[darcs-users] Re: cvs migration and legacy

David Brown darcs at davidb.org
Sun May 16 15:43:22 UTC 2004


On Sun, May 16, 2004 at 07:22:13AM -0400, Eric S. Johansson wrote:

> It seems to me that what you are doing is imposing a definition of 
> patches on CVS.  The definition is one set of changes from the 
> maintainer and you're also enforcing a policy of all changes go through 
> you and are pushed to CVS.  the implication is that there are no others 
> pushing data into CVS.

I am the only one who submits changes to the Vim CVS repository.  Having
changes submitted to CVS by multiple developers is chaotic, anyway :-),
since it doesn't group changes to multiple files together.

> I also should have said yesterday that you are pushing from your 
> comingled repository into CVS and darcs.  Are you pulling from the darcs 
> pool as well?

Not with the CVS tree, but there is no reason this couldn't be the case.
With a separate Perforce archive, I do migrate patches in both
directions.  P4 makes it a little easier, since it does have the notion
of atomically committing a group of changes.

> if one was going to actually deal with CVS on a more bidirectional 
> basis, it seems to me that one could impose a definition of patches on 
> CVS according to timestamp of check-in.  In other words, all the files 
> are check-in at the same time are considered the equivalent of a 
> "patch".  Or maybe they're the Kremlin of what you get when you "darcs 
> add".  If that definition works, then you might have a tolerable for 
> bidirectional operations.

Usually the grouping is done of similarly timed changes with the same
revision text.  You won't get exact times, since the checkin itself can
take some time.  The other approach is to have some use intelligence to
regroup the changes into darcs patches.

Dave




More information about the darcs-users mailing list