https://gcc.gnu.org/bugzilla/show_bug.cgi?id=43105
--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> --- (In reply to Jeffrey Yasskin from comment #0) > programs don't link when a -frtti class is derived from a -fno-rtti base > class. It's more complicated than that. The ABI defines when RTTI is emitted: https://itanium-cxx-abi.github.io/cxx-abi/abi.html#vague-rtti There's only a problem when compiling a key function with -fno-rtti, and other translation units compiled with -frtti have no way to know how the key function was compiled. For example, given: struct A { A(); ~A(); }; struct B { B(); virtual ~B() { } }; struct C { C(); virtual ~C(); }; A and B can always be derived from unconditionally. If C::~C() is compiled with -fno-rtti then classes that derive from C must also use -fno-rtti to compile their key function (and if they don't have a key function, then all uses of C that need a definition must be compiled with -fno-rtti). I don't think this can be fixed. Documenting it is non-trivial. I went to see what Clang does, and I can't even find -frtti / -fno-rtti in the manual, so we're doing better than that!
