fedora-meeting
LOGS
15:20:32 <adamw> #startmeeting Bugzappers 2010-02-02
15:20:32 <zodbot> Meeting started Tue Feb  2 15:20:32 2010 UTC.  The chair is adamw. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:20:34 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:20:41 <adamw> #topic sekrit meeting
15:20:50 <adamw> for the logs, we've been having a not-a-meeting-meeting for 20 minutes now
15:20:55 <adamw> so we're making it formal
15:21:15 <adamw> the only person with an agenda item who's here is me, so:
15:21:21 <iarlyy> adamw, akia is not able to write on channel
15:21:23 <adamw> #topic bugs filed on orphans
15:21:30 <adamw> I've done nothing!
15:21:36 <adamw> #info he's done nothing
15:21:37 <iarlyy> his told me now
15:21:44 <adamw> #topic open floor
15:21:49 <adamw> alrighty, there we go
15:21:55 <adamw> iarlyy: he needs to register with nickserv
15:22:08 <adamw> iarlyy: unregistered nicks can't post to public channels currently (to deal with the bot crap)
15:22:35 <azneita> i love the sekrit meeting
15:22:44 <adamw> :)
15:23:06 <adamw> #topic open floor - serega's abrt proposal
15:23:10 <adamw> serega: take it away
15:23:19 <iarlyy> adamw, thats the problem, thanks = )
15:24:04 <serega> ok. recently I have started to triage bugs for the empathy components. Initially (well, and still) the bug list was almost 100% abrt crash reports
15:24:37 <adamw> what kind of numbers are we talking here? in reports-per-day terms?
15:25:02 <serega> the backtrace abrt provides is quite hard to read due to the function argument expansion, so quickly I went to a python script which cleans the backtrace up, leaving only the function names stack
15:25:39 <serega> adamw: when I got to the triage there was 84 bugs already filed
15:26:21 <adamw> serega: can you provide a link for an example of that (script)?
15:26:51 <serega> adamw: sec, I'll find the bugreport with result
15:27:00 <serega> *a bugreport
15:27:08 <adamw> thanks
15:28:33 <serega> ah, here: bug #558883
15:28:49 <adamw> #link https://bugzilla.redhat.com/show_bug.cgi?id=558883
15:28:58 <adamw> #info example of serega's script
15:29:00 <adamw> carry on!
15:29:11 <serega> see the "cleaned backtrace" attachment
15:30:18 <iarlyy> serega, is other attachment the original?
15:30:29 <serega> the clean up backtraces are much easier to memorize, because you can look over the whole trace. Memorizing is dead critical for detecting dupes among abrt reports
15:30:42 <serega> let's move on
15:32:09 <adamw> ok
15:32:32 <serega> Finally, after 25 reports I got a feeling I can't hold all different things in my head. Now I'm thinking about the script which could compare two stack traces and compute their _similarity_
15:32:40 <serega> like... 90% similar
15:33:15 <serega> that's useful for reports that are different due to some symbol absense, but naturally the same
15:33:21 <mcepl> serega: I would suggest filing a bug against abrt... they may incorporate your script directly into abrt
15:34:26 <serega> mcepl: good point, but that is not the end
15:34:49 <mcepl> sure
15:35:51 * adamw waits with bated breath :)
15:36:08 <serega> finally, I keep the local copies of bug report backtraces I have triaged and search trough them with grep. Not very efficient and not multiuser either, heh :) I would like to have an ability to find similar bureports via bugzilla
15:36:25 <serega> adamw: breathe, please
15:37:10 * adamw turning blue
15:37:15 <adamw> hi akiase - working yet?
15:37:59 <adamw> serega: that would probably work best as some kinda jetpack thing
15:38:12 <mcepl> serega: there are plans to move all abrt backtraces to separate server, so as not to overload bugzilla (and bug triagers). Something similar to http://crash-stats.mozilla.com/
15:38:53 <serega> also, abrt when it files a report could add comments to the bugs which are seems to be similar
15:39:57 * serega apologies for his english :(
15:40:16 <adamw> mcepl: when you say 'there are plans'
15:40:30 <serega> how can I move on the idea? Wiki page?
15:40:41 <adamw> mcepl: is anyone actually working on it yet? :)
15:41:02 <adamw> serega: I would say, thread on the mailing list, wiki page in your personal space, code in one of the qa git repos
15:41:27 <serega> ok
15:41:44 <serega> I think I'm set
15:41:46 <adamw> if you mention it in your mailing list post jlaska or someone will probably help you set up a git repo on one of the qa trac spaces or smth
15:41:48 <akiase> (test)
15:41:51 <adamw> yay akiase!
15:42:06 <mcepl> adamw: I don't know, I am talking about it with jmoskovc on the internal IRC... and yes, there is a lot of work done there
15:42:16 <akiase> finally! :D It took the whole meeting...
15:42:37 <serega> well, open floor :)
15:43:18 <adamw> serega: this script of yours seems neat too
15:43:30 <adamw> mcepl: could it be incorporated into the jetpack and made into an idiot button?
15:43:35 <adamw> (click it and it works)
15:44:07 <serega> adamw: thanks
15:44:26 <mcepl> what? serega's script ... depends on what exactly it is. I do something similar myself (well, not similar, but I do parse abrt backtraces).
15:44:31 <serega> ah, I'm working on the chromium BugZapper plugin though
15:45:53 <serega> mcepl: it just strips out function arguments with regexp
15:45:54 <iarlyy> serega, good, let me know when alpha/beta version is available  heh
15:46:09 <mcepl> serega: this https://bugzilla.redhat.com/show_bug.cgi?id=543875#c2 is automatically generated
15:46:13 <serega> iarlyy: yup, I'll announce it on the test list
15:47:00 <mcepl> serega: well, I am not sure about it ... it seems to leaving too much stuff behind, which may improve (already pretty bad) misassignments of bugs to other components.
15:47:23 <serega> mcepl: looks like a plain cite of the original backtrace. what was changed?
15:48:12 <mcepl> serega: well, calling of functions like IA__g_type_check_instance_is_a (or almost any IA___g* function) is useless without parameters
15:48:16 <serega> mcepl: I don't propose to cut the abrt reports. Just provide an additional more human-readable form
15:48:52 <serega> sure, the arguments can be critical for bug resolution by maintainers
15:48:56 <mcepl> serega: still I hear some grumbling from developers about misassigned bug reports making more work for them.
15:49:23 <mcepl> and I am afraid that this wouldn't improve our understanding of backtraces.
15:50:19 <serega> matching similar backtraces would solve the problem of misassignment to a certain degree. with a reasonable similarity threshold.
15:50:52 <iarlyy> I guess the main problem with abrt reports is the dup, isn't?
15:51:02 <serega> in particular, triage of 60 more empathy abrt reports looks like a hell :)
15:51:37 <serega> iarlyy: yes, there are quite many dupes with different checksums
15:52:36 <serega> finally, reasonable fraction of empathy crash reports seem to be gtk2 bugs
15:53:27 <adamw> serega: anything you do that could sensibly become part of abrt, jmoskovc will be interested in
15:53:44 <adamw> he's very keen to get more accurate reports coming out of abrt
15:54:28 <serega> adamw: I'll try to sort the ideas carefully and then contact him. Thanks Adam
15:54:40 <mcepl> serega: please, keep me in the loop
15:54:47 <adamw> yup mcepl too
15:54:54 <adamw> mcepl: is there an abrt ml that you're on?
15:55:03 <mcepl> adamw: unfortunately, not
15:55:08 <adamw> maybe there should be...
15:58:58 <adamw> ok, so are you happy about having somewhere to go with this for now, serega?
15:59:12 <mcepl> adamw, serega: https://fedorahosted.org/abrt/wiki
15:59:35 <mcepl> https://fedorahosted.org/mailman/listinfo/crash-catcher
16:01:10 <adamw> #link https://fedorahosted.org/mailman/listinfo/crash-catcher
16:01:20 <adamw> looks like the place
16:02:06 <adamw> #info serega working on scripts to process backtraces for dupe detection and easy reading, will work with mcepl and abrt team to take it further
16:02:20 <adamw> #topic open floor
16:02:29 <adamw> did anyone have something else to discuss?
16:02:58 <iarlyy> adamw, no from me
16:03:24 <azneita> none from me either
16:03:25 <akiase> not for now from me...
16:04:08 <adamw> mcepl, anything?
16:04:21 <mcepl> nope
16:05:42 <adamw> alrighty
16:05:43 <adamw> #endmeeting