rsmith added inline comments.
================ Comment at: lib/CodeGen/CGClass.cpp:2183 ApplyInlineDebugLocation DebugScope(*this, GD); + RunCleanupsScope CleanupScope(*this); ---------------- Please add a test to ensure that we still destroy function parameters in the right order and at the right times (for both the exceptional and non-exceptional cleanup cases). ================ Comment at: test/CodeGenCXX/inheriting-constructor-cleanup.cpp:22 +// CHECK-LABEL: define void @_Z1fv +// CHECK: call void @_ZN1SD2Ev ---------------- rnk wrote: > The landingpad should be trivially dead, since T has nothing that throws in > it, right? Clang is usually pretty smart about not emitting unused > exceptional-only destructors, so I'd try to defend against it getting smart > in the future. I'd put a function call `foo` in `T(int, ...)`, then CHECK for > the invoke of it, and that it unwinds to a landingpad preceding this > destructor call. Maybe remove some of the function bodies so that we can't statically tell there is no possibility of unwind? Repository: rC Clang https://reviews.llvm.org/D44619 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
