unbreak push to sftp...

[John A Meinel]

Robert Collins wrote:
> Some cruft I snuck in broke push to sftp:
> 
> === modified file 'bzrlib/branch.py'
> --- bzrlib/branch.py
> +++ bzrlib/branch.py
> @@ -98,18 +98,6 @@
>      # this is really an instance variable - FIXME move it there
>      # - RBC 20060112
>      base = None
> -
> -    @staticmethod
> -    def create(base):
> -        """Construct the current default format branch in a_bzrdir.
> -
> -        This creates the current default BzrDir format, and if that 
> -        supports multiple Branch formats, then the default Branch
> format
> -        will take effect.
> -        """
> -        print "not usable until we have repositories"
> -        raise NotImplementedError("not usable right now")
> -        return bzrdir.BzrDir.create(base)
>  
>      def __init__(self, *ignored, **ignored_too):
>          raise NotImplementedError('The Branch class is abstract')
> 
> 
> The create method here has never been functional, and
> bzrdir.BzrDir.create_branch_convenience is much more useful IMNSHO so,
> I'm nuking this.
> 
> === modified file 'bzrlib/builtins.py'
> --- bzrlib/builtins.py
> +++ bzrlib/builtins.py
> @@ -504,10 +504,7 @@
>                          if new_transport.base == transport.base:
>                              raise BzrCommandError("Could not creeate "
>                                                    "path prefix.")
> -            if isinstance(transport, LocalTransport):
> -                br_to = WorkingTree.create_standalone(location).branch
> -            else:
> -                br_to = Branch.create(location)
> +            br_to =
> bzrlib.bzrdir.BzrDir.create_branch_convenience(location)
>          old_rh = br_to.revision_history()
>          try:
>              try:
> 
> 
> 
> This was an example of problematic layering - the replacement call here
> is thoroughly unit tested, and if cmd_push had been using such wouldn't
> have broken. 
> 
> The reason the test cases haven't caught this is that push with a
> non-local-fs is not tested currently - and IMO this is ok as long as
> push does not have such type-based dispatch within it... its underlying
> layers need the variety of testing.
> 
> 
> Desperately seeking Susan^W+1
> Rob
> 

Susan

Though I would still like to see push tested through sftp://

Also, does this change the fact that push can update local working
trees? I know I added that a long time ago. The original push never
updated the working tree, I added code so that if it could, it would try to.

I think it is important to keep that for current branch formats. Future
ones, where the working tree is decoupled don't have to ever update.

John
=:->


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060221/d72aebd0/attachment.pgp