Version 1.8 of ais/ai-00114.txt

Unformatted version of ais/ai-00114.txt version 1.8
Other versions for file ais/ai-00114.txt

!standard 12.03 (11)          00-08-31 AI95-00114/00
!class confirmation 96-04-04
!status work item 96-04-10
!status received 96-04-04
!priority Low
!difficulty Hard
!subject Editorial Comments on AARM
!summary
This AI serves as a holder for editorial comments on AARM-only annotations. Because the AARM has no official status as far as ISO is concerned, these will be considered low priority.
!question
!response
!appendix

!section 12.3(11)
!subject but the{y} can be in instances.
!reference AARM95-12.3(11.x)
!from Keith Thompson 95-11-10
!reference 95-5384.a Keith Thompson 95-11-11>>
!discussion

****************************************************************
From: Randy Brukardt (Editor)
Date: August 30, 2000

This issue was fixed in AARM 6.0 (of course, no one except Bob has that
version :-).

****************************************************************
!section 12.3(11)
!subject It doesn't matter that [an] one could imagine ...
!reference AARM-12.3(11.dd)
!from Keith Thompson 95-11-17
!reference 95-5391.a Keith Thompson 95-11-17>>

****************************************************************
!section 13.3(12)
!subject Address clauses for components
!reference AARM95-13.3(12)
!from Keith Thompson 96-02-22
!reference 96-5429.a Keith Thompson 96-2-22>>
!discussion

The referenced paragraph in the AARM says:

        12   {specifiable [of Address for stand-alone objects and for
        program units]} {Address clause} Address may be specified for
        [stand-alone] objects and for program units via an attribute_
        definition_clause.

The use of square brackets around the phrase "stand-alone" implies that it
is redundant and can be proven from other rules.  How does the restriction
to stand-alone objects follow from other rules?  In particular, why is
this illegal (adapted from a posting to comp.lang.ada by Bob Gilbert)?

   protected Discretes is
      procedure Write(Settings : in Setting_List);
   private
      HW_Control : Discrete_HW_Control;
      for HW_Control use at Some_Legal_Address;
   end Discretes;

I'm not suggesting that it should be, just asking how it can be proven
from other rules.


****************************************************************
!section 13.3(12)
!subject Address clauses for components
!reference AARM95-13.3(12)
!reference 96-5429.a Keith Thompson 96-2-22
!from Bob Duff
!reference 96-5456.a Robert A Duff 96-4-8>>
!discussion

> The referenced paragraph in the AARM says:
>
>       12   {specifiable [of Address for stand-alone objects and for
>       program units]} {Address clause} Address may be specified for
>       [stand-alone] objects and for program units via an attribute_
>       definition_clause.
>
> The use of square brackets around the phrase "stand-alone" implies that it
> is redundant and can be proven from other rules.  How does the restriction
> to stand-alone objects follow from other rules?

Good question.

Apparently, we thought that it was implied by 13.1(6.a):

        6.a   Ramification:  The ``statically denote'' part implies that it
        is impossible to specify the representation of an object that is not
        a stand-alone object, except in the case of a representation item
        like pragma Atomic that is allowed inside a component_list (in which
        case the representation item specifies the representation of
        components of all objects of the type).  It also prevents the problem
        of renamings of things like ``P.all'' (where P is an
        access-to-subprogram value) or ``E(I)'' (where E is an entry family).

But I must admit that I don't see how this follows from the RM.

In any case, the square brackets are AARM-only material, so the ARG can
safely believe them, even if they don't strictly follow from other
rules.  That is, this is a bug in the AARM's use of brackets, but not a
bug in the RM itself.

>...  In particular, why is
> this illegal (adapted from a posting to comp.lang.ada by Bob Gilbert)?
>
>    protected Discretes is
>       procedure Write(Settings : in Setting_List);
>    private
>       HW_Control : Discrete_HW_Control;
>       for HW_Control use at Some_Legal_Address;
>    end Discretes;
>
> I'm not suggesting that it should be, just asking how it can be proven
> from other rules.

The above was intended to be illegal, and *is* illegal by 13.3(12),
despite its use of square brackets in the AARM.

Anwyay, I think this can be taken as an editorial comment on the AARM,
rather than a comment on the RM itself.

- Bob

****************************************************************
!section 13.3(12)
!subject Address clauses for components
!reference AARM95-13.3(12)
!reference 96-5429.a Keith Thompson 96-2-22
!reference 96-5456.a Bob Duff 96-03-08
!from Keith Thompson
!reference 96-5458.a Keith Thompson 96-4-8>>
!discussion

Bob Duff wrote:
> In any case, the square brackets are AARM-only material, so the ARG can
> safely believe them, even if they don't strictly follow from other
> rules.  That is, this is a bug in the AARM's use of brackets, but not a
> bug in the RM itself.
[...]
> Anwyay, I think this can be taken as an editorial comment on the AARM,
> rather than a comment on the RM itself.

I agree.  I hadn't realized (though I probably should have) that the
phrase "stand-alone" does appear in the RM, without the brackets.

****************************************************************
!section 13.1(14)
!subject Errors in example
!reference RM95-13.1(14)
!from Jorgen Bundgaard 96-03-07
!keywords
!reference 96-5434.a Joergen Bundgaard 96-3-7>>
!discussion

The example 13.1(14.c-e) contains other violations of the legality rules than
those marked with "-- Illegal!": types A1 and A2 should be general access
types, and the accessibility levels of the Obj1(17) and Obj2(17) should not be
statically deeper than that of access types A2 and A1, respectively.

****************************************************************
!section 3.9.2(20)
!subject Illegal use of names in 3.9.2 example
!reference AARM95-3.9.2(20.e)
!from Ivan B. Cvar 96-09-19
!keywords dispatching operations
!reference 96-5703.a Ivan B. Cvar 96-9-19>>
!discussion

The example in the Annotated Ada95 Reference Manual 3.9.2(20.e) fails
to compile because it references 2 type names that are not visible at
the point of use.

However, the meaning of the example is clear, despite the errors, but you
may want to repair the example for future releases.

with P1; with P2;
procedure Main is
    X : T2;                  --<<< ERROR: Should be P2.T2
    Y : T1'Class := X;       --<<< ERROR: Should be P1.T1
begin
    P2.Op_A(Param => X); -- Nondispatching call.
    P1.Op_A(Arg => Y); -- Dispatching call.
    P2.Op_B(Arg => X); -- Nondispatching call.
    P1.Op_B(Arg => Y); -- Dispatching call.
end Main;

****************************************************************
!section 7.3(7)
!subject Semantic Errors in AARM 7.3 Examples
!reference AARM-7.3(7.f)
!reference AARM-7.3(7.h)
!reference AARM-7.3(7.m)
!from Ivan B. Cvar 96-08-01
!reference 96-5629.a Ivan B. Cvar 96-8-1>>
!discussion

There are 3 unintended semantic errors in Annotated Ada 95 Reference Manual
clause 7.3, paragraphs 7.f, 7.h, and 7.m.  They are:

In the example of 7.3(7.f), the closing designator at the end of the
body of procedure Foo is wrong.  It should be Foo, not A.

In the example of 7.3(7.h), the named association in the discriminant
constraint denotes a discriminant named A, whereas the correct name is D,
as declared in paragraph 7.g.

In the example of 7.3(7.m), the package Q should have the context clause
"with P; use P" to make the name of type Parent visible, and for consistency
with 7.3(7.n) since 7.3(7.n) contains a with clause that denotes Q.

****************************************************************

!topic      Change from Ada 83 omitted
!reference  AARM-95 13.7 (37.d-38.d)
!from       Wes Groleau  98 Oct 06
!keywords   System  Null_Address  incompatibility
<<reference as: 1998-15927.a W. Wesley Groleau x4923 1998-10-6>>
!discussion In LRM-83, no equivalent of System.Null_Address is required.
	    Many vendors added it as an extension, but not all used the
	    same name.  AARM mentions UI-0065 (which I am still trying to
	    locate) but doesn't specifically say it is a change.  (Karl
	    Grebyn's Ada 83 Annotations do not mention any Ada-83
	    standardization issue for this, so I presume there wasn't one
	    until the 9X effort.)

****************************************************************

!topic Object freezing and aspect clauses
!reference RM95 13.14(19), 13.1
!from Adam Beneschan 11-18-02
!discussion

I believe there's an omission from RM95 13.1 on the subject of an
aspect clause on an object after the object is frozen.  It seems clear
to me that this was intended to be illegal.  13.14(19) says:

   An operational or representation item that directly specifies an
   aspect of an entity shall appear before the entity is frozen (see
   13.1)

where "entity" includes objects as well as types.  ACATS test bde0008
confirms that this is supposed to be illegal:

   type Enum is (Red, White, Yellow);
   ObjA : Enum;
   ObjB : Enum := ObjA;

   -- The nonstatic expression ObjA in the initialization expression of ObjB
   -- above freezes ObjA.  The representation clause is therefore illegal.
   -- RM 13.14(4,8,19);6.0

   for ObjA'Size use System.Storage_Unit;                             -- ERROR:
                                   -- Representation clause of a frozen object.

The reason I think there's an omission is that the Legality Rules in
13.14(16-19) are all supposed to be repeats of rules that are stated
elsewhere in the RM, and they all refer to other sections.  The above
rule refers to 13.1.  However, there appears to be no language in 13.1
that makes it illegal for an aspect clause on an *object* to appear
after the object is frozen.  13.1(9) and 13.1(9.1) make it illegal for
types and subtypes, but there's no similar language in 13.1 that
pertains to objects.

This appears to me to be a simple typo, but it should be fixed.

****************************************************************

From: Randy Brukardt
Sent: Friday, November 22, 2002  11:16 PM

The above comment has raised during the Corrigendum process, and was handled
at that time. Note that the comment is only on the AARM, as there is no
such thing as the RM rule "that is intended to be a repeat". If it's a
normative rule (and this one is), it applies.

We discussed this during the Corrigendum discussions. See the !appendix of
AI-137 (toward the bottom).

But the whole question is a bit silly. I don't see anything in 13.14(16-19)
that indicates that it is some sort of repeat. Certainly, it is normative
text in the standard. It's true that the Ada 95 AARM has the text in square
brackets, but the AARM isn't normative in any way.

There were several such AARM errors discovered before or during the
Corrigendum work. The updated AARM that includes the Corrigendum changes
does not include square brackets around 13.14(19) for this very reason: it
is not redundant. (There are a couple of other, similar cases in the updated
AARM.)

In any case, this appears to be a comment on the AARM, not on the
standard -- and the AARM change has already been made. So there is nothing
that needs to be done here.

****************************************************************

From: Randy Brukardt (Editor)
Date: August 30, 2000

I've made the corrections needed to implement the presentation issues
above in the updated AARM, or explained why the suggested correction was
not (and will not) be made.

Newer items below this item (if any) have *not* been handled. (They should be
on the updated AARM).

****************************************************************

From: Dan Eilers [dan@IRVINE.COM]
Sent: Wednesday, October 18, 2000 8:29 PM

!topic AI-00077 is undocumented extension to Ada83
!reference RM95-10.01.04
!from Dan Eilers
!keywords extension
!discussion

The AARM should mention that AI-00077 is an extension to Ada83.

[Editor's note: AI-00077 is a confirmation.]

****************************************************************

From: Dan Eilers [dan@IRVINE.COM]
Sent: Wednesday, October 18, 2000 8:33 PM

!topic AI-00163 is undocumented incompatibility with Ada83
!reference RM95-4.5.5(18-20)
!from Dan Eilers
!keywords incompatibility
!discussion

The AARM should mention that AI-00163 is an incompatibility with Ada83.

[Editor's note: AI-00163 is a confirmation.]

****************************************************************

From: Dan Eilers [dan@IRVINE.COM]
Sent: Wednesday, October 18, 2000 8:53 PM

!topic undocumented inconsistency regarding 'size
!from Dan Eilers
!keywords inconsistency
!discussion

The AARM should mention that 'size is inconsistent with Ada83,
as noted by Robert Dewar in his email attached to AI95-00163,
as well as recent ARG email:

> ...   (we have found the incompatibilities in handling of Size to be FAR
> more worrisome, since in general they cause different results with no
> complaint at compile time).
  ...
> If one user bumping into one incompatibility using one vendors compiler IS
> enough justification for changing the language, I have a number of messages
> to transmit to this list, starting with a complaint about what to me is the
> gratuitous change in the handling of static expressions that makes LOTS of
> Ada 83 programs illegal.

----

>       ...                                               The pragmatic
> issues are that Ada 95 pinned down some things left undefined in Ada 83,
> and did it in a manner which, while formally allowed by the 83 RM did
> not correspond to the way most compilers did things (e.g. mandating
> Natural'Size = Integer'Size - 1, which was clearly allowed by the
> Ada 83 RM, but all Ada 83 compilers I worked with had Natural'Size =
> Integer'Size [I have heard that the Intermetrics compiler had the
> Ada 95 treatment, but I am not familiar with that Ada 83 compiler].

****************************************************************

!topic Most operations that yield[s] a String
!reference AARM95-A.4.3(2.a)
!from Christoph Grein 02-01-07
!discussion

This typo is not a RM bug, since it is only in the AARM. Nevertheless it should
be corrected.

****************************************************************

From: Robert I. Eachus
Sent: Wednesday, December  3, 2003  7:53 AM

>Feel free to send such suggestions to me. I added some index entries for the
>Consolidated RM, and I'll be happy to do so again.

I just ran into one that is not marginal at all. 3.8(12) should appear
somewhere under discriminants in the index, for example:

discriminant
   use in a record definition 3.8(12)

This is the paragraph that requires that a discriminant must appear
alone in a record declaration: "A name that denotes a noninherited
discriminant is allowed within the declaration of the type, but not
within the discriminant_part. If the discriminant is used to define the
constraint of a component, the bounds of an entry family, or the
constraint of the parent subtype in a derived_type_definition then its
name shall appear alone as a direct_name (not as part of a larger
expression or expanded name). A discriminant shall not be used to define
the constraint of a scalar component."

Of course I could recommend that we remove the restriction instead. ;-)
But I just don't see that as a big deal.

****************************************************************

From: Pascal Leroy
Sent: Friday, November 26, 2004  8:39 AM

1.1.2(23.c) should say "certification" rather than "validation", because
we should be consistent with the terminology of 18009.

****************************************************************


Questions? Ask the ACAA Technical Agent