Is this a bug in the conversion by FPC?

[dcoun]

Running the following code in Delphi 12 results to an exception. In FPC 3.2 gives a negative number.

Code: Pascal  [Select][+][-]

1. var m:integer; n:cardinal;
2. begin n:=high(cardinal);
3.   m:=n;
4.   if m<0 then showmessage('negative');
5. end;

The following code runs without exceptions and some results in both Delphi 12 & FPC 3.2

Code: Pascal  [Select][+][-]

1. var m:integer; n:cardinal;
2. begin n:=high(cardinal);
3.   m:=integer(n);
4.   if m<0 then showmessage('negative');
5. end;

To my way of thinking Delphi does it well. A Cardinal can not always fit to an integer. To force it to fit in the memory, it has now an other number that it is negative (-1).
In the second case, the programmer wants a conversion and is aware of the possible results.

What is your opinion?

[korba812]

Maybe range checking is enabled by default in your Delphi? Enabling range checking in FPC also causes an exception in this case.
https://www.freepascal.org/docs-html/3.2.0/prog/progsu65.html

[dcoun]

Maybe range checking is enabled by default in your Delphi? Enabling range checking in FPC also causes an exception in this case.
https://www.freepascal.org/docs-html/3.2.0/prog/progsu65.html

Running it from the IDE has range checking enabled. That it is not the problem.
The question if this is expected or not in conversion

[jamie]

I believe both integer and cardinal are the same size, ones an unsigned 32 bit while the other uses the MSB as the sign bit
so, both will interchange memory wise.

[dcoun]

Maybe range checking is enabled by default in your Delphi? Enabling range checking in FPC also causes an exception in this case.
https://www.freepascal.org/docs-html/3.2.0/prog/progsu65.html

Enabling the {$R+} I have the exception as in delphi. You have right.
So, it is as expected.

[jamie]

as a note, it only raises it as a runtime error, meaning, if the value exceeds the positive range of the integer when range is on.

[VisualLab]

I believe both integer and cardinal are the same size, ones an unsigned 32 bit while the other uses the MSB as the sign bit
so, both will interchange memory wise.

And aren't integers represented rather in 2's complement code?

https://en.wikipedia.org/wiki/Two%27s_complement

[dcoun]

as a note, it only raises it as a runtime error, meaning, if the value exceeds the positive range of the integer when range is on.

You have right, it is as expected.