| 10 Dec 2025 |
 lep | I'm havnig a bit of a strange issue. For some reason the wrong python binary gets executed despite not being in $PATH
~ $ whereis python
python:
~ $ nix-shell -p 'python3.withPackages( ps: [ ps.valkey ])'
bash: warning: setlocale: LC_COLLATE: cannot change locale (): No such file or directory
[nix-shell:~]$ whereis python
python: /nix/store/jlq2n516adkdbf7fgf1bja9h8l4gxk7j-python3-3.13.9-env/bin/python /nix/store/jlq2n516adkdbf7fgf1bja9h8l4gxk7j-python3-3.13.9-env/share/man/man1/python.1.gz
bash: warning: setlocale: LC_COLLATE: cannot change locale (): No such file or directory
[nix-shell:~]$ python -c 'import valkey'
Traceback (most recent call last):
File "<string>", line 1, in <module>
import valkey
ModuleNotFoundError: No module named 'valkey'
bash: warning: setlocale: LC_COLLATE: cannot change locale (): No such file or directory
[nix-shell:~]$ /nix/store/jlq2n516adkdbf7fgf1bja9h8l4gxk7j-python3-3.13.9-env/bin/python -c 'import valkey'
bash: warning: setlocale: LC_COLLATE: cannot change locale (): No such file or directory
[nix-shell:~]$
The python which runs when just typing python is just some pkgs.python3. nix why-depends says it's installed because of git.
| 15:59:23 |
| 11 Dec 2025 |
 Randy Eckenrode | I’ll do the MoltenVK PR this weekend when I rebase my staging branch (assuming the Vulkan SDK update has been merged). Trying to limit the number of bootstraps to only weekly …. | 03:27:48 |
| Randy Eckenrode | Also, boo. Swift Build isn’t finding the SDK again. They changed something that I’m going to have to printf debug again. | 03:28:08 |
|  @acidbong:envs.net left the room. | 04:45:32 |
 vcunat | We have a weird regression on 25.11 aarch64-darwin, blocking thousands of jobs. Perhaps someone here could have a look?
https://github.com/NixOS/nixpkgs/pull/468267#issuecomment-3636942915 | 17:06:19 |
| Randy Eckenrode | Anything in Console.app? Usually when something gets killed, it’s due to codesigning. | 17:46:56 |
| vcunat | Well, the problem is that I never really used Apple stuff. | 17:50:11 |
| vcunat | (though I do have remote access to some) | 17:50:42 |
 emily | man log is the CLI interface | 18:00:17 |
| emily | it looks like it might be the awful nondeterministic codesigning bug though | 18:00:36 |
| emily | my guess is that the autogen binary always gets SIGKILLed | 18:01:25 |
| emily | which should be easy to test | 18:01:31 |
| emily | shion:~
❭ /nix/store/9gk18p59s7fws5j7ykwhzmfaav4c8k7a-autogen-5.18.16-bin/bin/autogen
fish: Job 1, '/nix/store/9gk18p59s7fws5j7ykwh…' terminated by signal SIGKILL (Forced quit)
| 18:02:18 |
| emily | so it's that that probably needs rebuilding… which AIUI cannot be done without changing the hash because of Hydra/cache limitations… and will probably throw away a ton of builds since it's in autoreconfHook | 18:02:49 |
| vcunat | Changing the hash for a particular package-platform combination is easy enough, though, it will take lots of machine time, as you wrote. | 18:11:13 |
| vcunat | I don't know a way of rebuilding for cache.nixos.org without changing the hash. | 18:11:45 |
| emily | manually deleting from cache and rebuilding? | 18:11:59 |
| vcunat | Sounds risky. | 18:12:12 |
| vcunat | You'd be punching a hole. | 18:12:32 |
| vcunat | The cache is assumed to be closed on dependencies. | 18:12:46 |
| Randy Eckenrode | I wonder if we need a `no broken signatures’ hook on Darwin. | 18:12:50 |
| vcunat | * The cache is assumed to be closed on dependencies. (or whatever you'd call the property) | 18:12:55 |
| Randy Eckenrode | Like that symlinks one but for signatures. | 18:13:07 |
| emily | I see no alternative other than rerolling the hash for aarch64-darwin then | 18:13:27 |
| emily | though should try to rebuild it locally first to make sure that fixes it | 18:13:33 |
| emily | (cannot use --rebuild or --check, those break signature themselves) | 18:13:40 |
| emily | (need to just build it without substituting on a machine without it in the store) | 18:13:49 |
| vcunat | It does seem to fix ./result/bin/autogen --help on this machine. | 18:18:35 |
| vcunat | * It does seem to fix ./result-bin/bin/autogen --help on this machine. | 18:19:21 |
| vcunat | OK, pushed as e0d2da7967a3. Thanks. | 18:25:03 |
