[darcs-users] Colin Walters blogs on Arch changesets vs Darcs

[David Roundy]

On Sat, Nov 27, 2004 at 10:44:39AM +1000, Anthony Towns wrote:
> Andrew Pimlott wrote:
> >But what is the combination of two patches that both add (and probably
> >add different contents to) the same file?  The natural (if unintuitive)
> >answer is that neither version of the file exists.
> 
> How can something be natural but unintuitive? Anyway, the other
> possibilities are (obviously) only one of the files exists; or both
> versions of the file exist, with one getting the original name, or
> neither getting it.

The trouble is that what is in the "recorded" result (as opposed to the
conflict markers which are unrecorded) *has* to be uniquely determined.
The only uniquely determined result that I could come up with was doing
neither.  The "glump 0.9" which is used in marking conflicts was originally
intended to be the direct result of the conflict (i.e. what shows up in the
unrecorded state).  The trouble was that there doesn't seem to be any way
to choose a unique result.

Actually, we could have used patch IDs to break the symmetry, but that
would require reworking the code so the commutation of primitive patches
"knows" what patch IDs "own" the primitive patches, which is a bit ugly.
-- 
David Roundy
http://www.darcs.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.osuosl.org/pipermail/darcs-users/attachments/20041127/312e7dd3/attachment.pgp