David Allsopp

Thanks, @hongchangwu! We still need to address what to do about `opam-version` and ensuring that these fields are an error if `opam-version: "2.0"` was encountered. We're just in the process...

This PR is definitely still remembered! The machinery for validating fields by version (i.e. rejecting `etc_root` in `foo.install` if it has `opam-version: "2.0"`) is likely to be added as part...

We're not looking at file format changes in 2.2, so bumping this for consideration in 2.3

It's unfortunately obvious what's wrong when looking at the debug output: ``` + C:\Program Files\Git\cmd\git.exe "reset" "--hard" "refs/remotes/opam-ref-d35a6f5c0cc219e85bbcb822d6f19ed9b2ec3448" "--" (CWD=C:\Users\DRA\AppData\Local\opam\issue6472\.opam-switch\sources\flow_parser) fatal: cannot create directory at 'tests/long_directory_names/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path': Filename too long ```

I don't think this has anything to do with pinning a sha - the issue is the path at https://github.com/rescript-lang/flow/tree/rescript/tests/long_directory_names/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path/really_long_path. The fix here needs to be in the repo -...

We should probably surface the error message from Git more clearly, though!

```console Armoury+DRA@Armoury /cygdrive/c/Devel/msvs-tools $ opam admin compare-version 0.8 '0.8.0~dev' 0.8 < 0.8.0~dev ``` Happened to be on a machine which has opam 2.4~dev as its working opam copy - sometimes...

The .cmx format change is an improvement, so I wouldn't remove that (i.e. it would be better after this PR to be switching frame pointers and flat arrays to do...

> I haven't been following the PR closely, but compilation artifacts are incompatible between different versions of the compiler (different magic number), so am not sure that "not requiring changes...

I agree that if the implementation in the .cmx file is contentious (I hadn't actually reviewed that part - it was the concept I was commenting on above) then it...