On Wed, Apr 30, 2008 at 1:40 PM, Simon Marlow <[EMAIL PROTECTED]> wrote: > > Have > > you enabled the global cache mechanism? Have you experimented with the > > repository format upgrades? So many questions, so little time :) > > Not in production yet, no. I'm still using my old darcs-1 repositories > for now so that I can still use darcs-1 if anything goes wrong.
You can expect darcs-2 to do better for some of these operations when working on a hashed repository. It's pretty easy to run side-by-side comparisons (if you can afford the disk space for two ghc repositories...), since you can push and pull between them. There are some slowdowns when running darcs-2 on darcs-1-format repositories, which relate to refactoring the code to make atomicity easier and clearer, but which require workarounds on darcs-1-format repositories, since the format just doesn't allow for atomic operation. Other slowdowns relate to bug-fixing of certain "pending"-related functions. I ended up (after lots of bug-chasing) going with an inefficient-but-obviously-correct approach, which may be what's slowing down add and remove (and this will be the same regardless of repository format). On the plus side, this fixes bugs that were present in darcs-1. We could make things faster for certain operations (e.g. add/remove) if we added in special-case code for those commands, but then we'd also risk adding bugs back in. It's worth doing, but requires at least two developers (to be confident that the result is correct), one of whom probably needs to be me. Ultimately, I'm not particularly interested in regressions on the darcs-1 format. Anyone who wants to continue using that format can continue using darcs-1 as well. It's when darcs-2 on hashed format repositories behaves poorly (compared with darcs-1 on the old format) that there's really a problem that needs addressing. David _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
