Deprecated (legacy) string literal conversion to 'char *' causes strange overloading resolution

[• a.makarov]

Summary

It's a patch for PR28050. Seems like overloading resolution wipes out the first standard conversion sequence (before user-defined conversion) in case of deprecated string literal conversion.

[• a.makarov]

Ping.

[• a.makarov]

Reid, please, take a look.

[• a.makarov]

Ping.

[• a.makarov]

Friendly ping!

[rsmith]

LGTM

[• a.makarov]

Richard, thank you very much! =)

[DmitryPolukhin]

Committed as rL275970

include/clang/Sema/Overload.h

union {
       AmbiguousConversionSequence Ambiguous;
 
       /// When ConversionKind == BadConversion, provides the details
       /// of the bad conversion.
       BadConversionSequence Bad;
     };
 
     ImplicitConversionSequence()
-      : ConversionKind(Uninitialized), StdInitializerListElement(false)
-    {}
+        : ConversionKind(Uninitialized), StdInitializerListElement(false) {
+      Standard.First = ICK_Identity;
+      Standard.Second = ICK_Identity;
+      Standard.Third = ICK_Identity;
+    }
     ~ImplicitConversionSequence() {
       destruct();
     }
     ImplicitConversionSequence(const ImplicitConversionSequence &Other)
       : ConversionKind(Other.ConversionKind),
         StdInitializerListElement(Other.StdInitializerListElement)
     {
       switch (ConversionKind) {

lib/Sema/SemaOverload.cpp

TryUserDefinedConversion(Sema &S, Expr *From, QualType ToType,
   OverloadCandidateSet Conversions(From->getExprLoc(),
                                    OverloadCandidateSet::CSK_Normal);
   switch (IsUserDefinedConversion(S, From, ToType, ICS.UserDefined,
                                   Conversions, AllowExplicit,
                                   AllowObjCConversionOnExplicit)) {
   case OR_Success:
   case OR_Deleted:
     ICS.setUserDefined();
-    ICS.UserDefined.Before.setAsIdentityConversion();
     // C++ [over.ics.user]p4:
     //   A conversion of an expression of class type to the same class
     //   type is given Exact Match rank, and a conversion of an
     //   expression of class type to a base class of that type is
     //   given Conversion rank, in spite of the fact that a copy
     //   constructor (i.e., a user-defined conversion function) is
     //   called for those cases.
     if (CXXConstructorDecl *Constructor
if (ICS.isStandard()) {
     if (DeclType->isRValueReferenceType() && LValRefType) {
       // FIXME: This is the wrong BadConversionSequence. The problem is binding
       // an rvalue reference to a (non-function) lvalue, not binding an lvalue
       // reference to an rvalue!
       ICS.setBad(BadConversionSequence::lvalue_ref_to_rvalue, Init, DeclType);
       return ICS;
     }
 
-    ICS.UserDefined.Before.setAsIdentityConversion();
     ICS.UserDefined.After.ReferenceBinding = true;
     ICS.UserDefined.After.IsLvalueReference = !isRValRef;
     ICS.UserDefined.After.BindsToFunctionLvalue = false;
     ICS.UserDefined.After.BindsToRvalue = !LValRefType;
     ICS.UserDefined.After.BindsImplicitObjectArgumentWithoutRefQualifier = false;
     ICS.UserDefined.After.ObjCLifetimeConversionBinding = false;
   }
 

test/SemaCXX/pr28050.cpp

+// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc -std=c++11 %s -fsyntax-only
+//
+// expected-no-diagnostics
+
+class A {
+public:
+  A(char *s) {}
+  A(A &&) = delete;
+};
+
+int main() { A a("OK"); }