[clang] [clang][ExprConst] Fix crash on uninitialized array subobject (PR #67817)

[Takuya Shimizu via cfe-commits]

================
@@ -2411,10 +2411,15 @@ static bool CheckEvaluationResult(CheckEvaluationResultKind CERK,
                                   const FieldDecl *SubobjectDecl,
                                   CheckedTemporaries &CheckedTemps) {
   if (!Value.hasValue()) {
-    assert(SubobjectDecl && "SubobjectDecl shall be non-null");
-    Info.FFDiag(DiagLoc, diag::note_constexpr_uninitialized) << SubobjectDecl;
-    Info.Note(SubobjectDecl->getLocation(),
-              diag::note_constexpr_subobject_declared_here);
+    if (SubobjectDecl) {
+      Info.FFDiag(DiagLoc, diag::note_constexpr_uninitialized)
+          << true << SubobjectDecl;
+      Info.Note(SubobjectDecl->getLocation(),
+                diag::note_constexpr_subobject_declared_here);
+    } else {
+      // FIXME: We should add a test to check the output of this case.
+      Info.FFDiag(DiagLoc, diag::note_constexpr_uninitialized) << false << Type;
----------------
hazohelet wrote:

Finally I found a test case that can check the fallback diagnostic message: https://godbolt.org/z/Mc3nPq15n
Clang 16's output is a little weird in that it does not mention `new *char[3][1]` part, and this patch's fallback approach inherits this behavior.

https://github.com/llvm/llvm-project/pull/67817