NoQ added a comment. > Add a FIXME test case that demonstrates that automatic destructors don't fire > after lifetime extension through a POD field, even though lifetime extension > itself seems to work correctly.
This has always been broken - the CFG element for the automatic destructor is simply missing in this case. But previously we didn't inline the constructor, so the whole thing was invalidated anyway. I wonder how Sema lives without it. Now the interesting part here is https://reviews.llvm.org/rC288563 which seems to have removed the need to `skipRValueSubobjectAdjustments()` //three days after// i added that code in https://reviews.llvm.org/rC288263 ¯\_(ツ)_/¯ https://reviews.llvm.org/D43497 _______________________________________________ cfe-commits mailing list firstname.lastname@example.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits