Re: [v3] concept check macros, excess-errors

[Phil Edwards <pedwards at disaster dot jaj dot com>]

On Thu, Apr 19, 2001 at 04:58:19PM -0700, Benjamin Kosnik wrote:
> > ...when the user configures with --enable-concept-checks, we still need a
> > way to turn them off while building the library itself.  I have no problem
> > with not doing this in the makefiles, but we should decide on how.
> 
> Why? I am aghast at the idea that the library won't build itself with 
> concep checking turned on. This seems wrong.

Eeek, I didn't mean to imply that.  Let me clarify.  The library can
be built with no problem -- we're good to go as far as our own use
of library components goes, we're meeting the requirements -- but the
-fno-implicit-templates flag causes the glibcpp_function_requires<>()
calls to not be implicitly instantiated (see comment in src/Makefile.am
(actually the comment isn't that helpful, never mind)).

Thus libstdc++.so has a boatload of undefined references to various
instantiations of that function.  (At one point I thought it wasn't actually
being called, but that was for the old implementation.)

So, "hello world" fails to link because
__glibcpp_function_requires<YourFavoriteConcept>() isn't found, etc.


Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.