[Bug ada/25245] New: Discriminant is left uninitialized.

[Bug ada/25245] New: Discriminant is left uninitialized.

[listor1 dot rombobeorn at comhem dot se]

The attached file uninitialized_field.adb demonstrates a case where a 
discriminant of a record isn't initialized. I compile and run it like this:

$ gnatmake uninitialized_field.adb
gcc -c uninitialized_field.adb
gnatbind -x uninitialized_field.ali
gnatlink uninitialized_field.ali
$ ./uninitialized_field
Initialized with Unified_Encoding_Record aggregate:
With predefined "=" - A1a and A2a: equal
Initialized with Character_Encoding aggregate:
With predefined "=" - A1b and A2b: not equal
With redefined "=" - B1c and B2c: equal
OS of A1a: LINUX
OS of A2a: LINUX
OS of A1b: OS2
OS of A2b:

raised CONSTRAINT_ERROR : uninitialized_field.adb:117 invalid data

Correct output would be:

Initialized with Unified_Encoding_Record aggregate:
With predefined "=" - A1a and A2a: equal
Initialized with Character_Encoding aggregate:
With predefined "=" - A1b and A2b: equal
With redefined "=" - B1c and B2c: equal
OS of A1a: LINUX
OS of A2a: LINUX
OS of A1b: LINUX
OS of A2b: LINUX

The other attached file, convert_to_pointer.ada, contains relevant parts 
of the code in uninitialized_field.adb, but here the declarations are in 
a package and are referenced from the main program. This causes a very 
strange error message:

$ gnatmake convert_to_pointer_package.ads
gcc -c convert_to_pointer_package.ads
$ LANG=en_US gnatmake convert_to_pointer_main.adb
gcc -c convert_to_pointer_main.adb
convert_to_pointer_package.ads: In function `Convert_To_Pointer_Main':
convert_to_pointer_package.ads:37: error: cannot convert to a pointer type
gnatmake: "convert_to_pointer_main.adb" compilation error

Both these test cases fail on all of the following compilers:

$ LANG=C gcc -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.1-20051112/configure
--prefix=/home/bjorn/ada/gcc-4.1-20051112-inst --enable-libada
--enable-languages=ada,c
Thread model: posix
gcc version 4.1.0 20051112 (experimental)

$ LANG=C gcc -v
Using built-in specs.
Target: i386-redhat-linux
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man
--infodir=/usr/share/info --enable-shared --enable-threads=posix
--enable-checking=release --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-libgcj-multifile
--enable-languages=c,c++,objc,java,f95,ada --enable-java-awt=gtk
--with-java-home=/usr/lib/jvm/java-1.4.2-gcj-1.4.2.0/jre
--host=i386-redhat-linux
Thread model: posix
gcc version 4.0.2 20051125 (Red Hat 4.0.2-8)

$ gcc -v
Reading specs from
/home/bjorn/ada/gcc-3.4.0-inst/lib/gcc/i686-pc-linux-gnu/3.4.0/specs
Configured with: /home/bjorn/ada/gcc-3.4.0/configure
--prefix=/home/bjorn/ada/gcc-3.4.0-inst
Thread model: posix
gcc version 3.4.0

Note: I originally submitted these test cases to bug 17160, but that bug has
been fixed, and indeed that ICE no longer happens. These tests still fail
though, so I'm making a separate report for them.


-- 
           Summary: Discriminant is left uninitialized.
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ada
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: listor1 dot rombobeorn at comhem dot se
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[listor1 dot rombobeorn at comhem dot se]

------- Comment #1 from listor1 dot rombobeorn at comhem dot se  2005-12-03 17:54 -------
Created an attachment (id=10397)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10397&action=view)
demonstrates uninitialized discriminant


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[listor1 dot rombobeorn at comhem dot se]

------- Comment #2 from listor1 dot rombobeorn at comhem dot se  2005-12-03 17:55 -------
Created an attachment (id=10398)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10398&action=view)
demonstrates strange error message


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[laurent at guerby dot net]

------- Comment #3 from laurent at guerby dot net  2005-12-04 21:11 -------
The problem is that you are using "Unbounded_String" in your record, and this
brings in lots of dope information for Finalization purposes. You can see it
with:

gcc -c -gnatR3 uninitialized_field.adb
...
for Unified_Encoding_Record'Object_Size use 384;
for Unified_Encoding_Record'Value_Size use  (if (#1 != 1) then (if (#2 == 2)
then 384 else 208 end) else 200 end) ;
for Unified_Encoding_Record'Alignment use 4;
for Unified_Encoding_Record use record
   _Controller at  4 range  0 .. 159;
   Known       at  0 range  0 ..  7;
   Os          at  1 range  0 ..  7;
   Which       at 24 range  0 ..  7;
   Name        at 24 range  0 .. 191;
   Number      at 24 range  0 .. 15;
end record;
...

If you replace Unbounded_String with Character (or any other non controlled
like a subtype String5 is string (1..5)) your program does work without problem
(4.2.0 20051202).

I assume you used 'address and stuff in your test case but not in your original
program. Could you provide a test case without 'address that has a suspicious
behaviour (even if not all the times)?

Laurent


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[laurent at guerby dot net]

-- 

laurent at guerby dot net changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[listor1 dot rombobeorn at comhem dot se]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1753 bytes --]



------- Comment #4 from listor1 dot rombobeorn at comhem dot se  2005-12-06 22:21 -------
Subject: Re:  Discriminant is left uninitialized.

laurent at guerby dot net wrote:
> The problem is that you are using "Unbounded_String" in your record, and
> this brings in lots of dope information for Finalization purposes.

Your comment puzzles me. Do you really mean that the unbounded string is what 
causes the problem? Yes, I understand that it's involved in triggering the 
bug, but do you mean I should expect problems when using a controlled type 
inside a record type?

> I assume you used 'address and stuff in your test case but not in your
> original program.

Yes. That was only to make the test case reproducible.

> Could you provide a test case without 'address that has a 
> suspicious behaviour (even if not all the times)?

I'm attaching uninitialized_field_2.adb, which is the same program but without 
the address clauses. Here's what I get when I run it:

$ gnatmake uninitialized_field_2.adb
gcc -c uninitialized_field_2.adb
gnatbind -x uninitialized_field_2.ali
gnatlink uninitialized_field_2.ali
$ ./uninitialized_field_2
Initialized with Unified_Encoding_Record aggregate:
With predefined "=" - A1a and A2a: equal
Initialized with Character_Encoding aggregate:
With predefined "=" - A1b and A2b: not equal
With redefined "=" - B1c and B2c: equal
OS of A1a: LINUX
OS of A2a: LINUX
OS of A1b:

raised CONSTRAINT_ERROR : uninitialized_field_2.adb:92 invalid data

Björn Persson


------- Comment #5 from listor1 dot rombobeorn at comhem dot se  2005-12-06 22:21 -------
Created an attachment (id=10429)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10429&action=view)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[laurent at guerby dot net]

------- Comment #6 from laurent at guerby dot net  2005-12-09 19:51 -------
I see the same thing as you on trunk x86-linux with gcc version 4.2.0 20051202
(experimental).

I agree the "not equal" part is a bug:
  With predefined "=" - A1b and A2b: not equal
should print "equal".

But the constraint_error for "OS of A1b" looks correct, and if commented the
one on "OS of A2b", it is a language mandated discriminant check failing. But
may be I'm missing something, could you explain why you think an exception
shouldn't be raised?

Today Ada front-end commits may have fixed the bug, I'm retesting.


-- 

laurent at guerby dot net changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-12-09 19:51:33
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[listor1 dot rombobeorn at comhem dot se]

------- Comment #7 from listor1 dot rombobeorn at comhem dot se  2005-12-09 22:35 -------
Subject: Re:  Discriminant is left uninitialized.

> But the constraint_error for "OS of A1b" looks correct, and if commented
> the one on "OS of A2b", it is a language mandated discriminant check
> failing. But may be I'm missing something, could you explain why you think
> an exception shouldn't be raised?

The discriminant check fails because OS doesn't contain any of the valid 
values 0, 1 or 2 (for NT, OS2 and Linux). If there is random data in the 
memory I expect a 3/256 chance that it will get a valid value.

The OS field shouldn't be random. It's fixed to Linux in the definition of 
Character_Encoding_A:

   This_OS : constant Known_OS := Linux;
   type Character_Encoding_A (Known : Boolean := False) is
     new Unified_Encoding_Record (Known => Known, OS => This_OS);

All objects of this type should therefore have OS=Linux, so the discriminant 
would be valid and would pass the check.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245

[Bug ada/25245] Discriminant is left uninitialized.

[laurent at guerby dot net]

------- Comment #8 from laurent at guerby dot net  2005-12-10 13:26 -------
Ok I now agree on both counts :).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25245