* danilos -> lunch10:49
gary_posterbac benji danilos (gmb sick :-( ): call in 212:29
benjig*ry: RT is 4819612:38
danilosbac is fighting irc problems, he'll be unavailable on IRC while that's going on12:48
gary_posterdanilos, thanks for update on bac.  I need to take kids to school today, so I will be a bit late.  Will 20 minutes past the hour still work for you, or should we reschedule?12:50
danilosgary_poster, that's fine12:57
gary_posterthanks danilos.  ready any time13:18
danilosgary_poster, hey, still not used to non-existent notifications :)13:24
gary_posterbenji, when would you like a call this morning?  Any time is good for me.13:57
benjigary_poster: sure; how about now?13:58
gary_posterbenji, a fine idea.  Gimme a call!13:58
bachi danilos14:02
bacdanilos: the original failing test is bugs/doc/bugtask.txt at line 335.  the scenario presented there is the one i've tried to recreate in the unit test.14:03
danilosbac, right, let me take a look at as well14:12
danilosbac, hum, right, I think I see the point now14:25
bacdanilos: my understanding of conjoinedness is not deep, so i don't deny that i may have gotten them backwards but the doctest regression looks real14:26
bacbtw, anyone have an idea how long fsck should take on a 1TB disk?14:28
danilosbac, yeah, it all seems right, in this special case when a master is being created after slave, it copies stuff from slave back to master14:34
danilosbac, so, I think the problem is the storm validator: createTask creates a task with status set to NEW, which calls on the storm validator for status field, which changes the slave version back to the just-created master version14:36
bacdanilos: yes, i was looking at that...before my server died14:37
bacdanilos: but none of that code has changed...14:39
danilosbac, well, at least the column is being renamed from status to _status :)14:43
bacdanilos: yeah, there is that...14:44
danilosbac, fwiw, I can also see arbitrary SQL queries using ._status and others using .status which kind of confuses me14:48
danilosthat could be storm vs sqlobject stuff, but just very confusing14:49
=== bac`` is now known as bac
danilosbac, so, I could confirm that the NEW status is set by the validator, and what's else validate_status is called 9 times now, whereas stable did it 7 times for the same test (your earlier one)15:12
danilosbac, fwiw, doctest does seem to break with all those _status usages in plain-text SQL queries (as I expected) when validator is removed, so it seems the branch was more unfinished then you hoped for :)15:13
bacdanilos: yeah, that's my fear.  fix this blocking problem only to find a ton more15:14
danilosbac, ok, so the problem is that on stable, self.bug in validator is not set, yet in this branch it is15:41
danilosbac, basically, it seems that bug gets flushed earlier with all the changes15:43
benjiI'm going away now.  I suspect I'll keep meditating upon the visions of Python regexes which parse Perl regexes that are swirling around in my head though.16:10
bacdanilos: ah, interesting16:26
danilosbac, I actually dislike the extension to enum columns lifeless did, I find code that mixes two different DBEnums for a single value very confusing :/16:44
bacdanilos: how would you do it?16:44
danilosbac, usual class inheritance, like we do it everywhere else16:45
danilos(we can't have one class that derives from two DBEnumeratedTypes though, but we can have non-dbenum classes as base combining classes)16:47
* danilos -> off, tty tomorrow17:42
bachi gary_poster, i think (with danilos' help) i now understand what is going on...but i need to discuss it18:26
gary_posterbac, ok, cool.  now is fine, or later18:27
bacgary_poster: now would be good18:27
gary_posterwe can segue into our weekly call if you want18:27
gary_postercall me when ready bac18:28
bacgary i'll resume our call in a sec18:49
gary_posterbac ok, I'm back18:50
bacme too18:54

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!