Outstanding crashes

[Paul Aurich]

On 2010-10-07 04:18, Florian Quèze wrote:
> Libpurple has been upgraded from 2.6.6 (which was used for Instantbird
> 0.2) to 2.7.3 in nightly builds of Instantbird about two weeks ago, so
> the crash data that we collect may be relevant again to you.
> Our crash reports database can be queried at
> http://crash-stats.instantbird.com/query
> 
> I've checked the data this morning and if we except the uninteresting
> reports (very old versions, crashes in plugins that we know nothing
> about, ...) and the netsoul plugin crashes (already fixed, the bug was
> trivial), we are currently seeing at least 2 (or 3) crashes in
> libpurple that probably deserve some attention: 1 (or 2) new crashes
> in MSN (may be somewhat related to upnp) and one in oscar, that was
> already in libpurple 2.6.6.

> - the oscar plugin crashes in flap_connection_destroy_cb:
> http://crash-stats.instantbird.com/report/index/40bbf7ab-0ea3-4e96-9dc0-6dbf62101001
> For more similar stacks, see
> http://crash-stats.instantbird.com/report/list?product=Instantbird&query_search=signature&query_type=exact&query=&date=&range_value=2&range_unit=weeks&process_type=all&plugin_field=&plugin_query_type=&plugin_query=&do_query=1&signature=g_slist_remove
> Some variations:
> http://crash-stats.instantbird.com/report/index/d447ebe8-d025-4b3b-8b9a-964a32100924
> http://crash-stats.instantbird.com/report/index/5c95337a-21e3-4f75-938a-b25222100929
> The crashes that are in version 0.2 of Instantbird are with libpurple
> 2.6.6, the less-than-two-weeks-old crashes of version 0.3a1pre are
> with libpurple 2.7.3.
>>From what I've been told by users, this crash happens mostly when a
> laptop wakes up from sleep and is connected to a different network.
> It seems flap_connection_destroy_cb is called from a timeout after the
> connection has already been destroyed.
> I've already shown this crash in #pidgin a few months ago and it was
> suggested that https://hg.instantbird.org/instantbird/rev/f2d45147098b
> may fix it, but the crash reports we are still getting show it still
> isn't fixed.

That was me (both with the commit and suggestion in #pidgin).  It
definitely fixes one version of the patch, but the original ticket
predates the addition of SSL support for OSCAR, so clearly there's
another issue going on there.  (I'm staring at the code and just not
seeing how the crash is possible...)

I just committed bc9c4bb89eb6d36c548b61d8b1e6b8952b9c4c0e (trivial
printing out of pointer values), which will be helpful in correlating
where the timeout is coming from.  Logs+backtrace from this crash with
that patch would be awesome!

~Paul

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 897 bytes
Desc: OpenPGP digital signature
URL: <http://pidgin.im/pipermail/devel/attachments/20101029/1dc5f2b2/attachment.sig>