compiler-wrapper: unvendor, v1.1.0, make GCC determinstic

[haampie]

• Unvendor the cc.sh script, so we do proper releases
• Add v1.1.0 with -frandom-seed support to make GCC output deterministic.

[kwryankrattiger]

I agree this is the right call.

[haampie]

Sigh,

g++: fatal error: cannot execute '/usr/libexec/gcc/aarch64-linux-gnu/13/collect2': execv: Argument list too long

[tgamblin]

I love this -- but please hold off on this so that we can get complete history in the compiler-wrapper repo. There are a lot of contributors to that script and I don't want them to miss out on credit if it's going standalone.

I think that repo is worth some attention outside Spack -- our compiler wrapper is very much battle tested and the evolution / features, etc. that made it that way are IMO interesting.

[haampie]

History of https://github.com/spack/compiler-wrapper/ is restored (from 3 separate repos, using git-filter-repo on spack and spack-packages). I have also deleted the v1.1.0 and retagged v1.0 since it wasn't used in Spack yet.

The v1.1.0 release requires at a minimum spack/compiler-wrapper#8 to fix "Argument list too long".

[tgamblin]

Isn't 1.0 the version of the current compiler wrapper in Spack? I guess if this is the same as the current version it's fine. Labeling this 1.1 or something would make a clear break and would make it easier to run, e.g., spack find compiler-wrapper@1.1: to find only the ones from github.

> spack find compiler-wrapper
-- darwin-sequoia-m1 / no compilers -----------------------------
compiler-wrapper@1.0  compiler-wrapper@1.0

-- darwin-tahoe-m1 / no compilers -------------------------------
compiler-wrapper@1.0  compiler-wrapper@1.0

[haampie]

In short:

• v1.0 the same script as before
• v1.1.0 faster and makes GCC itself reproducible with -frandom-seed