[comp.sys.nsc.32k] Minix death and other issues.

markm@markm.intel.COM (Mark Mason) (06/14/91)

>From an article by Jordan:

> ...  I've also NEVER been able to stage 3 compile a gcc
> version later than 1.38 due to a mysterious problem where the stage2 cpp
> no longer generates output (it sucks it in but splits out nothing, simply
> exiting). 
> 

Back when I was doing one of the last GNU ports to the tektronix merlin
(a 32016 machine running utek - a bsd4.2ish system), I ran into this problem
using gas. Basically, the stage 2 binary was mis-assembled by gas so that
one of the common addressing modes (I forget which one, but it involved sb)
was assembled as a reference through the module link table. I happened in
such a way to make cpp exit silently. After about a week of trying to track
it and fix it, I gave up and went back to using the utek assembler (which has
its own problems).

I was using gcc 1.37, and whatever the latest version of gas was at the time (the
32000 port was *very* buggy at the time).

I'm not sure how much this helps, unless someone wants to try tracking it down.

mason
markm@markm.intel.com

BTW: actually, mine never read the input either, the bug occured in the
code that was processing the -I switches, but I can't quite remember which
function.

jkp@cs.hut.fi (Jyrki Kuoppala) (06/15/91)

In article <m0jo7O1-0003F9C@meepmeep.pcs.com>, jkh@meepmeep (Jordan K.
Hubbard) writes stuff about Minix & what should be done.

I'll add one thing.  It's real boooring to patch all the GNU & other
stuff for include files and cripple them with #ifdef MINIX #then
include <foo.h> #otherwise #include <bar.h> and such.

So, I hear 1.6.something is quite near POSIX.  How about getting the
1.6.something library running on top of 1.5h so everyone could be
doing something useful (like saying #ifdef POSIX instead of #ifdef
MINIX, otherwise doing the same as above ;-) instead of endless
munging with the include files to make them compatible with an
obsolete Minix version.

//Jyrki

arl@zen.cs.hut.fi (Ari Lemmke) (06/18/91)

Jyrki Kuoppala (jkp):
jkp>	So, I hear 1.6.something is quite near POSIX.  How about getting the
jkp>	1.6.something library running on top of 1.5h so everyone could be
jkp>	doing something useful (like saying #ifdef POSIX instead of #ifdef
jkp>	MINIX, otherwise doing the same as above ;-) instead of endless
jkp>	munging with the include files to make them compatible with an
jkp>	obsolete Minix version.

	I think the magical 1.6.something library is 1.5.x library,
	ast is _not_ going to support library in Minix itself (hahahaha),
	he's packing it with his new compiler, and you can buy it ...

	Here's some earlier (edited) articles from comp.os.minix:

From: ast@cs.vu.nl (Andy Tanenbaum)
Subject: MINIX 1.6.16 now available by FTP
ast>	One thing you must watch out for, and I apologize in advance,
ast>	is that I am now running the ACK ANSI C compiler, not the old
ast>	one. This means that the stuff has not been tested with the
ast>	old compiler.  In general, this is not a problem, since ANSI C
ast>	is stricter than K&R C.  However, I also use the ANSI C
ast>	libraries.  This means that you must not use a couple of my
ast>	headers unless you are using some flavor of ANSI compiler, and
ast>	even then.
ast>	You can ftp them.  Needless to say, MINIX itself is not FTPable,
ast>	just the cdiffs from 1.5 to 1.6.16.

From: sverre@lars.seri.gov (Sverre Froyen)
Subject: Re: MINIX 1.6.16 now available by FTP
sverre>	These diffs appear to be incomplete.  For instance, there are
sverre>	no diffs to the lib files nor to the files in the
sverre>	subdirectories of the command directory (note also that the
sverre>	commands directory has been renamed command). The file
sverre>	1.6.16.crc is also imcomplete.

	So, finally it seems that ast is trying to make more
	money out of Minix. He has right to do so, but I think
	using the net people is immoral. There are hundreds of
	people who thought their patches and work would help
	for the others. It helped ast to create so good Minix that
	he really would get money out of it.

	That's the reason why I try to put all my spare time
	for Mach (currently bug fixing SCSI driver code).

>	//Jyrki

	arl

phil@cs.wwu.EDU (Phil Nelson) (06/19/91)

Here is the most recent word from AST on the minix library.

>Date:     Tue, 18 Jun 91 10:55:04 MET DST
>From: Andy Tanenbaum <ast@cs.vu.nl>
>To: beta-minix@plains.nodak.edu
>Subject:  Re:  Library source for 1.6 and later
>
>I haven't quite decided what to do with the ANSI libraries yet.
>Last time they were part of MINIX, not the compiler package.  I'm
>not sure about this time yet.  If I could get 10 minutes rest some
>time, maybe I could think about it.
>
>Andy Tanenbaum

I hope he makes them the same way this time.

--Phil