On Wed, 2008-11-26 at 12:04 -0800, marcomaggi wrote: > On Nov 26, 10:39 am, Derick Eddington <[EMAIL PROTECTED]> > wrote: > > I don't mean to discourage you from making your build framework help > > figure things out for FFI-using libraries, but to encourage you to > > leverage the work already done. AFAIK, Felix's work was focused on > > getting struct offsets correct in spite of hidden packing/padding issues > > introduced by the C compiler, with the future goal of extending his work > > to help with more FFI things. I'm imagining that Autoconf-reported info > > should be used to generate libraries (or include files) which will be > > used by the rest of the Scheme FFI layers. > > > > Sorry but I have no time and will to inspect Felix's work. > But are we not talking about the "offsetof" operator of > the GNU C library, which can be implemented with (from > Tcl/Tk's header file): > > #define Tk_Offset(type, field) \ > ((int) ((char *) &((type *) 0)->field)) > > ? There is no problem in using in a GNU Autoconf test.
If you generate Scheme definitions that are imported or included and, based on whether the host is 32-bit or 64-bit, the appropriate definitions are used, then I think the Autoconf approach does seem promising. > And I can do it now (with neither porting nor reverse > engineering whatsoever, because I bet a coffee it is fully > undocumented and I want to see it working in that > environment with both 32b and 64b Ikaruses; first I > want to see it working, then I will believe it). Understandable. I hope you make it all work in a way other Schemers will use. -- : Derick ----------------------------------------------------------------
