@epage sorry if i am being a headache, but i couldn't understand why is this test is failing, could you help me, please.

 thread 'bash::complete_dynamic_empty_option_value' (3174) panicked at clap_complete/tests/testsuite/bash.rs:465:5:

---- expected: tests/testsuite/bash.rs:463:20
++++ actual:   In-memory
   1      - exhaustive --empty=     % exhaustive --empty=∅
        1 + exhaustive --empty=∅

the actual value, is not the expected behavior? i am a little confused.

I also want to know if its better to have '=' as default or ':' or both...

Looks like there are still design aspects to work out, so discussing them at #6280 (to avoid splitting the conversation between the two competing PRs).

On PR structure:

• We expect commits to reflect how they should be reviewed and merged and now how they are developed
• Commits should be atomic
• Preferably, tests are added in a commit prior to the change in behavior, with them passing, reproducing the undesired behavior. The fix would then update the tests to pass and the diff between them would represent how behavior changed

See also https://github.com/clap-rs/clap/blob/master/CONTRIBUTING.md#preparing-the-pr

sorry if i am being a headache, but i couldn't understand why is this test is failing, could you help me, please.

I'm unsure why CI and locally, these diverge. For now, I just make sure the output matches CI.

I may need help with a new test set, but it is very difficult because the behavior
is changed in bash rather than inside of the lib.

We have shell integration tests that run the completions inside of the shell. For the new completion system, we try to limit this to shell specific logic which fits here. We try to mirror the tests across the different shells so we can see where all stand, even if a fix only goes into one.