I hope that I have correctly checked that this is a problem with
the hugs distribution we have, rather than with the way that the
support staff here installed it. (And I hope I have not just made
a fool of myself too...)
It appears that - contrary to the specification in the manual page -
the
There seems to be an unwarranted assumption in iomonad.c that a value
of type fpos_t is a scalar which can be cast (in C) into an integer.
I have just had to install hugs98-Feb2000.tar.gz on a system (it is
a redhat 7.0, for what that's worth) where the fpos_t is a structure
containing the offs
Given a declaration
data S = S { a :: Int, b :: ! Int }
I expect
a (S { a = 0, b = undefined })
to force the undefined in the b field of the S (and ghci agrees with me).
However:-
__ __ __ __ ___ _
|| || || || ||
