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

[Denys Duchier]

John A Meinel <john at arbash-meinel.com> writes:

> I believe Aaron's original idea was deeper seated. He was thinking to
> change the actually ChangesetEntry objects, so that their "apply()"
> functions would change both the physical contents on disk, and the
> inventory.

That's more or less what I intended to do.  I am just moving in this
direction carefully - verifying with each modification that I am still
computing the same information as before.

> The nice part about doing it that way, is it maintains the separation
> between the lower and upper layers.
> It means that ChangeExec.apply() can set/unset the 'executable="yes"'
> flag, rather than having apply_changeset_to_inventory understand what
> ChangeExec is doing, and how to apply it to the inventory.

As I said, the separate function was merely a device convenient to me
during development.  It is only transitory.

While we are on the subject:  I find it odd that changeset entries
have explicit object-based representations for contents changes and
metadata changes, but not for tree changes.

Cheers,

--Denys