Re: PC's Locking up when in BPCS -- BPCS-L

Hi Dean,

The main issue we have at SSA, is that some customers get the 'hang' and the
lockup confused when calling in to HelpLine and frankly, since the causes
(and resolutions therefore) are different, we consider this to be fairly
important and not 'splitting hairs' !

Whether you have to get rid of duplicate records in a server file or apply a
server BMR to prevent a program from looping (causing a BPCS PC application
hang waiting for the server program) or apply the latest OPS Runtime Service
Pack because you are getting true lockups is a real issue for the end-user
and HelpLine. The OPS Runtime Service Pack is not going to resolve your
application looping issue, so that is why it is important to us that we know
the difference up front when analyzing the problem - we want to send you the
right fix! This is why we ask everyone to ask your end-users who are
complaining about a given PC 'hang' or 'freeze' problem to be as specific
and detailed as possible when calling the problem in to us.

Your other notes about the superiority of NT's operating system are very
true, and it is true that our product is not the only one suffering this
fate (I personally switched to using an NT workstation at work, and I do not
use BPCS except to do testing here and there on fixes I write - I had my old
Windoze 95 box lock-up when running simple 'Microserf' office applications
and other non-BPCS applications, which forced me to re-boot sometimes
several times in one day - I am much happier on NT because I have seen the
'blue screen of death' about 1 time in the year I have had the box).

All the same, we are committed to resolving this problem for the Win95
platform - it just is taking a long time to debug and fix due to the nature
of the problem and the lack of trails it leaves when it occurs.

Here is a snippet of information from one of our runtime developers that
goes into the technical details of why the OS on NT is a better design than
WIN95 in general at handling programming errors, and what is happening
'under the covers' on NT when the true 'lockups' occur:


"For the edification of  the thread members, I thought  I would add a short
explaination of why it's much more difficult to lock up the NT OS than the
Win 98/95.    Windows/NT is a true protected mode, entirely 32-bit
preemptive multitasking operating system. (ala AS/400 or UNIX).  Win 98/95
is merely an operating environment with a display subsystem based on Win 3.1
so it can still run Win 3.1 programs.

I could go on forever about the differences but as an example, what's one of
the important factors that makes a difference in the BPCS case ?

In NT, you get your own entire process space, including it's own hooks into
the Windows display subsystem as well as other  operating system services.
So if a process hangs or otherwise looses its integrity,  the operating
system acts essentially the same as an AS/400, protecting the rest of the
system from the unique process and in deleting the process as well as any
resources attached.

Win 95/98 has a 16-bit nonreentrant display subsystem.  In fact, when any
process is using the display system, a semaphore is set which literally
blocks every other task currently running which also utilizes the display
subsystem.  (As a bit of trivia, remember the original Intel Pro II's that
were optimized for 32-bit performance ...... They didn't perform on Win/95
machines because of the constant thunking to 16-bit mode).  Some might think
that this might be extremely inefficient, in fact,  crippling.  Imagine
having this screaming race car of a 450 MHz PC that is constantly slamming
on the brakes.  This could be why Linux screams on a box compared to Win
95/98.  I have also heard of people in the financial industry who say that
old war horse OS/2 screams compared to Win 95/98 but you can be the judge.

So let's say your program gets into that display code, sets the block every
task semaphore and than does something illegal as far as the OS is concerned
(unbeknownst to the programmer who programmed it, obviously).  Voila ....
what do you have is ..... PC Lockup.

I have also had other PC programmers show me just some normal programming
bugs such as illegal memory writes or stack overflows that will create PC
Lockup in Win 95/98 because of shared operating environment problems. These
are just a few examples. The runtime code on the BPCS client is very complex
and has numerous places where we have already found and fixed these sorts of
issues, obviously some still remain to be discovered.

If I was in a work environment without need for 16-bit emulation or games, I
would seriously consider Win/NT. "

Thanks

Genyphyr Novak
SSA


-----Original Message-----
From: DAsmussen@aol.com <DAsmussen@aol.com>
To: BPCS-L@midrange.com <BPCS-L@midrange.com>
Date: Tuesday, February 22, 2000 11:54 PM
Subject: Re: PC's Locking up when in BPCS


>Genyphyr,
>
>In a message dated 2/22/00 11:04:16 AM Eastern Standard Time,
novakg@ssax.com
>writes:
>
>> The reason NT is better has nothing to do with the Task Manager allowing
you
>>  to kill the BPCS job. NT simply does not lock up in BPCS. Please read
prior
>>  threads defining what a 'lock up' is versus what is a simple application
>>  'hang' on the PC. I still have found no reports of a real 'lock up' on
the
>>  NT platform. If you had a true lock up, the task manager would not even
be
>>  available because the keyboard does not respond. The PC lock up as
>described
>>  on BMR43360 is non-application related, whereas a hang can always be
traced
>>  to some sort of application problem, usually on the server job.
><<snip>>
>
>Well, let's not split hairs over what a lockup is.  I have the whole
history
>of Micro$oft operating systems on various machines at the office (two of
>which are so old I can't even _GIVE_ them away!) and, after DOS, NT is the
>most stable because it allows you an out instead of giving a hard freeze
>where even the keyboard won't work -- at least most of the time.  I did not
>intend to imply that the task manager was the reason that NT worked better,
>merely that it was available.
>
>BPCS is _FAR_ from the only software that runs better on NT than it does on
>the 95/98 suite.  NT (like OS/2 before it) seems to do a better job of
>isolating tasks from each other, so that memory "leaks" are less
>catastrophic.  Lotus SmartSuite, Visio, and ABC Flowcharter also run better
>under NT that they did under Windows 3.11 or 95/98.  Pretty good company to
>be sharing a problem with.  NT also seems to handle devices and their
drivers
>better than the lesser forms of Windows, which is odd given that NT drivers
>are harder to come by and 95/98 are often called upon to support a more
>diverse hardware configuration sampling in the home arena than does NT,
which
>is used primarily for businesses with more consistent hardware
configurations.
>
>While not technically a lockup, a non-respondent application is still a
>failed application, even if you can recover from it.  Contrary to your
>statement that most lockups are caused by server errors, my current client
>(full C/S, 6.1.00, latest patches) hangs several times daily and, with the
>exception of some poor code in ORD701B that's not checking SQLSTS and
causing
>a couple of the lockups with a decimal data error, the server is not at
fault
>over 90% of the time.
>
>My client had a long conversation with Helpline over this today, and you
guys
>are doing a fantastic (albeit late) job of trying to resolve this issue
>through the "OPS Runtime Service Pack" initiative.  Perhaps my original
>explanation of why NT works better was technically inaccurate.  But just
>because the system doesn't lock doesn't mean it's not the _SAME_ problem.
NT
>just isn't so picky...
>
>Regards!
>
>Dean Asmussen
>Enterprise Systems Consulting, Inc.
>Fuquay-Varina, NC  USA
>E-mail:  DAsmussen@aol.com
>
>"Jimi Hendrix was the first rocker on the Internet.  His modem was a purple
>Hayes." -- Anonymous
>+---
>| This is the BPCS Users Mailing List!
>| To submit a new message, send your mail to BPCS-L@midrange.com.
>| To subscribe to this list send email to BPCS-L-SUB@midrange.com.
>| To unsubscribe from this list send email to BPCS-L-UNSUB@midrange.com.
>| Questions should be directed to the list owner: dasmussen@aol.com
>+---
>

+---
| This is the BPCS Users Mailing List!
| To submit a new message, send your mail to BPCS-L@midrange.com.
| To subscribe to this list send email to BPCS-L-SUB@midrange.com.
| To unsubscribe from this list send email to BPCS-L-UNSUB@midrange.com.
| Questions should be directed to the list owner: dasmussen@aol.com
+---