[12:20] gary_poster: a church friend is taking Katie to the one doctor appointment so I just added a half vacation day for Thursday afternoon to take her to the other [12:27] benji, somehow missed that IRC msg, but I see it now, thanks. I'll go approve [12:27] thanks [12:28] approved [12:28] bac benji danilos gmb, call in 2 [12:28] ko [12:28] ok [12:30] gary_poster, actually I am on, headset doesn't work [12:31] gary_poster, yeah, you are on the speakers :) [12:31] hanging up so I fix that :) [12:32] danilos keeps rejecting my Skype overtures [12:33] sorry [12:33] my microphone still doesn't work [12:34] gary_poster, haha :) [12:34] gary_poster, incidentlog (linked from it)? [12:34] gary_poster, I'll check it up, ok [12:37] danilos, I have plenty of things to say! :) [12:37] gary_poster, ^ [12:38] gary_poster, nah, just kidding, thanks :) [12:50] ok, microphone fixed, I had to install libpulse0:i386 now that skype is installed in an entirely different way === benji___ is now known as benji [13:09] oops, I logged into the Canonical server twice [13:13] * gmb -> late lunch / errands [14:59] early early lunch (== taking care of baby while wife is out) [19:06] gary_poster: got a sec for a skype? [19:06] sure bac [19:06] logged on now bac [19:26] How does ec2 land communicate with PQM? I had an ec2 run that ended without me getting any notification. It at least started and ran for the expected amount of time. [19:34] gary_poster: https://pastebin.canonical.com/53482/ [19:34] benji, it sends an email [19:34] benji: like all PQM submissions it is via email [19:34] bac, looking [19:35] bac, try [19:35] !print rec [19:35] hmm, I recently changed my bzr whoami setting (because ec2 land stopped working because my GPG key has my benji@benjiyork.com address not my @canonical.com address); I wonder if that is the problem. [19:35] oh, but that gives you more than you want bac [19:35] benji, that sounds suspucious, certainly [19:35] ("p rec" is easier to type) [19:36] (and "pp rec" is pretty-printed) [19:36] the pretty print won't help here; it is a string [19:37] so in fact !print is nicer [19:37] (p will give you \n s in it and stuff) [19:37] sounds like you need ppp ;) [19:37] :-) [19:37] oh you're right, I always forget that "p" is really the repr, that's stupid [19:37] except when it's not :-) [19:38] gary_poster: hey, cool, the __repr__ is much nicer than the parts [19:38] yeah [19:38] let me paste that [19:38] and you don't need the ! :) [19:38] heh, I am paranoid in pdb [19:38] well, if "p" did print, then if you wanted the repr you could just do "!rec" [19:39] true, benji [19:39] https://pastebin.canonical.com/53484/ [19:41] I could just make an alias to override "p". I guess an economist would say that I therefore don't really care. [19:41] OK, so that's as we expected, to some degree, bac: the first thing that happens is the insert, then something else happens. I would personally just be stupid and put a pdb right before the flush now. Another approach would be to instrument the storm stuff and then put a conditional exactly when what we care about happens. I'll describe that, since it might be more practical [19:43] so the easiest thing to do to add instrumentation in these tracebacks is to make a variable that has the name __traceback_info__, with the values that you care about, bac [19:43] gary_poster: i've stepped into the flush at the point where we're concerned but didn't see anything suspicious, just the one newly created bugtask in the _dirty list not the one for generic_task [19:43] so in flush I might stash obj_info [19:44] ok [19:44] bac, right, what we want to know is when the _dirty is mutated, and then by whom [19:44] ok [19:45] so, just create a local var called __traceback_info__? [19:45] bac, then in _flush_one we could scribble expr in there I guess...thinking... [19:46] yeah, bac, it will be str'ed [19:47] bac, here's my crazy plan, I think: [19:48] you do that, then we will have the expression that is sent initially through _flush_one [19:49] then in _flush_one, right before _connection.execute, you set a pdb if that expression is there. That way we don't have to press a mind-numbing collection of "continue"s in pdb [19:49] then we look at _dirty. [19:49] then after each line we look at dirty, being particularly suspicious of _enable_change_notification and _add_to_alive [19:50] until we see when _dirty has been mutated [19:50] then we know what to step into [19:50] then we re-run, and we actually step into it. Or similar [19:50] bac, end of crazy plan. [19:50] ok, so in flush_one i've put in obj_info [19:50] into the traceback [19:50] also expr, bac [19:51] of [19:51] oh [19:51] that won't work [19:52] trying...see what happens [19:52] because it is an instance [19:52] bac, so maybe "changes"? ok [19:52] gary_poster: btw, for those following along at home, the endoscopy didn't work so the vet is doing real surgery. :( [19:53] bac :-( [19:53] bac, what is the object? [19:58] benji: over in #lp-dev sackett is complaining of pqm woes. maybe y'all should compare notes. [19:58] bac: thanks, I hadn't noticed [20:07] bac, about to get on call with flacoste [20:07] will check in with you after [20:47] bac, I'm off call [20:49] gary_poster: got distracted with some CHR stuff from abentley [20:51] np bac, I have my own mysteries (Trying to work with the new pgbouncer fixture)