Ancestry and Merging Problems - Your Analysis?

[Justin Wienckowski <Jw...@stacksafe.com>]

Hello everyone,

I'm the version control admin for a software engineering team, and  
one of the developers has come to me with a merging problem.  I  
*think* I understand what has happened, but I'm far from sure, and  
I'd really like to get your ideas on (a) what exactly is causing the  
problem, and (b) if anyone has an idea for a more efficient solution  
than a manual merge file-by-file.


Project Layout:

/project_foo
----- /trunk
----- /branches
----- /tags


Sequence of Events:

1)  Copy /trunk to /branches/b1

2)  Over a period of time, changes are made by developer to /branches/ 
b1 and by rest of the team to /trunk (approx. 40 revisions)

3)  Attempt to update /branches/b1 with the interim changes made to / 
trunk (ala svn merge -r XX:HEAD /trunk /branches/b1 where XX is the  
revision at which step #1 took place).  This merge fails.

4)  Copy /trunk to /branches/b2

5)  Attempt to merge /branches/b1 into /branches/b2 (ala svn merge -r  
XX:HEAD /branches/b1 /branches/b2).  This also fails due to a  
directory that is deleted and then re-added.

6)  Merge /branches/b1 into /branches/b2 using --ignore-ancestry,  
which succeeds.

7)  Continue work.

8)  At several other points along the way, merge changes from /trunk  
into /branches/b2 to "keep it up to date."


The Problem:

When attempting to merge /branches/b2 back into /trunk, the merge  
wants to change a bunch of files that were never actually modified  
in /branches/b2 except by updates from /trunk.  It's obvious that  
ancestry is broken for those objects and SVN isn't realizing that the  
version of the objects in /branches/b2 are in fact from an old  
revision from /trunk.

However, I'm not incredibly experienced with this stuff, and I'd love  
to understand the underlying issue to make sure we avoid it in the  
future.  Anyone have an analysis about exactly where ancestry is  
breaking here?

In addition, any ideas for how to merge /branches/b2 back into /trunk  
that would be more efficient than manually merging file-by-file only  
those files that are known to have changed?



Many thanks!

-Justin Wienckowski
Senior Software Engineer
jwinky@stacksafe.com

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