nitralime wrote:
On Mon, Mar 23, 2009 at 6:55 PM, David Matthews < I don't know how time consuming the implementation of XCOFF support would be. Excuse me my naive question: Is there any plan to implement it?
None at the moment. It shouldn't be a great deal of work especially if it is a variant of the PE/COFF. Look at the existing XXXexport.cpp files. You'd have to get the detailed specifications of XCOFF.
I am a little bit confused about your answer. What would it mean if you fix the bug?
The bug I've fixed (in SVN) is simply to allow this file to compile. However the effect of that will be to raise an exception when you try to export in native object file format. The Poly/ML Makefile does an export as part of building the "poly" file so the normal build process won't work. There is also a portable export format which is used for the files in the "imports" directory and the "polyimport" program is used to read and execute them. You would have to adapt the "exportPoly" file which is used to compile the basis library so that after the basis library is compiled the complete Poly/ML system is written out in the portable format rather than native object file format.
Do I get after that an interpreted byte code version of PolyML?
The issue of machine code or interpreted code is orthogonal to this. For example, i386 machine code may be exported as PE/COFF, ELF or MachO depending on whether the machine is running Windows, Linux or Mac OS X.
Regards, David.