diff mbox

PING: [patch] fix c++/33558 - references cannot be mutable

Message ID AANLkTikfecsvMadZsks_5RXyS+exkJmTTQhNNqMXBQRk@mail.gmail.com
State New
Headers show

Commit Message

Jonathan Wakely Jan. 13, 2011, 8:47 p.m. UTC 
On 1 January 2011 22:34, Jonathan Wakely wrote:
> On 22 December 2010 14:53, Jonathan Wakely  wrote:
>> This is a very small patch to fix a standard conformance issue.
>>
>> I've made it a permerror so broken code can use -fpermissive and
>> doesn't need to be fixed.
>>
>> I'd really like to get this in 4.6, so pinging for review
>
> ... again

I'd given up and was going to wait for 4.7, but I have been encouraged
to try to get this reviewed again.

cp/ChangeLog entry:

2010-12-18  Giovanni Funchal  <gafunchal@gmail.com>
           Jonathan Wakely  <jwakely.gcc@gmail.com>

       PR c++/33558
       * decl.c (grokdeclarator): Reject mutable reference members.

testsuite/ChangeLog entry:

2010-12-18  Giovanni Funchal  <gafunchal@gmail.com>
           Jonathan Wakely  <jwakely.gcc@gmail.com>

       PR c++/33558
       * testsuite/g++.dg/other/pr33558.C: New.
       * testsuite/g++.dg/other/pr33558-2.C: New.

Tested x86_64-linux, ok for trunk?


>> On 18 December 2010 18:57, Jonathan Wakely wrote:
>>> This is a slightly modified version of the patch attached to PR33558
>>> which was posted to gcc-patches but apparently never reviewed.  I
>>> don't know if the author of the original patch, Giovanni, has a
>>> copyright assignment but I think the change is obvious and I hope is
>>> sufficiently small to not need an assignment. I was in the process of
>>> making the same change when I found his patch on the PR and copied the
>>> wording of his diagnostic and testcase.
>>>
>>> Sun CC and g++ both incorrectly accept mutable references, so it might
>>> not be uncommon in the wild (I found a use in some production code
>>> yesterday.)  This version of the patch allows mutable on reference
>>> members when -fpermissive is used, giving old code a transition path.
>>> I'll add a note to changes.html if this is approved.
>>>
>>> tested x86_64-linux, OK for trunk?
>>>
>>>
>>> cp/ChangeLog entry:
>>>
>>> 2010-12-18  Giovanni Funchal  <gafunchal@gmail.com>
>>>            Jonathan Wakely  <jwakely.gcc@gmail.com>
>>>
>>>        PR c++/33558
>>>        * decl.c (grokdeclarator): Reject mutable reference members.
>>>
>>> testsuite/ChangeLog entry:
>>>
>>> 2010-12-18  Giovanni Funchal  <gafunchal@gmail.com>
>>>            Jonathan Wakely  <jwakely.gcc@gmail.com>
>>>
>>>        PR c++/33558
>>>        * testsuite/g++.dg/other/pr33558.C: New.
>>>        * testsuite/g++.dg/other/pr33558-2.C: New.
>>>
>>
>

Comments

Jason Merrill Jan. 14, 2011, 1:24 a.m. UTC | #1 
OK, sorry for the delay.

Jason
diff mbox

Patch

Index: cp/decl.c
===================================================================
--- cp/decl.c	(revision 167817)
+++ cp/decl.c	(working copy)
@@ -9210,6 +9210,12 @@  grokdeclarator (const cp_declarator *dec
 	  error ("const %qs cannot be declared %<mutable%>", name);
 	  storage_class = sc_none;
 	}
+      else if (TREE_CODE (type) == REFERENCE_TYPE)
+	{
+	  permerror (input_location, "reference %qs cannot be declared "
+	             "%<mutable%>", name);
+	  storage_class = sc_none;
+	}
     }
 
   /* If this is declaring a typedef name, return a TYPE_DECL.  */
Index: testsuite/g++.dg/other/pr33558.C
===================================================================
--- testsuite/g++.dg/other/pr33558.C	(revision 0)
+++ testsuite/g++.dg/other/pr33558.C	(revision 0)
@@ -0,0 +1,5 @@ 
+/* { dg-do compile } */
+
+class X {
+  mutable int &q; /* { dg-error "cannot be declared 'mutable'" } */
+};
Index: testsuite/g++.dg/other/pr33558-2.C
===================================================================
--- testsuite/g++.dg/other/pr33558-2.C	(revision 0)
+++ testsuite/g++.dg/other/pr33558-2.C	(revision 0)
@@ -0,0 +1,6 @@ 
+/* { dg-do compile } */
+/* { dg-options "-fpermissive" } */
+
+class X {
+  mutable int &q; /* { dg-warning "cannot be declared 'mutable'" } */
+};