	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	medium	Docs Contact:
Priority:	low
Version:	rawhide	CC:	i18n-bugs, mcrha, petersen
Target Milestone:	---
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2011-02-06 23:48:03 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Matthias Clasen 2010-08-27 19:17:29 UTC

I'm hitting these frequently in koji builds, both F14 and rawhide. The look like this:

+ msgmerge en_GB.po gnome-settings-daemon.pot
..../var/tmp/rpm-tmp.4Jz6es: line 63: 29819 Floating point exceptionmsgmerge $p gnome-settings-daemon.pot > $p.out


Here is an example: http://koji.fedoraproject.org/koji/taskinfo?taskID=2431661


This forces me to remove all uses of msgmerge from my spec files.

Comment 1 Jens Petersen 2010-08-30 05:03:45 UTC

I tried building on a f14 guest without problems.
Also tried mock builds for x86_64 and i386 f14 on RHEL6
and also succeeded.  There only seems to be one double in "msgmerge.c".

Comment 2 Jens Petersen 2010-08-30 05:48:46 UTC

Another example with nautilus: 
http://koji.fedoraproject.org/koji/taskinfo?taskID=2423378

Comment 3 Milan Crha 2010-09-01 06:38:09 UTC

And two from evolution:
http://koji.fedoraproject.org/koji/taskinfo?taskID=2434762
http://koji.fedoraproject.org/koji/taskinfo?taskID=2437502

Comment 4 Jens Petersen 2010-09-10 08:43:53 UTC

I tried rebuilding http://koji.fedoraproject.org/koji/getfile?taskID=2440952&name=gnome-settings-daemon-2.31.91-2.fc15.src.rpm
(from http://koji.fedoraproject.org/koji/taskinfo?taskID=2440951)
in dist-f15 x86_64 mock and it succeeded 10/10 times.
So this does not seem that easy to reproduce. :-|

Comment 5 Jens Petersen 2010-09-10 09:15:27 UTC

Has anyone seems the problem on F13?  (sorry I think I already asked Matthias)

I leaving this script running over the weekend:


#!/bin/sh
set -e

i=0
while true; do
for p in *.po; do   msgmerge -q $p gnome-settings-daemon.pot > $p.out; msgfmt -o `basename $p .po`.gmo $p.out; echo `basename $p .po`;  done
i=$(( $i + 1 ))
echo
echo $i
echo 

sleep 5
done


on my F13 box with gettext-0.18.1.1 to see if I catch catch a failure.

A simple reproducible testcase would really help a lot.

Comment 6 Jens Petersen 2010-09-10 09:50:27 UTC

Well it has done 200 iterations now...

Comment 7 Jens Petersen 2010-09-13 00:24:49 UTC

My f14 x86_64 guest appeared to do over 10000 loops without error
over the weekend.

Is this problem still happening?

Comment 8 Jens Petersen 2010-10-01 02:11:00 UTC

Is it fixed now?  If there are no updates or comments
I plan to close this in a week.

Comment 9 Milan Crha 2010-10-01 07:48:45 UTC

I tried to rebuild evolution with msgmerge enabled in rawhide and it didn't fail, so seems to be fixed, if it wasn't just a coincidence with a scratch-build.

[]1 http://koji.fedoraproject.org/koji/taskinfo?taskID=2505607