#More on (long)reals
6 messages in this thread
More on reals, longreals, and conversions to and from strings: M2S has a
real problem in the longreal and real conversion utilities. I haven't
looked lately (ie, in a couple of months), but as I recall, the conversion
from string to (long)reals is faulty, but the conversion back is not.
Anyway, the problems shows up on numbers where the exponent is less than
zero. As John Steiner reported (based on my message to him), the code
fragment
x := 0.5;
WriteReal(x,15); produces a value of 0.05, not 0.5.
Fixing the library modules will not solve the problem (although I will try
the fix noted in one of the other messages for longreals. There is a
similar fix for reals that I will try to upload later.) The compiler
itself used those modules. So, even if you use corrected modules, so your
code can read I/O correctly, the compiler itself cannot read its source
code correctly. If one had access to M2C.lnk, one could relink the compiler
with the correct libraries. Without that, we're stuck.
By the way; I notified PTI of this problem in January, and got repeated
promises of calls back, with no results. Finally, Dan Brookshire called
me, said they were working on the problem, and said he would get me a copy
of the fix. PTI still didn't answer, so I said answer or I'll contact the
magazines, so other people wouldn't get burned (it was not intended as a
threat; merely pointing out that I could not, in good conscience, let
others spend money on an unsupported product.) Leon Frenkel responded by
firing Dan Brookshire (for which I am very sorry), sending me 1.12 (without
documentation), even though PTI knew it did not solve the problem, and
telling me he didn't even owe me that much. His comments amounted to
1. You bought it from M2S, not PTI.
2. It was sold as is, just like most other software. I pointed out
that other software was not sold as is (in particular, TDI MII was not),
and sent letters to AC and AmigaWorld. AC's response is in the most
current Bug Bytes column. Leon Frenkel still has not replied, although he
should have the entire story.
Comment on GEnie is that he boA(Xt
Tom:
I appologize for any inconvenience you may have experienced. However,
contrarry to the contents of your message the problem is not exactly as you
describe it.
Dan Brookshire was never fired by me, because he never worked for me. Dan
worked for M2S and for a short time after M2Sprint was sold Dan continued
to provide some "support" for the product. In fact as was discovered later
he was doing more damage then any support. You mentioned talking to Dan
about fixing a bug, well that information was never passed on to PTI.
The first time I heard your name, was when your letter arrived here. Since,
you did not state the specific nature of your problem in the letter. I
decided the quickest way to resolve your problem would be to ship you an
upgrade to version 1.12 which would probably correct the prroblem, since
the original version was full of bugs. I am sorry that this did not solve
your problem, but I was trying to do you a favor and instead you make it
sound like I did something bad.
Let me put something else in a more proper context, you purchased M2Sprint
from M2S. M2Sprint was written by M2S not by PTI, any bugs in M2Sprint were
created by the programmers who designed it. If someone is having a problem
I am quite happy to help, but the problem has to be clearly identified. The
easiest way to do it is a letter descibing the exact problem and a sample
of code which causes the problem. I can't gurantee I can solve the problem,
but I'll give it my best shot.
Fair Enough?
>> Leon Frenkel <<
Tom,
I don't know the problem you are having 'cause I didn't see the letter you
sent to M2S/PTI/Benchmark or whoever is "dodging blame" this week. However
I found a probalem with the LongRealConversion and RealConversion libs.
There is a line in the ConvLongRealToStr and ConvRealToStr procedures that
reads:
num:=num+PowerOfTen(-D)*0.5D; It should read:
num:=num+PowerOfTen(-D-exp)*0.5D; This is in the ConvLongRealToStr. The
same line (single precision, of course) needs to be changed in the
ConvRealToStr procedure. Then recompile and move the output files into the
appropriate M2 subdirectory.
As for support, you should be specific about youtr problem eventhough you
will get no true help from the new wardens -er- owners. Basically they want
you to get so disgusted that you do two things: 1) buy BenchMark 2) come
to mistrust those truned out by the ownership change so that when their new
product hits you will not buy it.
When I furst found the problem, I thought it was in the compiler. I asked
for support here by publishing fragments of the troublesome code and desc
ribed the output. I got no response. I went to AmigaTech and posted in
"other languages" about how I felt like a victim of a business coup by
Benchmark (after hearing their explanation of how they came by M2S). I got
told my statements were "Scandalous" but received no support. So if you put
the full explanation of the problem before them, you take that excuse away
form them.
Then they wine (most exxagertedly, I might add) about M2S being "full" of
bugs. It really wasn't. But even if it was, how come "you-know-who" is
still selling it but not supporting it?
There was this company called Gulfstream. They bought the rights to a
design for the Commander 112 and 114 from Rockwell. These planes were
plastered with a wealth of Airworthiness Directives. Gulfstream tried to
wine that it was all rockwells fault and that they could not support the
owners of the aircraft in reworks to bring the craft into compliance with
[ MORE ]
[ continuation ]
FAA regs. The situation ended up in the courts and the judge told
Gulfstream essentially,"you bought the rights, you own the design and are
responsible for it." The same applies to Benchmark/PTI/Whoever, in my
opinion.
The reality is that there is no real support from any corporation for the
product. (I really am not sure why they are allowed to keep this forum
section, in fact.) So you best bet is to query AmigaTech other languages
and hope one of the other M2S owners can help you out.
wmc
Wayne, Thanks for the information. Just to get the record straight, I DID
send PT a letter with a complete explanation of the problem. I don't know
why Mr Frenkel claims he never saw the explanation of the problem. I agree
that his whining about not having to support software he bought from
somewhere else is just that; whining. I see no hope for M2S under the
present management; if they are unwilling to support it, it will go no
further. By the way; the fix you gave me should help my programs read data
files but that does not solve the entire problem. The compiler itself does
not correctly read its own source code. To fix that, one would need to
relink the compiler with the corrected libraries. Unfortunately, I don't
know how to get a hold of M2C.lnk to do that. Thanks for the help. Maybe
somebody, somewhere, will produce an MII compiler that works. Of course,
by that time I will be broke from buying lousy ones. Tom Gist