Erik <esigra@xxxxxxxxx> writes: > Ian Lance Taylor skrev: >> Erik <esigra@xxxxxxxxx> writes: >> >> >>> Bjarne Stroustrup wrote in his book The C++ programming language that; >>> while a dynamic_cast to a pointer type is a question (the result should >>> be checked), a dynamic_cast to a reference type is like an assertion. We >>> have followed this advice in our project. Now we need to disable these >>> assertions in release builds, so that dynamic_cast<Derived &>(base) is >>> just as efficient as static_cast<Derived &>(base). But I do not know how >>> to tell this to the compiler. I only found some other checks in gcc, >>> that can be disabled for release builds, such as the Ada checks (-gnatp). >>> >> >> -Ddynamic_cast=static_cast ? >> > That would of course not work, because it would affect all > dynamic_casts, not just those to references. > > >>> Or is the feature that I need missing in gcc? Maybe I have to use one of >>> the many hacks called down_cast, safe_cast and so on, that can be found >>> on the web? Like this one: >>> http://smolsky.net/index.php/2009/09/14/down_cast-v2 >>> If so, which of those hacks is the correct one that should be used? >>> >> >> Since you don't care about dynamic type safety you can just use >> static_cast. >> > > That would certainly not work, because then the assertions in the form > of dynamic_casts to reference types would not be checked in debug builds. #ifdef DEBUG_MODE template<typename T1, typename T2> T1& dynamic_reference_cast(T2& r) { return dynamic_cast<T1&>(r); } #else template<typename T1, typename T2> T1& dynamic_reference_cast(T2& r) { return static_cast<T1&>(r); } #endif Ian
