On Wed, Nov 28, 2007 at 09:04:04AM +0000, Simon Marlow wrote:
> Thanks David for taking the time to summarise the state of play.  I'm 
> pretty excited to hear about the developments, and eager to try the new 
> version!

I've been trying to fit in some darcs work in the cracks of my day, and
have gotten the tests to the point where all but four of the tests in the
test suite (these are the tests that actually call the darcs executable,
not the unit tests) pass.  Mostly it's been a process of ensuring that all
calls to darcs patch code go through class methods, which has been fun.

Most of the remaining bugs are a little more complicated: I haven't yet
implemented conflict-marking for the new patch type, which will be needed
for three of the four remaining failures.  It won't be too hard, but it
also won't be a 15-minute job.  The fourth remaining failure is a bizarre
parse error of some sort when running with no pristine cache (!).  Alas,
now I need to get back to "real" work (which means editing a grant
proposal, this week).

> Manuel M T Chakravarty wrote:
> >Maybe we should just invest that effort into helping David improve his 
> >code instead.
> 
> This idea occurred to me too, and there's probably a lot we could do to 
> improve darcs with a few days of Ian's and my time.  Something to think 
> about, definitely.
> 
> Cheers,
>       Simon

That would most definitely be appreciated! But not just yet: if you're only
going to devote a few days' time (which makes sense: none of us want ghc
neglected!), I'd rather pick the low-hanging fruit first, and then let you
handle the tougher cases.

At least to start with, the most helpful thing you could do with darcs is
just find test cases where the performance lags behind that of older darcs.
Or also cases where it has always been unacceptable, but such problems are
less likely to be quickly fixable.  This is all to say, in a week or two,
when you've got a shiny new version of darcs to play with...
-- 
David Roundy
Department of Physics
Oregon State University

_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc

Reply via email to