[PATCH 1 of 2] split: new extension to split changesets

[Martin von Zweigbergk]

On Tue, Jun 27, 2017 at 5:21 PM, Jun Wu <quark at fb.com> wrote:
> Excerpts from Martin von Zweigbergk's message of 2017-06-25 00:26:32 -0700:
>> Why is a2 still here? I would expect it to get stripped.
>
> Sorry, I missed this one (I probably want to do some email client changes to
> make sure I won't miss comments).
>
> It's a bug and fixed in V2.
>
>> Could you add another test for "Split a non-head with obsolete
>> descendant"? I'm curious what the behavior would be. If the descendant
>> had a successor, would we get divergence through the rebase?
>
> That feels like rebase's responsible to do it right. I'll try to fix
> rebase's code instead. But that's depending on a lot of things already so
> I'll probably defer sending split patches until rebase is good enough to
> use.

I agree that it's a problem that rebase can deal better with. However,
since it doesn't yet do that, I think it becomes split's problem. We
have the config to allow divergence. If that's not set, I suppose the
rebase will fail. Split should ideally detect that early so the user
won't have lost their work deciding how to split the commit up.
However, I don't think that feature needs to go into the first version
of split. We just need to get it fixed before the next release.