[darcs-users] Revision numbers and unpulling
Michael G Schwern
schwern at pobox.com
Sat Apr 2 10:41:05 UTC 2005
Had a nice little discussion on #darcs tonite about local revision numbers
and in particular how they work with unpull/undo'ing a patch. I think this
sums up the idea pretty well.
<beschmi> afaik (i haven't really looked at the code) they can be reordered when darcs tries to merge a patch http://azure.humbug.org.au/~aj/blog/2004/12/11?seemore=y#2004-12-11-darcs-merging
<Schwern> beschmi: When you say reorder... could you explain what that means?
<Schwern> beschmi: Does that mean that I have in my local repository changes A, B and C in that order in the change log and D is pulled down that I might wind up with something other than A, B, C and D?
<Schwern> Like A, B, D and C
<beschmi> i think so
<Schwern> Weird.
<Miteorite> can you not unpull B, and then pull it again and end up with A,C,D,B?
<Schwern> Miteorite: That's ok. It would be (to switch to numbers). 1,2,3,4. unpull 2. 1,3,4. pull 2. 1,3,4,5.
<Schwern> ie. The global "2" becomes your local "5".
<Schwern> Once you drop the requirement for global identifiers and make them just local identifiers things become easier.
<Miteorite> ah ha, that works. because they are.. idempotent.
<Schwern> Oooh, big word. :)
<Schwern> but exactly right.
<Miteorite> well as long as we remember that your #2 is my #5 :)
* Miteorite hugs math.
<Schwern> Miteorite: For cross-repository communication you'd continue to use patch names, dates and/or hashes.
<Schwern> Miteorite: You don't try to remember that your 2 is my 5.
More information about the darcs-users
mailing list