Ian Bicking: the old part of his blog

Re: Distributed vs centralized scm comment 000

I don't think that's the duplication Greg is talking about. What's he's referring to is this sequence:

  1. Branch trunk -> foo (cheap, no dup files)
  2. hack hack hack on trunk (modify 20 files)
  3. Merge trunk changes to foo (oops)

At stage 3, we've now got 20 modified files in foo, for which we are storing diffs in both the trunk and in foo. Instead, the modified trunk files could be "re-branched". You'd still need a new revision in the branch, obviously, but it would just be with the new pointers. This is probably yet another issue that really needs true branch and merge history tracking.

Comment on Distributed vs centralized scm comment 000
by Steve Greenland