Closed Bug 672913 Opened 14 years ago Closed 12 years ago

hangs on exit/shutdown with zero cpu [Mac]

Categories

(Thunderbird :: General, defect)

Product:
Thunderbird
Component:
General
Platform:
x86
macOS
Type:
defect
Priority:
Not set
Severity:
critical

Tracking

(Not tracked)

Status:
RESOLVED WORKSFORME

People

(Reporter: allan, Unassigned)

References

Details

(Keywords: hang, stackwanted, Whiteboard: [summary: comment 13][not proven in safe mode])

Attachments

(4 files)

sample from activity monitor
31.29 KB, text/plain
Details
crash backtrace and IMAP log
37.05 KB, text/plain
Details
gdb stack and imap.log
17.50 KB, text/plain
Details
Force-Quit log from TB 14.0 release channel on OS X 10.8
421.38 KB, text/plain
Details
Follow up on bug 508263. The issue remains the exactly the same for me. The only thing that is different now is that it no longer eats 100% cpu. Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:5.0) Gecko/20110624 Thunderbird/5.0
(In reply to comment #0) > Follow up on bug 508263. The issue remains the exactly the same for me. The > only thing that is different now is that it no longer eats 100% cpu. Is it now zero or low CPU? Does it happen only after sleep? (or was sleep never a factor for you even in bug 508263?) if it's more complicated than only after sleep, please list the symptoms per https://wiki.mozilla.org/Thunderbird:Testing:Shutdown_Hang
Severity: normal → critical
Keywords: hang
Zero cpu. I don't know for sure if sleep is involved, but my guess is that it is. I.e. all my comments in bug 508263 are still as valid. Nothing has changed for me, except for the CPU now not being 100%.
Allan, Lint, a stacktrace with thunderbird symbols is needed. thanks.
Keywords: stackwanted
Whiteboard: [closeme 2011-09-08][waiting on stack]
Using 7.0a2 Earlybird from Aug 11, I finally got this problem to reproduce. Took me a while, but the problems always run and hide when I use a nightly with symbols. Sheesh. File contains the backtrace and the IMAP log (and the console output fwiw). THe backtrace doesn't look super detailed, but I got the symbols using the fetch-symbols.py script.
Duplicated hang using 8/22 Earlybird 8.0a2. The hang stack is basically the same pattern, as is the imap log. Only thing I can say to reproduce is that the program has been running for at least a couple days (The longer the better), and mac has slept at least a few times. Not sure why the stack trace isn't coming out too detailed. Am following the instructions here to get symbols: https://wiki.mozilla.org/Thunderbird:Backtraces_On_OS_X And here to get the stack trace: https://wiki.mozilla.org/Thunderbird:Testing:Get_A_Debug_Thunderbird_Hang_Stack
ludo, can you help with the stack/symbols problem?
Whiteboard: [closeme 2011-09-08][waiting on stack] → [waiting on stack]
(In reply to Wayne Mery (:wsmwk) from comment #8) > ludo, can you help with the stack/symbols problem? Chris should be running nightlies to get stacks Ain't sure we produce them on aurora or beta.
bienvenu took a look at these and sez "wsmwk - no, nothing helpful for me there". So, if this is still happening to you please - test with daily to be sure it's not fixed in development ... - in safe mode ... - plus the rest of the info in https://wiki.mozilla.org/Thunderbird:Testing:Shutdown_Hang will be greatly appreciated
Whiteboard: [waiting on stack] → [closeme 2011-12-05][waiting on stack]
(In reply to Wayne Mery (:wsmwk) from comment #10) > So, if this is still happening to you please It is. Nothing has changed for me except for not spinning the cpu. > - test with daily to be sure it's not fixed in development ... Where are they again? It always takes me forever to find them
Comment on attachment 555606 [details] crash backtrace and IMAP log Hmm - the stack trace on Thread 8 in the attachment "crash backtrace and IMAP log" looks familiar to me - I had a shutdown hang a week or two ago running under the debugger, and spent a while poking around. What it looked like to me was that the underlying OS level socket had closed at some point in the past, but the async I/O subsystem hadn't notified all the way up to the nsPipeInputStream - so the necko thread was gone, but the imap thread was still waiting for a notification from the necko thread to break out of its PR_Wait.
(In reply to Wayne Mery (:wsmwk) from comment #10) > bienvenu took a look at these and sez "wsmwk - no, nothing helpful for me > there". > > So, if this is still happening to you please > - test with daily to be sure it's not fixed in development ... > - in safe mode ... > - plus the rest of the info in > https://wiki.mozilla.org/Thunderbird:Testing:Shutdown_Hang > > will be greatly appreciated The problem with this is safe-mode ... without the Nostalgy extension I waste time every day, and it takes time to reproduce this (takes either time and/or sleeps). So without safe-mode I can do.
I'm trying to set myself up with a nightly, but having trouble getting symbols from the server. First the fetch-symobls.py script had errors, now I'm getting 404 errors from the symbol server... Anyone else able to get symbols for a nightly?
> Anyone else able to get symbols for a nightly? Seems like the symbols server might not be giving up a proper set of symbols? I am getting failure at the end - the big 'thunderbird.dSYM.tar.bz2' bundle is showing at only 280 bytes, where it normally is many MB. The untar is failing in the python script because of this.
Mark ted are we uploading producing proper symbols for Thunderbird ?
(In reply to Ludovic Hirlimann [:Usul] from comment #17) > Mark ted are we uploading producing proper symbols for Thunderbird ? I have found the answer myself at https://developer.mozilla.org/en/Using_the_Mozilla_symbol_server#Downloading_symbols_on_Linux_.2F_Mac_OS_X Chris,Allan , could you guys try that to have symboled stacks ?
According to comment 15 and comment 16, Chris has tried this and it wasn't working. Chris: do you know which nightly build this was failing for? I can look at the files on the symbol server, if so.
> Chris: do you know which nightly build this was failing for? I can look at > the files on the symbol server, if so. I just tried it now for 12/16 nightly, and am still getting the same error. Maybe it matters that the fetch-symbols.py script that is linked to in comment #18 doesn't work as written - I have to remove a few lines up front to get it to run generate a URL that the symbol server can use. Once I modify it, then it runs until the failure on the big thunderbird.dSYM bundle as described above. I'd attach a diff and go into the fetch-symbols changes, but that seems like a different bug or something.
Chris, questions - does your issue go back as far as version 3.0 (where Allan reports it in bug 508263 - Allan, please comment if this is incorrect interpretation) - please refresh us about what factors of yours are similar to Allan's, other than zero cpu If there is sufficient divergence, then we should pursue your issue in a new bug, if it doesn't match any others - https://bugzilla.mozilla.org/buglist.cgi?type1-0-0=substring&list_id=1958355&short_desc=shut%20hang&field0-0-0=short_desc&type0-0-1=substring&field0-0-1=keywords&type1-0-1=allwordssubstr&resolution=---&classification=Client%20Software&classification=Components&query_format=advanced&short_desc_type=allwordssubstr&type0-0-0=anywordssubstr&field1-0-0=short_desc&product=MailNews%20Core&product=Thunderbird&field1-0-1=short_desc
Whiteboard: [closeme 2011-12-05][waiting on stack] → [need stacktrace]
I first noticed this enough to report on it 2 years ago, with 3.1. It probably started sometime before that. And it's been happening consistently since then. Not really much to say about this - if TB has been running for a while (usually takes at least a day, definitely with sleep cycles), I quit, and it hangs at some point before it's fully quit, sitting at zero cpu with ~9 threads. I have to force quit it. I don't get a spinning beach ball as the UI goes away and I'm switched to the most recent app (per OSX normal behavior) and I can't summon TB to the foreground at this point. It's hard to reproduce, as it doesn't happen if I just start, run for a short period, and then quit. I've been busy so I haven't been able to try a nightly with symbols recently.
I have had the same symptoms with Thunderbird for Mac for about 2 - 3 years. My Mac Book Pro stays on for months at a time, but goes to sleep a couple of times a day. I also change networks between work, home and offline. Thunderbird is always running, but sometimes it seems to get stuck trying to check mail. When that happens, I quit Thunderbird, but it hangs after the window disappears. It does not chew CPU. But I have to force quit to close the app. However, even when Thunderbird mail checking is working fine, and I quit, often it hangs in the same way. I keep Thunderbird up to date, and I am currently running 9.0.1 on Mac OS X 10.6.8. My IMAP email server was UW-IMAP, then dovecot, and now Exchange. The hang has occurred with all 3 servers. I have LDAP pointing to Exchange, but it doesn't work very well (I believe due to local SSL configuration issues), so I have LDAP address completion switched off. But the hang occurred with our previous LDAP server, which worked fine. Lately this hang on quit has become more annoying because I have to manually delete the .parentlock file after force quitting, in order to start Thunderbird again. Otherwise I get a dialog box saying another Thunderbird is already open. I'm sorry I don't have time to help with debugging this issue. But hopefully this report of similar symptoms might be helpful.
2 good things - getting symbols seems to be working again for me - I've been unable to reproduce using a recent nightly (after reading about bug 711786 and related, I thought it could have an effect on this). It will take a while to convince me that it's gone, though.
(In reply to Chris Andrichak from comment #24) > 2 good things > - getting symbols seems to be working again for me > - I've been unable to reproduce using a recent nightly (after reading about > bug 711786 and related, I thought it could have an effect on this). It will > take a while to convince me that it's gone, though. Good , chris let's just hope you'll get the symbols next time it happens to you.
Depends on: 698882
Attached file gdb stack and imap.log —
I actually got a hang, although after a couple weeks of trying I could live with it happening that infrequently. Stack and imap.log attached, although the stack doesn't seem particularly detailed. Using a Daily from 1-30-2012 with symbols.
irving, bienvenu, what does stack reveal?
there's not enough depth in the stacks to see anything useful. The protocol log part looks like we've cleaned up our imap connections.
Shouldn't we be getting better stacks from Daily builds? I thought we built those with gdb-friendly compiler options. I've seen a similar hang at shutdown once or twice under a debugger; I thought I had filed a bug on it but can't find it so perhaps I didn't. On the hang I saw, it looked like a connection had been closed by the networking thread but the IMAP thread hadn't received the notification; then, at shutdown time, the IMAP thread got stuck waiting for the networking thread to tell it that connection was closed. I'm all sad now because I remember having figured out quite a bit of detail about where the hang was, but I can't find any notes on it.
so you are not thinking of bug 698882?
(In reply to Wayne Mery (:wsmwk) from comment #30) > so you are not thinking of bug 698882? No, that one is in a different place.
(In reply to Irving Reid (:irving) from comment #29) > Shouldn't we be getting better stacks from Daily builds? I thought we built > those with gdb-friendly compiler options. > > I've seen a similar hang at shutdown once or twice under a debugger; I > thought I had filed a bug on it but can't find it so perhaps I didn't. perhaps usul or standard8 will know. > On the hang I saw, it looked like a connection had been closed by the > networking thread but the IMAP thread hadn't received the notification; > then, at shutdown time, the IMAP thread got stuck waiting for the networking > thread to tell it that connection was closed. > > I'm all sad now because I remember having figured out quite a bit of detail > about where the hang was, but I can't find any notes on it. remember any of the context? time frame? type of issue / bug? (In reply to Chris Andrichak from comment #22) > I first noticed this enough to report on it 2 years ago, with 3.1. It > probably started sometime before that. And it's been happening consistently > since then. Chris, which bug#?
(In reply to Wayne Mery (:wsmwk) from comment #32) > remember any of the context? time frame? type of issue / bug? I was working on something else (around December / January if I recall correctly), running under the debugger, and happened to have a hang on shutdown. Since I was in the debugger anyway, I spent a bunch of time looking at the system state and trying to figure out how I got there. Unfortunately after a while Xcode crashed on me so I couldn't continue digging.
I've been running the release channel for a few months now, at least, and I can't remember the last time I had this problem. I'd guess maybe once within the last month, but that's a guess. So this isn't really a problem for me anymore. As for bug #, this one and it's predecessor are the only ones I've reported it under.
Allan, then the question is back to you :) -- do you still see this when using newer versions?
(In reply to Wayne Mery (:wsmwk) from comment #35) > Allan, then the question is back to you :) -- do you still see this when > using newer versions? Let's just say that ever since I filed this bug, the way I quit Thunderbird is usually 'killall -9 thunderbird' :(
At running the risk of making a totally uninformed statement, is there possible to drop in a variable that can perhaps be read via a custom item you can drop into the menu bar on how many network online/offline transitions Tb has done. Perhaps in doing so with something "right in front" we can get a idea about how many times it takes to trigger the hang behavior.
1. caution - it remains to be seen that all reports of this have the same cause. 2. we just have to ride this out til irving's memory returns in full force :)
Keywords: stackwanted
Whiteboard: [need stacktrace] → [not proven in safe mode]
I never saw this problem until I upgraded from Snow Leopard (10.6.8) to Mountain Lion (10.8) last week. Now it happens very intermittently. This is on TB 14.0 on the Release Channel. The only extensions I am using are Import/Export Tools 2.7.2.2, Lightning 1.6 and the Provider for Google Calendar 0.13. If it happens again, is the log generated by OS X "Force Quit" of any use to anyone?
(In reply to Dave Lasker from comment #40) > Created attachment 653820 [details] > Force-Quit log from TB 14.0 release channel on OS X 10.8 I think this lacks Thunderbird symbols
Dave, I'd recommend updating to the latest version of Thunderbird first, and then if you see the hang again please start up Activity Monitor, select the Thunderbird process and click "Sample Process". You can save the output of that to a file and attach it to the bug.
(In reply to :Irving Reid from comment #42) > Dave, I'd recommend updating to the latest version of Thunderbird first, and > then if you see the hang again please start up Activity Monitor, select the > Thunderbird process and click "Sample Process". You can save the output of > that to a file and attach it to the bug. please use latest nightly from https://ftp.mozilla.org/pub/mozilla.org/thunderbird/nightly/latest-comm-central/
Flags: needinfo?(dave)
Flags: needinfo?(chris)
Flags: needinfo?(allan)
This bus is very intermittent. I keep up to date with the latest production releases and am currently on 24.1.0. I haven't seen this bug occur for a few months.
Flags: needinfo?(dave)
I'm running the beta channel on two Macs now and I definitely see this much less than before. I can't actually say the last time I ran into it.
Flags: needinfo?(chris)
Matthew, how does this look for you? Allan reports not using THunderbird in a long time, so he is out of the picture
Flags: needinfo?(allan) → needinfo?(maffew)
Keywords: stackwanted
Summary: hangs on exit/shutdown with zero cpu → hangs on exit/shutdown with zero cpu [Mac]
Whiteboard: [not proven in safe mode] → [summary: comment 13][not proven in safe mode]
I used to have this problem nearly every time I quit. But I haven't seen it now for at least 6 - 12 months. Thanks whoever fixed it! (In reply to Wayne Mery (:wsmwk) from comment #46) > Matthew, how does this look for you? > > Allan reports not using THunderbird in a long time, so he is out of the > picture
Flags: needinfo?(maffew)
Thanks all for your followup posts - not everyone is BMO is so conscientious! Based on feedback I think we can close this WFM. Please comment and reopen the bug if you see the problem again.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: