Re: Stupid question

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

"Markus E Leypold"
<development-2006-8ecbb5cc8aREMOVETHIS@ANDTHATm-e-leypold.de> wrote in
message news:y7zm5chxxf.fsf@hod.lan.m-e-leypold.de...
>
> "Randy Brukardt" <randy@rrsoftware.com> writes:
>
> > "Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> wrote in message
> > news:190oqb9t3eplx$.ffakhbeqreff$.dlg@40tude.net...
> >> On 13 Apr 2007 08:35:53 -0700, Adam Beneschan wrote:
> > ...
> >> This does not outlaw A.B, it just cannot. A trivial counter example is:
> >>
> >> with A.B;  -- This is 100% legal
> >>
> >> With clause is defined in 10.1.2(4) as "with <name>". Here <name> is
the
> >> same as in 12.7. It refers to 4.1 and is resolved to
<selected_component>.
> >>
> >> Any language lawyers here?
> >
> > Sorry, Adam is right. A generic package is not a package (that's a
standing
> > ARG joke, not one of the RM's finest hours of terminology). Context
clauses
> > have their own, unique visibility rules defined in 10.1.6; you can't use
> > them as a counterexample for anything. There is a very complex rule that
> > allows you to say A.B there, because it normally would not be.
> > Interestingly, the Ada 95 standard did not allow naming a grandchild
that
> > way (that is, A.B.C), which demonstrates how complex it is.
>
> So it is possible to say 'with A.B' in the context of a compilation
> unit if A is "only" a generic (as opposed to an instance)? Even in Ada
> 95? Is there any example for that somewhere?

It's necessary to make children of generic packages work. Recall that a
child of a generic must be a generic.  The original question included an
example. (Personally, I think generic children aren't worth the trouble,
both in the language definition and in implementation. YMMV.)

                                Randy.