pidgin evolution plugin, Evolution hang

Backtrace obtained from attaching to hung pidgin process with GDB. Killing evolution-data-server-2.26 caused an immediate revival pidgin.

Thank you for reporting this bug and helping make Ubuntu better.

I do remember hearing about this some time ago, and will search for it both here and upstream. Right now I am setting this bug as confirmed, and will either dup it against a previous, still open, Ubuntu bug, or move it to triaged, with a link upstream.

Here's a summary of what I remember:

the Automatic Contacts plugin (source ./plugins/bbdb) does have some problems (and needs some love).

There were some issues I remember, with the plugin enabled:

(1) synchronisation between Pidgin and Evolution is extremely simplistic, and the entries must match exactly. If they do not match, the plugin will re-add them;

(2) the plugin will drive resynchronisation every two minutes; this value is hardcoded in the plugin, and can only be changed in the source code;

(3) the search for a match is linear; as a result, the bigger your addressbook, the longer it will take. A match will be looked for each of the buddies in Pidgin.

So, if you add (2) and (3)... there is a potential for long loops...

Right now, I strongly suggest disabling Pidgin synchronisation.

On Sun, 2009-05-03 at 19:30 +0000, C de-Avillez wrote:
> Thank you for reporting this bug and helping make Ubuntu better.
>
> I do remember hearing about this some time ago, and will search for it
> both here and upstream. Right now I am setting this bug as confirmed,
> and will either dup it against a previous, still open, Ubuntu bug, or
> move it to triaged, with a link upstream.
>
> Here's a summary of what I remember:
>
> the Automatic Contacts plugin (source ./plugins/bbdb) does have some
> problems (and needs some love).
>
> There were some issues I remember, with the plugin enabled:
>
> (1) synchronisation between Pidgin and Evolution is extremely
> simplistic, and the entries must match exactly. If they do not match,
> the plugin will re-add them;
>
> (2) the plugin will drive resynchronisation every two minutes; this
> value is hardcoded in the plugin, and can only be changed in the source
> code;
>
> (3) the search for a match is linear; as a result, the bigger your
> addressbook, the longer it will take. A match will be looked for each of
> the buddies in Pidgin.
>
> So, if you add (2) and (3)... there is a potential for long loops...
>
> Right now, I strongly suggest disabling Pidgin synchronisation.

While I have done that, and will report back if it fixes the problem, I
don't think that's the source here, unless you got that from the
backtrace. I had a crash-free Pidgin after I disabled the Evolution
Integration plugin in _Pidgin_, not the Pidgin Integration plugin in
_Evolution_. Are you sure it's the same issue?

Marking as duplicate of bug 235250

Now I am not. I did not even know there was a evolution integration plugin on Pidgin. This is actually something to look at, and I will install Pidgin and check (even more because then we might have two synchronisation agents running concurrently).

Meanwhile, let's keep this one as it is; please report back on what you find.

Judging from the

>#0 0x00007f557f7cf2e9 in pthread_cond_wait@@GLIBC_2.3.2 ()

in the backtrace, two synchronisation agents running concurrently sounds very plausible if a deadlock is what's causing pidgin / evolution to hang. Since disabling the Evo->Pidgin plugin I haven't had any crashes, but since the crashes weren't coming at a fixed interval I wouldn't put much stock in that.

Either way, I don't think this is really a duplicate of #157207, as I'm not having any of the problems listed in that bug.

I reset the duplicate status. I am still not sure about not being a duplicate, though; the fact that disabling Evo's plugin (at least) alleviated the issue still suggests (2) and/or (3) of the three cases I pointed out -- if you have a large addressbook/Pidgin contact list, then there is a good chance this might be happening -- this might look as a freeze.

It is also interesting to note that this backtrace was related to the Pidgin plugin -- see frame #9.

I still do not really like having two plugins that do the same thing (or, more probably, a similar thing). I have not looked at the Pidgin code, but one could expect -- or hope -- it to be a bit more optimised, and only call for an add-to-addressbook if there is a new contact (while Evo drives a search every two minutes).

Nevertheless, we still should try to zero in the issue:

(1) conflict/deadlock between both plugins;
(2) Evo plugin issue;
(3) Pidgin plugin issue;
(4) EDS library issue.

I am not sure the culprit is EDS, not at all.

Back to incomplete status.

Alright, even with the Pidgin Integration plugin in Evolution disabled, Pidgin still crashed, and a checkin with Evolution confirmed it was crashed as well. I'm attaching simple backtraces from Pidgin, Evo, and e-d-s.

As usual, after killing e-d-s, both Pidgin and Evolution started back up, with Evolution giving me an error about my address book being inaccessible.

First, to Andrew: sorry, I mistakenly assigned to ubuntu-desktop instead of desktop-bugs. My fault.

Ted, thank you. I understand you took the backtraces at the time you saw it, with *only* the Pidgin plugin enabled.

* Unfortunately, the Evo BT is inconclusive -- you did not have the debug symbols loaded for Evo, it seems, and I do not know which thread was shown. No matter that, per the few identifiable frames, I would venture this specific thread was in wait.

* The EDS trace suggests this specific thread was trying to acquire a lock, and entered a wait for it.

* finally, the Pidgin trace shows what I think is part of the contacts plugin (the library gevolution.so) driving a call to the EDS libraries; this thread is waiting on a response.

It is difficult to say what is going on on the Evo and EDS BTs, since a single thread was traced.

I will try and install & enable the Pidgin Evolution plugin, and see if I can get a nice set of BTs. But, right now, I am tending to reassing this to Pidgin (or, having time, look upstream for similar bugs): the driver here is Pidgin, not EDS. It may well be a conflict on Evo (with the addressbook) and Pidgin, but I lack Pidgin knowledge to go there; anyway, it *works* when Evo calls on addressbook functions, so... until proved otherwise, it would be Pidgin...

I am also retitling this bug.

Here are new backtraces, this time with debugging symbols installed. The
Evolution one is about the same, but there's much more info in the
pidgin and eds ones.

thank you, Ted. For the record, I have been running with the pidgin plugin enabled, without any glitches.

Heh. Just noted I have been running Evo trunk, not Jaunty Evo. Will revert to Jaunty, and keep on trying.

These crashes are still happening (in evolution-data-server) without either the Evolution->Pidgin plugin or the Pidgin->Evolution plugin active. Attached is a trace of evolution-data-server.

i have the same bug with 9.10 and uptodate packages.

I did a ubuntu-bug report, but I'm not sure if it is attached.

If i can help with additional information - please feel free to ask.

[Expired for pidgin (Ubuntu) because there has been no activity for 60 days.]

Jaunty not longer supported
change status to invalid