[BUG] can't rename subtrees

[Martin Pool]

On 4/1/07, John Arbash Meinel <john at arbash-meinel.com> wrote:
> the 'tree_specific_string' is either a 'packed_stat' for working trees,
> or 'last_changed_revision' for revision trees.
>
> But 'fingerprint' is the bug. It is usually either a sha1 (files) or
> symlink target (symlinks), or '' (directories).
>
> I think for nested trees, it could either be '', or the
> last-committed-revision (but we could always just pull that out of the
> actual nested tree).

I think for the working tree (column 0) it should be empty, and for
the parent trees the referenced revision.  If it's None in the
inventory, it probably should be '' here.

There doesn't seem to be a very substantive test for
set_state_from_inventory at the moment; maybe we should add one that
does this.  That would be a bit more direct than test_revert_subtree,
though that is also welcome.

You could perhaps add to _validate the constraint that all the
components within each entry must be byte strings and not None.  It
might not have caught this but it would make the invariant concrete
and might help in future.

-- 
Martin