[darcs-users] A proposal: patch incoherence

Juliusz Chroboczek jch at pps.jussieu.fr
Sun Nov 21 18:54:15 UTC 2004


[I've reordered the wise followups.]

Andrew Pimlott :

> > Since unrecord creates a new empty patch that conflicts with the patch
                                                  ^^^^^^^^^
> ITYM "incoheres" with.

;-)

(Seriously, though, if you've got a better suggestion for the termino-
logy, I'm interested.  I'd rather avoid ``conflict'' in order to avoid
a terminological conflict, er, incoherence.)

> Can you think of a common reason for wanting the old patch?

Patch refusals, and, more generally, discrepancies between two
branches.

Suppose I've got a bug fix in my stable tree that doesn't apply to my
head tree; the patch that fixes the bug in the head tree is marked as
incoherent with the stable bug fix.  When I pull stable bug fixes into
head, I want the head patch to remain and the stable bug fix to be
ignored no matter how they are dated.

(I've got this very issue right now with the Polipo repositories, and
I have to manually exclude a bunch of patches whenever I merge stable
into head.)

> > 4.4 Arch-style merges

> I think that if you really want Arch-style merges, you have given up on
> darcs.

I realise that.  However, like it or not, I'm already doing Arch-style
merges in the Polipo tree, and I see no other solution.  I'm currently
doing them by hand -- excluding patches by hand --, and that's a
little backwards if you ask me.

> Consider that after doing one such merge, you can never use darcs on
> patches that depend on the patches in the merge.

Unfortunately.  Once you've started Arch-merging between two branches,
you're pretty much condemned to doing Arch merges between the two
branches for ever.

David Roundy:

> Your patch incoherence idea does seem like it could be a useful
> concept, but I have a few reservations.

I accept that.  I think that the *concept* is perfectly sound, but
that making it useful will require some serious thought about
user-interface issues.  (Which is why I'm not responding to your
user-interface comments, or Andrew's, for that matter -- both of you
are probably right, I need to think about it.)

> I would think there should be an --unsafe option to amend-record

That, at least, there's no doubt about.

> The trick of adding an incoherent empty patch when refusing a patch doesn't
> strike me as a good idea.  The problem is that it keeps you from changing
> your mind later.

Hmm... why can't you unpull the refusal later?  I think it's safe in
that case.

                                        Juliusz




More information about the darcs-users mailing list