| 25 May 2021 |
 sterni (he/him) | amirite | 15:53:00 |
 maralorn | In reply to @sternenseemann:systemli.org
sounds like a script which runs for five minutes It feels like another situation where the solution will query hydra a lot. | 16:03:58 |
| sterni (he/him) | well no | 16:04:14 |
| sterni (he/him) | well yes | 16:04:23 |
| sterni (he/him) | you need to run the query which takes 2min | 16:04:29 |
| sterni (he/him) | and then <500 additional queries for the logs | 16:04:40 |
| sterni (he/him) | depending on how much has failed | 16:04:46 |
| sterni (he/him) | oh much less even | 16:05:00 |
| sterni (he/him) | it would only be failed on its own right | 16:05:10 |
|  meet joined the room. | 16:05:17 |
| maralorn | In reply to @sternenseemann:systemli.org
https://docs.google.com/spreadsheets/d/1ZvqZOdOse1lIAJxccsWdyFNeDLyVmoCUvI12LJNFMks That's a cool table. Who is suggesting this? Is it likely to happen? | 16:08:16 |
| sterni (he/him) | graham and yes | 16:08:28 |
| sterni (he/him) | See #infrastructure:nixos.org | 16:08:57 |
| maralorn | In reply to @sternenseemann:systemli.org
oh much less even Yeah currently only 40 on Linux. | 16:09:04 |
| maralorn | Yeah, should actually be doable. | 16:09:22 |
| sterni (he/him) | I'm not sure what impact this will have, we'll have less of a share advantage over other jobsets, but also more percentage in shares overall | 16:09:34 |
| sterni (he/him) | idk never felt the impact of shares before so who knows | 16:09:49 |
| maralorn | Other idea btw: we can theoretically find packages requiring a jail break algorithmically. We could us that to a) mark them broken before they even failed once and b) mark them automatically unbroken as soon as they build again. So a temp incompatibility like now with random will not disable a package permanently. | 16:12:30 |
| sterni (he/him) | yeah that could be interesting | 16:12:54 |
 srid | Anyone fancy trying some Haskell stuff here?
Utilizing the power of Nix, every repl comes preloaded with over 30,000 packages. Adding a package is as simple as adding it as a dependency in the replit.nix file. https://lobste.rs/s/dy4x1c/how_we_went_from_supporting_50_languages
| 16:19:11 |
|  zwro changed their display name from z to yin. | 18:05:13 |
 utdemir | srid: seems to work as expected https://replit.com/@utdemir/nix-hs . My first impressions:
- There's not much magic going on, they pretty much only installed Nix with a mounted /nix/store.
- Run command simply runs
nix-shell .. --run '..', so I guess when the Nix expression becomes larger it'd be annoying to wait for Nix evaluation every time you run something.
| 23:46:14 |
| 26 May 2021 |
 cdepillabout | maralorn sterni (he/him) For haskell-updates, it looks like everything in maintained and mergeable is now building (or correctly marked broken). We have the choice of merging haskell-updates into master now (only after of course checking that merging would cause no eval errors), or merging master into haskell-updates and waiting for Hydra to work through all the builds again. | 01:44:31 |
| cdepillabout | I'm leaning towards merging haskell-updates into master. But I don't really have a good idea of how risky it is, or what peti would normally do. | 01:45:50 |
| cdepillabout | * I'm leaning towards merging haskell-updates into master. But I don't really have a good idea of how risky it is, or what peti would normally do. Do either of you guys have an opinion here? | 01:46:29 |
 bqv | > Other idea btw: we can theoretically find packages requiring a jail break algorithmically. We could us that to a) mark them broken before they even failed once and b) mark them automatically unbroken as soon as they build again. So a temp incompatibility like now with random will not disable a package permanently.
PLEASE fucking do it | 03:33:52 |
| bqv | This shit's a plague | 03:33:58 |
|  ashleyis joined the room. | 03:59:11 |
 joe (he/him) | I wonder what would happen if only breaking changes which don't change types required a major version bump... | 04:23:51 |
| joe (he/him) | obviously it's no good for build planning | 04:24:05 |
