[PATCH] more work on merge

[Robert Collins]

On Sun, 2005-07-10 at 15:08 -0400, Aaron Bentley wrote:
..
> But it's already tracing through all history...

You track all parents, not just the one committed against ?

> > Otherwise you could do some sort of "average" distance, rather than
> > minimum distance.
> 
> Anyhow, I'm not clear how we could use this information to pick a better
> base.

Right, this is the fundamental problem that trips up baz merge. Consider
the range of graph algorithms available!...

The real problem is defining 'closest common base'. Current baz does a
SPF sort of the graph from each revision should be done, then it is the
the first common node in each flattened side. In practice this is done
by a breadth first from each node and as soon as one search adds a node
in the others visited list, we stop at the end of that cycle, then
topoligcally sort all candidates (often more than one is reach
simultaneously) and choose from there.

If we can define 'closest common base' then we can code the right way to
pick it ;)

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/20050711/d86706ff/attachment.pgp