In ../source/dos/modes.inc

real32:
mov ax,7202h
push ax
popf
pushf
pop bx
cmp ax,bx
je processor_ok
call init_error
db 'required 80386 or better',24h
processor_ok:

My question is how does the address of the error msg get on the
stack? It is used by the call to init_error (see "pop dx" in the call)
yet it is declared, as shown above, after the call.

tia
Alan

The CALL instruction stores on the stack the address of the instruction that follows it, and then the RET instruction pops that value from the stack and continues execution from that point - this way calling the subroutines works.

In this case CALL is used differently, since the called routine never return control back, and there are no instructions defined after the CALL, but data (error message). So CALL stores on the stack the address of that data as if it was the next instruction.

Yes of course, for a non-returning call. Neat trick
that I didn't think through. Thanks.

While I have your attention <grin> does your
macro implentation include a variable number
of arguments? - like the 'vararg' in Masm.

..and a similiar 'vararg' for writing e.g. a 'printf'
function in asm?

Best Wishes.
Alan

p.s. I was weaned on Tasm and very much like what
I see in Fasm so far.

Yes, look in the section 2.3.3 of fasm's manual. My general advice for you is to read through all the materials from Documentation page, also the excellent vid's guide to preprocessor (the part of Tajga Tutorial).

Thanks Privalov.
I found it in the source e.g. CDECL.INC

Best Wishes,
Alan