> > > > BTW: in hbsetup.h you added: > > > ( defined(__POCC_TARGET__) && __POCC_TARGET__ == 2 ) > > > to detect HB_OS_WIN_CE builds. Maybe it's documented method in POCC > > > and in such case should be left but if not then I suggest to rather > > > use sth like: > > > ( defined( HB_OS_WIN ) && defined( _M_ARM ) ) > > Maybe I'm missing something, but _M_ARM doesn't seem to work in POCC 5, > > I've found no better solution back then, that's why I opted for this > route. > > --- test.c > > #if defined( _M_ARM ) > > #error defined > > #else > > #error not defined > > #endif > > --- > > >pocc -Tarm-coff test.c > > Hm, In such case this seems to be much more serious problem. > I do not have the newest POCC but in version 4.50 header files needs > _M_ARM, _ARM_ and _WINCE and UNICODE macros for PocketPC builds. > If they are not set then it's possible that we will generate wrong > binaries and probably we should set them manually probably using > -D<name> pocc.exe option. > I think it will be good to ask Pelle about it.
5.00.1 headers also seem to need _M_ARM. _ARM_ is automatically defined, so it's not needed to be set explicitly. I'll do it. Above __TARGET__* check isn't needed, as _WINCE is already there to detect WinCE. I'm doing some tests and commit. Slightly off: Now I know how to create x86/CE builds. But with this the number of combinations is so high we'd really need a separate CPU make layer, or some other solution like separating WinCE compilers to a distinct HB_ARCHITECTURE 'wce'. This is the way used by QT for example, and WinCE *is* a different OS, despite resembling to win in many aspects. So our new arch would look like: wce/mingwarm wce/msvcarm wce/poccarm wce/poccx86 If there is no strong opinion against it, I'd like to do this. It will also help excluding/including libs for WinCE. Brgds, Viktor
_______________________________________________ Harbour mailing list Harbour@harbour-project.org http://lists.harbour-project.org/mailman/listinfo/harbour
