[net.micro.amiga] 1) BASIC bug and Commodore / Response

bruceb@amiga.UUCP (Bruce Barrett) (01/20/86)

In article <1099@ecsvax.UUCP> urjlew@ecsvax.BITNET (Rostyk Lewyckyj) writes:
>Keywords:Microsoft BASIC,vendor attitudes,COMSPEC

Thank you Mr (or Ms?) Lewyckyj, for your bug report.

>I decided to test the accuracy of the arithmetic calculations... 

>    INPUT"Enter range";x,y 
>    INPUT"Enter steps";nx 
>    delta = 1.0/nx 
>    FOR i = x TO y STEP delta 
>      j = i*i : k=i^2 
>      IF j<>k THEN PRINT i,j,k,(j-k) 
>      NEXT i 

>Run ... from about 925 to 930 with nx=2,4,8,16 

For those of you who have not tried this:
  928*928 = 861184 correct
  928^2   = 861312 off by 128 (.01%)
Other calculations in this area seem to be off by .0625 (.000007%)

>On Thursday Jan. 2 I got through to Suzan Watlain at the 
>Commodore Product support center and attempted to report the 
>problem to her.  She kept asking me if I had declared the 
>variables and telling me that one should expect roundoff problems...

By the way... If you insert the following line at the start of the
program the problem "goes away":
	DEFDBL A-z

>... I asked her if the problem had 
>been reported to Pennsylvania and she told she would not 
>pass it on. 

	My opinion: This was a mistake.

>It appears to me that Microsofts BASIC was rushed out the door 
>to market before the ink was dry. Certainly without proper 
>testing. 

I am sorry but I must disagree with you on this point.
I feel that although bugs do still exist, that "Proper" testing was done.
I was involved with the testing of this product and have a good idea
A) What testing ws done, B) How long it took, and C) What bugs we
decided we could live with for the first release.  D) The quality
of the product before and after testing.

	a) All of the National Bureau of Standards (NBS) BASIC tests
		were run.  These include lots of arithmetic test
		although most of them are at the extreme ends of
		the numeric ranges.
	b) After several preliminary versions (< V1.0) Microsoft sent 
		us Version 1.0 "This is it good enough to ship" 
		about 10-Sept-85.  We had about 5 people working 
		from about the 10th of September to early November 
		(About 9 weeks or *10 man months*) testing this product.
	c) The bugs we were aware of at the time of the release were
		mostly cosmetic.  (We were not aware of this one.)
	d) The number and type of bugs found/reported/fixed during this
		time contributed significantly to the quality of the
		product.  What you have today is far better than what
		we received in September.

>			I think I understand the competitive 
>market pressures and teething troubles of a new machine. 
>What concerns me however is the attitudes of the vendors 
>support personnel. 

I agree that support personal should always report bugs they or their
customers find.

>		     This is exactly the time to be willing 
>to accept user inputs such as bug reports in order to shake 
>down the software as quickly as possible. 

Here! Here!  Send those bugs to me, or amiga!amiga.support.  Report 
Lattice-C bugs directly to Lattice.  My US Mail address is in the header.

>   Rostyk Lewyckyj 
>   Small Computer Systems Specialist    (5'3/4") 8-) 
>   1305 W. Main St. 
>   Carrboro NC 27510 
> 
>   urjlew@ecsvax.BITNET 

Bruce Barrett / Commodore-Amiga / Software Quality Assurance

acs@amdahl.UUCP (Tony Sumrall) (01/21/86)

In article <573@amiga.amiga.UUCP>, bruceb@amiga.UUCP (Bruce Barrett) writes:
> For those of you who have not tried this:
>   928*928 = 861184 correct
>   928^2   = 861312 off by 128 (.01%)
> Other calculations in this area seem to be off by .0625 (.000007%)
I THINK THIS IS A FIRST!--I can't remember the last time I heard a
"manufacturer" admitting to a problem like this.  It shows, to me at
least, that The folks at Amiga are open and willing to discuss their
"problems" in front of everyone!--Hooray!

> 
> Here! Here!  Send those bugs to me, or amiga!amiga.support.  Report 
> Lattice-C bugs directly to Lattice.  My US Mail address is in the header.
> Bruce Barrett / Commodore-Amiga / Software Quality Assurance
Okay, how about requests?  I would like to be able to access all of the
"devices" that are supported by the OS.  I'm able to open RAW: as random
but for the life of I can't figure out how to "poll" the window.  What I
wanted to do was implement an AmigaBASIC program to take advantage of
RAW:'s escape-seqence generation and remove AmigaBASIC's massaging of my
keystrokes (can't determine that the ALT key was pressed in conjuncion with
another character, etc.).  Maybe I shouldn't be using RAW: but I think I
should be able to; maybe RAW: buffers its input but I should be able to
unbuffer it if I want; maybe the standard AmigaBASIC keyboard handler
is nice because I have one keycode to correspond to one key but what if
I want to extend my programs to utilize *all* keycodes.

If the facility already exists, lemme know; if I'm wrong, lemme know;
if...well I think you get the idea.
-- 
Tony Sumrall                       ...!{ihnp4,hplabs,amd,sun}!amdahl!acs

[ This reflects no one's opinions. ]