Win32 libraries for Haskell

Simon Marlow simonmar at
Wed Feb 4 10:09:54 EST 2004

> > The situation seems to be this:
> >
> > * Sigbjorn is actively maintaining hslibs/win32 for Galois 
> reasons, and
> > is
> >    reluctant to transfer to libraries/win32 because it's 
> untested (I'm
> > assuming
> >    that the Green-Card dependency is soluble by putting the 
> GC output
> >    in the CVS tree).  Fair enough.
> >
> I have no motivation other than it is a useful library used by many.
> GreenCard output, and output by the FFI backend in particular,
> is pretty clunky(*), so much so that the *.hs files in 
> hslibs/win32 has been
> cleaned up quite a bit. Furthermore, GreenCard produces non-reentrant
> code, which is why I don't recommend it being used, libraries/Win32 or
> otherwise.
> hslibs/win32/ doesn't improve on the non-reentrancy problem, btw.
> --sigbjorn
> * - having written large chunks of the current GC codebase, I feel
>      entitled to criticize it :)

Interesting - in what sense is the greencard output non-reentrant?  Do
you mean that it won't work in a multithreaded app, or is it a
restriction on what you can do in a callback?


More information about the Libraries mailing list