File-ids including full path

[John Arbash Meinel]

At some point we changed the code path for new file ids to use their
full relative path rather than just their base name.

I don't think it was intentional.

The side effect is only that we are generating longer file ids than we
would normally.

The fix is just:
=== modified file 'bzrlib/workingtree_4.py'
--- bzrlib/workingtree_4.py     2007-04-25 22:07:09 +0000
+++ bzrlib/workingtree_4.py     2007-04-27 18:30:31 +0000
@@ -165,7 +165,7 @@
                     state.set_path_id('', generate_ids.gen_file_id(f))
                 continue
             if file_id is None:
-                file_id = generate_ids.gen_file_id(f)
+                file_id = generate_ids.gen_file_id(osutils.basename(f))
             # deliberately add the file with no cached stat or sha1
             # - on the first access it will be gathered, and we can
             # always change this once tests are all passing.


Do we want to do this? I don't know that we want tree implementation
tests about file id names (since SVN trees would use a different naming
convention).

The difference is:

bzr add foo/bar/baz

With WT3 this will get an id like:

  baz-20070602132324-aofoauba324ab-1

With WT4 it will be:

  foobarbaz-20070602132324-aofoauba324ab-1

John
=:->