[MERGE] new merge-type=lca

[Andrew Bennetts]

Aaron Bentley wrote:
[...]
> > I wonder if this merge is good enough that we don't need to warn users that
> > we've detected a criss-cross when merging?
> 
> In the case of criss-cross, they will sometimes see their merge
> resolutions shown as conflicts with alternative resolutions of the same
> conflict (this is what the 'conflicted-a' / 'conflicted-b' lines do).
> So I *think* that it still makes sense to warn them that the merge may
> be a little weirder than normal.  What do you think?

I guess we should see what the user feedback (ourselves included) is like.

[...]
> > It's tangential to your change, so don't feel that you need to do this to get my
> > approval for lca-merge, but some explanatory comments for this functional would
> > be extremely helpful.
> 
> Here's the explanation I'm planning to put:
>         """Remove changes from new_plan that came from old_plan.
> 
>         It is assumed that the difference between the old_plan and new_plan
>         is their choice of 'b' text.
> 
>         All lines from new_plan that differ from old_plan are emitted
>         verbatim.  All lines from new_plan that match old_plan but are
>         not about the 'b' revision are emitted verbatim.
> 
>         Lines that match and are about the 'b' revision are the lines we
>         don't want, so we convert 'killed-b' -> 'unchanged', and 'new-b'
>         is skipped entirely.
>         """
> 
> I don't feel like that explanation's crystal clear, but it's the best
> I've been able to come up with.  What do you think?

I think that's great, actually!

-Andrew.