Paul Rider wrote in a message to Mike Bilow:

 MB> but it does work.  Also, the Watcom compiler is a lot 
 MB> cheaper than the Borland compilers to get the same 
 MB> capabilities, since the Watcom compiler will make DOS, 
 MB> DOS extended, Windows, Windows NT, OS/2 16-bit, and 

 PR> What would be the advantage of a 16 bit os/2 app ? A little
 PR> bit faster code I guess ?

There are several advantages to 16-bit OS/2 code.  Some OS/2 components,
such as physical device drivers, must be 16-bit, so you need a compiler
that makes 16-bit code.  Certain system DLLs must also be 16-bit, as when
you register a custom replacement for a Vio, Kbd, or Mou function.  Any
code that itself calls 16-bit API functions, such as the Vio API, may run
faster than 32-bit code because it avoids thunking.  In general, 32-bit
code will probably be faster than 16-bit code, but a small model 16-bit
OS/2 program generated with a modern compiler such as Watcom might
outperform a comparable 32-bit program.

 PR> Well, I sure wish I'd of gotten this message about a week
 PR> ago... Oh well, if the newer Borland compilers are as lame
 PR> as the one you used, I can see me switching over to Watcomm
 PR> as well. borland never arrived on Friday as expected, now
 PR> won't be in till tuesday...Do you know if they allow labels
 PR> in their inline assembly yet ? or is it even supported in
 PR> the os/2 compiler...

I know that Watcom does allow labels in in-line ASM, and I believe that
Borland does although I cannot speak from experience about Borland. 
Borland will allow compiling via ASM, so that should in theory support
anything that the assembler itself (TASM) supports.
 
-- Mike


---