Re: [PATCH 13/13] contrib/subtree: Remove --annotate

From: James Nylen <jnylen@gmail.com>
Date: Feb 18, 2013, 19:46 UTC
Message-ID: <CABVa4NhyJ0FtYoFf9ZMnVt1eR9ezoCnd0oM5xGg1fgk0RwBUKg@mail.gmail.com>
In-Reply-To: <87sj4t4gar.fsf@waller.obbligato.org>

On Mon, Feb 18, 2013 at 1:39 PM,  <greened@obbligato.org> wrote:

Show 5 quoted lines

> James Nylen <jnylen@gmail.com> writes:
>>  - add "fancylib" as a subtree of "myprog"
>>  - commit to "myprog" repo: "fancylib: don't crash as much"
>>  - split these commits back out to "fancylib" main repo, and remove
>> the "fancylib: " prefix

> Should this really be a function of git-subtree?  It seems like it would
> fit better in a history-rewriting command.  Wouldn't rebase -i or even
> filter-branch be a better way to do this?

I'm not a git guru by any stretch, so I'm sure there are other ways to accommodate the example use case above. I really just want to be able to split and merge repositories while keeping meaningful commit messages with an appropriate level of detail. Can you suggest an alternative workflow?

> If there's no --annotate I don't see why git-subtree should have the
> --unannotate functionality.

Because they are not inverse operations - they both apply to `git subtree split`. I think that `--annotate` would only be useful as an option to `git subtree merge`. In that case it would be the inverse operation of `git subtree split --unannotate`, and then I would agree that if you remove one, you can/should remove the other.

> Again, I agree that your example is relevant, maybe even common, but I
> don't necessarily think git-subtree should be in the business of
> rewriting commit messages at all.

I'm willing to accept that. Junio seemed to be leaning that way too in earlier emails.

> I'd appreciate more thoughts from you on this.  I want to make sure we
> can support your use case.

I currently need to enable `git subtree` manually anyway, since it's not part of the main distribution. So it's not a burden for me to support this feature with a customized script, or learn a new way to do it.

Thanks for your consideration of this small and nit-picky issue.