New review comment by Logarithmus on void-packages repository

https://github.com/void-linux/void-packages/pull/31706#discussion_r662338735

Comment:
@q66
>CI for cross architectures is pointless when it comes to verifying whether things work, since tests do not run

Is it possible to use `qemu` to fix that? Just asking about theoretical possibility, and not the complexity.

>the linker still defaults to x86_64 mode when run without arguments (rather than explicitly erroring about unsupported architecture)

Are talking about `mold` specifically or about linkers in general? If about `mold`, I guess the error message can be added quite easily, since the author of `mold` is a very nice person and accepts contributions without much bureocracy.

>the usefulness of using an experimental linker to cross-link from some other arch to x86_64 is questionable

Please, don't decide for the other Void users, noone forces personally you to use `mold`. Like Arch, Void is a popular distro among Linux users who like experimenting, and having `mold` in a repos for other architectures certainly won't harm anyone. About "experimental": if you didn't know, `mold`'s author is also an original author of `lld`. Also read the release notes for `v0.9.0`: https://github.com/rui314/mold/releases/tag/v0.9#:~:text=We%20bumped%20the%20mold%27s%20version%20from%200.1.1%20to%200.9%20to%20show%20that%20mold%20is%20getting%20ready%20for%20production.

>so i'd still make it archs=x86_64*

You would make it `archs=x86_64*`, but I don't see a reason for this sort of gatekeeping. You sound like if Void is a super-stable Debian-like distro where every packages is supposedly needed to be very thoroughly tested before geting into the repo, but it's simply not true.