[darcs-users] Frustrations diffing against the last change to a file

Michael G Schwern schwern at pobox.com
Sat Apr 2 22:26:57 UTC 2005


On Thu, Mar 31, 2005 at 01:31:11PM +0200, Tomasz Zielonka wrote:
> Perhaps for most uses it would suffice to
> - name repositories

We're back to the issue of uniqueness.  How do you name a repository in a
globally unique manner without assigning it a huge hash?  I don't think
domain names are good enough.  Even user at domain.name isn't good enough,
what if I have the repo checked out twice (as with branching)?


> - create a script that will find the next version number not used
>   in this repository

Just stick an incrementing counter in _darcs that's not checked into the
repo.  This will track, per repo, what the next version number should be.


> - configure darcs to include such a version number (repo name + patch
>   number) in the patch name
> 
> Actually, it can be done with current darcs. Users would have to
> use some shell script instead of invoking 'darcs record' directly.
> 
> I have no time to go in the details, but I could prepare a
> proof-of-concept example it someone's interested.

Hmmm.  Thinking out loud (I've already rewritten this email a few times
throwing out objections)... 

The identifiers go into everybody's change log.  Now I see a patch as 
"foo.org-235" and another as "schwern.org-12" and another as 
"yarrow.bar.net-14".  If the identifiers can be that short and human readable
then it might work.  A domain name and a number isn't too hard to remember 
especially if that domain name repeats itself a lot.  

So I guess a lot comes down to how the repositories are named.





More information about the darcs-users mailing list