evolution crashed with SIGSEGV in strcmp()

StacktraceTop:strcmp () at ../sysdeps/x86_64/strcmp.S:29
IA__g_str_equal (v1=0x0, v2=0x7f8b6b72a2f2)
org_gnome_mail_new_notify (
epl_invoke (ep=0x1c3cde0,
e_event_emit (emp=0x1c3d2c0,

isn't this more like an evolution-indicator issue?

Hello,

I've got a bug fix that I have put in my PPA:

  https://launchpad.net/~ted-gould/+archive/bugfix

Could those who can recreate this test to see if this fixes the bug?

For sponsors/release team I've proposed a merge of that package here:

  https://code.launchpad.net/~indicator-applet-developers/evolution-indicator/packaging/+merge/5745

Thanks everyone!
Ted

hi,

I installed both packags. And did only one test sending myself on email and switching to another desktop. So no crash. But I didn't get a notification about new mail either.

the change has been uploaded now

oh, oh

First I don't get any notifcations from evolution about new mails at all now.

Second evolution crashed while look into my new mails with courser up/down. (Possible a bug in new-mail counter.) I'm trying to reproduce it.

One of the reasons that you might not be getting notifications of new
mails is that the notifications aren't sent if any Evolution window is
in focus. So, you need to make sure you're working in another window
when looking.

there is zero notification about emails in evolution. I tried diff workspaces, opened and focused other windows and min/maxed evolution.

Other notifications work well: evo-calender, printer, pidgin, volume...

On Wed, 2009-04-22 at 13:35 +0000, greenhunter wrote:
> there is zero notification about emails in evolution. I tried diff
> workspaces, opened and focused other windows and min/maxed evolution.
>
> Other notifications work well: evo-calender, printer, pidgin, volume...

Can you run evolution from the command line? The output there should
have information on why Evolution isn't sending notifications. Thanks!

right,

s@z:~$ evolution
** (evolution:4144): DEBUG: mailto URL command: evolution %s
** (evolution:4144): DEBUG: mailto URL program: evolution
** (evolution:4144): DEBUG: EI: SHELL STARTUP
** (evolution:4144): DEBUG: EI: mail_read_notify
** (evolution:4144): DEBUG: MAIL SERVER: Count changed: 0
** (evolution:4144): DEBUG: Evolution is focused
** (evolution:4144): DEBUG: Evolution is focused

regardless what its real status is: evolution always thinks it is focused.

hi

I was able to gather some new information:

Evolution only thinks it is on top, if "inbox" is currently selected. And it doesn't matter if evo is on another workspace or minimised.

If I select ex. trash folder in evo, create a new mail or do something else in another window. Notifications work properly.

The patch looks weird to me:

- if (g_str_equal (res_name, "evolution"))
+ if (g_strcmp0 (res_name, "evolution"))

g_str_equal() returns TRUE for identical strings, while g_strcmp0() returns 0 for identical strings, which is considered as "false". I don't see anywhere in the bug trail the intent to negate the condition.

Please try the updated package in the bugfix PPA. If you've still got
it installed, just update :)

Is the build only of the fix for amd only? I need 386.

On Thu, 2009-04-23 at 20:41 +0000, greenhunter wrote:
> Is the build only of the fix for amd only? I need 386.

No, i386, AMD and LPIA. The x86 deb is here:

https://launchpad.net/~ted-gould/+archive/bugfix/+files/evolution-indicator_0.1.15-0ubuntu1_i386.deb

nice,

finally it works like expected.

btw, why is there no notification-history about the last notifications? Further: clicking on not-hist should start some action.

At the moment notifications are totally useless if you don't watch the screen all the time.

And the number of new mails in evo inbox aren't shown in the not-applet.

Firefoxs old notification is really annoying among this new one.

Updates should be shown as well that way.

Ted, did you see my question in comment 14 about the inverted logic?

On Fri, 2009-04-24 at 10:16 +0000, Martin Pitt wrote:
> Ted, did you see my question in comment 14 about the inverted logic?

Yes, sorry. That's fixed in the new version.

Thanks, Ted. I rejected the current upload. Can you please send a new merge request for your fix?

This bug was fixed in the package evolution-indicator - 0.1.15-0ubuntu1

---------------
evolution-indicator (0.1.15-0ubuntu1) karmic; urgency=low

  * Upstream release 0.1.15: Brown paper bag. Missing a '!' in
    previous release. Now really fixes LP: #359658
  * Upstream release 0.1.14: Change to g_strcmp0 to check for a NULL
    string and not crash when getting it. LP: #359658

 -- Ted Gould <email address hidden> Mon, 11 May 2009 08:20:11 +0200

Bug 381144 isn't solved by this issue.

A month ago Evolution suddenly didn't want to start anymore. Nothing displayed. dmesg shows evolution[8484]: segfault at 0 ip 00007f6a831a5092 sp 00007fff97d1bf98 error 4 in libc-2.9.so[7f6a83121000+168000].

This is still true after the fix for this bug is applied.

I had tried reinstalling libc and evolution. Without success. evolution still segfaults. I haven't had local access to my e-mails for over one month now. Ubuntu Jaunty 9.04.

Please reinvestigate.

@Lars:
(1) what version of Evolution are you running?
(2) what version of Ubuntu are you running?

thanks.

Ubuntu 9.04 (jaunty)
Kernel Linux 2.6.28-11-generic
Gnome 2.26.1

latest updates as of today (all verses without backported) but I added Teds repo.

So I have evolution which came with Ubuntu 9.04 on the day it was released.

Lars, can you please open a new bug on that? I would like to see a stacktrace on it. You will need to enable apport: before trying to reproduce the issue:

sudo force_start=1 /etc/init.d/apport restart

then try to start Evolution. When you get the segfault, apport should kick in and ask you to file a new bug. Do it, and please subscribe myself (hggdh2).

On Tue, 2009-06-16 at 15:01 +0000, C de-Avillez wrote:

> Lars, can you please open a new bug on that? I would like to see a
> stacktrace on it.

The bug I reported and merged has a good backtrace given by apport. See
bug #385263.

--
Laurent Bonnaud.
http://www.lis.inpg.fr/pages_perso/bonnaud/

Laurent, thanks, but bug 385263 seems to be a real duplicate of this one.

Unless... you *did* apply Ted's fix, and it still fails... in this case I really need a new bug, with a new stacktrace, so that we can follow the code and see if it is still a special case of this one, or a brand new issue.

<assuming this is a duplicate of 363894 >

I am also hitting this after recent updates to ubuntu 9.04, here is the trace.

#0 0xb68de078 in strcmp () from /lib/tls/i686/cmov/libc.so.6
#1 0xb6a28c64 in g_str_equal () from /usr/lib/libglib-2.0.so.0
#2 0xb055db14 in org_gnome_mail_new_notify () from /usr/lib/evolution/2.26/plugins/liborg-freedesktop-evolution-indicator.so
#3 0xb7d378e1 in ?? () from /usr/lib/evolution/2.26/libeutil.so.0
#4 0xb7d37e29 in e_plugin_invoke () from /usr/lib/evolution/2.26/libeutil.so.0
#5 0xb7d305da in ?? () from /usr/lib/evolution/2.26/libeutil.so.0
#6 0xb7d30761 in e_event_emit () from /usr/lib/evolution/2.26/libeutil.so.0
#7 0xb5a4c248 in ?? () from /usr/lib/evolution/2.26/components/libevolution-mail.so
#8 0xb5a4d833 in ?? () from /usr/lib/evolution/2.26/components/libevolution-mail.so
#9 0xb5a4f8aa in ?? () from /usr/lib/evolution/2.26/components/libevolution-mail.so
#10 0xb6a01c81 in ?? () from /usr/lib/libglib-2.0.so.0
#11 0xb6a03b88 in g_main_context_dispatch () from /usr/lib/libglib-2.0.so.0
#12 0xb6a070eb in ?? () from /usr/lib/libglib-2.0.so.0
#13 0xb6a075ba in g_main_loop_run () from /usr/lib/libglib-2.0.so.0
#14 0xb73eecc3 in bonobo_main () from /usr/lib/libbonobo-2.so.0
#15 0x0805d563 in ?? ()
#16 0xb687d775 in __libc_start_main () from /lib/tls/i686/cmov/libc.so.6
#17 0x08050221 in ?? ()
(gdb)

The package listed in comment #17 reporting to fix this:

https://launchpad.net/~ted-gould/+archive/bugfix/+files/evolution-indicator_0.1.15-0ubuntu1_i386.deb

seems to be no longer available.

On another note, why was this simply not put into jaunty-updates? I wonder how many other bugs that i run into on a frequent basis are fixed by somebody and simply not given to the users. You really are doing Ubuntu a disservice fixing all of these bugs but not actually giving them to the users.

I find Ubuntu so frustrating some days with all the bugs that it makes me wonder, for a very dark moment if i should not be using Windows. At least I get bug fixes from them.

> On another note, why was this simply not put into jaunty-updates? I wonder how many other bugs that i run into on a frequent basis are fixed by somebody and simply not given to the users

Because every single fix can't be backported to stable, it would require lot of efforts which we would put at fixing new bugs and lower the quality of the next versions which create a circle of lower quality

Concerning the no bug fixes comment try to compare the number of updates you get in both system you will see that you probably get as many bug fixed, microsoft doesn't fix any single issue in their updates either

On Thu, 2009-10-08 at 13:55 +0000, Sebastien Bacher wrote:
> > On another note, why was this simply not put into jaunty-updates? I
> wonder how many other bugs that i run into on a frequent basis are fixed
> by somebody and simply not given to the users
>
> Because every single fix can't be backported to stable,

What backporting? The dude had a package already created and ready to
go. The only thing left to do was actually give it to the users.

> it would require
> lot of efforts which we would put at fixing new bugs and lower the
> quality of the next versions which create a circle of lower quality

So the current situation, a current release which most users are using,
rife with bugs (that have been fixed and are just floating around in
PPAs even!) is better?

I agree that the next release always needs to be solid too, but the
reality is that the next release is always full of "shiny new" with it's
own set of bugs, so the user only ever has the choice of trading in old
bugs for new ones. There's never an answer to the user that just wants
something that's stable and will receive bug fixes without getting a
whole raft of new bugs.

I guess ultimately, if an organization doesn't have enough manpower to
maintain the current release and develop the next release, then the
organization's goals are too lofty for the resources the organization
has to achieve them.

> Concerning the no bug fixes comment try to compare the number of updates
> you get in both system you will see that you probably get as many bug
> fixed, microsoft doesn't fix any single issue in their updates either

Yeah. Ultimately, I do believe Windows is just as bad as the linux
desktop and would not trade the transparency of linux for the blackbox
of Windows. Some days though, it's very frustrating to be a linux
desktop user.

Evolution is a pig that doesn't scale and plain just doesn't work
(vfolders are still broken). Daily, Firefox 3.5 starts out at a few
hundred MB of memory and by the end of the day (or three) it's consuming
2-3G of memory. And that's just the most visible apps/bugs.

But I digress.

Ok, this is very frustrating. I reported this identical behavior here:

https://bugs.launchpad.net/ubuntu/+source/evolution/+bug/424099

"Apport retracing service" reported the following:

> "This crash has the same stack trace characteristics as bug #359658.
> However, the latter was already fixed in an earlier package version
> than the one in this report. This might be a regression or because
> the problem is in a dependent package."

So, does the phrase "already fixed in an earlier package version" mean something different on Launchpad than it would pretty much anywhere else? What am I missing here? Because a few days later it was marked as a duplicate of this bug. Is it the same bug, and was simply never fixed? Was the info about it being fixed simply wrong?

Also, to add to the symptoms, I have definitely determined over the past month that, in my case anyways, Evolution *only* crashes when a) vlc is running, and b) a new email comes in. I tried turning off the notifier applet to see if that helped, but it still crashes randomly under those circumstances. It is usually within 30-45 minutes that it crashes, although of course that is dependent on whether or not I actually get an email in that timeframe. Sometimes apport generates a report on the crash, sometimes not.

Can anyone else confirm that they are experiencing the same symptoms? Or can anyone else say that they definitely do not have vlc and yet still get this issue?

I am available to help test, if someone wants to give me a scenario to set up, if it will help.

> What backporting? The dude had a package already created and ready to go. The only thing left to do was actually give it to the users.

The stable update process require testing and validation of the updates by different team to assure there is no regression in the change

> So the current situation, a current release which most users are using, rife with bugs (that have been fixed and are just floating around in PPAs even!) is better?

The situation is not ideal but that's what can be done with the current manpower knowing that ubuntu doesn't write most of the software it's shipping you get unhappy users either way either because they lack features available in new versions if you work on old version only by fixing bugs or because bugs are not fixed on all versions when working on updates

let's also not turn this bug in a discussion about updates, the bottom line will be a manpower limitation anyway and a choice to do between updates, the number of softwares you ship and the versions you work on...

unassigning from ted since he's not working on the jaunty update there

on a side note is somebody wanting to work on the jaunty update?

Brian J. Murrell [2009-10-08 14:40 -0000]:
> What backporting? The dude had a package already created and ready to
> go. The only thing left to do was actually give it to the users.

Where? There was a wrong patch sent here, but no jaunty package or
debdiff ever...

I will work on getting Ted's fix ported to Jaunty.

I have published evolution-indicator 0.1.13-0ubuntu2~ppa1 on my PPA (https://edge.launchpad.net/~hggdh2/+archive/ppa?field.series_filter=jaunty) exclusively for Jaunty.

I would like someone to test and confirm it is working. I will only provide a debdiff when I have such a confirmation.

Thank you.

I've installed evolution-indicator_0.1.13-0ubuntu2~ppa1_amd64.deb (http://ppa.launchpad.net/hggdh2/ppa/ubuntu/pool/main/e/evolution-indicator/evolution-indicator_0.1.13-0ubuntu2~ppa1_amd64.deb) on my workstation.

I will report back in a few days.

Why don't you use evolution-indicator_0.1.15-0ubuntu1? It works fine zero chrashes for me.

2009/10/12 greenhunter <email address hidden>:
> Why don't you use evolution-indicator_0.1.15-0ubuntu1? It works fine
> zero chrashes for me.

Dunno - aptitude never offered to upgrade. I'll check why.

Thanks for the tip.

Cheers,
Thiago.

2009/10/12 greenhunter <email address hidden>:
> Why don't you use evolution-indicator_0.1.15-0ubuntu1? It works fine
> zero chrashes for me.

Because e-i 0.1.15 is for Karmic. And it is obsolete, anyway.

Although you personally can grab a version for a newer release and use it, we cannot.

Unassigning myself from the Jaunty task. There does not seem to be a lot of interest in this fix for Jaunty.

C de-Avillez wrote on 2009-10-09:
> I have published evolution-indicator 0.1.13-0ubuntu2~ppa1
> on my PPA (https://edge.launchpad.net/~hggdh2/+archive/ppa?
> field.series_filter=jaunty) exclusively for Jaunty.

> I would like someone to test and confirm it is working. I will only
> provide a debdiff when I have such a confirmation.

> Thank you.

Sorry for not testing this sooner. I just updated tonight using this fix and tested, and I can definitely confirm that it did *not* work for me.

What I cannot confirm, however, is that I gave the same bug as what everyone else is reporting. Evolution *only* crashes for me if VLC media player is running, and a new email comes in. Under those circumstances it crashes about 1-2 out of every 3 times, and it still happens after this update.

If this is not what others are experiencing, then the bug I originally reported (https://bugs.launchpad.net/ubuntu/+source/evolution/+bug/424099) was marked as a duplicate in error. If it is what they are experiencing, then this fix does not work.

@mvandemar: thank you for your feedback. Indeed, bug 424099 seems to be this one -- which means the fix I cherry-picked did not do the trick. I will have to go back to it. :-(

Also, did the failure produce a crash (i.e., was apport triggered)? If so, I would like you to open a new bug via apport, and report here the bug #. I would like to see the backtrace.

Bug 424099 contains a backtrace that I submitted via apport. I can try and force another crash and see if it comes up again, but I think I quit seeing the apport window for this crash a while ago. Is it possible that it got flagged as "already submitted" on my machine, or that I inadvertently told it not to display the window anymore, and if so, how can I clear that?

Also, is there a way to show a more complete report of what is going on at the time of the crash? If the last apport didn't show it, I am not sure what a new one would have that would be differently. I am willing to install whatever debugging/tracing apps you need me to.

bug 424099 has the crash *before* applying my proposed e-i fix, so it will not help me here. I need the crash with my fix in place, so that I can follow the backtrace. I checked the build logs, and the fix is indeed being applied.

One thing you can do: please verify if under /var/crash you have *any* files matching the pattern '*evolution-indicator*'. If you do:

1. please remove these files;
2. run 'sudo force_start=1 /etc/init.d/apport restart' to enable apport crash triggering on Jaunty;
3. repeat the error. Apport should trigger, and offer to report the bug. Please do so, and give me the bug number.

Additionally, if you want to try local debugging: the easiest way would be, after the crash (and after submitting the new bug), to run 'sudo apport-retrace -g /var/crash/<whatever is the crash file name>'. This will open GDB, and you can then execute:

  bt full

and

  thread apply all bt full

and give me the output.

Thank you.

And now, of course, I cannot replicate the crash. I am starting to think that I may have had Evolution running when I did the update, failed to restart before testing, and therefore the process running when it crashed was still the pre-patch one.

I will continue to test over the next couple of days, but as I said before it was not difficult to get it to bomb before at all.

Thank you. This is actually one thing I thought of, but decided to wait for the backtrace. Yes, this might have something to do with it.

As you propose, let's wait a couple of days. Thank you, very much.

Zero crashes so far. Looks like it is working fine.

Good. I am attaching the debdiff for this. Thank you very much.

Thanks! Uploaded.

Accepted evolution-indicator into jaunty-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Hello there,

In Ubuntu 904x86_64, Evolution just crashed. Nothing unusual happened before the crash. The process was the following:

1- Turn on Computer.
2- Start Evolution and Empathy.
3- Check, write and send some email in Evolution.
4- Launch Xsane in another Workspace.
5- Before pressing the "Preview" in Xsane, the crash came out.

Other things happening in Evolution are:
1- It keeps Folders in the IN Folder marked as "has emails not read" when it has none until evolution restarts.
2- Every time the app starts, checking contact list and selecting 1 contact from the list makes the contact list scroll up and down from begining to end of list before showing the contact information on the Preview Pane.
3- If the Preview Pane is not shown, the contact list just does de scroll up and down for several times and then stops.

Thanks all.
Best regards

Xermán

Anyone who can test the proposed package? If not, the proposed update will be removed again. Thanks!

This update has been in jaunty-proposed for half a year or longer, without any testing feedback. I removed the jaunty-proposed package again.

I'm sorry, but I didn't encounter this problem again on my system.

My personal opinion as a developer is that changing bad code for good code is a good idea, and even if the original problem is not reproduceable and even if the patch is not being tested, that fix still makes sense and should go in - preventing SIGSEGV should be no. 1 priority of developers.

Here's the GNOME official documentation of g_str_equal() (the call being replaced by the patch):
---8<---
Note that this function is primarily meant as a hash table comparison function. For a general-purpose, NULL-safe string comparison function, see g_strcmp0().
---8<---

Which as I read it means: don't ever use g_str_equal() as a general string comparison method. IMHO no testing feedback should be required to apply this patch, because it makes sense the original code is bad.