| 19 Oct 2025 |
 dramforever | with the best line of code i've seen this week
size = INT_MAX; //2GB ought to be enough for anyone
| 16:48:29 |
 Grimmauld (any/all) | Yeah, i found that a few hours ago, alpine uses that: https://gitlab.alpinelinux.org/alpine/aports/-/blob/3.22-stable/community/accountsservice/musl-fgetspent_r.patch
But fgetpwent_r is broken the same way
| 16:49:07 |
| dramforever | unfortunately it does seem that you need to patch musl to get it | 16:49:08 |
| Grimmauld (any/all) | its just a shim, i can inline the shim | 16:49:27 |
| dramforever | okay you tricked me you wrote fgetspent_r twice 😆 | 16:49:49 |
| Grimmauld (any/all) | * Yeah, i found that a few hours ago, alpine uses something like that: https://gitlab.alpinelinux.org/alpine/aports/-/blob/3.22-stable/community/accountsservice/musl-fgetspent_r.patch
But fgetpwent_r is broken the same way
| 16:49:54 |
| Grimmauld (any/all) | OH | 16:49:59 |
| Grimmauld (any/all) | shit | 16:50:00 |
| Grimmauld (any/all) | sorry | 16:50:04 |
| Grimmauld (any/all) | * matthewcroughan: i have copied the nixosTests.login and added security.enableWrappers = false + pwaccess stuff, and the test is succeeding. Its screaming at me and i am definitely doing things wrong, but this might be progress?
Bad news is, pwaccess uses fgetspent_r / ~~fgetspent_r~~ fgetpwent_r which are unavailable on musl. Not sure yet how i'll fix that pain
| 16:50:18 |
| dramforever | okay but | 16:50:29 |
| dramforever | this alpine fgetspent_r is not _r | 16:50:42 |
| Grimmauld (any/all) | yeah.... | 16:50:51 |
| Grimmauld (any/all) | we love hacks | 16:51:13 |
| Grimmauld (any/all) | anyways, if matthewcroughan has an idea to make pwaccess work on musl (e.g. writing a similar shim for fgetpwent_r too) then we may have an suidless musl login | 17:11:50 |
| Grimmauld (any/all) | i won't poke the musl part of this too hard for now | 17:12:06 |
 matthewcroughan | if you wanted to, then matthewcroughan.cachix.org has a lot of cache | 17:12:25 |
| matthewcroughan | I want to host and sign a musl hydra at some point | 17:12:40 |
| Grimmauld (any/all) | this needs basically no cache, its like 25 packages (admittedly systemdLibs takes a few minutes, but still very manageable).
I don't have a setup to run musl VM tests, i am missing nix expressions for that, not necessarily cache | 17:13:21 |
| matthewcroughan | But do I even need to run hydra? Can't I just run something like https://git.snix.dev/snix/snix/src/branch/canon/ops/machines/snix-cache ? | 17:13:17 |
| matthewcroughan | There is no way to run a musl vm test because of only one thing | 17:13:38 |
| matthewcroughan | pytest breakage on staging-next | 17:13:45 |
| matthewcroughan | https://git.snix.dev/snix/snix/src/branch/canon/ops/machines/snix-cache | 17:13:52 |
| matthewcroughan | * https://github.com/MatthewCroughan/nixos-musl | 17:13:57 |
| Grimmauld (any/all) | fair | 17:14:06 |
| matthewcroughan | this repo has inside legacyPackages.vmtest-gnu-musl which would do it if pytest weren't broken by https://github.com/NixOS/nixpkgs/pull/451506 | 17:14:18 |
| Grimmauld (any/all) | anyways, you have suid-less login, if you make that musl-compatible then give me a shout | 17:14:24 |
| matthewcroughan | Where's the code for that exactly? | 17:15:02 |
| matthewcroughan | Ah here | 17:15:13 |
| matthewcroughan | If you want to test it too, you'll first need https://github.com/MatthewCroughan/nixpkgs/tree/mc/conditional-glibcsystemdinitrd because that is the only thing in nixpkgs that isn't overridable atm | 17:16:16 |
