Re: Unsigned Integer Restraint Errors

["Adam Beneschan" <adam@irvine.com>]

On Mar 12, 8:07 am, "frikk" <frik...@gmail.com> wrote:

> LASTLY:
> Please note that when I change the range value from 0 .. 5 to 0 ..
> 2**64-1, or 0 .. UNSIGNED_LONG_INT'Last, there is no constraint error
> raised. This is the same behavior as the first example.

I missed this question the first time.

The reason for this is probably that special things are needed to do
this sort of check.  When you declare a subtype in the range 0 .. 5,
the code generated by the compiler can do its normal arithmetic
operations, and then check to see if the result is less than 0 or
greater than 5.  But when you declare a subtype in the range
0..2**64-1, the code can't do this, because *all* 64-bit integers are
going to appear to be in this range.  In order to do the check, either
the compiler will need to generate additional special code, or it will
need to do something to enable the processor's hardware overflow
mechanism.  On GNAT and some other compilers, this requires that you
use a special flag when compiling.  This might be -gnato, but I don't
remember offhand, and I'm hoping someone else will tell us what it is
so that I don't have to look it up myself.

                                  -- Adam