hook concept checking into compiler checking? (was Re: feb-2002 libstdc++ outstanding)

[Phil Edwards]

For what it's worth:

On Tue, Mar 12, 2002 at 11:44:58AM -0500, Phil Edwards wrote:
> On Mon, Mar 11, 2002 at 07:03:24PM -0800, Benjamin Kosnik wrote:
> > 
> > > What if we make --enable-concept-checking pay attention to the top-level's
> > > --enable-checking option?
> > 
> > Hmmmmmmm....... sounds like a promising idea. First though, are 
> > testresults identical for --enable-concept-checking and 
> > --disable-concept-checking?
> 
> No, these two fail with c-c turned on:
> 
>     23_containers/list_modifiers.cc
>     23_containers/map_operators.cc
> 
> (although for some reason only the first is correctly reported as having
> failed to build an executable).
> 
> The first uses a 'class T' with a non-static const data member; this flunks
> the Assignable test.  The second uses a map instantiation (MapInt in test02)
> of a type (const int) which flunks the Assignable test.
> 
> The first would probably be solved by the proposed "testsuite library" and
> a counting class; I think this was in you list of January outstanding items.

I've now done this in my local tree.  The Assignable problem in itself
just involved tweaking the constness of Stephen's "T" class, but I've gone
ahead and moved the tweaked class into the testsuite library.  A patch
will be forthcoming after 3.1 ships.

> I don't think the second case is valid.  It's a useful idea (mapping to
> a const), but I don't think it meets the requirements.

This test still fails Assignable, of course (see reason above).


Phil

-- 
If ye love wealth greater than liberty, the tranquility of servitude greater
than the animating contest for freedom, go home and leave us in peace.  We seek
not your counsel, nor your arms.  Crouch down and lick the hand that feeds you;
and may posterity forget that ye were our countrymen.            - Samuel Adams