[darcs-users] Hang on pull

Andrew Pimlott andrew at pimlott.net
Wed Sep 22 19:46:09 UTC 2004


On Wed, Sep 22, 2004 at 08:43:24PM +0200, Tommy Pettersson wrote:
> I have tried this approach, and it does not work very
> well.  My thoughts about it is that it might just be too a
> complex problem to solve in the general way darcs examines
> (non)commutating patches.

That would be a shame....

> Especially if the vendor has
> included some of your changes in their new version there
> will be conflicts on all those lines, and with non-commuting
> patches on top of those patches and so on...
> 
> My solution is this:

If I follow, your solution differ's from Tommy's in that you collapse
all local changes into a single darcs patch, then pull the single darcs
patch that upgrades the vendor tree.  Tommy instead maintains local
changes in the "natural" way, as many darcs patches.

However, I don't understand why his method should give darcs trouble.
He still is pulling only one vendor patch into his tree, and no patches
depend on the vendor patch.  Even if the vendor patch conflicts with
every local patch, it seems the complexity should be only linear in the
conflicting local patches.  And as long as you resolve those conflicts
before pulling another vendor patch, the previous vendor patch should
not figure in new conflicts.

However, I don't understand the algorithms, so I may be making bad
assumptions.

Andrew




More information about the darcs-users mailing list