[darcs-users] User interface for conflicts: confusing enough to lose data

Juliusz Chroboczek jch at pps.jussieu.fr
Wed Jun 30 18:48:16 UTC 2004


DR> Would you also want to make whatsnew not display the conflict resolution?

Definitely not.  That would be dangerous.

DR> I don't like the idea that revert followed by whatsnew gives
DR> anything bug "No changes".

Okay.  (I don't necessarily share that feeling, but I understand where
you're coming from.)

DR>  And if we *did* make a change in whatsnew, then we'd have a weird
DR> situation where whatsnew gives "No changes", but you can still
DR> record...

As above, that would be dangerous.

DR> I think what you'd really want is something I wasn't able to
DR> figure out how to do, which is to have the contents of a repo with
DR> a conflict in it include the marker.

Yes, that would be ideal.

DR>  I definitely wanted to do this, and tried for about a year to do
DR> so, but failed.  The problem is that there's no way that I was
DR> able to figure out to mark conflicts in a manner that will be
DR> reproducible regardless of the order of merging.

Why do you want it to be deterministic?  As long as darcs behaves
deterministically when there are no unresolved conflicts, and it
always marks at least one conflict when there are, I'd be happy.

I think of unresolved conflicts as being an error situation.  In an
error situation, you're perfectly justified in behaving in any way you
desire as long as you (1) don't trash the user's data and (2) report
at least one error.  This is enough for the user to repair the
erroneous situation, and get his repo back to a consistent state.

DR> I guess you might use [switch] if you want to test a bugfix in two
DR> branches before recording?

Exactly.  (Before pushing, actually.)

                                        Juliusz




More information about the darcs-users mailing list