View Issue Details

IDProjectCategoryView StatusLast Update
0001292SOGoBackend Calendarpublic2011-12-08 15:05
Reportercorradofiore Assigned Towsourdeau 
PrioritynormalSeveritycrashReproducibilityrandom
Status resolvedResolutionfixed 
Product Version1.3.7 
Fixed in Version1.3.11 
Summary0001292: SOGo crashes (sometimes) when issuing PROPFIND
Description

I have set up a test user (maria.rossi@kiiama.om) with 3 empty calendars and iCal.app on Mac OS X 10.6 checking it via CalDAV every minute.

About 1 out of 10 times, SOGo crashes when iCal.app issues the PROPFIND command. This does NOT happen with other accounts having populated calendars.

This is the Nginx relevant log line:
2011/05/07 21:48:33 [error] 21951#0: *93501 upstream prematurely closed connection while reading response header from upstream, client: 81.174.46.78, server: , request: "PROPFIND /SOGo/dav/maria.rossi@kiiama.com/Calendar/ HTTP/1.1", upstream: "http://127.0.0.1:20000/SOGo/dav/maria.rossi@kiiama.com/Calendar/", host: "apps.drenalina.com"

The SOGo logfile can be seen at http://pastebin.com/LyLtnGVy

Additional Information

SOGo 1.3.7a on CentOS 32 bit, MySQL backend.

TagsNo tags attached.

Activities

corradofiore

corradofiore

2011-07-23 14:00

reporter   ~0002738

Still present in 1.3.8a

corradofiore

corradofiore

2011-07-28 09:43

reporter   ~0002751

I ran GDB and was able to get a backtrace. Please find it at http://pastebin.com/rzHWgMep

ludovic

ludovic

2011-07-28 12:58

administrator   ~0002753

Install the SOPE debugging symbols and produce a new stacktrace.

corradofiore

corradofiore

2011-07-30 14:07

reporter   ~0002769

Done. Please find it at http://pastebin.com/UAU0ZkX5

corradofiore

corradofiore

2011-10-11 12:36

reporter   ~0002885

Actually, it happens on populated calendars also (v 1.3.8b on Linux CentOS 32 bit)

ludovic

ludovic

2011-11-24 21:05

administrator   ~0003076

So, using just iCal with 3 empty calendars for 1 user, you make sogod crash randomly if it query for new events every minutes?

corradofiore

corradofiore

2011-11-25 07:43

reporter   ~0003080

Last edited: 2011-11-25 07:44

Yup, correct. I'll be glad to give you access to my server and / or arrange a paid debugging session.

I must admit that I haven't upgraded to 1.3.9 yet. I see in the changelog that 1.3.9 could possibly contain a fix for this issue.

wsourdeau

wsourdeau

2011-12-08 15:05

viewer   ~0003134

The bug was due to the use of a deallocated object occurring when GCSFolderDebugEnabled or GCSFolderManagerDebugEnabled were enabled
Fixed in 3c7a201fa0fd3707b904e5173f62765c592e904b

Issue History

Date Modified Username Field Change
2011-05-08 12:17 corradofiore New Issue
2011-07-23 14:00 corradofiore Note Added: 0002738
2011-07-28 09:43 corradofiore Note Added: 0002751
2011-07-28 12:58 ludovic Note Added: 0002753
2011-07-30 14:07 corradofiore Note Added: 0002769
2011-10-11 12:36 corradofiore Note Added: 0002885
2011-11-24 21:05 ludovic Note Added: 0003076
2011-11-25 07:43 corradofiore Note Added: 0003080
2011-11-25 07:44 corradofiore Note Edited: 0003080
2011-12-08 15:05 wsourdeau Note Added: 0003134
2011-12-08 15:05 wsourdeau Status new => resolved
2011-12-08 15:05 wsourdeau Fixed in Version => 1.3.11
2011-12-08 15:05 wsourdeau Resolution open => fixed
2011-12-08 15:05 wsourdeau Assigned To => wsourdeau