degenerate merge : false alarm

[Robey Pointer]

Aaron, when I properly cleaned up my pre-merge branch, and used the  
new bzr.dev merge, it only got 5 conflicts, like the old algorithm.   
But it was actually better because they were 5 contextual conflicts,  
where before 2 of the files had been "add" conflicts, so I'm guessing  
it picked a better merge base.

But just in case this scenario is useful for the discussions you all  
are having on picking merge bases, here's the scenario I was merging:

     M
     |
     |\      J   C
     | `-R   |   |    M = bzr.dev
     |   |\  |   |    R = me!
     @   | `.|   |    J = jam
     |\  |   *   |    C = robertc
     | `.|   |   |
     |   *   |\  |    * = merge
     |   |   | `.|    @ = chosen merge base
     |   |       *
     |   |      /|
     |   |    .' |
     |   |  .'
     |    .'
     |  .'
     |.'
     *   |
     |   |
     |\  |
     | `.|
         *  <-- my merge last night

Whew!

I can still post my branch, pre-merge, but I'm assuming it's not very  
interesting anymore.

robey