fix: Resolve conflicting name inference if from aliases

[SUPERCILEX]

This fixes 3 behaviors with inferring subcommands

• When name matches are for the same command, allow the match
• Don't panic when inferring long_flags because of a mix up between flag and name
• When long flag matches are for the same command, allow the match

Testing of long flags was covered in #4971

Fixes #5021

[epage]

Did you verify there are tests for the two other infer cases?

[SUPERCILEX]

Yeah: conflict test, simple, and exact.

[epage]

To be more precise, in #5021, I asked for a fix to the issue to also verify that longs and subcommand longs have tests to verify they don't conflict with their aliases when doing inferring.

[SUPERCILEX]

What do you mean by "longs and subcommand longs"? If longs means --foo, then tests were added in the last PR. And if subcommand longs means ./exe sub, then this PR adds a test for alias conflicts.

[epage]

If longs means --foo, then tests were added in the last PR.

So now the PR description links to that, saying so.

And if subcommand longs means ./exe sub, then this PR adds a test for alias conflicts.

clap supports subcommands to be flags, so you can use either foo or --foo. Named and long flag subcommands go through different code paths. It looks like long flag subcommands should work but we should make sure we have tests for each of these cases.

[SUPERCILEX]

clap supports subcommands to be flags, so you can use either foo or --foo

I had no idea that was a thing, thanks. Where is it documented?

[epage]

• Command::long_flag
• Command::long_flag_alias
• pacman cookbook entry

[SUPERCILEX]

Thanks, seems to all work!

[SUPERCILEX]

Sorry for completely dropping the ball on this, had to prioritize some other stuff. My nonsense tests should be fixed and long_flag inference actually works now.

[epage]

Thanks for the follow up!

As a nit, would you be ok moving all tests to a single commit at the start of the PR, making them pass on their own, and then have the 2 fix commits on top that update the test to pass again with their change? This will help show that (1) the test is testing the right thing and (2) this does fix the issue.

[SUPERCILEX]

As a nit, would you be ok moving all tests to a single commit at the start of the PR

Lol yeah I've even written about this before, done.

[epage]

Thanks for your patience through this whole series and for getting this implemented!

[SUPERCILEX]

Awesome, thanks!

[epage]

4.4.5 is now out