To: terra AT diku DOT dk
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Re: [ajh AT uwasa DOT fi: gmake371.zip Gnu Make 3.71 for MsDos (real/protected) w/ C src] 
Date: 	Thu, 30 Jun 1994 14:27:47 PDT
From: "Marty Leisner" <leisner AT sdsp DOT mc DOT xerox DOT com>


I got a copy of gmake371...it looks like:
	1) there's no info file...
	2) there's no djgcc binary (a turbo c binary?)

I really would rather not have to start by compiling sources on Ms-Dog
(I don't have this attitude with the gnu distribution on unix...)

Also, the convention arg0...argn is followed...it would be compatible
with the mks toolkit if a ~ was used...

Mks works as follows:
A command line of 
%echo "this is" a test
would result in an environment of:

~echo
~this is
~a
~test
rest of environment


I think the current diffs are going to be unpalatable to Roland McGrath
(gnu make's author).
A working pseudo-fork lets any program work without #ifdef MSDOS which 
followed the fork/exec/wait paradigm...

Here's what a want to see:
	1) a working pseudo-fork for djgcc
	2) an exec which automatically puts argv into ~ form in the environment
	3) no globbing in go32, but constructing argv from the environment using
~ convention.
	4) A copy of msh which follows this convention...

I changed the Holub shell and croot to follow this convention and it worked
fine... I had a working fork on dos 3.x and did:
	switch(fork()) {
		case 0:
			play with fds, environment and exec
		case -1:
			error	
		default:
			wait()
	}

It involved directly creating psps, copying the a data segment to another segment,
switching psps and running the child...

It definitely makes porting much cleaner to ms-dog...

If anyone wants to look at the code I had, feel free to ask...
I'm not sure how to map this into XMS...

marty