Comment 3 for bug 460702

John A Meinel wrote:

> A big concern here is that the "stable" branch breaks the import. Can
> you tell us what we broke so that we can evaluate if we really released
> stable bugfixes in 2.0.1?

Here was the stack trace from one of the selftest failures ...

> Traceback (most recent call last):
> File "/home/ian/.bazaar/plugins/fastimport/tests/test_generic_processor.py", line 1466, in test_copy_file_in_root
> handler.process(self.file_command_iter(src_path, dest_path))
> File "/home/ian/.bazaar/plugins/fastimport/processor.py", line 95, in process
> self._process(command_iter)
> File "/home/ian/.bazaar/plugins/fastimport/processors/generic_processor.py", line 280, in _process
> processor.ImportProcessor._process(self, command_iter)
> File "/home/ian/.bazaar/plugins/fastimport/processor.py", line 117, in _process
> handler(self, cmd)
> File "/home/ian/.bazaar/plugins/fastimport/processors/generic_processor.py", line 487, in commit_handler
> handler.process()
> File "/home/ian/.bazaar/plugins/fastimport/processor.py", line 208, in process
> self.post_process_files()
> File "/home/ian/.bazaar/plugins/fastimport/bzr_commit_handler.py", line 600, in post_process_files
> lambda revision_ids: self._get_inventories(revision_ids))
> File "/home/ian/.bazaar/plugins/fastimport/revision_store.py", line 236, in load_using_delta
> rev_id, basis_inv, inv_delta, present_parents, parent_invs)
> File "/home/ian/.bazaar/plugins/fastimport/revision_store.py", line 321, in _add_inventory_by_delta
> new_inv = basis_inv.create_by_apply_delta(inv_delta, revision_id)
> File "/home/ian/Projects/bzr/bzr/2.0/bzrlib/inventory.py", line 1204, in create_by_apply_delta
> new_inv.apply_delta(inventory_delta)
> File "/home/ian/Projects/bzr/bzr/2.0/bzrlib/inventory.py", line 1187, in apply_delta
> "New id is already present in target.")
> InconsistentDelta: An inconsistent delta was supplied involving '', 'TREE_ROOT'
> reason: New id is already present in target.
>

fastimport was providing a delta with TREE_ROOT already in it which is
correct for CHK repositories but not earlier ones. It had some
conditional processing deciding what to do based on whether
"create_by_apply_delta" existed. You added a method called that to
Inventory as part of the work to speed up "bzr log DIR", breaking my
logic. The fix was to change fastimport to test using "isinstance"
instead of "hasattr".

Ian C.