When I compiled the following simply programs using Lazarus 1.8 fpc 3.0.4 Win 64, run it, it will throws RunError(215) Arithmetic overflow error.

However, if I compile and run the same in Lazarus 1.8 fpc 3.0.4 win 32, it won't raise the RunError (215).
Why???

program test64bit;
{$mode objfpc}{$H+}
{$OVERFLOWCHECKS ON}
var
   LW1, LW2 : array[false..true] of LongWord;
   c : currency;
   aBoolean : Boolean;
begin
   LW1[false] := 5;
   LW1[true] := 5;
   LW2[false] := 10;
   LW2[true] := 10;
   c := 10;
   aBoolean :=  (LW1[true] + LW1[false]) - (LW2[false]+LW2[true]) > c ;
   Writeln('if compiled in 64 bit, a RunError(215) will have occurred
and it will not reach this line');
   readln;
end.

In our previous episode, Dennis said:

When I compiled the following simply programs using Lazarus 1.8 fpc 3.0.4 Win 64, run it, it will throws RunError(215) Arithmetic overflow error.

Which seems correct. The expression to compare with is entirely longword and it gets negative (5+5) - (10+10).

Marco van de Voort wrote

Which seems correct. The expression to compare with is entirely longword and it gets negative (5+5) - (10+10).

But shouldn't it then also give the same error if compiled for 32 bit?

Bart

Agree.

The weird thing is, in 32 bit, the compiler gives the Hint: (4079) Converting the operands to "Int64" before doing the add could prevent overflow errors.
but does not raise the RunError(215).

The 64 bit compiler does not give the Hint but raises the RunError(215).

I think both versions should be fixed.

Dennis