[Bug c++/67795] New: Wrong code generated for conditional expression with cast

[Bug c++/67795] New: Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

            Bug ID: 67795
           Summary: Wrong code generated for conditional expression with
                    cast
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: frankhb1989 at gmail dot com
  Target Milestone: ---

Case:
// g++ -std=c++1y
#include <iostream>
#include <string>
#include <experimental/string_view>

using namespace std;
using namespace experimental;

string_view&
erase_left(size_t n, string_view& s)
{
        return static_cast<string_view&>(n != 0 ? (s.remove_prefix(n), s) : s);
//      return n != 0 ? (s.remove_prefix(n), s) : s;
}

int main()
{
        string_view sv("abcde");
        cout << erase_left(3, sv).to_string() << endl; // expected "de"
}

The output is garbage, using i686-w64-mingw32-g++ from MSYS2. However,
x86_64-w64-mingw32-clang++ is OK. So I suspect it is a frontend bug.

[Bug c++/67795] Wrong code generated for conditional expression with cast

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |trippels at gcc dot gnu.org
         Resolution|---                         |INVALID

--- Comment #1 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
gcc even warns:

t.cpp: In function ‘std::experimental::fundamentals_v1::string_view&
erase_left(size_t, std::experimental::fundamentals_v1::string_view&)’:
t.cpp:9:73: warning: function returns address of local variable
[-Wreturn-local-addr]
   return static_cast<string_view &>(n != 0 ? (s.remove_prefix(n), s) : s);
                                                                         ^
>From gcc-bugs-return-498471-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Oct 01 11:17:20 2015
Return-Path: <gcc-bugs-return-498471-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 23409 invoked by alias); 1 Oct 2015 11:17:20 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 23374 invoked by uid 48); 1 Oct 2015 11:17:16 -0000
From: "renlin at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/66776] [AArch64] zero-extend version of csel not matching properly
Date: Thu, 01 Oct 2015 11:17:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: 6.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: renlin at gcc dot gnu.org
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: renlin at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 6.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: bug_status cf_reconfirmed_on everconfirmed
Message-ID: <bug-66776-4-hAQAwYwMRn@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-66776-4@http.gcc.gnu.org/bugzilla/>
References: <bug-66776-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-10/txt/msg00026.txt.bz2
Content-length: 390

https://gcc.gnu.org/bugzilla/show_bug.cgi?idf776

renlin at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2015-10-01
     Ever confirmed|0                           |1

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

frankhb1989 at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|INVALID                     |---

--- Comment #2 from frankhb1989 at gmail dot com ---
(In reply to Markus Trippelsdorf from comment #1)
> gcc even warns:
> 
> t.cpp: In function ‘std::experimental::fundamentals_v1::string_view&
> erase_left(size_t, std::experimental::fundamentals_v1::string_view&)’:
> t.cpp:9:73: warning: function returns address of local variable
> [-Wreturn-local-addr]
>    return static_cast<string_view &>(n != 0 ? (s.remove_prefix(n), s) : s);
>                                                                          ^

So why it is OK without 'static_cast'?
And actually G++ gives me no warnings with -Wall -Wextra -pedantic-errors.
>From gcc-bugs-return-498474-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Oct 01 12:36:52 2015
Return-Path: <gcc-bugs-return-498474-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 114578 invoked by alias); 1 Oct 2015 12:36:52 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 114533 invoked by uid 55); 1 Oct 2015 12:36:47 -0000
From: "olegendo at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/54236] [SH] Improve addc and subc insn utilization
Date: Thu, 01 Oct 2015 12:36:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: 4.8.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: olegendo at gcc dot gnu.org
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: olegendo at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-54236-4-ldUeW1f7kj@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-54236-4@http.gcc.gnu.org/bugzilla/>
References: <bug-54236-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-10/txt/msg00029.txt.bz2
Content-length: 451

https://gcc.gnu.org/bugzilla/show_bug.cgi?idT236

--- Comment #18 from Oleg Endo <olegendo at gcc dot gnu.org> ---
Author: olegendo
Date: Thu Oct  1 12:36:15 2015
New Revision: 228332

URL: https://gcc.gnu.org/viewcvs?rev"8332&root=gcc&view=rev
Log:
gcc/testsuite/
        PR target/54236
        * gcc.target/sh/pr54236-6.c: Fix assembler-no string.

Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/gcc.target/sh/pr54236-6.c

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

--- Comment #3 from frankhb1989 at gmail dot com ---
(In reply to Markus Trippelsdorf from comment #1)
> gcc even warns:
> 
> t.cpp: In function ‘std::experimental::fundamentals_v1::string_view&
> erase_left(size_t, std::experimental::fundamentals_v1::string_view&)’:
> t.cpp:9:73: warning: function returns address of local variable
> [-Wreturn-local-addr]
>    return static_cast<string_view &>(n != 0 ? (s.remove_prefix(n), s) : s);
>                                                                          ^

And if the diagnostics exist, they are definitely wrong.

First, here 's' is a local variable (in sense of C++ terms) indeed, but this is
perfectly valid. Note it is a _parameter_ referencing to an object whose
lifetime is unknown in the function/block scopes. This should not trigger such
warnings. Otherwise the identity functions for pointers/references are always
invalid. That's absurd.

Second, the "local variable" wording is somewhat misleading. Returning
reference to a local variable is generally wrong because the variable is an
_automatic object_ and it would be no sense to use the return value to
reference an end-of-life object which would cause undefined behavior.

Anyway, this is just not the case here.
>From gcc-bugs-return-498478-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Oct 01 12:48:18 2015
Return-Path: <gcc-bugs-return-498478-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 22977 invoked by alias); 1 Oct 2015 12:48:13 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 22903 invoked by uid 48); 1 Oct 2015 12:48:08 -0000
From: "trippels at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c++/67795] Wrong code generated for conditional expression with cast
Date: Thu, 01 Oct 2015 12:48:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: c++
X-Bugzilla-Version: unknown
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: trippels at gcc dot gnu.org
X-Bugzilla-Status: RESOLVED
X-Bugzilla-Resolution: INVALID
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: bug_status resolution
Message-ID: <bug-67795-4-CQ0qsCL0LW@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-67795-4@http.gcc.gnu.org/bugzilla/>
References: <bug-67795-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-10/txt/msg00033.txt.bz2
Content-length: 492

https://gcc.gnu.org/bugzilla/show_bug.cgi?idg795

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #4 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Return by value if you want to avoid undefined behavior.

[Bug c++/67795] Wrong code generated for conditional expression with cast

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
I would guess the issue is that ?: returns an rvalue (but that may not be 100%
correct if omitting the cast works and does not warn)

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

frankhb1989 at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|unknown                     |5.2.0

--- Comment #6 from frankhb1989 at gmail dot com ---
I forgot mentioning the versions I used ...
g++ -v
gcc version 5.2.0 (Rev3, Built by MSYS2 project)
clang++ -v
clang version 3.7.0 (tags/RELEASE_370/final)

[Bug c++/67795] Wrong code generated for conditional expression with cast

[glisse at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

Marc Glisse <glisse at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
            Version|5.2.0                       |unknown
         Resolution|INVALID                     |---

--- Comment #7 from Marc Glisse <glisse at gcc dot gnu.org> ---
Hmm, with the static_cast, the front-end produces:

<<cleanup_point return <retval> = (struct string_view &) (struct string_view *)
NON_LVALUE_EXPR <(struct string_view &) &(n != 0 ?
std::experimental::fundamentals_v1::basic_string_view<char>::remove_prefix
((struct string_view *) s, n);, *s; : *s)>>>;

while without it, I get:

<<cleanup_point return <retval> = (struct string_view &) (n != 0 ?
std::experimental::fundamentals_v1::basic_string_view<char>::remove_prefix
((struct string_view *) s, n);, (struct string_view *) s; : (struct string_view
*) s)>>;

I also find it fishy.

(note that you need -O2 or more for the warning)

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

--- Comment #8 from frankhb1989 at gmail dot com ---
(In reply to Markus Trippelsdorf from comment #4)
> Return by value if you want to avoid undefined behavior.

No. This is not the point. For something like 'std::move' or 'std::forward',
can you suggest such change?

The code of the case is reduced from a function template which return 'S&&',
where 'S' is a template type parameter (and 'S&&' is also in function parameter
list for perfect forwarding). Simply returning object type is not acceptable in
the original case because: 1. Even move construction for 'S' is not guaranteed
to be zero overhead. 2.It needs one more static_cast/std::move in many places
of the the client code which makes it bloat A LOT.

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

--- Comment #9 from frankhb1989 at gmail dot com ---
(In reply to Richard Biener from comment #5)
> I would guess the issue is that ?: returns an rvalue (but that may not be
> 100% correct if omitting the cast works and does not warn)

In C++ ?: can return lvalues.
I have tried inserting 'static_assert(std::is_same<string_view&,
decltype(..)>);' in the function and it does not complain.

[Bug c++/67795] Wrong code generated for conditional expression with cast

[frankhb1989 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

--- Comment #10 from frankhb1989 at gmail dot com ---
(In reply to Marc Glisse from comment #7)
> Hmm, with the static_cast, the front-end produces:
> 
> <<cleanup_point return <retval> = (struct string_view &) (struct string_view
> *) NON_LVALUE_EXPR <(struct string_view &) &(n != 0 ?
> std::experimental::fundamentals_v1::basic_string_view<char>::remove_prefix
> ((struct string_view *) s, n);, *s; : *s)>>>;
> 
> while without it, I get:
> 
> <<cleanup_point return <retval> = (struct string_view &) (n != 0 ?
> std::experimental::fundamentals_v1::basic_string_view<char>::remove_prefix
> ((struct string_view *) s, n);, (struct string_view *) s; : (struct
> string_view *) s)>>;
> 
> I also find it fishy.
> 
> (note that you need -O2 or more for the warning)

Thank you for the suggestion.

I find these diagnostics exist with -O2 or -O3. Then the output is empty ...

clang++ -Weverything with -O2/-O3 still behaves as I expect. (Only
[-Wmissing-prototypes].)

[Bug c++/67795] Wrong code generated for conditional expression with cast

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=64372

--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Looks fixed for GCC 10+.  I am suspecting by the patch which fixed PR 64372.

[Bug c++/67795] Wrong code generated for conditional expression with cast

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #15 from Martin Liška <marxin at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #14)
> Looks fixed for GCC 10+.  I am suspecting by the patch which fixed PR 64372.

Can you reproduce it on Linux? Can you please provide a pre-processed source
file?

[Bug c++/67795] Wrong code generated for conditional expression with cast

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67795

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING