dulwich-users team mailing list archive

[Jelmer Vernooij <jelmer@xxxxxxxxx>]

Am Thursday, den 13.09.2012, 15:39 -0700 schrieb Mark Mikofski:
> >>>  Here is the push code:
> 
> >>> 
> >>>      from dulwich.repo import Repo
> >>>      from dulwich.client import get_transport_and_path
> >>>      from getopt import getopt
> >>>      import sys
> >>> 
> >>> 
> >>>      def push(args):
> >>>          opts, args = getopt(args, "", [])
> >>>          opts = dict(opts)
> >>>          client, path = get_transport_and_path(args.pop(0))
> >>>          r = Repo(".")
> >>>          objsto = r.object_store
> >>>          refs = client.send_pack(path,
> >>>                                  objsto.determine_wants_all,
> >>>                                  objsto.generate_pack_contents,
> >>>                                  sys.stdout.write)
> >>> 
> >>>  and the traceback pushing to github over SSH
> >>>  (git@xxxxxxxxxx:mikofski/dulwichPorcelain) reads something like
> >>>  this ...
> >>> 
> >>>  in client.py at line 289:
> >>>      for refname in set(new_refs.keys() + old_refs.keys()):
> >>>      'List' has no attribute 'keys'
> >>>  from client.py at line 450-451:
> >>>      (have, want) = self._handle_receive_pack_head(proto,
> >>>          negotiated_capabilities, old_refs, new_refs)
> >>>  from dulwich_push at line 16:
> >>>      progress=sys.stdout.write
> >>>  from stdin at line 1
> >>>      push(['git@xxxxxxxxxx:mikofski/dulwichPorcelain'])
> >>> 
> 
> 
> >> So when I use send_pack, I use determine_wants_all from repo.object_
> 
> >> store, which is returning an empty list, but it should be a dictionary
> >> of {refs:shas}. When _handle_receive_pack_head is called it tries to get
> >> the keys and bam, AttributeError.
> >> 
> >> But I have 4 refs/shas in .git/objects by using bin/dulwich commit?
> 
> > So, the signature for determine_wants is different for send_pack and
> > receive_pack. determine_wants_all is meant to be used with the pack
> > receiver.
> 
> >
> > We should really rename the determine_wants argument for send_pack to
> > avoid this confusion.
> > 
> > Cheers,
> > 
> > jelmer
> >
> 
> 
> Hi Jelmer,
> So what is supposed to be used with send_pack? Should I modify or write
> my own determine_wants_all for send_pack that returns the dictionary as
> requested?
Yes, the easiest thing to do is to write your own. This should be fairly
simple if you're just updating a single branch or copying over the refs
of another repository.

> Is new_refs *all* of the refs in objects? or is it only the ones that are not
> _old_?
It's just the changes to the refs.

> For example, I have 4 ref:shas in objects. One commit is already on the
> remote, because it was there when I cloned it. So I assume that's old_ref.
> Now is new_refs the 3 remaining refs in objects, not including the one
> already on the remote? I really need to learn a lot more about git plumbing!
It's just the changes, though including refs that are already set (to
the same SHA1) shouldn't be harmful.

Cheers,

Jelmer

Attachment: signature.asc
Description: This is a digitally signed message part