| 28 Oct 2025 |
 emily | can we make stdenv.isFoo warn as well? 😆 | 17:56:16 |
| emily | maybe something for early 26.05 | 17:56:21 |
| emily | looking at the diff, runCommandNoCC, runCommandNoCCLocal, and maybe utillinux seem like the main things worth turning into warnings instead of throws | 18:00:22 |
 Artturin | You sure? | 18:00:30 |
| emily | might also be good to adjust the python throw error message | 18:00:47 |
| emily | since it currently recommends people use python2 😆 | 18:00:51 |
| emily | I'll PR those | 18:00:53 |
| emily | https://github.com/NixOS/nixpkgs/actions/runs/18828351853/job/53715047703?pr=456065 | 18:01:01 |
| emily | this is what happened when I forgot the commit reverting turning those into throws | 18:01:08 |
| emily | vcunat: btw, warning aliases used in NixOS tests won't block/fail anything on Hydra, right? because the warning check there is only in the release checks? | 18:01:40 |
| emily | is it possible to make Hydra do the equivalent of NIX_ABORT_ON_WARN instead of checking it in release checks, so that those are caught in the same way warnings in packages are? | 18:02:10 |
| Artturin | Ok I can take care of those | 18:03:18 |
| emily | no rush | 18:04:09 |
| emily | but I do have a PR to attach warnings to them | 18:04:24 |
| emily | for system it's kind of annoying, I had to do a __toString thing to stop it from spamming nix search :) | 18:04:45 |
| Artturin | :p tests shows only one *Platform attribute missing
platformEquality = «error: attribute 'buildPlatform' missing»; | 18:04:51 |
| emily | I can put a PR up for that and handling runCommandNoCC | 18:04:58 |
 hexa | We should just point python to python 3 | 18:05:23 |
| hexa | It will be a great surprise | 18:05:29 |
| Artturin | diff --git a/pkgs/test/top-level/default.nix b/pkgs/test/top-level/default.nix
index a5b44afb8522..e713b617f8dd 100644
--- a/pkgs/test/top-level/default.nix
+++ b/pkgs/test/top-level/default.nix
@@ -61,7 +61,7 @@ lib.recurseIntoAttrs {
pkgsLocal = map nixpkgsFun configsLocal;
pkgsCross = map nixpkgsFun configsCross;
in
- assert lib.all (p: p.buildPlatform == p.hostPlatform) pkgsLocal;
- assert lib.all (p: p.buildPlatform != p.hostPlatform) pkgsCross;
+ assert lib.all (p: p.stdenv.buildPlatform == p.stdenv.hostPlatform) pkgsLocal;
+ assert lib.all (p: p.stdenv.buildPlatform != p.stdenv.hostPlatform) pkgsCross;
pkgs.emptyFile;
}
| 18:05:29 |
| Artturin | Removing it completely might be better | 18:06:15 |
| Artturin | But not as good of a surprise | 18:06:28 |
| hexa | But where is the fun in that | 18:06:33 |
| emily | making it a throw is what my PR automatically did, because the alias is old | 18:07:42 |
| emily | but I will adjust the throw text to say "use Python 3 instead" | 18:07:47 |
| emily | freeing up the namespace will be nice, it's awkward that python is used as a generic-Python argument in places but also exists at the top level as an alias | 18:08:12 |
| Artturin | https://github.com/NixOS/nixpkgs/pull/456512 | 18:10:35 |
| emily | does anyone want to merge https://github.com/NixOS/nixpkgs/pull/456060 so I have fewer aliases.nix conflicts to resolve later 😆 | 18:14:50 |
| emily | do we want to keep forceSystem around? | 18:21:32 |
| emily | I've never seen anyone use it | 18:21:33 |
