[MERGE] InterCHKRevisionTree (brisbane)

[John Arbash Meinel]

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

Ian Clatworthy wrote:
> This patch adds a fast path for iter_changes between
> two RevisionTrees in a CHK repository. The code was
> written by Vincent and needs a review by someone more
> familiar than I am in this area.
> 
> Giving it a quick look, the embedded not_a_change()
> function is never used and can be removed. Or perhaps
> it needs to be used and the code doesn't do that yet?
> 
> Ian C.
> 

BB:tweak

Looking over the code, I think Vincent was thinking to use
'not_a_change()' as a different iterator, returning things in whatever
way that CHKMap.iter_changes() would have returned it. Which would then
require converting it to the second format.

Instead, we just do the work directly. So we should just remove the
'not_a_change' function completely.

...

+    # CHKInventory does not have an InterTree optimiser class (yet).
+    chk_tree_format = WorkingTreeFormat4()
+    chk_tree_format._get_matchingbzrdir = \
+        lambda:bzrlib.bzrdir.format_registry.make_bzrdir('development5')
+    test_intertree_permutations.append(
+        (InterTree.__name__ + "(CHKInventory)",
+         InterTree,
+         chk_tree_format,
+         chk_tree_format,
+         mutable_trees_to_revision_trees))


^- I think this is wrong, though. As we *do* have an inter-tree
optimizer format. And it should be properly set up with the appropriate
objects for testing. If it isn't, then we should quickly do so.

I remember the discussion on this, and I thought Vincent had already
done it. But perhaps it slipped through.

John
=:->

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

iEYEARECAAYFAknZVvEACgkQJdeBCYSNAAN2mQCdGDetPOadHsukqMDRfqYidzh8
9gIAoIAbuPPaBHBw/KO045lNFwZrJOpX
=0kXw
-----END PGP SIGNATURE-----