Reviving the issue-2897 branch

[Mark Phippard <ma...@gmail.com>]

All,

For a quick refresher, this branch was Kamesh's attempt to fix the
reflective merge problem.  We decided to postpone this for 1.5.0 and
go with the --reintegrate option.  Now that we have 1.5.0 out and on
our server, I think it would be worthwhile to revive this work based
on the current trunk.

It is still entirely possible that this approach will suffer all of
the problems that David Glasser predicted, but it should now be easy
to actually see if that is true on real code.  The reintegrate
approach clearly has a lot of limitations, as David also predicted it
would.  Some of those possibly are solvable, or can be improved, but
some clearly can't.  So it seems like it is worth seeing how this
other approach works.

I think we know in advance that the code will be complicated to
review, but if we can actually build a client and see how it works,
that potentially would provide the motiviation to do that review.

If the branch pans out and we like the result, it would be a nice
addition for 1.6.  Conversely, it should also be easy to see if the
approach just doesn't work well enough to be included in the code
base.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

RE: Reviving the issue-2897 branch

[Kamesh Jayachandran <ka...@collab.net>]

Hi,

I will start extracting the changes from current branch to a new branch in logical chunks that makes 
easier for people to review.

And will author a doc where this approach scores over/under the reintegrate approach.

With regards
Kamesh Jayachandran

-----Original Message-----
From: Mark Phippard [mailto:markphip@gmail.com]
Sent: Fri 6/20/2008 8:42 PM
To: dev@subversion.tigris.org
Subject:  Reviving the issue-2897 branch
 
All,

For a quick refresher, this branch was Kamesh's attempt to fix the
reflective merge problem.  We decided to postpone this for 1.5.0 and
go with the --reintegrate option.  Now that we have 1.5.0 out and on
our server, I think it would be worthwhile to revive this work based
on the current trunk.

It is still entirely possible that this approach will suffer all of
the problems that David Glasser predicted, but it should now be easy
to actually see if that is true on real code.  The reintegrate
approach clearly has a lot of limitations, as David also predicted it
would.  Some of those possibly are solvable, or can be improved, but
some clearly can't.  So it seems like it is worth seeing how this
other approach works.

I think we know in advance that the code will be complicated to
review, but if we can actually build a client and see how it works,
that potentially would provide the motiviation to do that review.

If the branch pans out and we like the result, it would be a nice
addition for 1.6.  Conversely, it should also be easy to see if the
approach just doesn't work well enough to be included in the code
base.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org