Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!swrinde!mips!daver!bungi.com!news
From: kls@ditka.UUCP (Karl Swartz)
Newsgroups: comp.sys.nsc.32k
Subject: Re: a first time for everything
Message-ID: <9009022058.AA13234@ditka.UUCP>
Date: 3 Sep 90 03:58:47 GMT
References: <<9009021947.AA16404@manatee.UUCP>>
Sender: news@daver.bungi.com
Lines: 75
Approved: news@daver.bungi.com

> Are you using the Minix uucp posted to comp.os.minix a while back ?

I don't know.  At the time I wasn't following comp.os.minix and didn't
hear about the uucp until the article(s) had expired.  I grabbed it via
ftp from nodak.edu; the comments in uucico.c say it was ported to Minix
by Peter S. Housel and later mods were done by F. van Kempen.  Does it
sound like the same beast?

> One thought, you might want to wait for version 1.5.10 before you
> do any serious code porting.  Makeing adjustments for version 1.3 
> /usr/include is can be a real pain, and will have to be undone
> for version 1.5.10.  Version 1.5.10  ANSI / POSIX compatibility 
> makes porting relatively easy.

So far I haven't run into any serious difficulties in this regard.  I
spent a lot of time figuring out that the kernel was silently tossing
my NULs into the abyss, then spent a lot more time trying to figure
out why transfers to Minix would die after about 12KB.  Still haven't
nailed that one but telling uucico to request a window size of 1 rather
than 3 provided a workaround, albeit with a massive performance penalty.
The behavior suggests poor error recovery by the uucico implementation,
perhaps compounded by problems with tty system.  (Several other issues
also smell like tty problems; this may be a weak spot at the moment.)

I also had a problem with the following:

    static char *tempname = "/tmp/fooXXXXXX";
        ...
    mktemp(tempname);

GCC makes the string in read-only, which led to a "bus error".  When I
changed the declaration to

    static char tempname[] = "/tmp/fooXXXXXX";

everybody was happy.  Other than this, having GCC and its libraries is
perhaps a big step towards making 1.3 easier to deal with.

Jove has been balky, but I expected this based on experience with other
systems.  The big chore here is figuring out what config options you
should be using -- and finding all the files that define them.  There
are at least 4 files that have options, and the documentation doesn't
point you to all of them.  There are also a fair number of outright
bugs which GCC justifiably gets indignant about.

One big hole is tty ioctls -- few are present on this version of Minix,
which has required a bit of work.

> I think you will be delighted with version 1.5.10.  In all respects it
> is a substantial improvement over 1.3:

Keep in mind that much of Bruce's code is 1.5 already.  As I recall
only one major piece is still at 1.3.

>         o  Much faster, block io uses an io_vector,

Interrupts?  It sure needs to use 'em.

>         o  tty driver is faster and handles binary data properly
>            (eg. zmodem works).

Performance seems reasonable (of course throwing horsepower at the
problem can't hurt!) and binary data has been working fine since I
took care of the NUL disposer.

> Also, I highly recommend that you tryout 'shh', shell with history

Sounds nice, though right now a better editor is my first priority.
Not that MicroEmacs is bad; jove is just better.

--
Karl Swartz			 |UUCP	uunet!apple!zygot!ditka!kls
1-408/223-1308			 |INet	zygot!ditka!kls@apple.com
"I never let my schooling get in |BIX	kswartz
the way of my education."(Twain) |Snail	1738 Deer Creek Ct., San Jose CA 95148