| 15 Sep 2025 |
 jade_ | codeowners should no longer hassle you on release branches | 07:58:05 |
 aloisw | In reply to @jade_:matrix.org
right, the codeowners plugin is set up in our configuration to fail closed That's not the entire problem as I understood it. Main branch has * as owners for meson.build but Marie didn't get any owners shown there. | 14:39:06 |
 emily | fyi lixStatic has linking issues on Darwin
Undefined symbols for architecture arm64:
"_libiconv", referenced from:
_psl_str_to_utf8lower in libpsl.a(libpsl_la-psl.o)
_mem_cd_iconveh_internal in libunistring.a(libunistring_la-striconveh.o)
(maybe you meant: __libiconv_version)
"_libiconv_close", referenced from:
_psl_str_to_utf8lower in libpsl.a(libpsl_la-psl.o)
_libunistring_iconveh_open in libunistring.a(libunistring_la-striconveh.o)
_libunistring_iconveh_close in libunistring.a(libunistring_la-striconveh.o)
"_libiconv_open", referenced from:
_psl_str_to_utf8lower in libpsl.a(libpsl_la-psl.o)
_libunistring_iconveh_open in libunistring.a(libunistring_la-striconveh.o)
ld: symbol(s) not found for architecture arm64
not sure whether a Meson problem or a pkg-config files omitting -liconv on static problem
| 16:00:23 |
|  laurents joined the room. | 16:05:59 |
 Qyriad | I thought stdenv automatically handled iconv now…? | 16:15:00 |
| Qyriad | hmm | 16:15:02 |
 raitobezarius | I remember aloisw knew things about this, but I'm so out of my depth on this topic | 16:17:23 |
| laurents | Hello! I am packaging a service for nixos, and I just hit https://git.lix.systems/lix-project/lix/issues/970 which completely prevents me from moving forward. (short of spending weeks trying to convince the upstream team to get rid of submodules) So I thought I'd try to fix that bug in lix. I write code for a living, but not C/C++, and I'm far from being a nix guru. Any thoughts on how unconscious it is of me to want to tackle it? | 16:17:56 |
| Qyriad | if no one has figured it out by Friday make an issue and assign it to usyo | 16:18:05 |
| Qyriad | In reply to @laurents:fsfe.org
Hello! I am packaging a service for nixos, and I just hit https://git.lix.systems/lix-project/lix/issues/970 which completely prevents me from moving forward. (short of spending weeks trying to convince the upstream team to get rid of submodules) So I thought I'd try to fix that bug in lix. I write code for a living, but not C/C++, and I'm far from being a nix guru. Any thoughts on how unconscious it is of me to want to tackle it? wait, why do you need self submodules for an upstream repo? | 16:19:14 |
| raitobezarius | I suppose it's an upstream repo that contains a flake.nix which use self submodules implicitly | 16:20:14 |
| Qyriad | I'd probably see if you can just consume the upstream repo with flake = false | 16:21:29 |
| raitobezarius | If you do that, you won't pull any relevant submodules, will you? | 16:21:52 |
| laurents | it's a bit convoluted:
- the app itself is a rails app, and I am packaging that with a flake. the app uses git submodules, so I managed to make it work with
self.submodules = true in the flake
- I have a separate flake to configure a server which uses that app. So I use flake1 as an input here. and this is where I get stuck
| 16:22:10 |
 K900 | Just use fetchFromGitHub or similar with fetchSubmodules = true | 16:22:49 |
| K900 | Instead of another flake | 16:22:52 |
| Qyriad | yeah recentl | 16:23:42 |
| Qyriad | * yeah exactly | 16:23:48 |
| raitobezarius | laurents: if you want, I can help you take a look at how to get rid of self submodules and look at your packaging; otherwise, we can work together how to fix flake-self-attrs | 16:24:39 |
| laurents | my problem is that the app (https://github.com/mysociety/alaveteli/ for context) requires a fairly complex localization to work, it's a "theme" which customises a number of things, including by overriding some of the ruby code. so I need to inject stuff in the upstream, I can't just pull the built package | 16:24:46 |
| raitobezarius | Do you have pointers to the various repositories with flake.nix and so on? | 16:25:16 |
| raitobezarius | With a reproducer recipe so we can test on our end | 16:25:22 |
| laurents | ok, let me push my code and I'll come back here in a moment | 16:26:52 |
| emily | aloisw: will you be up for rebasing https://github.com/NixOS/nixpkgs/pull/434186 after the toml11 bump? | 16:26:55 |
| aloisw | I am aware of the static Darwin breakage but don't know what causes it. The Darwin stuff I did for the prelinking change was pretty much strictly about not breaking dynamic linkage. | 16:44:33 |
| aloisw | Generally yes, but I can't guarantee a specific turnaround time (probably less than 1 day though). | 16:45:48 |
| aloisw | (Not to be understood in the "I will be slow" sense, but more like "please don't consider me paged in the middle of the night".) | 16:49:17 |
| laurents | raitobezarius Ok, I think it's showable-ish. My code is at https://gitlab.com/madada-team/dada-core/-/tree/add-theme-flake?ref_type=heads. Running a nix develop --no-pure-eval (no pure eval because I use devenv) should show you the bug. | 17:02:19 |
| raitobezarius | thx | 17:02:37 |
| raitobezarius | laurents: it uses a local path for alaveteli | 17:33:47 |
