[WWW PATCH] [PR14618] Document new access checking performed by g++

["Giovanni Bajo" <giovannibajo at libero dot it>]

Hello,

as suggested by Gaby, this patch documents the new access checking performed by
g++ while binding rvalue to const references. This is strictly standard
conformant, but no compilers in my personal collection enforce this, not even
EDG-based ones. Since we already got two or three reports in a couple of weeks,
I'm adding this to both the common invalid bug list and changes.html.

Validated XHTML 1.0, OK to commit?

Giovanni Bajo


Index: bugs.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/bugs.html,v
retrieving revision 1.82
diff -c -3 -p -r1.82 bugs.html
*** bugs.html 13 Jan 2004 17:36:40 -0000 1.82
--- bugs.html 18 Mar 2004 02:03:36 -0000
*************** a parse error before the character <code
*** 685,690 ****
--- 685,735 ----
  <p>The simplest way to avoid this is to write <code>std::vector&lt;
  ::X&gt;</code>, i.e. place a space between the opening angle bracket
  and the scope operator.</p></dd>
+
+
+ <dt><a name="known_cxx_rvalbind">Copy constructor access check while
+ initializing a reference.</a></dt>
+
+ <dd><p>Consider this code:</p>
+
+ <blockquote><pre>
+ class A
+ {
+ public:
+  A();
+
+ private:
+  A(const A& );   // private copy ctor
+ };
+
+ A makeA(void);
+ void foo(const A& a);
+
+ void bar(void)
+ {
+  foo(A());       // error, copy ctor is not accessible
+  foo(makeA());   // error, copy ctor is not accessible
+
+  A a1 = A();
+  foo(a1);        // OK, a1 is a lvalue
+  A a2(makeA());
+  foo(a2);        // OK, a2 is a lvalue
+ }</pre></blockquote>
+
+ <p>Starting from GCC 3.4.0, binding a rvalue to a const reference requires
+ an accessible copy constructor. This might be surprising at first sight,
+ especially since most popular compilers do not implement correctly this
+ rule.</p>
+
+ <p>The C++ Standard says that a temporary object should be created in
+ this context and its contents filled with a copy of the object we are
+ trying to bind to the reference; it also says that the temporary copy
+ can be elided, but the semantic constraints (eg. accessibility) of the
+ copy constructor still have to be checked.</p>
+
+ <p>For further information, you can consult the following paragraph of
+ the C++ standard: [dcl.init.ref]/5, bullet 2, sub-bullet 1, and
+ [class.temporary]/2.</p></dd>
  </dl>

  <h3><a name="upgrading">Common problems when upgrading the compiler</a></h3>
Index: gcc-3.4/changes.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-3.4/changes.html,v
retrieving revision 1.104
diff -c -3 -p -r1.104 changes.html
*** gcc-3.4/changes.html 15 Mar 2004 08:19:12 -0000 1.104
--- gcc-3.4/changes.html 18 Mar 2004 02:03:37 -0000
***************
*** 534,539 ****
--- 534,571 ----
      <pre>
      int* a = new (int)[10];    // error, not accepted anymore
      int* a = new int[10];      // OK</pre></li>
+
+     <li>When binding a rvalue of class type to a reference, the copy
+  constructor of the class must be accessible. For instance, consider
+  the following code:
+  <pre>
+  class A
+  {
+  public:
+    A();
+
+  private:
+    A(const A& );   // private copy ctor
+  };
+
+  A makeA(void);
+  void foo(const A& a);
+
+  void bar(void)
+  {
+    foo(A());      // error, copy ctor is not accessible
+    foo(makeA());  // error, copy ctor is not accessible
+
+    A a1 = A();
+    foo(a1);        // OK, a1 is a lvalue
+    A a2(makeA());
+    foo(a2);        // OK, a2 is a lvalue
+  }</pre>
+
+  This might be surprising at first sight, especially since most
+  popular compilers do not implement correctly this rule. Please,
+  consult <a href="http://gcc.gnu.org/bugs.html#known_cxx_rvalbind";>
+  this page</a> for a more exhaustive explanation of this non-bug.</li>
    </ul>

  <h3>Objective-C</h3>