Compatibility for cmdliner 2.0.0

[psafont]

The type Arg.conv was made abstract and now needs to be handled with specialized functions. I've tested that it still works with cmdliner 1.3.0. and verified that the function was introduced in 1.0.0. Unfortunately 1.1.0 is not available in opam anymore, so I did not test that.

[Leonidas-from-XIV]

Thanks, looks good! Can you bump the required version in the metadata? It probably does not need a changelog entry even as it probably isn't a user visible change.

[Leonidas-from-XIV]

@mtelvers got the CI to work again, so we can run the tests in CI again :)

@psafont Turns out that Cmdliner does not accept "prefix" options anymore:

+  Usage: dune-release distrib [--help] [OPTION]…
+  dune-release: unknown option '--skip-test'. Did you mean '--skip-tests'?
+  [124]

I think our tests need to be updated to use the full wording and unfortunately (my apologies!) I have to get back on my previous statement, as this is a user-visible change it needs a changelog entry, something akin to this:

- Updated to use Cmdliner 2.0. This means that option switches have to be specified completely, prefix-matches don't work anymore (e.g. `--skip-tests` instead of `--skip-test`) (#512, @psafont)

[psafont]

We can solve this in other ways rather than changing user-facing behaviour:

• Limit the version to 1.3.0, and defer on what should be done for 2.0.0
• Add the common prefix options as explicit aliases, unsure if this would be compatible with 1.3.0
• Limit the version to 2.0.0, behaviour is changed, and there is no ambiguity on the behaviour of the tool depending on the library version used.

[Leonidas-from-XIV]

These are good points. Personally I am not a fan of ambiguous behavior and multiple ways to do the same, so I prefer the behavior of Cmdliner 2.0. Technically the old behavior can still be enabled, but it sounds like this isn't the way to go forward.

The issue I see is that we don't really know which prefix options we should introduce explicitely. I guess all the ones that end in a plural, but technically nothing prevented users from writing --skip-te.

@samoht what do you think? Should we bite the bullet and break some invocations or try to work around it?

[avsm]

Could we advance this one? Just go with the cmdliner 2.0 option of not autocompleting; scripts are easy to fix.

[psafont]

Just go with the cmdliner 2.0 option of not autocompleting; scripts are easy to fix.

Just in case it helps getting it merged: this is what the PR is currently doing

[samoht]

@samoht what do you think? Should we bite the bullet and break some invocations or try to work around it?

I've completely missing the ping... but yea, let's just fix the scripts to call full invocations.

[avsm]

thanks, merged and released in ocaml/opam-repository#29307