Re: AI-344, generics/instances...

["Randy Brukardt" <randy@rrsoftware.com>]

"Martin Dowie" <martin.dowie@baesystems.com> wrote in message
news:41498c90$1_1@baen1673807.greenlnk.net...
> Just a thought that popped into my head and I wanted a 'sanity' check on
if
> first...
>
> With AI-344 generic packages that define a tagged type no longer have to
be
> instantiated at library level. I was wondering if it would be useful to
> define a mechanism (pragma?) that allowed an author to insist that a
generic
> package be instantiated at library level.

We tried that (to solve a different problem), and it led to no end of
troubles. (Mostly contract problems in generic bodies.) We gave up on it in
favor of defining Restrictions of the packages in question. See older
versions of AI-303 for details. But we weren't really trying to do that for
user code.

Probably the best way to do that for user code would be to define a
Restriction similar to No_Dependence. But that probably would be
controversial; we ended up limiting No_Dependence to only language-defined
units because of concerns with supporting user-defined units. (And in any
case it is too late for the standard; the scope of the Amendment is set, and
the hope is to approve all of the amendments at the next meeting -- which
starts tomorrow.)

                    Randy.