[darcs-devel] announcing darcs 2.0.0pre1, the first prerelease for darcs 2
droundy at darcs.net
Mon Dec 17 14:23:03 UTC 2007
On Mon, Dec 17, 2007 at 12:20:28AM -0500, Max Battcher wrote:
> On Dec 16, 2007 4:45 PM, Alexander Staubo <alex at purefiction.net> wrote:
> > I'm a bit confused about the new repository formats. There's the new
> > hashed format, and also the "darcs-2 format"; but I don't see an
> > explanation anywhere as to why Darcs 2 needs two new formats or why
> > users should not simply go ahead use the "darcs-2" format.
> It's in the original announcement as well as the wiki page:
> Here's my 3rd party, possible inaccurate, summary of the differences:
> * The hashed repository format is just a change to the way inventories
> are stored and so you can still easily move patches between old
> non-hashed repositories and new hashed repositories. I believe you
> need someone with Darcs 2 to act as a mediator between the two
> repository styles, but it's possible. It seems that you can convert
> your personal working copies of repositories and get some (all?) of
> the benefits of hashed repository even if "upstream" repositories
> aren't using it, and you can still darcs push and darcs send to the
> non-hashed repositories with no worries.
> * The Darcs 2 format makes backwards-incompatible changes to the patch
> format (specifically the conflicts/mergers bits) and patches generated
> in Darcs 2 format can't be handled by Darcs 1 or Darcs 1 repositories
> at all.
Right. So, for instance, darcs itself cannot yet switch to --darcs-2
without forcing all darcs contributors to switch immediately to darcs 2.
So I'm using the --hashed format as well as a mirror repository in the old
format to allow multiple versions of darcs to interact.
Also, the darcs-2 format is still marked as experimental, specifically
because I may need to make changes. This is one of the reasons why this is
a "prerelease" rather than a "release candidate". Ultimately the darcs-2
format will be in use, and --hashed is merely a transitional format.
However, this is very useful, as the majority of the performance-affecting
code is identical between the hashed format and the darcs-2 format.
Department of Physics
Oregon State University
More information about the darcs-devel