I don't know if it is a bug, but 100000000000000000000000000000000000000000000 = 0 (mod 2^32), so, if GHC considers exponent as Int (not Integer), then it's the same as 1e0 = 1. In the same way, 999999999999999999999999999999999999999999999 = -1 (mod 2^32), and 1e-(-1) = 1e1 = 10. Note also that the function "exponent" from the RealFloat class has the type exponent :: a -> Int On 5 Apr 2008, at 12:00, Vladimir Reshetnikov wrote:
The float literal 1e-100000000000000000000000000000000000000000000 in GHC evaluates to 1.0, and 1e-999999999999999999999999999999999999999999999 evaluates to 10.0. Is it a bug, or a documented overflow behavior?
What it the correct place to submit bug reports concerning GHC? -- Thank you,
Vladimir Reshetnikov (aka nikov), Microsoft MVP _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe