From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,99ab4bb580fc34cd
X-Google-Attributes: gid103376,public
From: bobduff@world.std.com (Robert A Duff)
Subject: Re: Q: access to subprogram
Date: 1996/07/26
Message-ID: <Dv4M4o.6GH@world.std.com>#1/1
X-Deja-AN: 170582774
references: <4rb9dp$qe6@news1.delphi.com> <4t4siv$bh2@goanna.cs.rmit.edu.au>
 <ROGOFF.96Jul24161355@sccm.Stanford.EDU> <4t7dvt$cbo@mulga.cs.mu.OZ.AU>
organization: The World Public Access UNIX, Brookline, MA
newsgroups: comp.lang.ada
Date: 1996-07-26T00:00:00+00:00
List-Id: <comp.lang.ada>


In article <4t7dvt$cbo@mulga.cs.mu.OZ.AU>,
Fergus Henderson <fjh@mundook.cs.mu.OZ.AU> wrote:
>>However, a wrapper cannot be built for 'Accesses in the generic body
>>inside the generic body. 
>
>Can you explain in a bit more detail when this would be necessary,
>and why it is not possible?

It would be necessary when you say P'Access inside a generic body, and P
is in that body (and not declared in the spec), and you then let the
pointer escape outside the generic.  This is illegal in Ada 95.

I believe we're talking about full generic sharing here.  Most
implemtnations of sharing share *sometimes* and not other times,
depending on what sort of generic it is, and what the formals are.  But
here, we *always* share.  Well, when you call through the pointer, the
called subprogram needs a pointer to the generic data -- an area of
memory where variables declared in the generic package body live.
There's one such are per instance.  But since you have only one copy of
the generic code, that code doesn't know which instance is the current
one.

You could generate code to do this at run time, as gcc does, but Randy
said that doesn't work on some OS's.  Generating code to do it at
compile time is impossible, since there's only one piece of code, but
many instances.

If you *always* represent access-to-subprogram values with extra data
telling where the generic instance data is, then it can work, but that's
distributed overhead for the case where you didn't use generics.

Make sense?

I know nothing of Mercury, in particular how its version of generics
relates to Ada's.  Could you outline how you did it, and why you say it
has no distributed overhead?

- Bob