docs: add alternative Clang invocation

[mati865]

Clang accepts -fuse-ld=wild if ld.wild is present in PATH.

[mati865]

Maybe we should also put ld.wild symlink in prebuilt binaries and somehow recommend this symlink to distro packagers?

[davidlattimore]

It'd seem a shame to use double the disk space by having two identical binaries. But maybe we can include a tiny extra binary that invokes the other from the same directory. Recommending that distros include a symlink seems sensible.

[mati865]

You can include symlink in the tarball, IIRC hard-links as well, so there would be no additional disk usage.
Obviously, it has to be created manually by calling ln in release job and won't work with cargo install.

Recommending that distros include a symlink seems sensible.

I don't think we've got packager recommendations anywhere. Where should it go?

[davidlattimore]

You can include symlink in the tarball, IIRC hard-links as well, so there would be no additional disk usage. Obviously, it has to be created manually by calling ln in release job and won't work with cargo install.

That's true, I was mostly thinking about cargo install. I wonder if cargo binstall would pick up the symlink from the tarball.

Recommending that distros include a symlink seems sensible.

I don't think we've got packager recommendations anywhere. Where should it go?

I suspect we should perhaps have a new markdown file, since it's a pretty different audience. Perhaps PACKAGING.md?

[mati865]

From the docs it seems cargo binstall would not install the symlink:

Then it will concat the dir with "{ bin }{ binary-ext }" and use that as the final bin-dir.

https://github.com/cargo-bins/cargo-binstall/blob/main/SUPPORT.md#defaults

I suspect we should perhaps have a new markdown file, since it's a pretty different audience. Perhaps PACKAGING.md?

Sure, should I add it here or as a separate PR?

[davidlattimore]

Probably a separate PR. I'd say that it deserves its own entry in the changelog