@RossSmyth I understand what you are saying. At the same time, it doesn't make it any less stupid.
Why don't we define in another file what the tbb from all-packages.nix is as well. And in that other file it's still not what it means because it's defined in another 2 files both with something different, so anyone that looks at the mold package.nix cannot know what the things in it mean because they are defined to something else somewhere else.

@RossSmyth I understand what you are saying. At the same time, it doesn't make it any less stupid. Why don't we define in another file what the tbb from all-packages.nix is as well. And in that other file it's still not what it means because it's defined in another 2 files both with something different, so anyone that looks at the mold package.nix cannot know what the things in it mean because they are defined to something else somewhere else.

Because overriding a generic attribute is easier than overriding a specific "versioned" attribute, which is a hack by the way. This is pretty much a nitpick because we don't have anything about this in contribution guidelines, so I'll merge it as is.

@azahi Is it a check problem or is it because the darwin build doesn't work in the first place?
That check is supposed to test the basic functionality of showing the program version and I doubt mold shows something on linux, but nothing on darwin and I have a feeling it didn't work before either, be that a nix problem or an upstream one, tho this is just a theory since I have no way to test darwin to see what is the problem.
Do I add darwin as being broken or just remove the check?

https://logs.ofborg.org/?key=nixos/nixpkgs.402472&attempt_id=9d18c985-8cc3-45a5-b264-0dbf5a17344c

Compilation works fine, it clearly fails to find the version string.

Compilation working != program working
Who here has a darwin and can test this program to see why it fails to find the version because I think it doesn't work even tho it compiles; I'm guessing paveloom since you clearly don't have darwin.

The below program builds completely fine, but it doesn't work

#include <stdio.h>

int main() {
    int *ptr = NULL;
    *ptr = 1;
    printf("",*ptr);
    return 0;
}

I also think it would be a good to set hydraPlatforms to unix except darwin since no one will use mold on macos and if someone does, it's cross-compilation where many packages need to be compiled and mold even compiles faster than glibc

I'll merge this as-is, Darwin users are welcome to try to fix this if needed.

FYI, it's the switch to tbb_2022_0 that broke mold on darwin, not the mold upgrade itself. It still works with the old tbb. I also tried upgrading tbb to 2022.1.0 and that didn't solve the issue. But it's the same version homebrew uses, so it's probably something wrong with either how we package tbb_2022_0, or the wider darwin stdenv.

How would mold package maintainers feel about downgrading tbb, at least for darwin?