star or parallele merge

[David Allouche]

On Wed, 2005-10-05 at 08:54 -0400, Aaron Bentley wrote:
> David Allouche wrote:
> > That is what star-merge does. But in practise, it often conflicts in
> > criss-cross situations (for reasons that are not entirely clear to me).
> 
> You shouldn't get conflicts if you do star-merge --three-way, but the
> default diff/patch doesn't handle this case: THIS = "A", BASE = "B"
> OTHER = "A"

Star-merge does not conflict in this simple criss-cross case, but it
more often conflicts than not in real-life criss-cross situations in
Launchpad development.

The typical bug report (or user complaint) associated with a criss-cross
situation is "star-merge gives 42 conflicts in files I have not
modified". As I said, the mechanics of exactly why these conflicts occur
are not clear to me.

> > Smarter merge base selection algorithms properly pick the closest
> > unambiguous ancestor as merge base and cause a conflict. Aaaron Bentley
> > designed an algorithm with a tongue twisting name that has this
> > property.
> 
> I don't remember doing that.  If you mean "shortest-longest path", or
> "perhipheral merge", those didn't do it.  I did have a variant of
> star-merge that could at least *detect* criss-cross and give up...

Oh, my bad then.

All my dumbfounded apologies to Jan Hudec.

Is there any reason in principle why that would not be possible, or is
just an issue of insufficient pretzel?

-- 
                                                            -- ddaa
-------------- 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/20051006/4720495e/attachment.pgp