R. Boujbel

The new mechanism is good for me! clearer and more local

@feapef Do you still have the issue?

hej! For `opamroot-versions`, `BASEDIR` should be replaced in the output. Is it the only test failing like that? For `pin.unix`, the tests works for some distribution (tweaked for CI, cf....

Yes, the message need to be updated for tis case to be more clear

But using `working-dir` with no argument is not an error, it is a no-op.

Agree on the idea, but what if user set its jobs variable to that value? Should we add a warning to inform that we are removing that value?

In order to update sandbox scripts, a `reinit` is needed : `opam init --reinit -ni`. It is advised in all release blog posts (at the bottom). Is the problem occurs...

It is possible if there is changes at opam root (cf. new field `opam-version` at the top of the `.opam/config`), but it won't help you about sandbox scripts. To be...

Note that there is a [fix](https://github.com/ocaml/opam/pull/4092) of `-ni` option in 2.1.0~alpha.

Something that may help you for integration: in `2.1.0~alpha`, we added a hash of sandbox script, to avoid overwriting user's modifications (cf. #4020). That hash (md5) is stored in `.opam/opam-init/hooks/sandbox.sh.hash`....