Note: I only added the 'opam compatible' check in the depends field, but not the similarly formed conflicts and depopts. For the latter it seems to make sense, but the former?

@ElectreAAS We use conflicts as inputs to the solver IIRC so they should also be validated the same as depends.

I just rebased the PR on main, sorry for the noise.

I have updated the expect part of test/blackbox-tests/test-cases/pkg/invalid-version.t, but not the text yet, as I'm quite confused as to what is the difference in tested behaviour in this test and test/blackbox-tests/test-cases/invalid-package-version.t.

The two tests are slightly different:

• One of them is about the fact that during dune build the package depends field is never validated. Allowing invalid names to go into the .opam file generation.
• The other is about dune lock not giving a good error when such a name is encountered.

However should we care about (depends) if we're not generating opam files and never use the data?

Not sure how many packages specify depends without generating opam files, because without it it is fairly pointless at the moment.