José Pedro Magalhães wrote:
bits is just pattern-matching on a constructor which no longer
exists. Replacing it with the new constructor solves the problem.
I meant genUpTo and not bits. I'm attaching a patch (genUpTo.patch) to
update the test.
The other 3 seem to be related to
http://hackage.haskell.org/trac/ghc/ticket/2759. Apparently this
caused gread to fail to read Float and Double types. I'll look
further into it.
Actually, strings and tree were related to #2759, but bits was also
related to #2875.
The problem with #2759 was that it caused readConstr to fail for Float
and Double types, since reads was being used without being told what
type to read. This was fixed by forcing the type to Double. I'm
attaching amend2759.base.patch.
#2875 caused showConstr (toConstr 'p') to return "'p'" instead of "p".
This in turn makes bits return a different (larger) result (but
deserializing after serializing is still the identity). However, I think
this result is 'ok'. Does anyone disagree?
patches pushed, and I've updated the output for bits (I didn't hear any
objections, and it looks reassonable to me).
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
