| 18 Jul 2025 |
 emily | well | 21:03:41 |
| emily | the package we are trying to fix is sort of one of them | 21:03:47 |
| emily | kind of? | 21:03:48 |
| emily | it's weird | 21:03:49 |
| emily | I really think the GNU target-specific cross model ruined everything forever. | 21:04:24 |
| emily | if Plan 9 had won everyone would be much less confused about cross. | 21:04:36 |
 dramforever | it always bothers me how gnu toolchains just does this | 21:05:32 |
| dramforever | i guess it was designed in an era where every different platform is weird | 21:06:27 |
| dramforever | like -z pack-relative-relocs, "Supported for i386 and x86-64." | 21:07:08 |
| dramforever | how do you even make this architecture specific | 21:07:15 |
| dramforever | there's literally one RELR format for all of ELF | 21:07:24 |
| dramforever | two if you have to split 32 and 64 | 21:07:37 |
| dramforever | anyway </rant> | 21:07:41 |
| emily | I assume from their perspective they were innovating by not having entirely separate compiler codebases for every architecture/ | 21:10:56 |
| emily | * I assume from their perspective they were innovating by not having entirely separate compiler codebases for every architecture? | 21:10:57 |
| emily | though I don't know what the pre-GCC landscape was like | 21:11:07 |
 colemickens 🏳️🌈 | looks like ruby broke | 22:56:04 |
| colemickens 🏳️🌈 | I wonder why my system even uses ruby 🤔 | 22:56:13 |
| 19 Jul 2025 |
 Colin | since doCheck = true and doInstallCheck = true act as no-op when cross compiling to an incompatible architecture, what's the right way to express a check phase when authoring a cross-aware package? | 01:39:10 |
| emily | what kind of check are you doing that works with cross? | 01:44:49 |
| Colin | run shellCheck, ideally even $out/bin/$pname --help under an emulator as a smoke test | 01:47:22 |
| Colin | i just deployed some totally invalid shell scripts to a aarch64 host & didn't realize until after the deploy. i'm using the oils shell, and it's not 100% backward compatible across updates; in this case a script i wrote a year ago suddenly became invalid. would have caught that with any trivial checkPhase. | 01:50:44 |
 Artturin | In reply to @colin:uninsane.org
since doCheck = true and doInstallCheck = true act as no-op when cross compiling to an incompatible architecture, what's the right way to express a check phase when authoring a cross-aware package? postInstall / passthru.tests https://github.com/NixOS/nixpkgs/blob/aa030ff8678c65967b09472893bab8f35c643d15/pkgs/by-name/si/siketyan-ghr/package.nix#L41 | 02:01:50 |
| Colin | is there an easy way i can build all passthru.tests for every package i have on my system? | 02:03:52 |
| Artturin | Let's add doInstallCheckCross? | 02:04:23 |
| Colin | In reply to @artturin:matrix.org
Let's add doInstallCheckCross? hmm, there are some old TODOs about not forcing doInstallCheck to be false on cross. idk if those might still be relevant | 02:05:12 |
| Artturin | In reply to @colin:uninsane.org
is there an easy way i can build all passthru.tests for every package i have on my system? I'm not aware of an easy way | 02:07:12 |
| Colin | Redacted or Malformed Event | 02:08:04 |
| Colin | In reply to @colin:uninsane.org
hmm, there are some old TODOs about not forcing doInstallCheck to be false on cross. idk if those might still be relevant https://github.com/NixOS/nixpkgs/blob/aa030ff8678c65967b09472893bab8f35c643d15/pkgs/stdenv/generic/make-derivation.nix#L322-L324 | 02:08:07 |
|  missbehaves changed their display name from thewholeworldisburning123 💕 to missbehaves. | 19:49:29 |
