On Tue, Oct 07, 2008 at 16:21:03 +0100, Eric Kow wrote: > > Personally, I think we should ship Petr's bugfix. The performance hit > > isn't all that big, and this makes repair usable for windows folks > > with annoying histories.
Ok, so I've had a chat with Petr on this on #darcs and have come to a
final decision: we will be accepting the performance hit and shipping
with Petr's code as is, no rollbacks.
Details: Petr agrees that there is a memory-related issue after all. He
has reproduced the slow-down on a machine with 1G of RAM. We both agree
that trying to repair the GHC repo on a 2G machine works fine (and
faster actually). So the problem is pretty well understood and
contained.
Following Petr's advice, I intend to document this performance
regression in the form of release notes, pointing out
(a) the bad news: repairing a GHC-sized repository requires a machine
with over 1G of RAM [2 is known to work], or in the worst case,
darcs 2.0.2
(b) the good news: the cause of the regression is pretty well
understood and we are working hard to fix it for the next release
(c) the better news: now you can repair nasty case-sensitive
repositories on your Windows or Macintosh, yay!
Thanks, everyone!
[now all we need is to get issue1095 sorted, and we're good to go]
--
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
signature.asc
Description: Digital signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
