[darcs-users] Benefits of patch theory

Guillaume Hoffmann guillaumh at gmail.com
Thu Aug 19 16:37:14 UTC 2010


I agree that in my case I don't need such intensive bug-tracking in my
repositorie, so I'm not representing the full spectrum of VCS uses :-)

>  (Many users of
> history-DAG-oriented (hDAG) systems claim that knowing the context of
> a change is an aid in debugging.  That's a YMMV kind of thing, of
> course.  Legal departments are also often very interested in such
> history for some reason.)

If such thing is needed, do like the trac-darcs plugin: prevent patch
obliteration and the command "darcs optimize --reorder" on the central
repo of a project. So, in such a repository, the order on patches
never changes, you have your history.

For non-central repositories (those in which people actually work),
Delta Debugging would at least automatize the problem of combinatorial
explosion you mention. (Maybe the next major version of Darcs will
have it?)

> Somebody mentioned "semantic patches" earlier.
> [...]

It as a heuristic and nothing more. Darcs is not supposed to
understand your code or to guarantee anything. It is just a very good
heuristic, that works well enough in most cases, and works better when
you write "necessary and sufficient patches".

I'm glad a lot of commands in Darcs are interactive, thus for instance
I can inspect the patches I'm pulling before accepting or rejecting
them. For me it reflects the fact that you are encouraged to validate
what you are letting Darcs do.

Guillaume


More information about the darcs-users mailing list