GNOME Bugzilla – Bug 344656
Evolution 2.6.2 hogs the CPU
Last modified: 2006-06-19 05:19:28 UTC
Please describe the problem: Orginally I reported this bug in Redhat's bugzilla here: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=194673 But I am not getting any answer. There are a couple of forums posts here: http://forums.fedoraforum.org/forum/showthread.php?t=112708&highlight=evolution http://forums.fedoraforum.org/forum/showthread.php?t=112886&highlight=evolution Basically, the Exchange module got very unusable after the latest Fedora Core 5 update to 2.6.2. Previously I was using 2.6.1 and it was working very well. Everytime I try to read new messages on the exchange server, evo blocks consuming 100% of the CPU trying to fetch the message. The funny thing is that the GUI doesn't block, I can use the menus and everything, I can select a different message from the list but it won't show the content because it is blocked with the previous unread message. The only way to make it work is quitting evolution and starting it back. If I click through read messages it will work fine. I tried everything I could, I reinstalled evolution, I deleted the exchange account and redid it completely but I got no improvement. I hope this bug is reproducible upstream. Steps to reproduce: Actual results: Expected results: Does this happen every time? Other information:
when it blocks, can you get the gdb traces for all the threads of evolution and also evolution-exchange-storage processes and paste it here. Since it is not reproducible for me, without the traces can't work on this. See http://live.gnome.org/GettingTraces for more details.
Created attachment 67253 [details] This is the full evolution backtrace when it blocks As anyone on the forums said, the problem happens when you try to see mails with attachments, it doesn't matter if the message is "read" or "unread". The following backtraces are without debugging symbols, I'm crossing my fingers for them to be good enough. Just remember that this might be a distro specific bug.
Created attachment 67254 [details] backtrace of evolution-exchange-storage before the block This was the backtrace of evolution-exchange-storage, I attached gdb to an already running process with "gdb --pid=<PID>", then I tried to reproduce the bug and evolution blocked as always but it wasn't consuming the CPU at all, then I got this backtrace. Finally, as soon as I detached from the process, it started to consume the CPU... stay tuned. I don't know if this procedure is right.
Created attachment 67255 [details] backtrace of evolution-exchange-storage after the block This backtrace was obtained after evo-exchange-storage started to consume 100% of the CPU when I detached gdb from the running process in the previous backtrace. It looks a bit different so I think it must be important. If there is any need to install debugging packages, please just tell me what packages are needed, just remember that I am running Fedora Core 5 if that helps you to give me more specific instructions. Thanx for your help.
I have the same problem. Can i help?
That's great!, sure you can help us... may be you can post some backtraces with debugging information enabled, I am just too lazy to install the debuginfo packages from Fedora. By the way, what distribution are you using?? If you have any doubts don't hesitate to ask here.
I'm also having this problem when using Evolution with our Exchange Server. The reason it's hard to debug however is that my Evolution is not crashing. Here is some information. The following are versions of RPMs that I had installed while experiencing this current problem. # uname -a Linux new-host-6 2.6.16-1.2122_FC5smp #1 SMP Sun May 21 15:18:32 EDT 2006 i686 i686 i386 GNU/Linux #rpm -qa | grep evolution-2.6.2-1.fc5.2 evolution-connector-2.6.2-1.fc5.2 evolution-debuginfo-2.6.2-1.fc5.1 evolution-data-server-1.6.2-1.fc5.1 ======== Observed ======== Evolution loads and begins to download mail for 'offline' viewing. At random points in the download during different runnings of Evolution, it hangs and begins to use 50% (100% of 1 CPU, dual core system) of my cpu. After that Evolution is unuseable. It does not crash, just continues like this. ======== Expected ======== Evolution loads, begins to download mail for offline viewing, and completes successfully. =================== Steps to reproduce =================== As far as I can tell, and I've done this several times, using the above mentioned packages on my system allows the bug to occur. I've reverted to the following packages: # uname -a Linux new-host-6 2.6.16-1.2122_FC5smp #1 SMP Sun May 21 15:18:32 EDT 2006 i686 i686 i386 GNU/Linux #rpm -qa | grep evolution evolution-connector-2.6.1-1.fc5.1 evolution-2.6.1-1.fc5.2 evolution-debuginfo-2.6.1-1.fc5.2 evolution-data-server-1.6.1-1.fc5.2 evolution-webcal-2.4.1-3.2 There is/are still bug(s) in these versions. ======== Observed ======== Starting Evolution and clicking on the 'Contacts' switcher button and going to my 'Global Contacts' causes the above message to be shown: Unable to open this addressbook. This either means you have entered an incorrect URI, or the server is unreachable. I know my settings are correct because I have an FC4 machine with the same settings that works perfectly with the Exchange Global Address List When running Evolution from the command-line here is what is shown during this process: evolution CalDAV Eplugin starting up ... ** (evolution-2.6:9546): WARNING **: LDAP authentication failed (0x31) evolution-shell-Message: Killing old version of evolution-data-server... (evolution-2.6:9546): camel-WARNING **: camel_exception_get_id called with NULL parameter. Yes, I see the LDAP failure, but I'm not having this on another system with different packages and the *exact* same configuration. I've been through this many, many times. =========== Expected =========== Evolution loads. Click on 'Contacts>Global Address List' the contacts are displayed. ===========-====== Steps to reproduce ================== Use the same packages as I've mentioned above on FC5 I'd be happy to include some debugging information, however, Evolution isn't crashing. Let me know what to do and I'll do it. I've got to get this working for my work email using the Exchange server - so you've got a guinea pig! NOTICE: these are the versions I've had to revert to in order to get my Evolution *partially* working again. These were the versions I had before:
Created attachment 67352 [details] Backtrace of when Evolution was hanging. My Evolution locked up...and then I noticed that the GDB console said: Program received signal SIGPIPE, Broken pipe. [Switching to Thread -1209063744 (LWP 7537)] 0x00368410 in __kernel_vsyscall () so I ran: thread apply all bt and got the backtrace I've attached. hope this helps.
William Lovaton, from comment #4, i believe it is evolution-exchange-storage using 100% cpu and not evolution. when it uses, attach gdb only to evolution-exchange-storage and get the trace of all the threads. If it same as the trace in comment #4, then it is not of much use :(
Dave E, are you observing the cpu hog in Evolution or evolution-exchange-storage? and how often? and LDAP error 0x31, means Invalid credentials, how does your username and owa url look like?
Created attachment 67411 [details] Another backtrace of evolution-exchange-storage Sushma Rai, following your suggestion from comment #9 I have to say that the backtrace in comment #4 is exactly from evolution-exchange-storage. What I didn't specified is that at first, when triggering the bug, is evolution the one that is consuming the CPU, after you kill evolution in the command line, evolution-exchange-storage takes over from evolution and starts consuming the CPU. The backtrace in comment #4 represents the state of e-e-s when evolution was still running. I managed to get a different backtrace when I killed evolution and e-e-s started to consume the CPU which is a different situation that previous attachments... I hope this helps. In another attachment I'll post a portion of the output of "strace -p <e-e-s PID>" when consuming the CPU.
Created attachment 67415 [details] strace -p from evolution-exchange-storage This is the output of strace -p while e-e-s is consuming 100% after killing the evolution process.
I got aditional new information: Evolution is not always hanging on emails with attachments, I can view them without any problem. It seems that for some reason evolution manage to get those messages into the local store and when that happens it will work fine only with those messages. But if I select a message with attachment that is not in the local store, evolution has to talk with the exchange server and it's there where the hanging is happening. This is all a simple guess of corse... I hope this helps.
This is a random crash/hang observed in 2.6.2 From the Evolution traces in comment #2, this is duplicate of 344196. *** This bug has been marked as a duplicate of 344196 ***