[Coco] A bug in Basic09/RunB

[Wayne Campbell]

Gene,

As I said before, I can't be certain this problem didn't exist before the modifications. For all the work I did developing DCom, I don't recall looking that deeply into what was being passed. I assumed it would be correct.

Therefore, the bug may go all the way back to Basic09 1.00.00 and was never caught. It is possible.

Wayne




________________________________
From: Gene Heskett <gene.heskett at verizon.net>
To: CoCoList for Color Computer Enthusiasts <coco at maltedmedia.com>
Sent: Sun, December 6, 2009 7:28:51 PM
Subject: Re: [Coco] A bug in Basic09/RunB

On Sunday 06 December 2009, Wayne Campbell wrote:
>James,
>
>I would love to do that, but I have had a problem with booting the original
> OS-9 Level 2. I get a black screen with a magenta cursor. I can see that
> the system is booted and the startup file is being displayed by the
> movements of the cursor. I can set date and time, though I can't see what
> I'm typing. Using display commands from the command line yields no change.
> Until I can fix that, I can't test anything.
>
>In addition to the previous post, there is a third bug. I had almost
> entirely forgotten it. It also involves parameter passing, but deals with
> passing fields of records. You can pass one level deep:
>
>RUN GFX2("FCOL",col.bl)
>
>but you cannot pass:
>
>RUN GFX2("FCOL",col.fg.bl)
>
>Basic09 does not return a compiler error, but the value passed during
> execution is incorrect.
>
>Since all 3 of these bugs are related to parameter passing, I think the
> routines dealing with establishment of the pointers in the list should be
> looked at first, and then the parameter counter for receiving parameters
> into called procedures to see why it's miscounting them.
>
>Wayne
>
>
>
>________________________________
>From: "jdaggett at gate.net" <jdaggett at gate.net>
>To: CoCoList for Color Computer Enthusiasts <coco at maltedmedia.com>
>Sent: Sun, December 6, 2009 5:56:32 PM
>Subject: Re: [Coco] A bug in Basic09/RunB
>
>Wayne
>
>One way of determing whether the original Basic09 has this bug or was
> introduced in later modifications is to run you tests on the stock version
> Basic09. Instead of booting with the newer NitrOS9, just boot up your
> system with the orignal OS9 Level II disks and with Basic09
>
>just my thoughts
>james
>
>On 6 Dec 2009 at 17:49, Wayne Campbell wrote:
>> When I first learned Basic09, I did not possess the knoweledge and
>> experience that I now have. There were many things that were overlooked
>> in DCom, simply because I did not understand the function. Now, I know
>> more about the function.
>>
>> I believe I have found a bug in Basic09. With all of the modifications
>> that were made to Basic09 between 2002 and 2006, I cannot be certain that
>> this bug did not exist prior to the modifications. However, if experience
>> is any guide, it may well have gotten broken during the modification
>> process and was never discovered.
>>
>> The bug is in the way parameters are passed to called procedures. First,
>> any simple variable type (other than boolean) can be passed by value
>> using +0 on the byte, int and real variables, and +"" on strings. To pass
>> a boolean value, use TRUE or FALSE.
>>
>> Second, as long as the receiving variable is of the correct size, it
>> doesn't matter what type or order you specify the parameter list in. It
>> just has to match the size.
>>
>> The first bug occurs when you pass a byte by value:
>>
>> a:BYTE
>> a:=1
>> RUN myProc(a+0)
>> END
>>
>> PROCEDURE myProc
>> PARAM a:BYTE
>>
>> PRINT a
>>
>> END
>>
>> The result should be
>>
>> 1
>>
>> Instead,, it is
>>
>> 0
>>
>> The wrong byte is being retrieved. All BYTE variables are expanded into
>> INTEGERs before being processed. In this case, the first byte of the
>> integer is being read, not the second. I can never remember which order
>> they are in, so I'm not certain which one is LSB and which one is MSB. To
>> be clear:
>>
>>   +---+---+
>>
>>   | 1 | 2 |
>>
>>   +---+---+
>>
>> Byte 1 is being read, when byte 2 should be being read.
>>
>> The second bug occurs when you pass multiple variables into a single
>> variable. Based on the fact that it works correctly, under the following
>> conditions, on the first 2 elements of the receiving array, it should be
>> working correctly for all of the parameters needed to equal its size. The
>> test procedure passed 3 INTEGER variables into 1 3-element INTEGER array.
>>
>> PROCEDURE testParams
>> DIM a,b,c:INTEGER
>> a:=1 \b:=2 \c:=3
>> RUN recPars(a,b,c)
>> END
>>
>> PROCEDURE recPars
>> DIM cntr:INTEGER
>> PARAM d(3):INTEGER
>>
>> FOR cntr:=1 TO 3
>> PRINT d(cntr)
>> NEXT cntr
>> END
>>
>> Results:
>>
>> 1
>> 2
>> 56 (Parameter Error)
>>
>> In addition, I added a fourth parameter, a second reference to a, and a
>> 4th element to d. The error occurred in the same position in the array as
>> before. Therefore, the accumulator that is supposed to count the
>> parameter sizes is exiting before the final parameters are being
>> processed.
>>
>> I do not know enough about assembly language to track down these bugs.
>> Maybe someday, not not any time soon. If someone on this list wants to
>> see if they can figure it out, the source to the modified Basic09 and
>> RunB are on sourceforge.
>>
>> http://nitros9.cvs.sourceforge.net/viewvc/nitros9/nitros9/3rdparty/packag
>>es/basic09/?sortdir=down
>>
>> I'll report anything else I find out as I go.
>>
>> Wayne

Wayne;  AFAIK, only 1 byte has been changed is both basic09 and in runb since 
the original distribution by Tandy.  I don't have access to it at the moment 
but a cmp should spit out the diff.

If this is no longer true, then Robert or Boisy please pipe up and clarify.

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
The NRA is offering FREE Associate memberships to anyone who wants them.
<https://www.nrahq.org/nrabonus/accept-membership.asp>

"It is better to have tried and failed than to have failed to try, but
the result's the same."
- Mike Dennison

--
Coco mailing list
Coco at maltedmedia.com
http://five.pairlist.net/mailman/listinfo/coco