Hi all, I'm currently using large parts of booos::mpl and boost::fusion in a large project. Alas, I keep triggering some unwanted ADL betwenn some of my code and mpl code. Here is a simplified but significant code that showcases the problem : #include <iostream> #include <boost/mpl/vector.hpp> using namespace std; namespace bar { namespace foo { template<class T, class LST> struct some_class { int size() const { return sizeof(T); } }; template<class T,class LST> static inline int size( const some_class<T,LST>& src ) { return src.size(); } } using foo::some_class; using foo::size; } using namespace bar; int main() { some_class<int,int> f; cout << size(f) << endl; some_class<int, boost::mpl::vector<int> > g; cout << size(g) << endl; } This doesn't compile and outputs the following error (using mingw 4.1) : In function 'int main()':

error: 'template<class Sequence> struct boost::mpl::size' is not a function error: conflict with 'template<class T, class LST> int bar::foo::size(const bar::foo::some_class<T, LST>&)' error: in call to 'size'

What can I do to prevent this ADL while not forcing the client code to explicitely write bar::size(g) ? Thanks in advance -- Joel FALCOU Research Engineer @ Institut d'Electronique Fondamentale Université PARIS SUD XI France

Steven Watanabe a écrit :

To my knowledge there is no way to block ADL without changing MPL. (I assume that you want ADL for your size function)

Foudn it the hard way. I think MPL just need to use its ADL barrier around meta-functions too to prevent such behavior. Maybe it's worth some bug report or features request. Anyway, I found a decent workaround to prevent the passing of mpl types in my templates argument list. So problem solved. Thanks again -- Joel FALCOU Research Engineer @ Institut d'Electronique Fondamentale Université PARIS SUD XI France