re: access all definitive

re: access all definitive

[fintan]

re: access all query

It seems that my interpretation of the 95 LRM was
correct and that the gnat implementation is yet to 
be correctly implemented. i.e. 

  an access all type can only designate an aliased
  variable and cannot designate a dynamic variable.

  an access variable is a la 83 and can only designate 
  a dynamic variable and cannot designate an aliased
  variable.

fintan

_____________________________________________________________
Fintan Culwin PhD
Senior Lecturer Software Engineering
School of Computing
South Bank University
Borough Road
London SE1 0AA

Phone (+44) (0)171 815 7434
Fax   (+44) (0)171 815 7499

fintan@vax.sbu.ac.uk

Re: access all definitive

[Michel Gauthier]

In article <1995Jan24.170441.8957@vax.sbu.ac.uk>, fintan@vax.sbu.ac.uk wrote:

> re: access all query
> 
> [...]
> 
>   an access all type can only designate an aliased
>   variable and cannot designate a dynamic variable.
> 
>   an access variable is a la 83 and can only designate 
>   a dynamic variable and cannot designate an aliased
>   variable.

A dynamic variable a la 83 is intrinsicly aliased, with all
that can be implied by this property.

-- 
Michel Gauthier - Laboratoire d'informatique - 123 avenue Albert Thomas
F-87060 Limoges - fax +33()55457315
----- Are the messages that objects exchange also objects ? -----

Re: access all definitive

[Cyrille Comar]

fintan@vax.sbu.ac.uk writes:
: re: access all query
: 
: It seems that my interpretation of the 95 LRM was
: correct and that the gnat implementation is yet to 
: be correctly implemented. i.e. 
: 
maybe you could check the latest distributed GNAT (i.e. v2.02 ;-)
-- 
------------------------------------------------------------------------
Cyrille Comar,                                  E-mail: comar@cs.nyu.edu
Gnat Project                                    US phone: (212) 998-3489

Re: access all definitive

[Robert Dewar]

Just so Fintan doesn't lose too many sleepless nights worrying about
GNAT failing to diagnose this error :-)

     1. procedure a is
     2.    type x is access integer;
     3.    vx : x;
     4.    q : aliased integer;
     5. begin
     6.    vx := q'access;
                  |
        >>> result must be general access type
        >>> add "all" to type "x" declared at line 2

     7. end;

Actually we anticipate that a lot of users of GNAT version 2.02 will
be surprised by this new check (just wait till you get version 2.03
with the accessibility checks fully implemented :-)

Re: access all definitive

[Robert Dewar]

fintan said:

"an access all type can only designate an aliased
variable and cannot designate a dynamic variable."

that's very wrong. an access all type can point to all things (why do you
think it uses all>? :-)

in particular, you can of course have a access all designate a dynamic
variable (i.e. one generated by an allocator).

one coding style that will develop in Ada 95 is simply to use all on all
access type declarations. However, I am not sure that is such a good idea,
one ought to minimize the use of aliased variables (there is a good reason
that Ada 83 did not have this feature, its overuse is definitely damaging)

Re: access all definitive

[Michael Feldman]

In article <3ge71t$h4r@gnat.cs.nyu.edu>, Robert Dewar <dewar@cs.nyu.edu> wrote:
>fintan said:

>one coding style that will develop in Ada 95 is simply to use all on all
>access type declarations. However, I am not sure that is such a good idea,
>one ought to minimize the use of aliased variables (there is a good reason
>that Ada 83 did not have this feature, its overuse is definitely damaging)

And the Ada 95 Rationale (I think that is where I read it!) points out
that distinguishing the two types leads to potential optimizations,
because whereas a general access type (ACCESS ALL) will typically be
represented as a full address, a pool-specific (Ada 83) access type
might be able to use a more compact rep, since only a specific, probably
smaller, storage pool needs to be designated.

Mike Feldman