Why does this happen?

Why does this happen?

[svesse]

Hello,
I got a question regarding the enclosed code
which compiles on Rational Apex and on old
GNAT versions (at least on 3.11p). However
in GNAT 3.14p the following error message is
shown :

gnatmake test.adb
gcc -c test.adb
test.ads: In function `test__proc':
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
test.ads:6: prior parameter's size depends on `ns'
gnatmake: "test.adb" compilation error

if any of the two commented lines are used instead
it also compiles on GNAT 3.14p

Can someone explain why this would be illegal ?

---- test.ads ---
package test is

    Type Mtr_Type is array (Positive range <>,
                            Positive range <>) of integer;

    type T_typ (Ns : Positive) is record
       M : Mtr_Type (1..2, 1 .. ns);  -- does not compile
--      M : Mtr_Type (1..Ns, 1..Ns); -- compiles
--      M : Mtr_Type (1..Ns, 1..2);  -- compiles
    end record;

    function Func (Left : Mtr_Type) return Mtr_Type;
    procedure Proc (T : in out T_Typ);

end test;

----------

---- test.adb ----
package body test is

   function Func (Left : Mtr_Type) return Mtr_Type is
   begin
      return Left;
   end Func;

   procedure Proc(T : in out T_Typ) is
   begin
     T.M := Func(T.M);
   end Proc;
end test;
-------

Re: Why does this happen?

[Dennis Lee Bieber]

svesse fed this fish to the penguins on Tuesday 15 October 2002 03:12 
pm:


> 
> Can someone explain why this would be illegal ?
> 
> ---- test.ads ---
> package test is
> 
>     Type Mtr_Type is array (Positive range <>,
>                             Positive range <>) of integer;
> 
>     type T_typ (Ns : Positive) is record
>        M : Mtr_Type (1..2, 1 .. ns);  -- does not compile


        I know Ada is supposed to be case insensitive, but if this is a direct 
cut&paste from your file... You have "Ns" in the descriminant (hope I 
have the proper term), but "ns" in the array dimension. You /also/ have 
spaces surrounding the "..". Neither of these differences show in the 
following two lines.

> --      M : Mtr_Type (1..Ns, 1..Ns); -- compiles
> --      M : Mtr_Type (1..Ns, 1..2);  -- compiles

        The only other idea is that the calculation of offsets for "1..2" can 
not be done immediately -- it may be that the parsing requires variable 
ranges to be leftmost in the subscript order.

--  
 > ============================================================== <
 >   wlfraed@ix.netcom.com  | Wulfraed  Dennis Lee Bieber  KD6MOG <
 >      wulfraed@dm.net     |       Bestiaria Support Staff       <
 > ============================================================== <
 >        Bestiaria Home Page: http://www.beastie.dm.net/         <
 >            Home Page: http://www.dm.net/~wulfraed/             <

Re: Why does this happen?

[tmoran]

>   Type Mtr_Type is array (Positive range <>,
>                           Positive range <>) of integer;
>
>   type T_typ (Ns : Positive) is record
>      M : Mtr_Type (1..2, 1 .. ns);  -- does not compile
>   end record;

Gnat 3.14 on my Windows system compiles those lines without complaint.

Re: Why does this happen?

[Simon Wright]

svesse <svesse@linux.nu> writes:

> Can someone explain why this would be illegal ?

I suspect it isn't illegal, rather looks like a compiler bug; 3.15a1
gives similar results to yours (only 4 lines of complaint!) whereas
GCC-3.2 (and 3.1, 3.3) give bombs:

   smaug[21]$ gnatgcc -c test.adb
   +===========================GNAT BUG DETECTED==============================+
   | 3.2 20020814 (release) (i686-pc-linux-gnu) GCC error:                    |
   | Internal compiler error in make_decl_rtl, at varasm.c:835                |
   | Error detected at test.adb:14:13                                         |
   | Please submit a bug report, see http://gcc.gnu.org/bugs.html.            |
   | Include the entire contents of this bug box in the report.               |
   | Include the exact gcc or gnatmake command that you entered.              |
   | Also include sources listed below in gnatchop format                     |
   | (concatenated together with no headers between files).                   |
   +==========================================================================+

   Please include these source files with error report

   test.adb
   test.ads

   compilation abandoned

I think this is worth a PR, shall I do it?

Re: Why does this happen?

[Anders Wirzenius]

<tmoran@acm.org> wrote in message news:D64r9.59604$rz6.10428@sccrnsc02...
> >   Type Mtr_Type is array (Positive range <>,
> >                           Positive range <>) of integer;
> >
> >   type T_typ (Ns : Positive) is record
> >      M : Mtr_Type (1..2, 1 .. ns);  -- does not compile
> >   end record;
>
> Gnat 3.14 on my Windows system compiles those lines without complaint.

... but not on my Gnat 3.14, Windows NT.
I get the same error message as OP.

instead
       M : Mtr_Type (-1..ns, -1..ns);
compiles without errors.

Could it be a question of elaboration, because a compilation, having
       M : Mtr_Type (-1..ns, -1..ns);
in test.ads,
of a test_test.adb

with test;
with Ada.Text_IO;
procedure test_test is
   A: test.T_typ(2);
begin
   null;
end;

gives the warning

test.ads:7:22: warning: static value out of range of type
"Standard.positive"
test.ads:7:22: warning: "constraint_error" will be raised at run time

Anders Wirzenius

Re: Why does this happen?

[tmoran]

> > Gnat 3.14 on my Windows system compiles those lines without complaint.
>
> ... but not on my Gnat 3.14, Windows NT.
> I get the same error message as OP.
  Interesting.  The declarations compile OK.  An assignment to T.M
in the body causes the error message to appear in the spec.  Dropping
Func, the body:

package body test is

   junk:mtr_type(1 .. 2, 1 .. 2);

   procedure Proc(T : in out T_Typ) is
--   junk : mtr_type := T.M;
   begin
     T.M := junk;
   end Proc;
end test;

causes the error message, but if you switch which "junk" is commented out,
Gnat gives its "I crashed" message.  Clearly a confused compiler.

Re: Why does this happen?

[Frank J. Lhota]

With GNAT 3.14p under NT, I get the same problem, even after adjusting the
case of "Ns" and the white space. ObjectAda seems to handle this just fine,
however. Just a weird compiler bug.

Re: Why does this happen?

[Jerry van Dijk]

Simon Wright <simon@pushface.org> writes:

> I suspect it isn't illegal, rather looks like a compiler bug; 3.15a1
> gives similar results to yours (only 4 lines of complaint!) whereas
> GCC-3.2 (and 3.1, 3.3) give bombs:

Yes, the gcc GNAT 3.1 and 3.2 that come with the mingw system are not very
useful. However, after the freezing of the gcc 3.2 sources, a big update was
done to gcc GNAT, and it is now a LOT better.

If you want to try for yourself, goto

        http://prdownloads.sf.net/mingw

and download

        gnatgcc-3.3-exp-20021006-notes.txt.gz
        gnatgcc-3.3-exp-20021006.tar.gz        
        (and gnatgcc-3.3-exp-20021006.tar.gz if you need the sources)

and follow the instructions in the notes file. REALLY as it is not installed
under the current mingw distribution.

Note: the is currently no such thing as an gcc 3.3, and this is strictly
an experimental build. Although it is, IMO, much better then the 3.2 GNAT
version on Mingw, an ACVC run still shows some problems.

-- 
--  Jerry van Dijk   | email: jvandyk@attglobal.net
--  Leiden, Holland  | web:   users.ncrvnet.nl/gmvdijk

Re: Why does this happen?

[Gautier]

> Hello,
> I got a question regarding the enclosed code
> which compiles on Rational Apex and on old
> GNAT versions (at least on 3.11p). However
> in GNAT 3.14p the following error message is
> shown :
[...]

You found a nice bug!
It compiles with ObjectAda 7.2.1 SE, GNAT/DOS 3.10p
and also an Ada 83 compiler.
________________________________________________________
Gautier  --  http://www.mysunrise.ch/users/gdm/gsoft.htm

NB: For a direct answer, address on the Web site!

Re: Why does this happen?

[Colin Paul Gloster]

In article <ur8eqwfb9.fsf@attglobal.net>, Jerry van Dijk wrote:

"Yes, the gcc GNAT 3.1 and 3.2 that come with the mingw system are not very
useful. However, after the freezing of the gcc 3.2 sources, a big update was
done to gcc GNAT, and it is now a LOT better.

If you want to try for yourself, goto

        http://prdownloads.sf.net/mingw

and download

        gnatgcc-3.3-exp-20021006-notes.txt.gz
        gnatgcc-3.3-exp-20021006.tar.gz        
        (and gnatgcc-3.3-exp-20021006.tar.gz if you need the sources)
  
[..] Although it is, IMO, much better then the 3.2 GNAT
version on Mingw, an ACVC run still shows some problems."

Does the -gnat83 switch still allow non-Ada83 programs with -1 .. 100
literal loop constraints pass without any warnings?

Re: Why does this happen?

[Jerry van Dijk]

Colin_Paul_Gloster@ACM.org (Colin Paul Gloster) writes:

> Does the -gnat83 switch still allow non-Ada83 programs with -1 .. 100
> literal loop constraints pass without any warnings?

If it does in the current compiler, than it probably still does.

Note that by better, I mean that it functions better on Windows (less
blowups), not that there are additional frontend bugs resolved. 3.14p is,
IMHO, still the better compiler.

-- 
--  Jerry van Dijk   | email: jvandyk@attglobal.net
--  Leiden, Holland  | web:   users.ncrvnet.nl/gmvdijk