http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57350
--- Comment #11 from David Krauss <potswa at mac dot com> --- No, that code wasn't right either. I'll just leave it at this and a caveat to the reader: inline void *align( std::size_t alignment, std::size_t size, void *&ptr, std::size_t &space ) { std::uintptr_t pn = reinterpret_cast< std::uintptr_t >( ptr ); std::uintptr_t aligned = ( pn + alignment - 1 ) & - alignment; std::size_t padding = aligned - pn; if ( space < size + padding ) return nullptr; space -= padding; return ptr = reinterpret_cast< void * >( aligned ); }