Add linting check for restricted prefixes

[benmandrew]

Fixes #315.

[raphael-proust]

The list of conflict-class (that we want the linter to help keep in sync with the prefixes) is discussed on ocaml/opam-repository#25861 (review)

The discussion is not complete (as time of writing of this comment).

[benmandrew]

The current prefix to conflict class mapping according to my understanding is:

• arch- to ocaml-architecture-setting
• system- to ocaml-system-setting
• host-arch- to ocaml-host-archicture
• host-system- to ocaml-host-system
• msys2- to msys2-environment

This can easily be changed as the discussion evolves.

[shonfeder]

Looking good! I have one question and related suggestion about the new-package detection logic. My other suggestions are minor documentation and testing-related things.

[benmandrew]

Recommendations have been integrated and the PR has been rebased onto master; now we're waiting for the finalisation of the discussion on ocaml/opam-repository.

[shonfeder]

The list of conflict-class (that we want the linter to help keep in sync with the prefixes) is discussed on ocaml/opam-repository#25861 (review)

The discussion is not complete (as time of writing of this comment).

@raphael-proust could you confirm whether the prefix/conflict classes here match what you are expect from a maintenance angle?

[raphael-proust]

The prefix (presented as a glob) / conflict classes (presented as a literal string) that were settled on are:

mysys2-* / "msys2-env"
arch-* / "ocaml-arch"
ocaml-env-mingw* / "ocaml-env-mingw"    -- NOTE: the globbing pattern for file names doesn't have a dash
ocaml-env-msvc* / "ocaml-env-msvc"      -- NOTE: same
host-arch-* / "ocaml-host-arch"
host-system-* / "ocaml-host-system"
system-* / "ocaml-system"

[shonfeder]

@punchagan since I've touched this a fair bit and you're looking at the linting stuff anyhow, would mind reviewing this when you have a chance?

[punchagan]

@punchagan since I've touched this a fair bit and you're looking at the linting stuff anyhow, would mind reviewing this when you have a chance?

This looks good to me. I ended up re-ordering and splitting up the commits slightly to split out the refactor commits and the changes that add the new lint checks.

We could move the "analysis" to see if a package is a newly published one to the analysis module, as discussed yesterday. But, it could also come as a separate change.

[shonfeder]

Thank you for the review and the cleanup, @punchagan! I've made a no-op change just to sign the commits so I get the green label XD

We could move the "analysis" to see if a package is a newly published one to the analysis module, as discussed yesterday. But, it could also come as a separate change.

I think that would be the right way, but locating that check in the linting rather the analysis was already inherited in this work. If anything, that should be a followup I think, as it would possibly affect other parts of the pipeline. However, I think we may just postpone this entirely, as we are liable to be replacing all of this with the propotype work soon.