10 Aug 2024 |
ghpzin | It's a feature rustup added in that PR Alyssa bisected failure to. | 14:29:22 |
ghpzin | * It's a feature rustup added in that PR Alyssa bisected failure to. It is called "test" and not included in default features. | 14:29:49 |
ghpzin | * It's a feature rustup added in that PR Alyssa bisected failure to. It is called "test" and not included in default features. Unless I misread, they just added it to all their CI, ie:
cargo check --all --all-targets --features test | 14:30:41 |
emily | right, that makes sense. isn't there a feature set on by default for test builds though? oh, I guess that's #[cfg(tests)] ? confusing… | 14:31:55 |
emily | ok, I will check if it is that simple :) | 14:32:04 |
emily | I'm used to habitually skipping over CI changes when checking PRs for Nixpkgs, bad habit. | 14:32:35 |
emily | …but then why do tests pass on aarch64-linux ??? | 14:33:48 |
emily | ohh | 14:34:34 |
emily | doCheck = !stdenv.isAarch64 && !stdenv.isDarwin;
| 14:34:42 |
emily | 🤪 | 14:34:45 |
Alyssa Ross | ohhh | 14:56:21 |
emily | I'm working on testing it now | 14:58:10 |
emily | I suspect we might be able to run the checks now | 14:58:15 |
emily | (on those platforms) | 14:58:21 |
emily | btw, PSA for anyone interested: https://github.com/NixOS/nixpkgs/issues/333702 | 14:58:29 |
emily | Alyssa Ross: you said that the individual crates would have to be essentially source packages because of Cargo's demands, but is that really true? I don't know how much compilation time it'd actually save us, but we could compile libraries as rlib s, couldn't we? hell, I think we could even get away with dylib – ABI compatibility doesn't matter to us given the whole Nix thing | 19:40:11 |
emily | I think there is at least one other distro that dynamically links Rust crates (Fedora?) | 19:41:24 |
emily | I guess there is the potential issue of packages that support multiple major SemVer versions getting locked to different things at dependency resolution time 🤔 | 19:43:17 |