Ben Collins-Sussman wrote:
> Sure... the hard part here is: how are you going to divide r96? You
> need to 'subtract' /trunk:50-95 from r96, and capture the leftover
> diff. Or, as Ph. Marek has suggested, maybe you can generate the
> special diff by simply comparing trunk@95 with branch@96. Or, as Karl
> is pondering, maybe we want to avoid this situation altogether, and
> force r96 to be two separate commits.
How does SVK handle this? I haven't delved into that portion of the SVK
codebase, but I know from personal experience that this situation is
completely handled with SVK (since I do exactly this sort of back and
forth development all the time). Even if Subversion doesn't go the same
way, it would be worthwhile investigating "the competition" for good ideas.
The one (and as far as I can tell only) think that causes SVK to fall
over is where a file is renamed *and* changed in a single revision.
That never merges properly between the mirror and local branches, but if
I redo the merge alone on the local branch, I can successfully merge
from the mirror again (though I have to resolve the conflict on the file
that was modified).
John
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Nov 29 16:18:24 2007