Two more External Core bugs

[Kirsten Chevalier]

Two more bugs, both of which I was able to fix (patches attached):

- MkExternalCore.lhs didn't properly handle character literals greater than
0xff. It would generate code that contained something like (1010101::Char),
which made no sense. I changed it to use type Int instead of Char for these
literals.

- LexCore.hs ignored the sign of floating-point literals (so if a program 
contained -1.0, it would be parsed as 1.0). I fixed it.

-- 
Kirsten Chevalier * krc at uclink.berkeley.edu * Often in error, never in doubt
-------------- next part --------------
Index: LexCore.hs
===================================================================
RCS file: /cvs/fptools/ghc/compiler/parser/LexCore.hs,v
retrieving revision 1.6
diff -r1.6 LexCore.hs
69c69
< 		| isDigit c -> cont (TKrational r) rest'
---
> 		| isDigit c -> cont (TKrational ((fromInteger sgn) * r)) rest'
-------------- next part --------------
Index: MkExternalCore.lhs
===================================================================
RCS file: /cvs/fptools/ghc/compiler/coreSyn/MkExternalCore.lhs,v
retrieving revision 1.20
diff -r1.20 MkExternalCore.lhs
166,167c166,167
<     MachChar i | i <= 0xff -> C.Lchar (chr i) t
<     MachChar i | otherwise -> C.Lint (toEnum i) t
---
>     MachChar i | i <= 0xff  -> C.Lchar (chr i) t
>     MachChar i | otherwise -> C.Lint (toEnum i) (make_ty (literalType (MachInt (toInteger i))))