| 29 Jul 2021 |
 hexa | that is the case | 10:25:53 |
 Vladimír Čunát | So the main remaining regression is the PerlMagick problem that seems stuck? | 10:35:55 |
| hexa | https://shells.darmstadt.ccc.de/~hexa/1690057.html | 10:51:53 |
| hexa | mostly aiohttp | 10:51:56 |
| hexa | weirdly enough, if you have restarted failed jobs, they failed again even though aiohttp has succeeded | 10:53:40 |
| hexa | * weirdly enough, if you have restarted failed jobs, they failed again with a reference to aiohttp, even though aiohttp has succeeded | 10:54:26 |
| Vladimír Čunát | No, I only did it now. | 11:00:57 |
| hexa | ah ok, was surprised that home-assistant was still queued, I thought it finished earlier | 11:08:53 |
| Vladimír Čunát | It failed, so I restarted it individually. | 11:20:43 |
| Vladimír Čunát | Sigh, pytorch gets repeatedly scheduled on wendy again: https://github.com/NixOS/nixos-org-configurations/issues/146 | 12:05:24 |
 Sandro | wendy hit again 😏 | 13:27:04 |
| hexa | wendy 🎉 | 13:46:44 |
 FRidh | mass rebuild on master? | 14:17:10 |
| FRidh | oh never mind, it's just because it's a different hydra project | 14:19:04 |
 sterni | FRidh: looks like it to me? https://hydra.nixos.org/jobset/nixpkgs/trunk | 16:03:06 |
| sterni | there's also no cache for current master, so seems like it | 16:03:21 |
| sterni | oh hm seems to be in order mostly | 16:07:30 |
| sterni | I screwed up with haskell-updates though since I missed the staging-next merge today | 16:10:50 |
| sterni | so that's 20000 builds on master, maybe revert time? | 16:11:57 |
| 30 Jul 2021 |
| FRidh | In reply to @sternenseemann:systemli.org
I screwed up with haskell-updates though since I missed the staging-next merge today haskell-updates were always merged directly into master. While in principle all larger rebuilds go via staging I can get there are reasons to send it to master though. | 06:25:19 |
| sterni | In reply to @FRidh:matrix.org
haskell-updates were always merged directly into master. While in principle all larger rebuilds go via staging I can get there are reasons to send it to master though. we have specifically updated our workflow to minimize the rebuilds by adding all relevant platforms to the haskell-updates jobset and making it a policy to merge master before merging and waiting for any rebuilds, so normally that shouldn't be an issue | 09:21:32 |
| sterni | just forgot to check if there was anything significant on master between merging haskell-updates and the last merging in of master yesterday which was the screw-up | 09:21:59 |
| sterni | but seems like the rebuilds finished over night, so not too bad | 09:22:20 |
| Vladimír Čunát | Multiple mass-rebuild branches are difficult. I think it's historically the reason why the haskell jobset has such high priority on Hydra (very many scheduling shares). | 09:22:47 |
| sterni | I can imagine it was struggling to get build shares with 5000-6000 jobs as it was back then | 09:29:35 |
| sterni | nowadays a rebuild is like 20000 jobs since we build darwin and aarch64 as well | 09:29:53 |
| 10 Aug 2021 |
|  happysalada joined the room. | 15:27:09 |
|  chvp joined the room. | 15:33:58 |
|  asbachb joined the room. | 15:50:40 |
|  Jan Tojnar joined the room. | 17:57:00 |
