Yes, I believe that explicitly qualifying unnamed and named rvo should be the norm and would clarify a lot of confusion on these topics. I guess it makes sense

Yes, I believe that explicitly qualifying unnamed and named rvo should be the norm and would clarify a lot of confusion on these topics. I guess it makes sense that an optimization that GCC is not able to do at all should not be part of the standard ; I'm pretty sure it should be possible to write cases that clang wouldn't be able to optimize either.