5 Jun 2020 |
Bubu | gnu parallel | 17:30:47 |
jochensp | uh, I had some fun things with parallel and git already | 17:31:07 |
jochensp | could we schedule all commits after the parallel run? | 17:31:25 |
jochensp | maybe even make it one commit? | 17:31:33 |
Bubu | it's much nicer having this each app in a seperate commit. for reviewing and reverting stuff | 17:32:36 |
cdesai | http://dustin.sallings.org/2008/12/30/git-alternates.html | 17:32:50 |
cdesai | this could help with network and io | 17:33:15 |
jochensp | Bubu: agreed. Having one job doing the commits could still work, or? | 17:33:57 |
Bubu | docker was pretty dead after that first attempt | 17:35:29 |
jochensp | :D | 17:35:45 |
Bubu | I had to force-shutdown the vm | 17:36:32 |
Bubu | so let's try again (2 parallel jobs) | 17:36:44 |
Bubu | 2 parallel seemed to work, let's try with 10 | 17:50:04 |
Bubu | (I wish there was a cancel all pending pipeline button :-/) | 17:50:22 |
jochensp | pkill on the gcc VM ;) | 17:52:12 |
Bubu | The runner will still pick up new jobs until they time out :) | 17:54:56 |
jochensp | Oo lol | 17:56:00 |
Bubu | _hc: can we switch fdroiddata CI to shallow clones? It's a project wide setting so it'll affect all the other jobs as well. | 18:00:57 |
Bubu | the machine is totally IO starved :-/ | 18:10:53 |
Bubu | in the screenshot I posted above the yellow colored cpu bars mean io-wait | 18:11:10 |
_hc | Bubu: thanks for the branch protection, makes sense. didn't know that feature existed. once issuebot is reliable on fdroiddata merge requests, I'll remove test-issuebot . Or as soon as it is causing issues | 18:11:32 |
Bubu | In reply to @bubu:bubu1.eu in the screenshot I posted above the yellow colored cpu bars mean io-wait or maybe not? investigating... | 18:14:37 |
Bubu | it seems to be guest cpu time which seems better | 18:15:27 |
_hc | Bubu: I think using buster for checkupdates makes sense. | 18:16:53 |
_hc | In reply to @bubu:bubu1.eu _hc: can we switch fdroiddata CI to shallow clones? It's a project wide setting so it'll affect all the other jobs as well. the CI jobs need some history to be able to tell what was changed in the merge request | 18:17:26 |
Bubu | I'll probably prepare an image with the required dependencies that we don't spend 10 minutes installing bzr and hg every job | 18:17:38 |
Bubu | In reply to @eighthave:matrix.org the CI jobs need some history to be able to tell what was changed in the merge request makes sense | 18:17:57 |
_hc | you want to use docker/gitlab-runner still in gcc148? I was thinking we could just provision a VM with everything pre-installed | 18:18:23 |
Bubu | In reply to @bubu:bubu1.eu I'll probably prepare an image with the required dependencies that we don't spend 10 minutes installing bzr and hg every job the image size is largely irrelevant as it's local after the first job | 18:18:30 |
_hc | like a buildserver VM without the Android SDK/NDK/gradle | 18:18:38 |