Hi Viktor, | I have downloaded Hugs98 now and compiled it with GCC on Linux and | DJGPP on DOS. However, the version for DOS does not seem to be very | stable, | I encounter internal errors from time to time, and sometimes it reboots | my computer when I execute a shell command. I am using Windows95, have | Cyrix166 CPU and have compiled Hugs with make.exe that comes with | Borland Delphi (version 5.1). I'd suggest that you try using the binaries that we distribute as part of the standard Hugs distribution; they should run ok on Windows 95. I used to build the Hugs distribution quite regularly using DJGPP2, and never had any problems, but I don't think anyone has tried it for a long time, preferring to use cygwin instead for gcc. I guess it's possible that something in recent changes to Hugs has broken with DJGPP, but I suspect that the problems you've encountered are caused by something else. | I also have a specific question: | what does the following type error mean: | | ----- | ERROR "comb.hs" (line 39): Inferred type is not general enough | *** Expression : p2 | *** Expected type : a (Either b c) (d e) | *** Inferred type : _80 (Either _81 _82) (_83 _84) | ----- The _n notation indicates a free (i.e. unquantified) type variable, in contrast with lowercase letters like a, b, c, ... which represent type variables that are (implicitly) bound by a universal quantifier. Here's a small expression that illustrates the same kind of problem: let f x = let g :: a -> [a]; g y = [x,y] in g x in f True The type signature provided for g here is not accurate; in the context in which it appears, g will not accept a value of any type "a"; it will only accept values that have the same type as "x", which is bound in an enclosing scope, and hence can't be generalized over. In Haskell, there is no way to give a type signature for g in this example. In Hugs, at least in -98 mode, you can get round this problem by using an explicitly scoped type variable: let f (x::a) = let g :: a -> [a]; g y = [x,y] in g x in f True Hope this helps! All the best, Mark ---------------------------------------------------------------------------- [EMAIL PROTECTED] Pacific Software Research Center, Oregon Graduate Institute Looking for a PhD or PostDoc? Interested in joining PacSoft? Let us know!
