:::: www.foddex.net ::::

Contact me

Stuff

BLOG: C++ syntactical weirdness

posted Sat 07-02-2009 17:06:47, in the c++ () category

I thought I had seen it all. Syntactical weirdness in C++ like templated friend definitions, template template parameters, method pointer definitions, but no... there was one more and it dazzled me all day.

I had the following code (simplified for the example), and it worked just fine in Visual Studio 2008, but it barfed in gcc:

template<typename _S, typename _T>
struct foo {
        static _S& bar( _S& s, const _T& t ) {
                t.test<_S>( s );
                return s;
        }
};

The error gcc spewed was:

a.cpp: In static member function 'static _S& foo<_S, _T>::bar(_S&, const _T&)':
a.cpp:4: error: expected primary-expression before '>' token

But why? It shouldn't be complaining about the templated method call on t, as it has no idea what t is, right? It complains during template compilation, not during instantiation (as there is none in that snippet of code!) Visual Studio didn't complain, and for once I thought it might be right....

But ofcourse it wasn't ;) There is actually a way to satisfy a compiler like gcc in this particular scenario. For you C++ coders out there, do you know what it is? :-)

I'll put the answer in here but invisible. To see it, select the black box below.

template<typename _S, typename _T>
struct foo {
        static _S& bar( _S& s, const _T& t ) {
                t.template test<_S>( s );
                return s;
        }
};

-- Foddex

[ Back to blog listing ]