I noticed two overflow cases where my fpk (on PC-Linux), performed less satisfactory than…
Veselic wrote 59 lines
Dear fpk friends,
I noticed two overflow cases where my fpk (on PC-Linux),
performed less satisfactory than expected and in particular
less satisfactory than GNU pascal in the same environment:
Case 1. The program
program test2(input,output);
var
b,c:single;
begin {main}
c := 1e+33;
c := sqr(c);
b := c - c;
writeln(b,' ',c);
end.
was normally compiled and upon execution produced
Run time error 200 at 0x0804F02F
0xBFFFF894
0x080480A4
0x00000000
On the other hand, after compiling the same with GNU pascal,
the execution yielded
NaN Inf
as expected.
Case 2. The program
program test1(input,output);
var
c:single;
begin {main}
c := sqr(1e+33);
writeln(c);
end.
the compiler produced the message
Target OS: Linux for i386
Compiling test1.p
Assembling test1
Linking test1
FPE error inside compiler
and the execution yielded a curious output
4.357678781E-34
Again, GNU pascal gave
Inf
Question: Can one teach freepascal to make equally friendly
outputs in such cases?
I will add it next week.
Michael.