ruote tmp/log_2013-03-21.html

2013-03-21 01:41:29 utc phaeron jmettraux: hello
2013-03-21 02:15:00 utc phaeron jmettraux: what would happen to a concurrent iterator if an error is raised from one branch ? (answer any time you like :))
2013-03-21 02:44:00 utc jmettraux phaeron: hello, feel free to use the mailing list when there is no answer here
2013-03-21 02:44:24 utc jmettraux the branch with the error stops, the other branches go on
2013-03-21 04:40:09 utc jmettraux the error is listed in the process status
2013-03-21 09:26:40 utc phaeron jmettraux: thanks for the answer, I prefer IRC for small non-urgent things like this , and will definitely use the mailing list if there's a worthy question :)
2013-03-21 09:50:51 utc jmettraux ok, hopefully the conversation is eaten by the goog spiders via the logs
2013-03-21 21:51:02 utc jmettraux blackdog: hello, welcome to #ruote
2013-03-21 21:51:06 utc blackdog jmettraux: danke
2013-03-21 21:51:12 utc jmettraux bitte schoen
2013-03-21 21:51:34 utc blackdog we are completely abusing your tool, i fear. for us it's pretty much a way of running single tasks
2013-03-21 21:51:46 utc jmettraux the worker has its own thread, and by default, each participant, upon receiving a workitem, processes it in a dedicated thread
2013-03-21 21:52:23 utc jmettraux no worries, happy to get such issue reports
2013-03-21 21:52:58 utc jmettraux my instinct would be to wrap @seen and @log in a mutex
2013-03-21 21:52:59 utc blackdog so if you had one participant, you'd have two threads total?
2013-03-21 21:53:08 utc blackdog yeah, i think that's a sane solutio
2013-03-21 21:53:39 utc jmettraux one permanent thread for the worker, and the time the participant "performs", a second thread
2013-03-21 21:54:10 utc jmettraux I will add jruby to the ruote CI, sorry, it's been a while since someone talked about jruby + ruote
2013-03-21 21:54:24 utc blackdog cheers
2013-03-21 21:54:33 utc blackdog it's been popping up in production a few places
2013-03-21 21:54:51 utc blackdog i had a long test run recently - came up once in a thousand times
2013-03-21 21:55:01 utc blackdog which means reproduction is a colossal pain :/
2013-03-21 21:56:37 utc jmettraux argh
2013-03-21 21:57:02 utc jmettraux when I saw your profile, I was telling to myself, this guy must be in #polyhack
2013-03-21 21:57:04 utc blackdog jmettraux: am slightly worried that there are more of the same kind, though - the usual advice in jruby is not to share mutable objects across threads, right?
2013-03-21 21:57:11 utc blackdog heh. why is that?
2013-03-21 21:58:20 utc jmettraux sorry, I have been careless, cushioned by the MRI array implementation
2013-03-21 21:59:45 utc jmettraux jruby 1.7.0 should be ok, right?
2013-03-21 22:00:30 utc blackdog it's ok for me :) i'm running head, but that's pure cussedness
2013-03-21 22:11:46 utc jmettraux grrr, the tests seem to run fine, going for a second run and then a third, I was hoping the problem to show up
2013-03-21 22:18:40 utc jmettraux love Jruby, second run is way faster than first, but still error-less
2013-03-21 22:19:34 utc jmettraux I think I will wrap @seen and @log anyway, in theory it may happen, in production it happens (at your place), it's just that it doesn't happen here in test
2013-03-21 22:19:56 utc jmettraux if I CI it it would probably happen from time to time
2013-03-21 22:26:28 utc jmettraux blackdog: does that pass your sanity check?
2013-03-21 22:36:28 utc blackdog jmettraux: yes, i think so
2013-03-21 22:37:20 utc jmettraux ok, I will push that then, muchas gracias, if you spot any more of those, please ping me
2013-03-21 22:37:29 utc blackdog shall do.
2013-03-21 22:45:55 utc jmettraux linking conversation to modification:
2013-03-21 22:48:04 utc jmettraux and to issue:
2013-03-21 22:50:34 utc blackdog thank you :)
2013-03-21 22:51:23 utc jmettraux you're welcome