Y2038 compliance broken on 32bit

General
1

Hi,

regarding Bug report

Wondering if there is any chance of this being fixed in 16.3?

Leads to another question. What is the future, for gnat on 32bit?

Regards,

Peter

2

I suppose 32bit era comes to past. Except embedded systems, but usually they are crosses hosted on 64bit

3

Having looked into this a bit further, it seems to me that time routines are generally OK,

but dates after Y-2038 are broken on 32 bit. (Y2038 is fine on 64bit)

4 
--  Test y2038 compliance with
--  faketime -f '+20y' ./test

with Ada.Calendar, Ada.Text_IO;
procedure test is
   Y : Ada.Calendar.Year_Number;
   M : Ada.Calendar.Month_Number;
   D : Ada.Calendar.Day_Number;
   S : Duration;

begin
   Ada.Calendar.Split (Ada.Calendar.Clock, Y, M, D, S);
   Ada.Text_IO.Put_Line (Y'Img & M'Img & D'Img);
end test;
5

The most recent 32-bit GNAT I have is 14.2.0 on armhf gives wrong answer:

1910 8 3

6

Indeed. Itā€™s not a recent regression.
Iā€™ve tested 32bit gnat versions, 13, 14, 15, 16 and its broken on all of them.

7

This topic should be tagged ā€˜gnatā€™ rather than ā€˜releasesā€™.

Is there any way to retag it?

8

Hi, the category is ā€˜releaseā€™ and there is no ā€˜gnatā€™ category. I can change it to ā€˜helpā€™ if you want.

For now I have added the gnat tag but didnā€™t change the category (unless you want me to).

9

Thanks for adding the ā€˜gnatā€™ tag.
Yes, please remove ā€˜releaseā€™ category. As the bug exists in 13-16, and probably always existed,

it clearly has nothing to do with any recent release.

10

And, if possible , ā€œY2038 compliance broken on 32bitā€,

would be a better title.