Void type and :m

[Lauri Alanko]

Two bugs. Firstly, the notorious void type

newtype Void = Void Void

makes ghc -O go in an infinite loop in the strictness analysis phase.
Without optimization it works fine. (It would be very nice if Void could be
eliminated _completely_, ie. Void -> a would have the same representation as
a alone, but this is just wishful thinking...)

Secondly, the :m command in ghci does not accept underscores in module names.

Both are still present in CVS as of May 29th.


Lauri Alanko
la at iki.fi