On Sat, May 17, 2008 at 04:17:01PM +0000, [EMAIL PROTECTED] wrote:
> issue855 may open up an interesting question : can we do repository locking 
> any
> better?

Incidentally issue855 only affects push --dry-run on old-fashioned
repositories, since we don't need a lock in order to push from hashed repos 
(becaues they're more atomic).

> For example, we could just modify things centrally so that we never lock the
> repository when the --dry-run flag is present, but then each command would 
> have
> to guarantee that dry runs really are dry (which may be actually quite
> reasonable).

I think doing this centrally makes sense, we can do it right in the
lock-getting code, I believe.  And we could also ensure that commands never
violate the dry-run rule with some well-placed assertions, also in the
Darcs.Repository (well, probably both would go in
Darcs.Repository.Internals), in commands that modify the repository.

> I'm guessing at the Windows thing ( http://buildbot.darcs.net )

All right.

Applied.  Thanks!

David
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to