Re: 3-way merge considered harmful (was Re: [Monotone-devel] merge weirdness...)

[K. Richard Pixley]

Nathaniel Smith wrote:

> To expand: I'll assume everyone knows about the criss-cross merge
> case, which forces us to choose more distant ancestors in some cases,
> or else risk silently corrupting code.
I don't.  Is there a reference?

> (This problem seems specific to the semantics of "rename"; there
> doesn't seem to an analogue for file content, because file content
> cannot be moved, only created or destroyed.
>  
How is this different from the following content case...

A =
y
B =
x
y
C =
y
x
D =
x
y

--rich
ps, is there a reference on the codeville merge?