[MERGE] Root entry has a revision id

[John Arbash Meinel]

Aaron Bentley wrote:
> Hi all,
> 
> This is another patch in the make-root-less-weird series.
> 
> This patch ensures that the root entry has a revision id in all
> RevisionTrees and BundleTrees.
> 
> Since the stored inventory doesn't always have its revision id recorded,
> this is determined when the RevisionTree or BundleTree is generated.
> 
> Updates to commit were necessary to ensure that the revision id was
> appropriately set there, as well.
> 
> The inventory change is due to the commit changes.  It
> 1. Allows inventories to be generated that don't have root entries
> 2. Allows root entries to be added the same way other entries are added.
> 
> Although commit can now call record_entry_contents on a root entry, this
> does not create a weave or knit for the root entry.  Similarly, other
> CommitBuilders may choose to ignore the root entry if appropriate.
> 
> I have tried to maintain backwards compatibility two ways:
> 1. commit should work with old CommitBuilders (e.g. the bzr-svn one)
> 2. other CommitBuilder clients (are there any?) should work with new
> CommitBuilders.
> 
> Oh, and I nuked a bare except and added some comments.
> 
> Aaron

...

+
+
+class BundleInfo08(BundleInfo):
+    def _update_tree(self, bundle_tree, revision_id):
+        bundle_tree.note_last_changed('', revision_id)
+        BundleInfo._update_tree(self, bundle_tree, revision_id)

Why does BundleInfo08 need to exist, rather than just updating BundleInfo?



=== modified file 'bzrlib/commit.py'
--- bzrlib/commit.py	2006-08-05 22:33:36 +0000
+++ bzrlib/commit.py	2006-08-08 17:43:56 +0000
@@ -308,8 +308,11 @@
                 raise PointlessCommit()

             self._emit_progress_update()
-            # TODO: Now the new inventory is known, check for conflicts
and prompt the
-            # user for a commit message.
+            # TODO: Now the new inventory is known, check for conflicts and
+            # prompt the user for a commit message.
+            # ADHB 2006-08-08: If this is done, populate_new_inv should
not add
+            # weave lines, because nothing should be recorded until it
is known
+            # that commit will succeed.
             self.builder.finish_inventory()
             self._emit_progress_update()
             self.rev_id = self.builder.commit(self.message)
@@ -502,10 +505,13 @@
         # in bugs like #46635.  Any reason not to use/enhance
Tree.changes_from?
         # ADHB 11-07-2006
         mutter("Selecting files for commit with filter %s",
self.specific_files)
-        # at this point we dont copy the root entry:
         entries = self.work_inv.iter_entries()
-        entries.next()
-        self._emit_progress_update()
+        if not self.builder.record_root_entry:
+            warnings.warn('CommitBuilders should support recording the
root'
+                ' entry as of bzr 0.10.', DeprecationWarning, stacklevel=2)
+            self.builder.new_inventory.add(self.basis_inv.root.copy())
+            entries.next()
+            self._emit_progress_update()
         for path, new_ie in entries:
             self._emit_progress_update()
             file_id = new_ie.file_id

I'm guessing this should actually be stacklevel=1. Since the direct
caller is the one that is being warned. stacklevel=2 is generally used
for wrappers, IIRC.


...

v- Same thing about stacklevel here.

     def finish_inventory(self):
         """Tell the builder that the inventory is finished."""
+        if self.new_inventory.root is None:
+            warn('Root entry should be supplied to
record_entry_contents, as'
+                 ' of bzr 0.10.',
+                 DeprecationWarning, stacklevel=2)
+            self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))

...

v- I'm pretty sure we're using 'assertEqual(expected, actual)'

+    def test_root_entry_has_revision(self):
+        tree = self.make_branch_and_tree('.')
+        tree.commit('message', rev_id='rev_id')
+        self.assertEqual(tree.basis_tree().inventory.root.revision,
'rev_id')
+        rev_tree =
tree.branch.repository.revision_tree(tree.last_revision())
+        self.assertEqual(rev_tree.inventory.root.revision, 'rev_id')
+

 class TestCaseWithComplexRepository(TestCaseWithRepository):


So the NewCommitBuilder => _CommitBuilder stuff needs to happen. And a
couple other small things. Otherwise +1 from me.

John
=:->


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060814/bf6ca8f7/attachment.pgp