[MERGE] Perform creation of BzrDirMetaFormat1 control directories using an RPC where possible. (Robert Collins)

[Andrew Bennetts]

Robert Collins wrote:
> This changes creation of .bzr control directories on smart transports to
> use the existing RPC when possible.
> 
> This drops 13 round trips of initial-push use cases, dropping the
> ratchet for initial push scenarios and also adds a close-to-the-code
> unit test to make sure this specific case doesn't regress.

Yay!

bb:tweak

> === modified file 'bzrlib/bzrdir.py'
[...]
> @@ -1707,7 +1707,29 @@
>  
>      def initialize_on_transport(self, transport):
>          """Initialize a new bzrdir in the base directory of a Transport."""
> -        # Since we don't have a .bzr directory, inherit the
> +        try:
> +            # can we hand off the request to the smart server rather than using
> +            # vfs calls?
> +            client_medium = transport.get_smart_medium()
> +        except errors.NoSmartMedium:
> +            return self._initialize_on_transport_vfs(transport)
> +        else:

Hmm.  Will this work right over vfs-only HTTP?  Perhaps this needs to do a
similar dance to probe_transport, which checks client_medium.should_probe()
etc...  Not that writing to non-smart HTTP is a very useful thing to do, but
it'd be good to check that this case still gets a sane error.

The “tweak” I want is:

  a) verify that an attempt to initialize_on_transport with a plain HTTP server
     gives a sane error, and
  b) add a NEWS entry for the improvement :)

-Andrew.