Strange err

[Pawel Kusmierek]

select case atcfg.blasteroverride <error here
case 0
?"No"
?:?:?:?
case 1
?"Yes"
?hex$(atcfg.baseaddr)
?atcfg.irq
?atcfg.lodma
?atcfg.hidma
end select

error 11: Division by zero

Any hints?

------------------


[This message has been edited by Pawel Kusmierek (edited August 06, 2002).]

[Mel Bishop]

I have been told that using periods in a label is not a good
idea. With PB3.5, I have bumped against this a couple of times,
but byin-large, it works okay. Try deleting the period in the
label(s) and see what happens.


------------------


[This message has been edited by Mel Bishop (edited August 06, 2002).]

[Pawel Kusmierek]

Actually, it is record.field, not a label with period.
In the line just above there was [something]/1e3
I changed it to /1000 and the program was OK.

Previously, /1e3 was fine, the error showed up after changes
somewhere far away in the code. Now, with /1000, errors
show in other strange places. It looks like PB/DOS fails
with large code (((.



------------------

[Tom Hanlin]

If you're not using $ERROR ALL ON ... do. It's exactly designed to help
you find program errors.

It looks like PB/DOS fails with large code

If I had a nickel for every time someone blamed the compiler for their
own mistakes...! C'mon, Pawel, let's not jump to conclusions. PB/DOS
has been through many, many years of testing by many, many thousands
of customers. It surely still has bugs but, it's a quite mature product,
and it's almost certain that the problem is in your code. If you can
show us otherwise, of course, please notify mailto:[email protected][email protected]</A>
so we can fix it. In the meantime, vague allegations of doom are not
greatly appreciated.

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

[Pawel Kusmierek]

Thanks Tom, but I have:
$error bounds on, numeric on, overflow on, stack on

Maybe my conclusions went to far, however I have been receiving
obviously strange error messages since my code exceeded 200K
or something. These errors can be overcome by 'magical' methods,
such as changing order of procedures, or, as above, replacing
something by its equivalent.

I am absolutely aware that my program is rather poorly written,
overcomplicated and so on. But I do not like to receive
mysterious error messages at a place in the code when I make
minor changes in a far, unrelated (or at least not directly
related) part of the code.

I apologize if my words are too bitter, I am just slightly
desperate. Since the code size exceed about 200K, I have had
to fight with magical methods for almost every extension
of the program.

Sorry for bad English and for my bad mood.

PS. Tom, you said 'PB/DOS has been through many, many years
of testing by many, many thousands of customers.'

Has this testing produced any patches to PB 3.5. Maybe I
missed something?

------------------




[This message has been edited by Pawel Kusmierek (edited August 06, 2002).]

[Tom Hanlin]

Use Compile / Get Info to check on your segment sizes. If any of them
are too close to maximum, try adjusting your $SEGMENT statements.

There have been a number of minor updates to PB/DOS 3.5. See the Support
page for the latest version info:

http://www.powerbasic.com/support/

If you do not have the latest, ask mailto:[email protected][email protected]</A> for an update.
Don't forget to let them know your serial number.

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

[JOSE ROCA]

Tom,

I have asked [email protected] for an update twice, but they
don't answer.

Here is a copy of my last @mail, dated 7-7-2002:

Dear Sirs,

In January 13th, 2002 I send to you the following @email:

-----------------------------------------------------------------------------------

From a reply posted by Lance Edmonds in the PB forums:

"Finally, if your copy of PB.EXE is dated earlier than 19-12-97 (dd/mm/yy) then you should send an email to [email protected] (with your serial number!) and request an update to that edition."

Mine is dated 17-12-97. Would you please send me the updated version?
My serial number is xxxxxxxxx.

Regards.

José Roca
Daman Software, S.L.L.
----------------------------------------------------------------------------------

I'm still awaiting for a response.
Regards


------------------

[Neil Bertz]

Originally posted by JOSE ROCA:
I have asked [email protected] for an update twice, but they
don't answer.

Actually... I have record of your email on 1/13/02
and my subsequent reply on the 14th which I will not post here.

I will send my inquiry to you again via email and attach all
previous correspondence for clarification.

For future reference - Although we are awfully busy here,
it will never take us a month to respond to an email of that
nature.



------------------
Neil Bertz
[email protected]
PowerBASIC Staff

[Pawel Kusmierek]

Thank you, Tom.

Use Compile / Get Info to check on your segment sizes. If any of them
are too close to maximum, try adjusting your $SEGMENT statements.

How much is too close? 60K? 63K?

If you do not have the latest, ask mailto:[email protected][email protected]</A> for an update.
Don't forget to let them know your serial number.

I will as them as soon as I find my serial number .

Pawel

------------------

[Michael Mattias]

Use Compile / Get Info to check on your segment sizes. If any of them
are too close to maximum, try adjusting your $SEGMENT statements.

Agree; concur; assent; aye; yes.

MCM

[Tom Hanlin]

64 Kb would be 65,535 bytes. The maximum you ever want to use is 65,520.
I'd suggest a more comfortable margin-- say, 65,000 bytes maximum.
More, if you expect to add more code to that section in the future.

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

[Michael Mattias]

say, 65,000 bytes maximum.
More, if you expect to add more code to that section in the future

I'd say less, maybe in the 50 Kb range.

I've had problems over the years with all kinds of strange things happening when segments start getting even that close to the 64 Kb magic number.

Besides, it costs nothing to add a $SEGMENT, as $SEGMENTs may not occur within a procedure, and all calls to procedures are FAR (full segemnt ffset) calls anyway (AFAIK).

(Unless you are doing something really dopey, like getting a CODESEG value,making it a SHARED variable and assuming it never changes).

MCM

[Tom Hanlin]

Again, please avoid vague allegations about "I thought I sometime saw
something." If you can demonstrate a problem, tell us. If you can't,
don't assume it has anything to do with PowerBASIC.

There is no great harm in limiting segments to 50K or less, as long
as you don't run out of segments.

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

[Davide Vecchi]

I wouldn' t consider the 64 KB a magic number, it' s right the size limit for code segments.
I must say that it happened to me sometimes to get rid of unexplicable runtime errors by just moving or adding $SEGMENT or $CODE SEG metastatements, but unfortunately i didn' t keep the code to demonstrate it (so i can' t even tell that the cause wasn' t in my code ).
On the other hand, my current PB/DOS project is made of several segments one of which is 62K, and this never never gave absolutely any trouble.
So my conclusion was that simply when there is some "problem" in my code which corrupts memory somewhere, or when i' ve met one of those darned memory manager problems, then if the segment size is very big this makes the problem more likely to show up. I don' t know whether this is true anyway, it' s just my impression.

------------------
Davide Vecchi
[email protected]

[Lance Edmonds]

Over the past 12-odd years I've written a fairly large number of commercial application suites, utilizing every version of PowerBASIC/DOS that has ever been released, from v2 through v3.5.

Each of these suites contains large numbers of program and chain modules, containing anything from 1 to 16 segments each, and each segment pushed very close to the edge of the 64K segment limit.

In all this time, I've never experienced the problems alluded to here. I've seen similar effects with faulty memory managers though, so such symptoms can indeed be misleading and difficult to pin down.

Hence, to investigate this problem, we need to enlist <U>your</U> help folks, and you can do this by submitting a detailed report and source code/data directly to Tech Support. If/when we receive the code necessary to duplicate the problem, believe me, we will take the report very seriously and investigate.

Therefore, if you do happen to encounter such a problem, please be sure to immediately save a copy of the code (and any support/data files necessary to run the code), and then continue on with the original code to determine if moving the $SEGMENT metastatements sorts out the problem. If it does, you can then easily email the saved copy of the code/data to Tech Support for investigation.

Naturally, we'll also need to see details of your OS version, memory manager version(s), memory settings (conventional/EMS/XMS), and also the time/date of your PB.EXE file.

Thanks!


------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>