🚀 go-pugleaf

>>>>> On 13 Oct 92 15:33:27 GMT, resnicks@netcom.com (Steve Resnick) said:
Steve> In article <JT.92Oct13145625@fizyk1.fuw.edu.pl> jt@fuw.edu.pl (Jerzy Tarasiuk) writes:
>>>>>> On Tue, 13 Oct 1992 03:17:03 GMT, draper@ais.org (Patrick Draper) said:
>PD> In article <9210061742.AA05172@deepthought.cs.utexas.edu> STDN%MARIST@VM.MARIST.EDU (Dan Newcombe) writes:
>>What I want to do is this:
>>  Generate pure 80x86 code from a compiler that could be used without
>>DOS.  I don't want stuff like INT 21 and all of those functions.

Steve> You don't need source to the compiler. All you need is to write your own
Steve> startup code (which I have done) which sets up whatever runtime environment is
Steve> necessary. main (or something else) can be called to run your C program.
Steve> Most of the RTL is not going to work, since there are either DOS or BIOS calls
Steve> used in the RTL.

Steve> I have not encountered a real compiler which, as part of its code generation,
Steve> made DOS or BIOS calls. That's not legitimate code generation.

I suppose almost any compiler makes them. Otherwise, where it would put
the code produced ? Usually it puts code on .OBJ file.

But I remember there is one exception: TURBO PASCAL 3. It allows to
edit program in memory, compile in memory and then run using its
build-in RTL. Once it is loaded in memory, it almost don't need DOS!
Few DOS uses it does: get DOS version, parse filename, read (or try
read) .PAS file (on request can also write), exit. Perhaps all!

Steve> As for why you would use a compiler without accessing DOS? Embedded systems,
Steve> operating systems, any 80x86 application which needs to do things its own
Steve> way...

Not I. Dan Newcombe asked for it. I only put my 3 US$/14000/100 on it.