ac>> FWIW, Microsoft QuickC 2.5 for DOS is a cut-down version of
 ac>> Microsoft C 6.0.

 > Almost everything except the optimiser.

 ac> QC has an optimiser, albeit not a particularly configurable one.

Perhaps you forgot the quotes around the word "optimiser". It
does an abysmal job, and you'll often get better code from leaving it off.
The one in QC is *not* the same one as in MSC (nor the code generator for
that matter), and certainly not in the same class.


 > files, output assembler from C and do inline assembler (there were
 > probably more - I never used QC, so it was never an issue).

 ac> Right - except QC can do inline assembler with the best of them. :-)

Must have changed at some point. But - I've never seriously used QC and I'm
relying on memory cells which have remaind unused for some years, so
perhaps I'm mistaken.

 ac> Unfortunately there doesn't seem to be a way (eg. a
 ac> pragma) to tell the QC compiler that the function
 ac> should return whatever value happens to be in AX; the
 ac> compiler will always report a "No return value"
 ac> warning.  The online help doesn't hint at anything to
 ac> disable this warning on a function-by-fun ction basis.
 ac> I could always disable all warnings, but that's
 ac> somewhat less than useful.

Or simply create a variable on the stack. and move your return value there
and return it.

 ac> Incidentally, you can't do inline assembler in WATCOM
 ac> C, right?  Or is that what #pragma aux is for?  

Yes, that's exactly what it is for. Not that I ever used much inline
assembler either until recently. And the only reason there is because C++
name mangling makes it difficult to use separate assembler modules without
resorting to separate functions outside of class members with C linkage.

--- MaltEd/2 1.0.b6