https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93645
--- Comment #15 from Fangrui Song <i at maskray dot me> ---
-- is definitely rare, but not non-existent.
In GCC, there is {-,--}specs.
In Clang, there are --cuda-path, --ptxas-path, --hip-path, --classpath, etc.
(In reply to Martin Liška from comment #14)
> >
> > I think it may be useful to simply allow -fuse-ld=word (`word` cannot
> > include a separator).
>
> Sure, but Jakub had some concerns:
> https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573833.html
I do not see an objection to -fuse-ld=word.
For --ld-path, -- is definitely rare, but not non-existent.
In GCC, there are {-,--}specs --sysroot.
In Clang, there are --cuda-path, --ptxas-path, --hip-path, --classpath, etc.
-fuse-ld= users mostly care about whether another linker can build their
programs, not whether the option can bootstrap GCC. I actually think ld.lld is
quite sufficient in bootstrapping GCC but if there are edge-case extensions
which not supported, ld.lld developers may not want to the project with more
obscure options...
> >
> > If that may be troublesome, having -fuse-ld=mold in GCC 12 is still nice.
> >
>
> I've just done that:
> https://gcc.gnu.org/pipermail/gcc-patches/2021-December/587426.html
Thanks
