[darcs-users] Re: Onto which other patches does a patch depend?

Michael Conrad conradme at email.uc.edu
Sun Jan 9 22:54:57 UTC 2005


On Sunday, January 09, 2005 5:58 AM, Thomas Zander wrote:
> I maintain the trees at work and occasionally someone commits something
> that is incorrect (he forgot to update another repository and 'fixed' his,
> for example).  So, his patch is incorrect, but its already been pushed
into
> the core repository.  I could unpull it from there, but changes are others
> already pulled it.
> The correct way to solve this is to really rollback his patch.  In order
to
> do that without actually walking up to the person who did it wrong, I want
> to be able to type a reason for the rollback and I want my email to be
> registered for that rollback, not the original authors email.

Well, it sounds like there are actually three totally separate use cases for
the rollback command in this thread:

1)  Temporarily cancelling a local patch to save the effort of branching, as
convenience.
2)  Expanding idea #1 to hold the patch in limbo until it might be needed
again.
3)  Using rollback to broadcast an official undo for a shared patch.

The first and second ideas could be accomplished with my previous post about
copying the patch to a suspension directory.  The third idea could be done
by having darcs generate the inverse of a patch (which I would think should
be a "darcs invert" command?) and then recording it and letting the usual
mechanisms for distribution handle the rest.

-Mike





More information about the darcs-users mailing list