#random
- m
mammoth-pharmacist-54215
04/04/2016, 10:15 PMyou have a link? I saw it because of your tweet. - h
happy-kitchen-89482
04/05/2016, 12:39 AM - w
wide-energy-11069
04/05/2016, 12:46 AMupvoted - m
mammoth-pharmacist-54215
04/05/2016, 12:53 AMupvoted - h
happy-kitchen-89482
04/05/2016, 1:08 AMThanks! simple smile - h
happy-kitchen-89482
04/15/2016, 8:10 AM - h
happy-kitchen-89482
04/15/2016, 8:10 AMI especially liked Hieronymus Bosch's “A visual guide to the Scala language” - d
dry-analyst-73584
04/15/2016, 4:42 PM - d
dry-analyst-73584
04/15/2016, 4:57 PMI wonder what deep dream would make of dali? - d
dry-analyst-73584
04/15/2016, 4:57 PMOh, the internet already tried that: - d
dry-analyst-73584
04/15/2016, 4:58 PMAnd the answer is dogs. - d
dry-analyst-73584
04/15/2016, 4:58 PM - c
cold-lizard-88792
06/21/2016, 9:09 PM - a
ancient-stone-50795
06/22/2016, 3:46 PMSo, some of the points in this article don't make a lot of sense. If you have the shader configured correctly, it merges together stuff in META-INF/services and the behavior of choosing the first thing on the classpath is in general the same way the classloader works. - a
ancient-stone-50795
06/22/2016, 3:47 PMThe process of combining jars is slow... unless you use the copy jar trick that we do in jar-tool and in that case it is pretty darn fast. Yes, the upload time is slow, but... if we shipped separate jars but still wanted a hermetic build we'd have to ship all of those files in separate jars anyway. - a
ancient-stone-50795
06/22/2016, 3:48 PMIt would be cool to do the "SlimFast" thing, but it adds a layer of complexity. Our build times are already under their "fast"times. Way faster. - a
ancient-stone-50795
06/22/2016, 3:49 PMOur largest project takes 14 minutes to build package and upload - a
ancient-stone-50795
06/22/2016, 3:49 PMMost projects are under 5 minutes - a
ancient-stone-50795
06/22/2016, 3:49 PMThat includes running tests - w
witty-crayon-22786
06/22/2016, 4:49 PManother point: our hadoop team begged us to push people off fatjars, but now they're complaining that deploy bundles push ~2000 jars to the hadoop distributed cache - w
witty-crayon-22786
06/22/2016, 4:50 PMstill the right thing to do in this case, but ironic maybe - m
mammoth-pharmacist-54215
06/28/2016, 12:53 AM@limited-country-20626: @witty-crayon-22786 when you get a moment would you mind taking a look at https://rbcommons.com/s/twitter/r/4015/ and let me know if you are okay with the change. - l
limited-country-20626
06/28/2016, 2:35 PMI will do that today - thanks matt - r
ripe-easter-25668
07/01/2016, 10:10 PM@bored-art-40741: hi Patrick, I am from Twitter's CI team. We had an issue when multiple Pants instances on one Jenkins slave are sharing the Ivy cache. The Ivy lock may not work effectively sometime, causing the cache to corrupt. I heard from @rough-minister-58256 and @witty-crayon-22786 that you are investigating Pants-Jenkins integrations now. Is fixing the Ivy lock on your plan? - r
ripe-easter-25668
07/01/2016, 10:11 PMThanks - b
bored-art-40741
07/01/2016, 10:40 PMHi @ripe-easter-25668. Is there a bug report/repro for this lock issue? I don't know of any problems with the lock. Our open source CI generally avoids sharing the cache with other workspaces. On the other hand, our internal CI never had issues with the lock, and it did share a cache - r
ripe-easter-25668
07/01/2016, 10:43 PMHow did you "avoid sharing the cache with other workspaces"? Did you do it by putting cache folder inside workspace and never do "git clean -fdx" after each build? Or something more smart? - b
bored-art-40741
07/01/2016, 10:43 PMWell, two things - b
bored-art-40741
07/01/2016, 10:43 PMMore or less "yes" to the first question. We do a
and agit clean -fd
, but we don't blow away the workspace's artifact cache. This is for pants' open source CI./pants clean - b
bored-art-40741
07/01/2016, 10:44 PMInternally, we share a global cache, but we also wrote our own resolver, so ivy resolution is rare and only happens when bootstrapping tools for pants