using per file merge bases ?

[Robert Collins]

On Wed, 2006-01-04 at 21:45 -0500, Aaron Bentley wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Aaron Bentley wrote:
> | Robert Collins wrote:
> | | Hi,
> | |     got an interesting bad base selection here:
> | |
> | | Merging http://bzr.arbash-meinel.com/branches/bzr/storage/;revno=1254
> | | into http://people.ubuntu.com/~robertc/baz2.0/storage;revno=1253
> |
> | Could you post the graph-ancestry output?
> 
> Actually, I've done my best to reproduce this:
> http://panoramicfeedback.com/opensource/bzr/badbase2.png
> 
> This looks a bit like a criss-cross merge, in that there are two
> candidate bases:
> d0e03, and rR1252.
> 
> I'm not sure that picking either one (even on a per-file basis) would
> necessarily give a satisfactory merge.  But you'll note that the
> selected merge base is not in the revision history of either branch.

Yah. I think the ancestor selection algorithm did the right thing as far
as it can tbh - its that we need a virtual tree for this: on a per file
basis, all of the new file versions have common ancestors, theres no
'new files' at all.

Rob


-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060109/4811f4eb/attachment.pgp