387 required? *and* Memory errors? Mothra sucks?

387 required? *and* Memory errors? Mothra sucks?

[Daniel]

In article <95Aug5.024030edt.46619@colossus.cse.psu.edu>,
 <9fans@cse.psu.edu> wrote:

>The "no physical memory" messages can be removed by turning on swapping.
>Choose a swap partition and point swap at it; see swap(8). It is not enabled 
>by default because the system cannot safely identify a partition to use; you 
>must set one up.  The announcement about how much swap space reflects the size
>of an internal data structure.

Neato!  I will do this, thanks (though I will have to wait for the CD).  
People who have done this, what happens when some process starts chewing up 
VM?  Can you set process limits in Plan 9?

>A `urlfs' would be folly, because URLs are not a real name space and the 
>semantics are utterly unmanageable.  

What is meant by this, in particular?  Maybe I will re-read some of the papers
to try to understand more fully the concept of a Plan 9 "name space"...

>The argument is analogous to that which prevented us from putting network 
>names in the file name space.  As for mothra, the interface may be lame but 
>I wonder why you think that, given that you have not got the CD
>and therefore probably haven't seen it.  

A version of mothra comes with the four disk set.

>Its bugs, as said in the manual, are because it is a very new program.  

*nod* -- "early alpha".

Dan

387 required? *and* Memory errors? Mothra sucks?

[philw]

Look at the manual page for swap. You might try reading the
documentation before posting.

phil

387 required? *and* Memory errors? Mothra sucks?

[mkc]

Daniel Egnor writes about Plan 9 dealing poorly with memory errors and
gives this example of rc code:

>   fn ls { lc -F }

> (If you're clever you'll figure out why this loops -- it took me a while, since
> I wasn't used to the system) will very quickly cause an 8MB machine to scroll
> "No physical memory" messages all over the screen.  

I did that too, under AIX, the IBM (and Bull) Unix. My rc was over 15
MB before I killed the process. Oops. The memory in the machine is 288
MB, though, so I didn't have to go through what you describe. Anyway,
it's not just Plan 9 that deals poorly with out-of-memory
situations. Look at your vanilla Unix system and see what it does...

Mikael Cardell, Dharma Hacker <mkc@bull.se>
/* No, I don't speak for Bull */

387 required? *and* Memory errors? Mothra sucks?

[Steven]

> The system will run fine without a 387 until you need a floating point
> instruction to be executed.  The kernel and most utilities do not use
> floating point.

Good, then maybe I still have a shot at getting this working.  I have the=

4-floppy distribution installed (after swapping my floppy controller), an=
d
configured for 640x480x1.

The kernel loads okay, but the system hangs at the "init: starting /bin/r=
c"
message.  After it prints that message, it waits a few seconds, then some=

multicolored horizontal bars flash once on the screen (still in text mode=
),
then the display stays at the "starting /bin/rc" screen until I whap the =
reset
button.

I assumed the screen flash was the attempt to change to VGA graphics mode=
, so
I swapped in an old Oak VGA card.  No flash, but it hung in the same plac=
e
anyway.

I've fiddled with the BIOS settings, turned shadowing and caching on and =
off,
etc, etc.  I even swapped mice.  Always get the above result.

Here's my setup:

no-name ISA motherboard with AMD 386DX/40, UMC 491F chipset, and AMI BIOS=

32MB RAM (tried with only 16MB too)
Adaptec 1542CF (IRQ 11, DMA 6, mem dc000, port 330)
Quantum ProDrive 105S (SCSI 1, only device on chain)
STB PowerGraph X-24 (I've tried all the different jumper settings)
Logitech C-9 serial mouse
generic multifunction I/O card (2S,1P,1G) (mouse is the only thing connec=
ted)

Now I'm left to suspect the 1542 or my motherboard.  Anybody have any ide=
as?
_________________________________________________________________________=
______
Steven Plite  <splite@wdni.com>       Open Systems Eng. & Support, Weyerh=
aeuser
    "This is the roller coaster of endless and violent vomit."	-- Jason F=
ox

387 required? *and* Memory errors? Mothra sucks?

[rob]

The system will run fine without a 387 until you need a floating point instruction
to be executed.  The kernel and most utilities do not use floating point.

The "no physical memory" messages can be removed by turning on swapping.
Choose a swap partition and point swap at it; see swap(8). It is not enabled by
default because the system cannot safely identify a partition to use; you must
set one up.  The announcement about how much swap space reflects the size
of an internal data structure.

A `urlfs' would be folly, because URLs are not a real name space and the semantics
are utterly unmanageable.  The argument is analogous to that which prevented
us from putting network names in the file name space.  As for mothra, the interface
may be lame but I wonder why you think that, given that you have not got the CD
and therefore probably haven't seen it.  Response from outsiders who have seen
it is that its interface is an improvement.  Its bugs, as said in the manual, are
because it is a very new program.  New releases will likely appear on the net.
I disagree with your assessment of the program - it's a perfectly fine interface
to a perfectly awful network.  It's not a marvel, but it does its job.

-rob

387 required? *and* Memory errors? Mothra sucks?

[Daniel]

In article <199508050446.AA26993@interlock.wdni.com>,
Steven Plite  <9fans@cse.psu.edu> wrote:
>Is a 387 required to run the 4-floppy dist on a 386?  The hardware 
>requirements in the install docs don't say so, but the section on the 386 
>compiler in "The Various Ports" says "...the compiler assumes i387-compatible 
>hardware...".

I have successfully installed and booted Plan 9 on a 386 sans 387.  It seems
to work, except that floating-point operations fail with bogus results -- the
only immediate result of this is that the 8 1/2 clock gizmo does not draw the
hands correctly (or at all), since doing so requires trig.  :)

I assume other things break, too.

This same installation crashes after being alive for more than a minute or
two.  I believe this is due to a faulty memory configuration (gah) but have
not verified -- it *could* be due to lack of a coprocessor, but I rather 
doubt it.

Additional topics:

It appears (on a more stable system!) that Plan 9 deals very poorly with memory
errors.  For example, a simple recursive rc function, which a beginner can
easily make accidentally, for example:

  fn ls { lc -F }

(If you're clever you'll figure out why this loops -- it took me a while, since
I wasn't used to the system) will very quickly cause an 8MB machine to scroll
"No physical memory" messages all over the screen.  Unless one is very quick
to kill the offending shell (which is difficult with messages coursing all
over one's screen!) the system crashes.  This is poor.  Have I misconfigured
somehow?  Perhaps I should add some swap space or something?

Along similar lines, when Plan 9 boots, it prints a message about the amount
of memory available, and invariably (even when booting from floppy!) prints
some _large_ amount of swap available (like 50MB).  Why does it do this?

OK, an unrelated gripe: Why is Mothra just an ordinary old Web browser,
except with a lame interface (and a lot of bugs)?  Given Plan 9, I'd think
you'd have something like a urlfs, which would be terribly handy, and an HTML
browser on top of that... am I missing something?

(Don't get me wrong, I love a lot of the concepts in Plan 9, I'm just wondering
about some of the smaller issues.  I do not yet have the CD, BTW.)

Dan