Re: concept checking for templatize member function?

From: Eric Ford (eford_at_[hidden])
Date: 2001-09-19 22:25:03

> One of the other grad students here at IU (Jeremiah Willcock) just
tried
> adding "inline" (both to function_requires and
> ignore_unused_variable_warnings) and that seemed to do the trick.
Did you
> add "inline" to ignore_unused_variable_warnings?

Strange, I don't beleive that fixed it for me.  But it does eliminate
the code when NDEBUG is defined, so I'll just use that, since that
avoids the comman in macro problem.  Not a big deal.  Thanks.

Another problem...  When I do a BOOST_CLASS_REQUIRES for a concept
that needs it's namespace explicitly mentioned, I get compile errors
(g++ 2.95 and g++ 2.96).  Is this the known compiler problem that
prevents BOOST_CLASS_REQUIRES from getting more frequent use? The
error messages look like... 
DomainInfo.h:84: parse error before `::'
DomainInfo.h:84: parse error before `::'
DomainInfo.h:84: missing ';' before right brace
DomainInfo.h:84: syntax error before `::'

Also, a stylistic question...  Is it considered good style to do
something like template<class T> struct FooConcept { void
constraints() { function_requires< BarConcept<typename T::return_type>
>(); }; };?  Even if nested those several deep?  They can result in
long error messages, but most of the lines are easy to skip over.  If
that's considered ok, then I have a work around to the above problem.

Thanks,
E