Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!rutgers!ames!amdahl!pyramid!uccba!hal!ncoast!kent
From: kent@ncoast.UUCP (Kent Williams)
Newsgroups: comp.os.minix
Subject: Version 1.2
Message-ID: <3652@ncoast.UUCP>
Date: Mon, 27-Jul-87 16:33:51 EDT
Article-I.D.: ncoast.3652
Posted: Mon Jul 27 16:33:51 1987
Date-Received: Sat, 1-Aug-87 08:56:46 EDT
Distribution: comp
Organization: Cleveland Public Access UN*X, Cleveland, OH
Lines: 48
Keywords: Notes on upgrading


This is a report on Minix 1.2, and on yet another attempt to use a
DOS compiler for Minix Development.

I had all 5 parts, and got everything compiled and poked into it's
proper home with relatively few problems. My Notes:

1. Several programs in /commands need more than the 2K the makefile
gives them: diff, fix, make, find, and probably more that I haven't
tried yet.  Running fix with too little space had the amusing effect
of entering 'panic' over and over again until I rebooted.

2. When you use the minix CC compiler, Kernel memory usage goes from
roughly 91K to 105K.  I didn't notice things being any slower.  My
general reaction to the compiler and assembler is that they are at
least as good as the ones that come free with PC-DOS ;-)

3. The 1.2 'cc' program is broken.  It passes a bogus flag
(apparently '-L' ) to one pass or another, which causes asld to kick
out a silly message about 'illegal word size.'

4. I took a broad whack at using the Datalight Optimum-C compiler for
cross development, and would appreciate any help anyone could give me
in doing this.  I got the library all compiled, and the assembly
files converted.  DOS2OUT compiled fine (except that dos.h had to be
included and the open() flags needed to be changed).  Programs seem
to convert to a.out format OK.

A simple hello.c type program will print out argv and envp properly,
so I can (almost) assume that message passing works.

BUT, when I tried to compile ls.c the resulting program behaves in a
highly icky manner under Minix. It doesn't lock up or trap to Mars or
anything like that -- it just doesn't work right, prints out stupid
stuff, etc., etc.

My questions are these:
1. Do you have to zero Udata yourself?  How exactly does Minix decide
how big Udata is? Do the 'stock' minix program assume that Udata is zeroed
2. Are structures byte or word aligned? (I tried both, with the same results).
3. What am I doing wrong?

If anyone has solved the porting problems for MSC, Aztec C86,
Datalight C, or Metaware HighC, email to me, Por Favor.

Kent Williams 
  {{harvard,mit-eddie}!necntc,well!hoptoad,sun!cwruecmp!hal}!ncoast!kent
ARPA: necntc!ncoast!kent@harvard.harvard.edu