I'm probably about to get my head bitten off for this, but here goes: would it be possible to move H/Direct in the direction of using the FFI libs? For a few reasons: - deprecated features are going to be removed at some point. In fact, I've removed them all in the new Core Libraries. - there's lots of duplicated functionality. - one might be able to mix H/Direct and hand-written FFI code. I realise this would probably be a lot of work, but then of course H/Direct will benefit from improvements to the FFI libs, like stack-allocation (if I ever get around to doing it). Cheers, Simon

-----Original Message----- From: Sigbjorn Finne [mailto:sof@galconn.com] Sent: Thursday, July 12, 2001 3:46 PM To: Simon Peyton-Jones; ahey@iee.org Cc: ffi@haskell.org Subject: Re: Trying to build HDIRECT with ghc-5.00.2

No, it was just a bug that was introduced somewhere on the road to GHC5. The version of Int in the repo does the right thing now.

--sigbjorn

----- Original Message ----- From: "Simon Peyton-Jones" <simonpj@microsoft.com> To: "Sigbjorn Finne" <sof@galconn.com>; <ahey@iee.org> Cc: <ffi@haskell.org> Sent: Thursday, July 12, 2001 01:04 Subject: RE: Trying to build HDIRECT with ghc-5.00.2

...

| (*) - ghc-5.00.2 give Int.sizeofInt{8,...} the type Int32; | HDirect generated sources assume Word32 (as it should be). => | you may need to tweak the generated code in a select few | places to make ghc-5.00.2 happy.

Sigbjorn:

Are you suggesting a change to the Int module? It would be nice to settle ethis so that no tweaking was reqd.

I'm ccing the FFI list as that seems the right place to decide.

Simon

_______________________________________________ FFI mailing list FFI@haskell.org http://www.haskell.org/mailman/listinfo/ffi