| 12 Oct 2025 |
 matthewcroughan | Is this just a flaky test, or is it actually broken on musl? | 20:53:15 |
| matthewcroughan | ( FAILED ) job 12 t/integration/failed.t0m
< REASON > job 12 Test script returned error (Err: 1)
< REASON > job 12 Errors were encountered (Count: 1)
< REASON > job 12 Assertion failures were encountered (Count: 1)
< REASON > job 12 Subtest failures were encountered (Count: 1)
( FAILED ) job 13 t/integration/help.t
< REASON > job 13 Test script returned error (Err: 1)
< REASON > job 13 Assertion failures were encountered (Count: 1)
< REASON > job 13 Subtest failures were encountered (Count: 1)
| 20:54:20 |
 Sergei Zimmerman (xokdvium) | Has been for some time. Noticed this in the Nix repo | 20:58:58 |
| Sergei Zimmerman (xokdvium) | I think I've seen this on i686? | 20:59:42 |
| matthewcroughan | hmm, I only notice it on this staging-next, a few weeks ago it wasn't failing | 21:38:25 |
| matthewcroughan | Does anyone know why pytest fails with ┃ error: build of '/nix/store/3haz9vxnx4v0fl5vadyzspy3kbx4rp05-python3.13-pytest-8.4.2-aarch64-unknown-linux-musl.drv' on 'ssh-ng://nix-ssh@m2u' failed: output '/nix/store/1vh3vhp6kf04350d6sjs… ┃ /nix/store/d29jvsi1x76l0bbwd704srmmsijfbl8a-python3-3.13.8 on musl? | 23:51:07 |
| matthewcroughan | I guess hooks sometimes produce bad references in outputs? | 23:52:02 |
| matthewcroughan | specifically in the case of cross | 23:52:37 |
 Ihar Hrachyshka | I just noticed it and asked in main nixpkgs channel. we disabled wrapper hook on pytest and it was supposed to replace shebangs to make use of host | 23:52:55 |
| Ihar Hrachyshka | https://github.com/NixOS/nixpkgs/commit/84dced5a94ed901362dc6e285d0ac0c08ba671ff | 23:53:13 |
| matthewcroughan | This blocks musl pretty hard | 23:54:21 |
| Ihar Hrachyshka | I have absolutely no clue how I can meaningfully
test this aside from kind of yoloing it into staging.
WELP
| 23:54:23 |
| matthewcroughan | at least with the default pkgs.nixos | 23:54:25 |
| Ihar Hrachyshka | yeah a lot of broken pieces... | 23:54:36 |
| matthewcroughan | Yeah it's unfortunate that they didn't test adequately | 23:54:37 |
| matthewcroughan | Did they just not have a big enough machine to perform testing? They could have asked? | 23:54:53 |
| matthewcroughan | Can we revert it? | 23:56:03 |
| Ihar Hrachyshka | maybe?
I am not sure they were aware this could affect cross. So even if they'd rebuild the main variant tree, they would not catch it. | 23:56:41 |
| matthewcroughan | It seems to effect native too | 23:56:58 |
| Ihar Hrachyshka | have a repro? | 23:57:18 |
| matthewcroughan | Ah no just nom playing tricks | 23:57:33 |
| Ihar Hrachyshka | I doubt pytest is broken in main... | 23:57:39 |
| matthewcroughan | nix build .#legacyPackages.aarch64-linux.pkgsMusl.pkgsLLVM.python3Packages.pytest -L | 23:58:14 |
| matthewcroughan | that would do it though | 23:58:19 |
| Ihar Hrachyshka | a revert would get cross scenario fixed... but I assume they were dealing with some other legit issue, so it would be unfortunate to regress there. | 23:58:27 |
| Ihar Hrachyshka | the dontWrap thing seems like a footgun. I tried some other packages using it - and surely enough they are broken for cross. | 23:59:23 |
| 13 Oct 2025 |
| matthewcroughan | it's pkgsLLVM that breaks actually | 00:00:28 |
| matthewcroughan | pkgsMusl is fine | 00:00:31 |
| Ihar Hrachyshka | pkgsStatic is broken. | 00:01:30 |
| matthewcroughan | pkgsLLVM is not static? | 00:02:11 |
