Log Message:
Update to 0.40.  Tested in NetBSD-current/amd64 and OS X 10.4.

Changes

- The vim merger has been improved and now uses diff3 to merge
  non-conflict changes automatically before executing vimdiff.

- Values used with the --depth option used to control recursion with
  node and path restrictions have changed. Using --depth=0 now means
  exactly the specified directories and *not* their children. Using
  --depth=1 now means the specified directories and their immediate
  children. Previously --depth=0 included children and --depth=1
  included grandchildren and it was not possible to exclude children
  using --depth.  The simple fix for anyone using --depth is to add 1 to
  the values they are using.

- Document that ssh: and file: sync transports are not supported on
  native Win32.

Bugs fixed

- `commit' now uses keydir specified in _MTN/options

- duplicate name conflicts now show a proper error message, even if
  a parent directory got renamed as well. In that case, the error
  message now shows both names for the directory and the offending
  file name.

New features

- The bare parent selector 'p:' can now be used in a workspace to
  query the parent(s) of the workspace' base revision. This is
  equivalent to "mtn au select p:`mtn au get_base_workspace_revision`".

- push, pull, and sync can be run with a single argument, which looks
  like
     mtn://hostname?include_pattenr/-exclude_pattern
  or
     mtn://hostname?include=include_pattern/exclude=exclude_pattern

Internal

- Update Botan to 1.7.4.

- Usage of the internal app_state object has been reduced, objects
  are better encapsulated now. The database interface has been
  enhanced to ease reduction of locking contention in the future.

- Merged the two indexes on revision_certs into a single one.

- The database schema has been changed so that it now stores
  binary SHA1 hashes rather than their hexadecimal encoding,
  in most places where these are used.  This reduces the
  database size and speeds up operations a little.

  Users who like to fiddle with the database directly are
  advised to use the sqlite functions hex() and quote() to
  print columns that store hashes (including IDs), and the
  hexadecimal literal notation x'DEADBEEF' to input them.

- Binary SHA1 hashes are also used for most in-memory
  processing, avoiding conversions and saving memory.