Marc Khouzam

Thanks for the contribution @nclaeys. I want to make sure we don't have a backwards-compatibility issue here. Could you provide an example of the bad behavior and of the improved...

I think this PR is giving problems: 1. it removes indentation while #1444 adds indentation. When both are joined in #1525 we are not consistent with the indentation between `completion...

> We'd be changing the completions API to include a new `Init` that was once private but not changing the existing API that would constitute a breaking change? I'm not...

I like this a lot @scop! It keeps the user-experience identical for programs that don't have subcommands and still provides them with completion. I haven't looked at the test failures,...

I've looked into the test failures and I believe we have a potential backwards-compatibility issue with this change. This approach of creating the "completion" command even if there are no...

I've looked into this and there is indeed a problem. So, as I started explaining in my previous comment, a program that normally has no sub-commands will accept arguments. But...

I've posted what I think might be a working solution in #1559 . Please let me know what you think.

Thanks for the PR @scop. I now realize that there is value in a `completion` command even for tools that don't have subcommands. It would allow flags to be completed,...

Still valid

I believe this has been replaced by #1450