type inference failure in ghc/hugs extensions

[guest]

Oleg recently posted on the mailing list haskell@… a magic coercion that makes pairs of numbers (x,y) behave as complex numbers. Aside from context, the type should be (a,a). The trick works with option -fglasgow-exts in ghc releases at least since 6.8.3. However in hugs Sept 2006 under option -98, it fails type inference. As I understand the documentation, option -98 is supposed to support the necessary extensions in common with ghc.

Is the inconsistent handling of the attached fragment a bug, a feature, or perhaps an undefined corner of the semantics?

Doug McIlroy? doug@…

[ross]

duplicate of #89.