[MERGE] Implement generic stacking rather than pack-internals based stacking.

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Robert Collins wrote:
| This is now ready for final review. This patch contains:
|  - a Development1 format which is stackable on any repository with
| matching serialiser and model
|  - various tests and refactoring to support that..
|
| Attached is a full bundle, and a plain text diff of the incremental
| changes (which show the removal of the changes made to be able to use
| the internals of packs, back to doing purely interface based stacking).
|
| AFAIK this is all reviewed and good to go except for the incremental
| change that I've attached.
|
| -Rob
|
|
|

- -
- -        total_revisions = self._local_revision_index().key_count()
- -        total_packs = len(list(collection for collection, sizes in
- -            self._names.values() if collection is self))
+        total_revisions = self.revision_index.combined_index.key_count()
+        total_packs = len(self._names)


^- This seems to be counting all revisions in all packs, but only using
the names from the local packs. Seems a little odd.

Also, in general having the api be:

~ self.revision_index.combined_index

Is a bit strange. What is self.revision_index if it isn't the combined
index? I suppose if this is all just a plain "revert" to rip out all the
places that you were stacking in the internal structures, then there
isn't much for you to really do about it.

So overall

BB:approve

(Though this was only reviewing the "internals-to-generic.patch".

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkhiUC0ACgkQJdeBCYSNAAPeYgCgtzF7x2aHyAnX4WUWFEzhqbTs
6TEAoKn/FWdEXHErFyXZ01MwAC3riqGM
=+RL3
-----END PGP SIGNATURE-----