Hi,

this issue is by and large a duplicate of C++/79790, which I already fixed. There is a minor remaining nit: for the testcase, after the correct:

    error: cannot deduce template arguments of ‘C<S>’, as it has no viable deduction guides

we also emit the meaningless:

    error: too many initializers for ‘<type error>’

only because in finish_compound_literal we don't check - as we do in most other places - the return value of do_auto_deduction for error_mark_node and it filters through until reshape_init. Tested x86_64-linux.

Thanks, Paolo.

////////////////////

/cp
2017-10-23  Paolo Carlini  <paolo.carl...@oracle.com>

        PR c++/80449
        * semantics.c (finish_compound_literal): Check do_auto_deduction
        return value for error_mark_node.

/testsuite
2017-10-23  Paolo Carlini  <paolo.carl...@oracle.com>

        PR c++/80449
        * g++.dg/cpp1z/class-deduction46.C: New.
Index: cp/semantics.c
===================================================================
--- cp/semantics.c      (revision 254005)
+++ cp/semantics.c      (working copy)
@@ -2711,8 +2711,12 @@ finish_compound_literal (tree type, tree compound_
 
   if (tree anode = type_uses_auto (type))
     if (CLASS_PLACEHOLDER_TEMPLATE (anode))
-      type = do_auto_deduction (type, compound_literal, anode, complain,
-                               adc_variable_type);
+      {
+       type = do_auto_deduction (type, compound_literal, anode, complain,
+                                 adc_variable_type);
+       if (type == error_mark_node)
+         return error_mark_node;
+      }
 
   if (processing_template_decl)
     {
Index: testsuite/g++.dg/cpp1z/class-deduction46.C
===================================================================
--- testsuite/g++.dg/cpp1z/class-deduction46.C  (nonexistent)
+++ testsuite/g++.dg/cpp1z/class-deduction46.C  (working copy)
@@ -0,0 +1,6 @@
+// PR c++/80449
+// { dg-options -std=c++17 }
+
+template<class S> struct C;
+template<> struct C<int> { C(int, int) {} };
+auto k = C{0, 0};  // { dg-error "cannot deduce" }

Reply via email to