| 4 Jan 2025 |
 emily | (I saw some talk about maybe not needing all the servers that were ordered in the infra room) | 21:29:03 |
 Janne Heß | hexa probably knows more about that ;) | 21:29:16 |
| emily | right now I think most contributors have basically no way to test something like Firefox on aarch64-linux, which is unfortunate, as it's our second-best supported platform. | 21:29:27 |
| emily | so is ofborg using the GHA eval outputs now? | 21:32:37 |
| Janne Heß | No, it's only parsing the commit messages and PR messages and parses @ofborg commands | 21:32:54 |
| Janne Heß | GH API is still horrible and I will try to get around it as long as I can | 21:33:09 |
| emily | it, uh, tried to build the package named treewide on my PR :) | 21:33:46 |
| Janne Heß | well. Simon Hauser could you take a look on tuesday? I dropped the logic that filters actual attributes to get rid of all evals, but looks like that has downsides as well | 21:34:35 |
 Gaétan Lepage | In reply to @emilazy:matrix.org
I think 256 gigs of RAM btop was reporting 250GB for sure. | 21:45:40 |
| Gaétan Lepage | In reply to @emilazy:matrix.org
would be nice if we can make a shared Ampere happen for ofborg + community builder, maybe +1: This would be very useful indeed :) | 21:58:29 |
| 5 Jan 2025 |
 hexa | the foundation is in the netherlands, so 21% taxes apply | 01:39:44 |
| hexa | 205 EUR for an RX170, 265 EUR for an RX220 | 01:39:58 |
| hexa | if you want that to happen, send a funding request to the foundation issue tracker | 01:41:33 |
| hexa | but frankly, a community builder should be hosted with nix-community ideally | 01:41:58 |
| hexa | I don't want to have to deal with that | 01:42:07 |
| emily | well, I was thinking it makes more sense to use one machine for both ofborg and community builds (since it was like that before, and the machines are so huge) | 01:43:21 |
| emily | it seems like nix-community doesn't have the funding though sadly | 01:43:29 |
 Jonas Chevalier | done! | 09:13:11 |
| Jonas Chevalier | the refresh might take a while, the TTL was set to 10800 | 09:13:32 |
|  @5sqrgsrdwk:matrix.org joined the room. | 11:53:24 |
| @5sqrgsrdwk:matrix.org left the room. | 11:55:05 |
| emily | more fun package names :) | 15:22:26 |
| emily | 
Download image.png | 15:22:32 |
|  Ainz Ooal Gown joined the room. | 19:30:52 |
 @infinisil:matrix.org | Janne Heß: We never really communicated much around that, but what's exactly the issue with ofborg and evals right now? | 20:06:01 |
| @infinisil:matrix.org | I've heard that you need something from the evals, but I can't see what, because afaik ofborg never built anything other than what commit messages implied (and from comment commands) | 20:07:04 |
| emily | (FWIW it was also trying to rebuild nixos/foo which sparked this discussion because I was under the misapprehension(?) that eval somehow factored into what old ofborg decided to build) | 20:10:36 |
| @infinisil:matrix.org | Janne Heß: Here's how to get to the workflow run from a PR number, assuming that's what you need:
prHead=$(gh api /repos/NixOS/nixpkgs/pulls/371217 --jq .head.sha)
gh api --method GET /repos/NixOS/nixpkgs/actions/workflows/eval.yml/runs -f head_sha="$prHead" -f event=pull_request_target --jq '.workflow_runs | sort_by(.run_started_at) | .[-1]'
| 20:14:12 |
| @infinisil:matrix.org | From there you can then use the artifacts API | 20:14:36 |
|  collares joined the room. | 22:15:37 |
