| benji | 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 | 12:20 |
|---|---|---|
| gary_poster | benji, somehow missed that IRC msg, but I see it now, thanks. I'll go approve | 12:27 |
| benji | thanks | 12:27 |
| gary_poster | approved | 12:28 |
| gary_poster | bac benji danilos gmb, call in 2 | 12:28 |
| bac | ko | 12:28 |
| gmb | ok | 12:28 |
| danilos | gary_poster, actually I am on, headset doesn't work | 12:30 |
| danilos | gary_poster, yeah, you are on the speakers :) | 12:31 |
| danilos | hanging up so I fix that :) | 12:31 |
| gary_poster | danilos keeps rejecting my Skype overtures | 12:32 |
| danilos | sorry | 12:33 |
| danilos | my microphone still doesn't work | 12:33 |
| danilos | gary_poster, haha :) | 12:34 |
| danilos | gary_poster, incidentlog (linked from it)? | 12:34 |
| danilos | gary_poster, I'll check it up, ok | 12:34 |
| danilos | danilos, I have plenty of things to say! :) | 12:37 |
| danilos | gary_poster, ^ | 12:37 |
| danilos | gary_poster, nah, just kidding, thanks :) | 12:38 |
| danilos | ok, microphone fixed, I had to install libpulse0:i386 now that skype is installed in an entirely different way | 12:50 |
| === benji___ is now known as benji | ||
| benji | oops, I logged into the Canonical server twice | 13:09 |
| * gmb -> late lunch / errands | 13:13 | |
| gary_poster | early early lunch (== taking care of baby while wife is out) | 14:59 |
| bac | gary_poster: got a sec for a skype? | 19:06 |
| gary_poster | sure bac | 19:06 |
| gary_poster | logged on now bac | 19:06 |
| benji | 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. | 19:26 |
| bac | gary_poster: https://pastebin.canonical.com/53482/ | 19:34 |
| gary_poster | benji, it sends an email | 19:34 |
| bac | benji: like all PQM submissions it is via email | 19:34 |
| gary_poster | bac, looking | 19:34 |
| gary_poster | bac, try | 19:35 |
| gary_poster | !print rec | 19:35 |
| benji | 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 |
| gary_poster | oh, but that gives you more than you want bac | 19:35 |
| gary_poster | benji, that sounds suspucious, certainly | 19:35 |
| benji | ("p rec" is easier to type) | 19:35 |
| benji | (and "pp rec" is pretty-printed) | 19:36 |
| gary_poster | the pretty print won't help here; it is a string | 19:36 |
| gary_poster | so in fact !print is nicer | 19:37 |
| gary_poster | (p will give you \n s in it and stuff) | 19:37 |
| benji | sounds like you need ppp ;) | 19:37 |
| gary_poster | :-) | 19:37 |
| benji | oh you're right, I always forget that "p" is really the repr, that's stupid | 19:37 |
| gary_poster | except when it's not :-) | 19:37 |
| bac | gary_poster: hey, cool, the __repr__ is much nicer than the parts | 19:38 |
| gary_poster | yeah | 19:38 |
| bac | let me paste that | 19:38 |
| bac | and you don't need the ! :) | 19:38 |
| gary_poster | heh, I am paranoid in pdb | 19:38 |
| benji | well, if "p" did print, then if you wanted the repr you could just do "!rec" | 19:38 |
| gary_poster | true, benji | 19:39 |
| bac | https://pastebin.canonical.com/53484/ | 19:39 |
| benji | I could just make an alias to override "p". I guess an economist would say that I therefore don't really care. | 19:41 |
| gary_poster | 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 | 19:41 |
| gary_poster | 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 |
| bac | 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 |
| gary_poster | so in flush I might stash obj_info | 19:43 |
| bac | ok | 19:44 |
| gary_poster | bac, right, what we want to know is when the _dirty is mutated, and then by whom | 19:44 |
| bac | ok | 19:44 |
| bac | so, just create a local var called __traceback_info__? | 19:45 |
| gary_poster | bac, then in _flush_one we could scribble expr in there I guess...thinking... | 19:45 |
| gary_poster | yeah, bac, it will be str'ed | 19:46 |
| gary_poster | bac, here's my crazy plan, I think: | 19:47 |
| gary_poster | you do that, then we will have the expression that is sent initially through _flush_one | 19:48 |
| gary_poster | 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 |
| gary_poster | then we look at _dirty. | 19:49 |
| gary_poster | then after each line we look at dirty, being particularly suspicious of _enable_change_notification and _add_to_alive | 19:49 |
| gary_poster | until we see when _dirty has been mutated | 19:50 |
| gary_poster | then we know what to step into | 19:50 |
| gary_poster | then we re-run, and we actually step into it. Or similar | 19:50 |
| gary_poster | bac, end of crazy plan. | 19:50 |
| bac | ok, so in flush_one i've put in obj_info | 19:50 |
| bac | into the traceback | 19:50 |
| gary_poster | also expr, bac | 19:50 |
| gary_poster | of | 19:51 |
| gary_poster | oh | 19:51 |
| gary_poster | that won't work | 19:51 |
| bac | trying...see what happens | 19:52 |
| gary_poster | because it is an instance | 19:52 |
| gary_poster | bac, so maybe "changes"? ok | 19:52 |
| bac | gary_poster: btw, for those following along at home, the endoscopy didn't work so the vet is doing real surgery. :( | 19:52 |
| gary_poster | bac :-( | 19:53 |
| gary_poster | bac, what is the object? | 19:53 |
| bac | benji: over in #lp-dev sackett is complaining of pqm woes. maybe y'all should compare notes. | 19:58 |
| benji | bac: thanks, I hadn't noticed | 19:58 |
| gary_poster | bac, about to get on call with flacoste | 20:07 |
| gary_poster | will check in with you after | 20:07 |
| gary_poster | bac, I'm off call | 20:47 |
| bac | gary_poster: got distracted with some CHR stuff from abentley | 20:49 |
| gary_poster | np bac, I have my own mysteries (Trying to work with the new pgbouncer fixture) | 20:51 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!