rewriter: allow reading dict-valued kwargs; improve key-value pairs in CLI; minor improvements

[bgilbert]

• Fix kwargs info for dict-valued kwargs. We can't modify these yet but it's useful to be able to read them.
  • Issue for modification support: Rewriter doesn't support dict-valued kwargs #14739
• The default-options delete and kwargs delete subcommands required key/value pairs, where the key was deleted and the specified value was ignored. This matches the JSON script mode interface but is unpleasant as a CLI. Since the UI is documented to be unstable, we can change this. Have the CLI delete commands accept a list of keys instead.
  • Fixes meson rewrite kwargs delete project / <option> raises unhandled exception unless a dummy second value is passed #13234
• Many rewriter subcommands expect key/value pairs as adjacent arguments. If the last value is missing, show a user-friendly error instead of a backtrace.
  • Reported in meson rewrite kwargs delete project / <option> raises unhandled exception unless a dummy second value is passed #13234
• Support the license_files kwarg to project().
• Minor cleanups.

[bonzini]

Fixes: #14739

[bgilbert]

@bonzini Nope, this PR allows reading dict-valued kwargs but not writing them.

[dcbaker]

This all looks good to me.

I'm not sure if we can land the drop unused command-line arguments patch. That changes the API in a way that has the potential to break users who are depending on that behavior. Of course, I'm not really sure what to do about it, because the behavior we have right now is less than ideal.

[bgilbert]

Well, the documentation says:

The rewriter itself is considered stable, however the user interface and the "script mode" API might change in the future. These changes may also break backwards compatibility to older releases.

We are also open to suggestions for API improvements.

No doubt users are depending on it anyway, but at least there's guidance we can point to.

[bonzini]

There was even a bug that wasn't closed, so I think it is doable... Maybe detect an empty even-position value and ignore it with a warning, so that it's at least possible to write a script that works for both old and new? Does it make sense for any rewriter use case to use empty keys?

[bgilbert]

Maybe detect an empty even-position value and ignore it with a warning, so that it's at least possible to write a script that works for both old and new? Does it make sense for any rewriter use case to use empty keys?

Good idea, done. Empty keys don't make sense; kwargs currently rejects them but default-options will agree to remove a nonsensical default option like '=foo'. However, I've opted to conservatively ignore-with-warning only when all even-position values are blank.

[dcbaker]

LGTM me. Thanks for dealing with the backwards compat questions, I don't like asking them and I know no one likes dealing with them :D

[bonzini]

@dcbaker this was approved months ago, anything blocking the merge?

[dcbaker]

Just my memory :(