| 17 Jul 2024 |
 jade_ | In reply to @qyriad:katesiria.org
we haven't enabled clang-format in our pre-commit yet, so if you want to clang-format some code then you can just run clang-format yourself uh wrong channel qyriad lol | 23:11:54 |
 Qyriad | OOPS | 23:36:21 |
| Qyriad | asdgjkalshgja | 23:36:23 |
 Las | I found out, I just have to remove my file from maintainers/flake-module.nix | 23:46:13 |
| 18 Jul 2024 |
 fzakaria | totally should clang-format it | 03:54:42 |
 emily | really hope the team can take another look at this PR that I discovered recently after talking about trying to balance cores vs. jobs for large builds in the macOS room and discovering that the functionality that made it pleasant in the past was gone from Nixpkgs https://github.com/NixOS/nix/pull/6855#issuecomment-2212554612 | 10:45:49 |
 fricklerhandwerk | In reply to @emilazy:matrix.org
really hope the team can take another look at this PR that I discovered recently after talking about trying to balance cores vs. jobs for large builds in the macOS room and discovering that the functionality that made it pleasant in the past was gone from Nixpkgs https://github.com/NixOS/nix/pull/6855#issuecomment-2212554612 Oh yes, I had looked at it when prompted, and wanted to merge my suggestions and fix the conflicts because it's obviously a good thing, but then something got in the way and I never returned. If you happen to have the keyboard hands to do that, I'd merge. | 10:55:51 |
| fricklerhandwerk | And otherwise it's still in my queue, it's just that the queue constantly gets reshuffled | 10:59:38 |
|  stick left the room. | 11:00:13 |
| emily | thank you! I'll happily fix the conflicts and open a new PR :) | 11:00:24 |
| emily | just wasn't sure if it was blocked on waiting for a more long-term solution, but I've seen a lot of people ask for it so getting this merged and used in Nixpkgs would be great | 11:00:55 |
| emily | (would it be acceptable to backport to older release branches too since it's just setting a single environment variable btw? it would be nice for the out-of-the-box experience with the default Nixpkgs-supplied Nix version to be better, and https://github.com/NixOS/nix/issues/11098 unfortunately makes new versions quite painful on macOS right now) | 11:16:14 |
| fricklerhandwerk | In reply to @emilazy:matrix.org
(would it be acceptable to backport to older release branches too since it's just setting a single environment variable btw? it would be nice for the out-of-the-box experience with the default Nixpkgs-supplied Nix version to be better, and https://github.com/NixOS/nix/issues/11098 unfortunately makes new versions quite painful on macOS right now) Sounds reasonable | 11:53:41 |
| emily | what's the deal with the cores default being different from cores = 0? it seems like Settings::getDefaultCores() will never return 0? is it just a legacy holdover from before that was implemented? or related to remote builders or something? | 12:51:55 |
| fricklerhandwerk | In reply to @emilazy:matrix.org
what's the deal with the cores default being different from cores = 0? it seems like Settings::getDefaultCores() will never return 0? is it just a legacy holdover from before that was implemented? or related to remote builders or something? In this context, 0 has no special meaning for Nix. It just happens that make --jobs 0 runs all available cores, but so does make --jobs foo, but that Make is ever called is entirely Nixpkgs business. I don't like that example very much, because these Nixpkgs specifics may change. I don't even like the make -j example particularly, but the feature seems to have been pretty much built for Make, so it doesn't make much sense to show anything else. | 13:29:49 |
| emily | pkgs/stdenv/generic/setup.sh actually specially handles 0 (it calls out to ncpus etc.) | 13:30:12 |
| emily | and yeah, I'm adjusting the examples etc. | 13:30:23 |
| emily | but, in the distributed builds setting, Settings::getDefaultCores() doesn't actually make sense, right? it yields the number of cores on the evaluating machine. so I think supporting cores = 0 (and load-limit = 0) is actually important. | 13:31:02 |
| emily | it makes me question the default, though – but maybe I'm mixing things up? maybe -j on the evaluating machine is ignored for distributed builds and it'll just be using the cores from the daemons? | 13:31:27 |
| emily | did anyone notice the test suite getting a lot slower on x86_64-darwin recently? I can't tell if it's just frozen or what | 19:22:39 |
| emily | a run with --cores 1 suggests to me that either tests/functional/flakes/flakes.sh somehow became incredibly slow (although I don't see any commits to it since the last time I was making a PR?), or it no longer terminates at all. the build also didn't even go through until I added a missing #include (but the macOS CI seems fine?). is there anyone on x86_64-darwin who can actually succeed at building the current HEAD? | 22:06:07 |
| emily | okay, doing another run with just the header fix suggests that I'm the one who broke the tests :) | 22:26:00 |
| 19 Jul 2024 |
|  C0deboy joined the room. | 04:19:22 |
|  easel changed their profile picture. | 14:17:20 |
| 20 Jul 2024 |
|  @phileas:asra.gr left the room. | 16:53:17 |
|  @felixfahrbahn:matrix.org left the room. | 17:51:26 |
 abathur | A question I saw on fediverse this morning reminded me I have an issue that might merit a re-triage: https://github.com/NixOS/nix/issues/8192
Théophane did query for more detail after an initial triage attempt last year; I provided some, but it may be worth gut-checking my response(s) in case there's something I need to address before it goes back on the full team's plate. | 18:26:22 |
| Las | In reply to @abathur:matrix.org
A question I saw on fediverse this morning reminded me I have an issue that might merit a re-triage: https://github.com/NixOS/nix/issues/8192
Théophane did query for more detail after an initial triage attempt last year; I provided some, but it may be worth gut-checking my response(s) in case there's something I need to address before it goes back on the full team's plate. I don’t think this is in the realm of what Nix should do IMHO | 19:52:09 |
| Las | just wrap it in bubblewrap | 19:52:25 |
| emily | I don't think abathur is asking for runtime sandboxing | 19:53:04 |
