#general
- r
rough-minister-58256
10/30/2015, 12:09 AM(I know Jenkins is planned, but this could alleviate near-term pain in the meantime) - b
bored-art-40741
10/30/2015, 12:19 AMUnless things have changed, this is not possible with an open source project - r
rough-minister-58256
10/30/2015, 12:23 AMah - h
happy-kitchen-89482
10/30/2015, 12:24 AMHi all, I'm about done with my profiling, and attempts to speed up a no-op build. We're now down to most of that time (close to 1 sec) spent on loading plugins. - h
happy-kitchen-89482
10/30/2015, 12:24 AMSomething to keep in mind for the new engine - h
happy-kitchen-89482
10/30/2015, 12:30 AMwe should find a way to separate registration code from the target/task code it registers, so that we only load the latter's modules when actually needed. - w
witty-crayon-22786
10/30/2015, 12:44 AMyep - w
witty-crayon-22786
10/30/2015, 12:44 AMone advantage will be that tool bootstrap will be handled natively - w
witty-crayon-22786
10/30/2015, 12:46 AM@happy-kitchen-89482: saw that one of the test abort tests flaked on you after your last merge... just tightened up the timings a bit there. - w
witty-crayon-22786
10/30/2015, 12:47 AM(in
)eb35f4 - h
happy-kitchen-89482
10/30/2015, 12:49 AMthanks! - b
bored-art-40741
10/30/2015, 2:16 AMTwitter has some tricks for dealing with git scaling these days, right? - b
bored-art-40741
10/30/2015, 2:16 AMAside from shallow clone / partial checkout sort of stuff - b
bored-art-40741
10/30/2015, 2:16 AMJust like, better fetch rules to prevent reference bloat? - a
ancient-stone-50795
10/30/2015, 2:17 AM@bored-art-40741: the biggest trick that you can ‘easily’ do is split into 2 git servers. One serves master, the other one you push branches to - a
ancient-stone-50795
10/30/2015, 2:18 AMI spent some time talking to folks at twitter about this. They have some patches to git that they have made to the client and server to speed up git pull. - b
bored-art-40741
10/30/2015, 2:18 AMYeah, that split is exactly what I'm looking into right now - b
bored-art-40741
10/30/2015, 2:19 AMI want to make sure I can provide automatically configured fetch rules for users though - a
ancient-stone-50795
10/30/2015, 2:19 AMThe idea is that if your master repo has no branches, part of the server side computation they have to do for pulls is not needed. Stu said they probably would not open source the server side changes that speed up git pull. - b
bored-art-40741
10/30/2015, 2:22 AMI'm also trying to figure out how to configure it such that I don't pull down tags by default - i
important-piano-50289
10/30/2015, 2:46 AMhey, just to be clear, we do have technology that's separate from just doing sane splitting of branches and stuff - i
important-piano-50289
10/30/2015, 2:46 AMit hasn't been accepted upstream, and it's operationally difficult to deal with - i
important-piano-50289
10/30/2015, 2:47 AMit works well, but it's not like "oh yeah, just flip the switch and presto! things get faster" - i
important-piano-50289
10/30/2015, 2:47 AMwe're working on getting it accepted upstream but it's slow going 😕 - b
bored-art-40741
10/30/2015, 2:48 AMI'm curious how you do the sane splitting, concretely - i
important-piano-50289
10/30/2015, 2:48 AMwe haven't - b
bored-art-40741
10/30/2015, 2:48 AMOh - i
important-piano-50289
10/30/2015, 2:48 AMwell - i
important-piano-50289
10/30/2015, 2:48 AMtbh - i
important-piano-50289
10/30/2015, 2:49 AMthat's not totally true