[lazarus] Compiler issues.

[Liebrecht Venter]

Michael.VanCanneyt at Wisa.be wrote:

> Hello,
>
> People that have installed FPC 1.0.6 using an RPM from our download site
> or one of the mirrors, should check their installation. There has been an
> error when generating the RPMs due to some misconfiguration of the RPM
> build system.

Hi Michael.
Apologies for not have prepared the bug report...as promissed at AEI.....unf

ortunately the workaround gives the correct result and I am pressed for time
to continue. I will get back to *that* specific  bug...later..


However,
I had a bit of a ding-dong with the compilers.
Verion 1.0.4 had several small bugs which partially was resolved with the
compiler I downloaded now (1.0.6).    "Beta" also had the bug.

I picked up some extras with 1.0.6.
Whether I can call (1) a  bug I dont know, but it is probably not correct.
(2) may be of serious consideration although I really do not want to point
fingers.

=================================================================
1) If a writeln contains functions which generates a lot of writeln output,
then the result of the writeln is lost.
e.g
 writeln('Integral=',Generalised_EllipticE(y,k)[1],'
',Generalised_EllipticE(y,k)[2],' i');
output is NOT the last writeln  on the terminal if the functions themselves
contain writelns.
Somehow this does not seem as intended (or correct to me)....Did you intend
it to do that?
If this is not exactly clear I will send an example although this does not
bother me at the moment.

Then for a curious one
=================================================================
2)  The question is,
Was it your intention that operators such as   "=" be independent of the
datastructure?
e.g.   in double precission vs extended  ?

An example:
==========
In attached  "bug1.tar.gz"  you will find two directories.
case_double
and
case_extended

In both directories there is a file named.
test.bug.pas
Run each of them in turn by CD'ing into the relevant directory...
Please compile them in their respective directory if you do not want to use
my binaries.

Description of problem:
The example programs is simple and self explanatory.
All it does is to evaluate y=1/k  with double and extended precission.

Cd to case_double ansd execute  test.bug
[Liebrecht at linfloat case_double]$ ./test.bug
k>1
y>1/k
y<=1
k>1
y>1/k
y<=1
y-1/k= 7.4918370118748356E-0017
re= 0.000000000000000E+000  im= 0.000000000000000E+000
[Liebrecht at linfloat case_double]$

Then,  Cd to  case_extended and execute  test.bug
[Liebrecht at linfloat case_extended]$ ./test.bug
k>1
y<= 1/k
k>1
y<= 1/k
y-1/k=-5.4210108624275222E-0020
re= 0.0000000000000000E+0000  im= 0.0000000000000000E+0000
[Liebrecht at linfloat case_extended]$

Note the programs in the two directories are IDENTICAL except for DOUBLE
replaced with EXTENDED.

>From the output above, you will see that the "=" operator behaves
differently under double and extended precission.
The extended precission example is correct as it yields  the following
switches k>1  y<= 1/k.

However,  the double precission example yields  the following switches  k>1
y>1/k  y<=1 which is not correct as operators such as "=" should be
independent of precission used.

=================================================================



-------------- next part --------------
A non-text attachment was scrubbed...
Name: gz00000.gz
Type: application/octet-stream
Size: 115683 bytes
Desc: "bug1.tar.gz"
Url : http://localhost/pipermail/lazarus/attachments/20020523/9f926bb9/gz00000.obj