On 07/17/15 15:59, Nathan Sidwell wrote:
On 07/17/15 15:42, Jason Merrill wrote:
On 07/08/2015 10:50 AM, Nathan Sidwell wrote:
On 06/30/15 19:21, Nathan Sidwell wrote:
On 06/30/15 00:19, Jason Merrill wrote:
On 06/29/2015 06:57 PM, Nathan Sidwell wrote:
    * method.c (synthesized_method_walk): Skip virtual bases of
    abstract classes in C++14 mode.

Let's not limit this to C++14 mode; most DRs apply to earlier
standards as well.

curiously opening it up leads to some test failures related to
determining the exception specifier for implicit ctors and dtors.  Not
had time to investigate that yet ...

If C++98 mode is problematic, we can limit this to C++11 and up.

I'm not yet sure.  The failure mode I saw surprised me, and suggests there's
something wrong with the patch.  Sadly, I've got interrupted by other stuff.

Ok, this patch fixes things up. The previous version was a little too lax, extending the logic of DR1611 to all synthesized functions. However, this broke virtual synthesized dtors, in that an abstract class's synthesized dtor's exception specification would not take account of any virtual base dtor exception specs. This would mean that a non-abstract derived class's synthesized dtor might end up with a throwing exception spec (because the virtual base's dtor did), and that would be looser than the exception spec on the abstract base's non-callable synthesized dtor. And that fails the virtual overriding checks.

So this restricts the skipping to exactly what DR 1611 discusses -- default constructors of virtual bases of an abstract class. Those can never be virtual, so we don't end up with the above problem in that case.

Jason, WDYT?

nathan
2015-08-01  Nathan Sidwell  <nat...@acm.org>

	cp/
	PR c++/66443
	* init.c (emit_mem_initializers): Do not emit initializers for
	virtual bases of abstract classes.
	* method.c (synthesized_method_walk): Skip virtual bases of
	abstract classes in C++14 mode.

	testsuite/
	PR c++/66443
	* cpp1y/pr66443.C: New test.

Index: cp/init.c
===================================================================
--- cp/init.c	(revision 226444)
+++ cp/init.c	(working copy)
@@ -1140,9 +1140,7 @@ emit_mem_initializers (tree mem_inits)
 	}
 
       /* Initialize the base.  */
-      if (BINFO_VIRTUAL_P (subobject))
-	construct_virtual_base (subobject, arguments);
-      else
+      if (!BINFO_VIRTUAL_P (subobject))
 	{
 	  tree base_addr;
 
@@ -1156,6 +1154,8 @@ emit_mem_initializers (tree mem_inits)
                               tf_warning_or_error);
 	  expand_cleanup_for_base (subobject, NULL_TREE);
 	}
+      else if (!ABSTRACT_CLASS_TYPE_P (current_class_type))
+	construct_virtual_base (subobject, arguments);
     }
   in_base_initializer = 0;
 
Index: cp/method.c
===================================================================
--- cp/method.c	(revision 226444)
+++ cp/method.c	(working copy)
@@ -1506,7 +1506,13 @@ synthesized_method_walk (tree ctype, spe
   vbases = CLASSTYPE_VBASECLASSES (ctype);
   if (vec_safe_is_empty (vbases))
     /* No virtual bases to worry about.  */;
-  else if (!assign_p)
+  else if (!assign_p
+	   /* DR 1611 ignores virtual bases of abstract classes for
+	      the default constructor.  Such ctors can never be
+	      directly called.  Thus their deletion should not affect
+	      whether they are deleted in this class.   */
+	   && (!ABSTRACT_CLASS_TYPE_P (ctype) || sfk != sfk_constructor
+	       || inherited_parms))
     {
       if (constexpr_p)
 	*constexpr_p = false;
Index: testsuite/g++.dg/other/pr66443.C
===================================================================
--- testsuite/g++.dg/other/pr66443.C	(revision 0)
+++ testsuite/g++.dg/other/pr66443.C	(working copy)
@@ -0,0 +1,21 @@
+// { dg-do compile }
+
+class A {
+public:
+    A( int ) {  }
+};
+
+// B's virtual base is ignored for default ctor determination as B is
+// abstract.  DR1611 &  DR1658
+
+class B: virtual public A {
+public:
+    virtual void do_something() = 0;
+};
+
+class C: public B {
+public:
+    C(): A( 1 ) {  }
+    virtual void do_something() {  }
+};
+

Reply via email to