Re: Ada83 equivalents for Ada95

Re: Ada83 equivalents for Ada95

[tmoran]

In <31927190.35AA@csehp3.mdc.com> James A. Squire asked:
>Has anybody gone to the trouble of translating code that uses features
>new to Ada95 back into its Ada83 equivalent?  If anyone knows of such
>examples that are available on the Internet, please let me know (email
  I too would be interested in seeing examples of techniques others
have used to approximate Ada 95 features in Ada 83.  Sometimes I'd
like to design something using Ada 95 constructs, but there is as
yet no adequate Ada 95 compiler for the target, while there is a
good Ada 83 compiler.  It would be nice to code with an eye toward
easing future changeover.  Surely this must be a common problem,
for which lots of folks have developed a set of techniques.

Re: Ada83 equivalents for Ada95

[Tucker Taft]

In <31927190.35AA@csehp3.mdc.com> James A. Squire asked:
>Has anybody gone to the trouble of translating code that uses features
>new to Ada95 back into its Ada83 equivalent?  If anyone knows of such
>examples that are available on the Internet, please let me know (email

Norm Cohen's new book on Ada includes some descriptions of how
to use Ada 83 features to accomplish the same goal as certain of
the new Ada 95 features.

-Tucker Taft   stt@inmet.com   http://www.inmet.com/~stt/
Intermetrics, Inc.  Cambridge, MA  USA

Re: Ada83 equivalents for Ada95

[James A. Squire]

On Fri, 10 May 1996 14:15:04 GMT, Robert A Duff <bobduff@WORLD.STD.COM>
wrote:

> In article <31927190.35AA@csehp3.mdc.com>,
> James A. Squire <m193884@CSEHP3.MDC.COM> wrote:
> >Has anybody gone to the trouble of translating code that uses features
> >new to Ada95 back into its Ada83 equivalent?  If anyone knows of such
> >examples that are available on the Internet, please let me know (email
> >or comp.lang.ada).
>
> I don't know of any such effort.
>
> I'm curious as to why you want this.  Do you want to develop Ada 95
> code, but then run it on a system where Ada 95 does not yet exist, but
> Ada 83 compilers do exist?

No.  I wanted to be able to show "what you have to go through to do
<this> in Ada83" and do comparisons.
--
James Squire
MDA Avionics Tools & Processes
ja_squire@csehp3.mdc.com
"one of these days I'm going to better myself by going to Knight school"
"You'll be a web knight instead of a web page!"

Re: Ada83 equivalents for Ada95

[Robert A Duff]

In article <3193BA1E.31C1@csehp3.mdc.com>,
James A. Squire <m193884@CSEHP3.MDC.COM> wrote:
>No.  I wanted to be able to show "what you have to go through to do
><this> in Ada83" and do comparisons.

Oh.  Sorry I missed the point.

I think the Rationale has some comparisons that might be of interest.
It's at:

  ftp://sw-eng.falls-church.va.us/public/adaic/docs/standard/95lrm_rat/v6.0

Also, it's not too hard to come up with such examples.  Write something
using a protected type, and then write the same thing using rendezvous.
Use a conditional entry call to illustrate one nice thing about Ada 95.

Write a hierarchy of tagged types, with dispatching operations, and
compare it to the equivalent variant record with case statements.

Write a package, with a private type, and a child package.  Compare to
the equivalent Ada 83 code, which either combines everything into one
package, or else exports private information to clients.

Write some code that deals with money, using decimal types, and try to
write the same thing in Ada 83.

Etc.

- Bob

Re: Ada83 equivalents for Ada95

[tmoran]

>On a different but related topic, has anyone looked at designing Ada 83
>code so that the transition to using the *new* features of Ada 95 is as
>easy as possible.
  I'm currently trying to write some code that will use types with Access
discriminants in Ada 95, using Integers in Ada 83 where the integer is an
index into a table of entries of the Access-ed type.  Essentially doing
my own heap management for that type.  It's a bit kludgey in internal
implementation, but for the user of the package and type it seems a) not
too weird, and b) easily convertible.  If anyone has a better way,
though, I'd sure like to hear about it.  :)
  The problem with coding Ada 83 in the presence of Ada 95 is the pain
caused by seeing a simpler, less error-prone way to do something, while
knowing you must do it the clunky way.  :(

Ada83 equivalents for Ada95

[James A. Squire]

Has anybody gone to the trouble of translating code that uses features
new to Ada95 back into its Ada83 equivalent?  If anyone knows of such
examples that are available on the Internet, please let me know (email
or comp.lang.ada).

Thank you.
--
James Squire
MDA Avionics Tools & Processes
ja_squire@csehp3.mdc.com
"one of these days I'm going to better myself by going to Knight school"
"You'll be a web knight instead of a web page!"

Re: Ada83 equivalents for Ada95

[Robert A Duff]

In article <31927190.35AA@csehp3.mdc.com>,
James A. Squire <m193884@CSEHP3.MDC.COM> wrote:
>Has anybody gone to the trouble of translating code that uses features
>new to Ada95 back into its Ada83 equivalent?  If anyone knows of such
>examples that are available on the Internet, please let me know (email
>or comp.lang.ada).

I don't know of any such effort.

I'm curious as to why you want this.  Do you want to develop Ada 95
code, but then run it on a system where Ada 95 does not yet exist, but
Ada 83 compilers do exist?

It seems pretty difficult, to me.  For example, how do you translate
access-to-subprogram types into Ada 83?  It could be done.  Represent an
access-to-subprogram value as an integer number.  Generate a single
subprogram for each access-type-subtype type.  This subprogram does a
case statement on the integer, and has one branch for every possible
subprogram that might have been called.  The translation tool would have
to do a global analysis of the entire program, to track down every
'Access.  For every subprogram P, if somebody wrote P'Access anywhere in
the program, assign a unique integer to P, and put a call to P in that
case statement mentioned above.

Every such P would have to be declared in the visible part of its
package, even if in the original source, P is declared in the body.

I'm not sure how you would preserve elaboration order, since that
case-statement subprogram will need extra with-clauses to be added.

In any case, such a tool would be approximately as complex as an Ada
compiler -- the same order of magnitude anyway.  If you've gone that
far, why not just write an Ada 95 compiler?

Also, efficiency would suffer.

- Bob

Re: Ada83 equivalents for Ada95

[Joe Gwinn]

We have the Ada83 versus Ada95 problem as well.  In short, we plan to 
use the Ada83 subset of Ada95, so that in some future, we will be able 
to transition to Ada95, should the customer so desire.  This will 
happen, if at all, long after the present project is complete.  We are 
avoiding Ada95 for now, because we couldn't find sufficiently mature 
compilers, tools, runtimes, for our targets, and Ada83 is sufficient.  
If Ada83 is any guide, it will be three years before Ada95 reaches 
anything like Ada83's present level of maturity.  Ada83 took more like 
six years, but started from scratch.

The Ada95 standard defines this Ada83 subset, by what it doesn't mention 
as having been broken by Ada95.

Joe Gwinn

Re: Ada83 equivalents for Ada95

[JP Thornley]

In article: <31A66DAA.7418@sud.ed.ray.com>  Joe Gwinn 
<gwinn@sud.ed.ray.com> writes:
> 
> We have the Ada83 versus Ada95 problem as well.  In short, we plan to 
> use the Ada83 subset of Ada95, so that in some future, we will be able 
> to transition to Ada95, should the customer so desire. 
> [snip]

We are in the same position - but with (almost) total upward 
compatibility it isn't too difficult.

On a different but related topic, has anyone looked at designing Ada 83 
code so that the transition to using the *new* features of Ada 95 is as 
easy as possible.  For example, I can't use child packages in Ada 83, 
but perhaps I can make greater use of sub-packages, so that when the 
switch is made I can easily pull out the children and get the benefit in 
subsequent development (although I'm not sure about how much I would be 
forced to change the 'with's).

Similar ideas for tagged types, protected objects, .... would be very 
valuable for those of us who aren't going to be using Ada 95 for a few 
years yet.

[This may have been asked before, but I missed any replies (if there 
were any).]

Phil Thornley

-- 
------------------------------------------------------------------------
| JP Thornley    EMail jpt@diphi.demon.co.uk                           |
------------------------------------------------------------------------

Re: Ada83 equivalents for Ada95

[Brad Balfour]

In article <31A66DAA.7418@sud.ed.ray.com>, Joe Gwinn
<gwinn@sud.ed.ray.com> wrote:

>We have the Ada83 versus Ada95 problem as well.  In short, we plan to 
>use the Ada83 subset of Ada95, so that in some future, we will be able 
>to transition to Ada95, should the customer so desire.  
[snip]
>The Ada95 standard defines this Ada83 subset, by what it doesn't mention 
>as having been broken by Ada95.
>
>Joe Gwinn

You are looking for a report on the Upward Incompatibilities that Ada 95
has with respect to Ada 83. The short answer is that there are *very few*
upward incompatibilites. The long answer is that an exhaustive document
has been produced by Bill Taylor. It can be found at 
http://sw-eng.falls-church.va.us/AdaIC/docs/compat-guide/compat-guide6-0.txt

There are even tools available that you can purchase which will
automatically detect these problems in your Ada 83 code. May people have
found that just running the code through an Ada 95 compiler, such as GNAT,
will detect many (but not all) problems.

You may also want to reconsider using Ada 95 now. My organization has been
supporting Ada 95 projects for the past 18 months. We've found that Ada 95
can definitely be used for production projects to be delivered to the
field.

Brad
Brad Balfour                         SIGAda WWW Server
CACI, Inc.                              http://www.acm.org/sigada/
703/277-6767                         and also try:
bbalfour@std.caci.com                   http://lglwww.epfl.ch/Ada/
*new address: 3930 Pender Drive * Fairfax, VA 22030 * 703/277-6767**

Re: Ada83 equivalents for Ada95

[Robert A Duff]

In article <bbalfour-2805961408500001@stmac0088.std.caci.com>,
Brad Balfour <bbalfour@std.caci.com> wrote:
>There are even tools available that you can purchase which will
>automatically detect these problems in your Ada 83 code.

Which tools are those?

>... May people have
>found that just running the code through an Ada 95 compiler, such as GNAT,
>will detect many (but not all) problems.

Running the code through an Ada 95 compiler, and then executing it,
ought to find the vast majority of problems, I would think.

And as Brad says, there aren't many problems.  The two *worst* problems
are probably:  (1) There are now 256 Characters, instead of 128, and (2)
generic formals need a (<>) to accept an unconstrained array.

- Bob

Re: Ada83 equivalents for Ada95

[Norman H. Cohen]

In article <Ds4yxt.7wJ@world.std.com>, bobduff@world.std.com
(Robert A Duff) writes: 

|> And as Brad says, there aren't many problems.  The two *worst* problems
|> are probably:  (1) There are now 256 Characters, instead of 128, and (2)
|> generic formals need a (<>) to accept an unconstrained array.

A third problem right up there with those is that package bodies that are
not required are now forbidden: 

   package Data is
      Table: array (1 .. 100 ) of Float;
      ...
   end Data;

   package body Data is -- legal but dangerous in Ada 83, illegal in Ada 95
   begin
      for I in Table'Range loop
         Table(I) := ...;
      end loop;
   end Data;

There are simple tricks that can be used in Ada 83 to make the package
body required, and hence legal.  Here's one that was posted here
recently: 

   package Data is
      Table: array (1 .. 100 ) of Float;
      ...
   private
      type Package_Body_Needed;
   end Data;

   package body Data is -- legal but dangerous in Ada 83, illegal in Ada 95
      type Package_Body_Needed is record null; end record;
   begin
      for I in Table'Range loop
         Table(I) := ...;
      end loop;
   end Data;

(One can also use a dummy subprogram rather than a dummy type.)

This should be done even in Ada-83 code that will never be ported to Ada
95.  Otherwise, it's too easy to accidently build a system that includes
the declaration but not the body of package Data, meaning that
Data.Table never gets initialized.

--
Norman H. Cohen    ncohen@watson.ibm.com

Re: Ada83 equivalents for Ada95

[Ken Garlington]

Robert A Duff wrote:
> 
> In article <bbalfour-2805961408500001@stmac0088.std.caci.com>,
> Brad Balfour <bbalfour@std.caci.com> wrote:
> >There are even tools available that you can purchase which will
> >automatically detect these problems in your Ada 83 code.
> 
> Which tools are those?

Two non-compiler tools promoted at TRI-Ada '95 that look
for Ada 95 incompatibilities in Ada 83 code:

* Ada-ASSURED 3.0 from GrammaTech (info@grammatech.com)

* Ada9X Transition Aid from Software Compositions (swcinfo@aol.com)


-- 
LMTAS - "Our Brand Means Quality"

Re: Ada83 equivalents for Ada95

[Jon S Anthony]

In article <31A66DAA.7418@sud.ed.ray.com> Joe Gwinn <gwinn@sud.ed.ray.com> writes:

> If Ada83 is any guide, it will be three years before Ada95 reaches 
> anything like Ada83's present level of maturity.  Ada83 took more like 

Well, since this is wrong, I guess your guide is poor...

This really is a misleading and confusing statement at best.  I have
been using Ada95 on real stuff for over a year under Gnat.  We are
also looking into the Thompson ObjectAda for NT (though we have not
yet used this, information indicates that it is pretty much "there").

What platform are you using?  If it is VMS, then, yes, I suppose you
will not get something until the VMS port of Gnat that ACT is doing
for Digital is released.  Otherwise, IMO, you are just plain in the
weeds - big time.

/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com

Re: Ada83 equivalents for Ada95

[Ken Garlington]

Jon S Anthony wrote:

> What platform are you using?  If it is VMS, then, yes, I suppose you
> will not get something until the VMS port of Gnat that ACT is doing
> for Digital is released.  Otherwise, IMO, you are just plain in the
> weeds - big time.

Just as a clarification - the Gnat port will be to Alpha/VMS. As far
as I know, VAX/VMS will be "in the weeds" for an indefinite period.

-- 
LMTAS - "Our Brand Means Quality"

Re: Ada83 equivalents for Ada95

[Robert Dewar]

Ken Garlington said

"Just as a clarification - the Gnat port will be to Alpha/VMS. As far
as I know, VAX/VMS will be "in the weeds" for an indefinite period."

This of course depends on customer interest in a VMS port for VAX. There
is a GCC port for VAX/VMS, and once the GNAT port for Alpha/VMS exists,
putting 2+2 together is certainly practical. So if you are interested in
a Vax/VMS port, let us know!

Robert Dewar
President, Ada Core Technologies

Re: Ada83 equivalents for Ada95

[Ken Garlington]

Robert Dewar wrote:
> 
> Ken Garlington said
> 
> "Just as a clarification - the Gnat port will be to Alpha/VMS. As far
> as I know, VAX/VMS will be "in the weeds" for an indefinite period."
> 
> This of course depends on customer interest in a VMS port for VAX. There
> is a GCC port for VAX/VMS, and once the GNAT port for Alpha/VMS exists,
> putting 2+2 together is certainly practical. So if you are interested in
> a Vax/VMS port, let us know!
> 
> Robert Dewar
> President, Ada Core Technologies


That's why I used the term "indefinite period," as opposed to "forever." I assume
sufficient interest will make the "indefinite" period a "definite" one! :)

I also assume that this statement might hold true for other compiler vendors -
sufficient interest in VAX/VMS would cause them to generate a port. However,
I don't think this statement would necessarily be cause to announce multi-vendor
support of VAX/VMS just yet!

-- 
LMTAS - "Our Brand Means Quality"

Re: Ada83 equivalents for Ada95

[Robert Dewar]

iKen Garlington said

"I also assume that this statement might hold true for other compiler vendors -
sufficient interest in VAX/VMS would cause them to generate a port. However,
I don't think this statement would necessarily be cause to announce multi-vendor
support of VAX/VMS just yet!"

The difference in the case of GNAT is that the additional work required is
relatively small, given the effort that will already have been put in to
create the VMS port for Alpha.