[PATCH] D28705: [Sema] Fix bug in handling of designated initializer

[Reid Kleckner via Phabricator via cfe-commits]

rnk added a comment.

What happens with virtual bases?

  struct B { int x; };
  struct D : virtual B { int y; };
  void test() { D d = {1, .y = 2}; }



================
Comment at: lib/Sema/SemaInit.cpp:2250
 
+    if (auto *CXXRD = dyn_cast<CXXRecordDecl>(RT->getDecl()))
+      FieldIndex += CXXRD->getNumBases();
----------------
Do this before counting fields, IMO it's more intuitive.


================
Comment at: test/SemaCXX/designated-initializers-base-class.cpp:1
+// RUN: %clang_cc1 %s -std=c++1z -ast-dump | FileCheck %s
+
----------------
This might be less fragile as an IRGen test. Alternatively, this is a good expected-no-diagnostic test:
  void test() { D d = {1, .y = 2}; }

Right now we emit `warning: initializer overrides prior initialization of this subobject`, which is obviously wrong.


https://reviews.llvm.org/D28705