| 16 Oct 2025 |
 matthewcroughan | Is it any more or less WIP than the kmod patch? | 16:04:31 |
 dramforever | the kmod module-dir.patch doesn't apply because it's nixos specific | 16:04:34 |
| dramforever | so we can't not ship it in nixpkgs | 16:04:44 |
| matthewcroughan | i would argue that the musl patch is nixos specific too, at least I'm sure when you upstream it someone might say something like that | 16:04:54 |
| matthewcroughan | It's a bit of a blurry line | 16:05:12 |
| matthewcroughan | I mean, I wonder how many other edge cases are encountered in musl in nixpkgs because of the lack of the patch you just made | 16:05:51 |
| matthewcroughan | I'm sure it shows up in other places | 16:06:03 |
| dramforever | when upstream says they don't want it then we'll talk | 16:06:28 |
| matthewcroughan | Maybe need to find more and provide them as examples to upstream | 16:06:32 |
| matthewcroughan | Or maybe the systemd kmod context example is good enough :D | 16:06:52 |
| matthewcroughan | But yes, very excited about it working, so good | 16:07:01 |
| dramforever | anyway the real point is that i wrote the patch and it's wip from me so i'm going to say no | 16:08:12 |
| matthewcroughan | Yeah makes sense, I respect that | 16:11:04 |
| matthewcroughan | Damn, why is perl so unstable with musl | 16:11:13 |
| matthewcroughan | nix used to build just fine, now it's all broken because of perl deps | 16:11:21 |
| matthewcroughan | ┃ ┌─ ⏸ unit-systemd-tmpfiles-resetup.service
┃ │ ┌─ ⏸ nixos-rebuild-ng-0.0.0-aarch64-unknown-linux-musl waiting for 2 ⏵
┃ │ │ ┌─ ⏵ perl5.40.0-DBI-1.644-aarch64-unknown-linux-musl (buildPhase) ⏱ 4m41s
┃ │ │ ┌─ ⏸ nix-perl-aarch64-unknown-linux-musl-2.31.2
┃ │ │ ├─ ⏵ nix-expr-aarch64-unknown-linux-musl-2.31.2 (buildPhase) ⏱ 23s
┃ │ │ ┌─ ⏸ nix-aarch64-unknown-linux-musl-2.31.2
| 16:11:30 |
| matthewcroughan | perl5.40.0-DBI-1.644-aarch64-unknown-linux-musl be broken, as well as nix-expr-aarch64-unknown-linux-musl-2.31.2 due to function names | 16:11:55 |
| matthewcroughan | /nix/store/bvp4jwn1ibkjh7yhr2mlcxxd908y4x4v-perl-5.40.0/lib/perl5/5.40.0/aarch64-linux-thread-multi/CORE/perl.h:3358:22: error: unknown type name 'off64_t'; did you mean 'off_t'?
3358 | # define Off_t off64_t
|
| 16:12:04 |
| matthewcroughan | Alyssa Ross: O.o (https://github.com/NixOS/nixpkgs/pull/452642#discussion_r2436725816) how do you upstream to perl? | 16:57:10 |
| matthewcroughan | I'm young | 16:57:15 |
| matthewcroughan | Looks like https://github.com/perl5-dbi/dbi actually | 17:00:08 |
 Alyssa Ross | it's not to perl, it's just a perl package | 17:00:09 |
| matthewcroughan | Yeah I meant to the perl ecosystem | 17:00:18 |
| matthewcroughan | like it's not always clear where the sources or appropriate upstream path is | 17:00:33 |
| matthewcroughan | sometimes github, sometimes mailing lists, sometimes nowhere | 17:00:40 |
| matthewcroughan | and since we use mirror:// that further masks it | 17:01:00 |
| Alyssa Ross | looks like they merge PRs in this case | 17:01:12 |
| matthewcroughan | NIce! | 17:01:16 |
| matthewcroughan | * Nice! | 17:01:22 |
| Alyssa Ross | BTW: if something is going to be broken on every musl distro (like this fix), please try to take the fix upstream before working around it in Nixpkgs. If we all do this (and the other distros usually do), it's less work for all of us than if we all have to separately apply the same workarounds downstream. | 17:02:29 |
