#general
- r
rough-minister-58256
11/04/2015, 9:35 PMare we looking at the same wrapper? mine is only 88 lines 😮 - a
ancient-stone-50795
11/04/2015, 9:38 PM-.txt - f
fast-megabyte-33579
11/04/2015, 9:38 PMooh cute idea - a
ancient-stone-50795
11/04/2015, 9:39 PMour script is mostly gather some info, copy it into .pants.d/bug-report then 'tar-jcvf ...some files in .pants.d...' and file a JIRA ticket - a
ancient-stone-50795
11/04/2015, 9:39 PMOur
wrapper detects bug-report as $1 and invokes this script so it looks like you are invoking a pants goal (but you're not)./pants - h
happy-kitchen-89482
11/04/2015, 9:49 PMThe
flag captures your .pants.d and other stuff, for reproducing bugs, is that what you're referring to?--repro-capture - r
rough-minister-58256
11/04/2015, 9:50 PMnice. we have plans to stand up Sentry internally and wire pants up for automated exception reporting - guessing others could benefit from that too. - r
rough-minister-58256
11/04/2015, 9:51 PM - r
rough-minister-58256
11/04/2015, 9:52 PMwhich solves the ‘please repro with —print-exception-stacktrace’ support problem - we just automatically collect full stack traces and other environmental data - h
happy-kitchen-89482
11/04/2015, 10:52 PMOooh, I should add full stack trace to the repro - h
happy-kitchen-89482
11/04/2015, 10:52 PMcapture it, stick it in a file - r
rough-minister-58256
11/04/2015, 10:53 PM🚢 - e
enough-analyst-54434
11/04/2015, 11:10 PMAnyone have experience with this or other multiprocessing.Pool variants?: https://pypi.python.org/pypi/billiard/3.3.0.20 I implemented a means of detecting internal pool errors, namely pickle errors in its internal queues, but it requires polling AsyncResult objects - not ideal. in python 3.2+ there is an error_callback that would be just the ticket, and billiard above gives me that, but before going either path wanted to see if folks had wisdoms here - e
enough-analyst-54434
11/04/2015, 11:10 PMMy only other thought was pre-pickling outside the pool, but that also would be nice to avoid - e
enough-analyst-54434
11/04/2015, 11:17 PMerm, lied on billiard - its error_handler is not invoked for internal errors - e
enough-analyst-54434
11/04/2015, 11:18 PMI'll use the polling since this is experimental still and if anyone knows what I'm talking about / has ideas I'll circle back - r
rough-minister-58256
11/04/2015, 11:38 PMmaybe look at the newer concurrent.futures.{Future,Executor} stuff? iirc, it has better exception handling - e
enough-analyst-54434
11/04/2015, 11:41 PMAh yes, thats it - ty. Alec Thomas used this at Twitter way back, forgot there was a backport lib - m
mammoth-pharmacist-54215
11/04/2015, 11:43 PMwas the futures stuff backported to 2.7? - b
bored-art-40741
11/04/2015, 11:43 PMYes - m
mammoth-pharmacist-54215
11/04/2015, 11:43 PMooohh shiny, I didn’t realize that. - b
bored-art-40741
11/04/2015, 11:43 PMNot in the standard library - b
bored-art-40741
11/04/2015, 11:44 PMBut on pypi - m
mammoth-pharmacist-54215
11/04/2015, 11:44 PMah, that makes more sense… still cool though. - r
rough-minister-58256
11/04/2015, 11:46 PMyeah - e
enough-analyst-54434
11/05/2015, 12:36 AMNicer interface for waiting on a set of Futures using their condition variables - no polling, but same underlying issue, does not handle pickle errors - e
enough-analyst-54434
11/05/2015, 12:37 AMNice idea though - e
enough-analyst-54434
11/05/2015, 12:37 AMThank you - m
mammoth-pharmacist-54215
11/05/2015, 12:41 AMI’ve heard good things about dill for problematic pickles. https://pypi.python.org/pypi/dill Not sure what specific problems you are hitting. - m
mammoth-pharmacist-54215
11/05/2015, 12:42 AMbtw, it looks like you can already do —repro-capture on clean-all already, no work needed by me. Anyone mind if I add an option to ignore some more dirs?