NMI's and PB3.5 : redirection?

[Criss French]

I have a neat little ancient handheld DOS computer, a Husky FS/2.

It's an 8mhz NECv25+ (8088 enhanced) with 1mb NVRAM. Cool machine for
survey data collection. Waterproof, bulletproof, mono-LCD is good
from full sunlight to dark (backlit), 2 COM ports, goes approx. 30hrs
on 3AA's. You might be suprized at how hard it is to find something
with these specs in the modern world.

In the 'Systems Developer's Guide' there is a section about PB
compatability. It says :

"NMI's: Spectra Publishing's POWER BASIC, a derivative of Borland's
TURBO BASIC, redirects the NMI handler, causing FS/2 to hang or
crash as described above. To run POWER BASIC on the FS/2, you must
patch the compiler to disable the NMI redirection as follows:
-Obtain a copy of the PBPREP.EXE program ... and run it..."

So, obviously very out of date. Google can't even find PBPREP.EXE.

But the question remains (and these forums don't seem to contain
an answer): Does PB3.5 redirect the NMI handler?

If not, I will re-wire myself a serial connector and start programming
for this nifty little beast to be a remote (via RF modems) data-entry
terminal.

If PB3.5 does 'redirect the NMI handler', then is there a way to
fix it? (I'm not expecting you good folks to rewrite the mythical
PBPREP.EXE for PB3.5)

I am just hoping that someone here can still remember this factum,
and put the issue to rest (on the don't bother side) if they are
reasonable sure that it won't work. Otherwise, I will spend a day
playing, and let you all know if it worked.

Cheers, Criss.



------------------
What can go wrong will go wrong.
Anything can go wrong.
What hasn't?!?!

[Tom Hanlin]

NMI, as in Non-Maskable Interrupt? I'd forgotten the PC ever had one
of those. Hmmm, well... I'll see what I can find out. It may take a
few days.

------------------
Tom Hanlin
PowerBASIC Staff

[Criss French]

Thanks for checking into it Tom. Here is some more info from the
manual, to help put the issue in context. Oh yeah, NMI = non maskable interrupt.
Please don't mind any typo's.

17.1.1 Problems with interrupts

NMI's: on IBM PC's, NMI's are used for RAM parity errors or not at all,
but the FS/2 uses NMI's for several other purposes. While this is
generally not a problem, it can affect the use of certain packages,
described here.

If an application redirects the NMI vector, it will crash or hang
the FS/2. Programs that do this include Codeview and TurboBasic (PowerBAsic).
To use such a program, you must patch the code to either prevent
redirection or cascade the application's NMI handler from the FS/2 one.
An application's NMI handler is normally used as "catch all"
error handlers, eg to catch numeric co-processor errors, on a
PC-AT system.

While co-operative task schedulers pose no special problems, pre-
emptive task schedulers will also crash or hang the FS/2. They
switch tasks by working off a timer tick, passing a different
return cs c and flags to the processor. The FS/2 timer tick is
actually generated within the NMI handler, so the return address
on the stack is not to the application's code segment, but to the
BIOS segment.

Maskable iterrupts: Finally, applications which rely on the use of
the CLI instructions to mask (disable) all interrupts other than
NMI may also cause problems, since NMI's are issued regularly on
the FS/2, unlike the PC (where it indicates a memory error and
halts the system).

[Tom Hanlin]

The news is, PB/DOS already cascades the NMI... so, if that's the only
qualification, it should simply work as-is. Try it.

If it doesn't work, I'm afraid there's no simple solution. The NMI
handling is tied to floating point support. It's an intrinsic part
of the PB/DOS runtimes and can't readily be modified or removed.
You could, possibly, contract for a custom version of PB/DOS but,
it would not be an inexpensive solution.


------------------
Tom Hanlin
PowerBASIC Staff

[Criss French]

Thanks Tom. From what it says in the manual, cascading
the NMI's should work just fine. Thank's for checking it out.

I now deem it very worthwhile to build a serial cable, and
proceed with trying out programming for this cool little
dinasour.

Now I'll see if I can figure out how to supply correct voltage
for the recharging circuit, just got the NiCad A cells.

I'll share my success with you all, or beg for more help
if things seem to be within reach but out of my grasp.

Thanks, Criss