[clang] Add test for CWG418

[Endill]

Summary

P1787: CWG418 is resolved by trivial rephrasing along with that necessary for the new using-declaration interpretation.
Wording: see changes to [dcl.fct.default]/9 and [over.match.best]/4.

[over.match.best]/4 includes an example that is not properly diagnosed by Clang.

[cor3ntin]

I'm not sure how I feel about this.
Clang does not, implement correctly that paragraph at all, so I think the best course here is to create an issue on the llvm github and mark the dr resolved as part of fixing the bug in clang.
The fact the previous wording was unclear may be related to clamg non-conformance.

[shafik]

I think the relevant change from p1787 is to dcl.fct.default p4, specifically:

... Declarations that inhabit different scopes have completely distinct sets of default arguments ...

[shafik]

In D139429#3974491, @cor3ntin wrote:

I'm not sure how I feel about this.
Clang does not, implement correctly that paragraph at all, so I think the best course here is to create an issue on the llvm github and mark the dr resolved as part of fixing the bug in clang.
The fact the previous wording was unclear may be related to clamg non-conformance.

I agree, I don't believe we are conforming and there should be a bug report.

[Endill]

I think the relevant change from p1787 is to dcl.fct.default p4, specifically:

... Declarations that inhabit different scopes have completely distinct sets of default arguments ...

It previously said "Declarations in different scopes". To me it's no more that adjusting the wording to the new language of lookup, keeping the intent intact. Am I missing more serious consequences of this change?

[shafik]

In D139429#3974745, @Endill wrote:

I think the relevant change from p1787 is to dcl.fct.default p4, specifically:

... Declarations that inhabit different scopes have completely distinct sets of default arguments ...

It previously said "Declarations in different scopes". To me it's no more that adjusting the wording to the new language of lookup, keeping the intent intact. Am I missing more serious consequences of this change?

I think your correct on this.

[Endill]

I opened #59363 on bug tracker.
What should I do about this patch then? I believe availability no should come with a failing test. Should I include example from [over.match.best]/4 as another example in the test?

[shafik]

In D139429#3975028, @Endill wrote:

I opened #59363 on bug tracker.
What should I do about this patch then? I believe availability no should come with a failing test. Should I include example from [over.match.best]/4 as another example in the test?

Was chatting with @aaron.ballman and he recommends adding a FIXME comment along the lines of:

// FIXME: this shouldn't pass, it should fail!

[Endill]

@shafik does this imply that example from [over.match.best]/4 should be included in this patch?

[shafik]

In D139429#3975223, @Endill wrote:

@shafik does this imply that example from [over.match.best]/4 should be included in this patch?

Yes, I believe we should. We are only conforming if we get all of the cases correct.

[Endill]

Add example from [over.match.best]/4, and mark CWG418 as not available

[shafik]

LGTM, thank you for all your efforts.

[Endill]

Thanks to all of you for your time to review this.

clang/test/CXX/drs/dr4xx.cpp

namespace dr417 { // dr417: no
   struct G;
   using N::H;
   namespace M {
     struct dr417::G {}; // expected-error {{namespace 'M' does not enclose}}
     struct dr417::H {}; // expected-error {{namespace 'M' does not enclose}}
   }
 }
 
+namespace dr418 { // dr418: no
+namespace example1 {
+void f1(int, int = 0);
+void f1(int = 0, int);
+
+void g() { f1(); }
+} // namespace example1
+
+namespace example2 {
+namespace A {
+void f2(int); // #dr418-f2-decl
+}
+namespace B {
+using A::f2;
+}
+namespace A {
+void f2(int = 3);
+}
+void g2() {
+  using B::f2;
+  f2(); // expected-error {{no matching function}}
+  // expected-note@#dr418-f2-decl {{requires 1 argument}}
+}
+} // namespace example2
+
+// example from [over.match.best]/4
+namespace example3 {
+namespace A {
+extern "C" void f(int = 5);
+}
+namespace B {
+extern "C" void f(int = 5);
+}
+
+using A::f;
+using B::f;
+
+void use() {
+  f(3);
+  f(); // FIXME: this should fail
+}
+} // namespace example3
+} // namespace dr418
+
 namespace dr420 { // dr420: yes
   template<typename T> struct ptr {
     T *operator->() const;
     T &operator*() const;
   };
   template<typename T, typename P> void test(P p) {
     p->~T();
     p->T::~T();

clang/www/cxx_dr_status.html

<tr id="417">
     <td>CD1</td>
     <td>Using derived-class qualified name in out-of-class nested class definition</td>
     <td class="none" align="center">No</td>
   </tr>
   <tr id="418">
     <td><a href="https://wg21.link/cwg418">418</a></td>
     <td>CD6</td>
     <td>Imperfect wording on error on multiple default arguments on a called function</td>
-    <td class="none" align="center">Unknown</td>
+    <td class="none" align="center">No</td>
   </tr>
   <tr class="open" id="419">
     <td><a href="https://wg21.link/cwg419">419</a></td>
     <td>open</td>
     <td>Can cast to virtual base class be done on partially-constructed object?</td>
     <td align="center">Not resolved</td>
   </tr>
   <tr id="420">