[rfc] [merge] removal of support for reverse changeset application

[Denys Duchier]

Aaron Bentley <aaron.bentley at utoronto.ca> writes:

>> It also simplifies the task of directly computing the resulting inventory
>> from the changeset (so I have a vested interest in this
>> simplification).
>
> I still don't agree that computing the inventory from the changeset is
> the right approach.  Since you want to mark the predecesors for files
> that have had merges, it would make sense to do it when you actually
> perform the merges, i.e. as metadata changes in apply_changeset.

Perhaps I'm missing your point, but you don't seem to be disagreeing
with the approach but with the timing.  My goal is to fuse
apply_changeset and apply_changeset_to_inventory.  The latter function
was merely a convenient stepping stone in the interest of modular,
staged development.

Cheers,

--Denys