Bug 10646 - [3.3/3.4 regression] non-const reference is incorrectly matched in a "const T" partial specialization
Summary: [3.3/3.4 regression] non-const reference is incorrectly matched in a "const T...
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: c++ (show other bugs)
Version: 3.2
: P3 normal
Target Milestone: 3.4.1
Assignee: Not yet assigned to anyone
URL:
Keywords: rejects-valid
: 10647 (view as bug list)
Depends on:
Blocks:
 
Reported: 2003-05-06 22:46 UTC by notbob
Modified: 2004-05-05 07:11 UTC (History)
3 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2003-07-05 23:24:13

Attachments
is_const.cpp (208 bytes, application/octet-stream)
2003-05-21 15:17 UTC, notbob 		Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description notbob 2003-05-06 22:46:00 UTC 
I apologize in advance if I'm not reporting this "bug"
absolutely correctly. I haven't done this before. I
appreciate your patience. Thank you.

I believe my template should report references and functions as NOT const.

Is this "pilot error" or a compiler bug?

Release:
gcc version 3.2 20020927 (prerelease)

Environment:
in bash via cygwin on windows 98 se

How-To-Repeat:
g++ is_const.cpp
./a.exe
Comment 1 notbob 2003-05-06 22:46:00 UTC 
Fix:
As a fix, you can add another specialization that matches
references and set the result to false. Alternatively,
use boost::is_const that works correctly (has workarounds).
Comment 2 Giovanni Bajo 2003-05-07 00:17:50 UTC 
State-Changed-From-To: open->analyzed
State-Changed-Why: The bug for function types is confirmed in 3.2 but it's
    fixed in 3.3 (20030503) and 3.4 (20030503). It might be
    fixed in 3.2.3 as well, I don't have that version.
    
    There still is a bug with references, though:
    
    ---------------------------------
    template <typename T>
    struct is_const
    {
        typedef int type;
    };
    
    template <typename T>
    struct is_const<const T>
    {};
    
    is_const<bool&>::type a;
    ---------------------------------
    
    This should compile (but I'm not 100% sure because this
    very part of the standard is still a bit misterious for me),
    and it does on MSVC71 and Comeau/EDG, but fails with GCC
    3.3 and 3.4. Since it worked in 2.95, I rate this as a
    regression.
Comment 3 Giovanni Bajo 2003-05-07 02:23:22 UTC 
From: "Giovanni Bajo" <giovannibajo@libero.it>
To: <gcc-gnats@gcc.gnu.org>,
	<gcc-bugs@gcc.gnu.org>,
	<notbob@tessellation.com>,
	<nobody@gcc.gnu.org>,
	<gcc-prs@gcc.gnu.org>
Cc:  
Subject: Re: c++/10646: [3.3/3.4 regression] non-const reference is incorrectly matched in a "const T" partial specialization
Date: Wed, 7 May 2003 02:23:22 +0200

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10646
 
 As an additional note, the same bug exists for any combination of
 cv-qualifier. So, you can construct a template called is_volatile like the
 above to reproduce the same bug.
 
 Giovanni Bajo
Comment 4 janis187 2003-06-23 15:55:03 UTC 
The regression in 10646 was introduced or exposed by this patch:

--- gcc/gcc/cp/ChangeLog ---

> 2002-02-22  Nathan Sidwell  <nathan@codesourcery.com>
>
>       PR c++/2645, DR 295
>       * cp-tree.h (tsubst_flags_t): Add tf_ignore_bad_quals,
>       tf_keep_type_decl.
>       (make_typename_type): Use tsubst_flags_t.
>       * decl.c (make_typename_type): Adjust. Return non-artificial
>       TYPE_DECLs, if required.
>       (grokdeclarator): Simplify CVR qualification handling. Allow bad
>       qualifiers on typedef types.
>       * decl2.c (handle_class_head): Adjust make_typename_type call.
>       * parse.y (nested_name_specifier): Likewise.
>       (typename_sub0): Likewise.
>       (typename_sub1): Likewise.
>       * pt.c (convert_template_argument): Adjust make_typename_type
>       return value.
>       (tsubst): Adjust cp_build_qualified_type_real calls.
>       (check_cv_quals_for_unify): Cope with alowing bad qualifications
>       on template type parms.
>       (instantiate_decl): Recheck substitutions to give warnings on bad
>       qualifications.
>       * tree.c (cp_build_qualified_type_real): Use tf_allow_bad_quals.

The regression hunt used the smaller test case from comment #2.
Comment 5 Nathan Sidwell 2003-07-13 14:10:48 UTC 
Not a bug.
CV qualifiers are ignored on reference type when introduced via a typedef
or template type parameter.
Comment 6 Andrew Pinski 2004-05-05 03:52:49 UTC 
Nathan, this now compiles on the mainline, this is correct right?
Comment 7 Nathan Sidwell 2004-05-05 07:11:07 UTC 
yes this is now working