philnik777 wrote: > I originally modified type_info.comparison.unmerged.sh.cpp to build a shared > library which I don't think is feasible for static configurations: > generic-static, generic-llvm-libc, android, picolibc, windows. Instead I now > simulate unmerged RTTI instead: where I construct two type_info objects whose > names are equal strings at different addresses. This does the same comparison > code.
We _really_ shouldn't use the constructor for `type_info`. It's undocumented, so not considered an extension by libc++ and should be removed instead (as I do in #235739). I think we should simply disable the test for configurations where it doesn't work. If we can't generate two different typenames in a normal way there is no point in testing it. https://github.com/llvm/llvm-project/pull/220864 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
