[Bug ipa/62015] New: ipa-cp-clone uses a clone that is too specialized for the call context

[Bug ipa/62015] New: ipa-cp-clone uses a clone that is too specialized for the call context

[oremanj at mit dot edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

            Bug ID: 62015
           Summary: ipa-cp-clone uses a clone that is too specialized for
                    the call context
           Product: gcc
           Version: 4.10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ipa
          Assignee: unassigned at gcc dot gnu.org
          Reporter: oremanj at mit dot edu

In the C++ program below, interprocedural constant propagation cloning creates
a clone of A::adjust() specialized for adjust(Side::Invalid, false). This clone
is then used for an adjust(side, false) call when there is no way to prove side
== Side::Invalid. This bug is present in gcc 4.8.2, is not present in 4.7.3,
and is still present in a version compiled from today's SVN. Using an
optimization level less than -O3 or passing -fno-ipa-cp-clone prevents the bug
from manifesting. The onset of the bug was bisected to commit r193410.

$ cat t.cc
extern "C" int printf(const char *fmt, ...);

struct Side {
    enum _Value { Left, Right, Invalid };

    constexpr Side() : _value(Invalid) {}
    constexpr Side(_Value value) : _value(value) {}
    operator _Value() const { return (_Value)_value; }

  private:
    char _value;
};

struct A {
    void init();
    void adjust(Side side, bool final);
    void move(Side side);
};

void A::init()
{
    adjust(Side::Invalid, false);
}

__attribute__((noinline))
void A::adjust(Side side, bool final)
{
    printf("adjust: side is %d, final %d\n", (int)side, final);
}

void A::move(Side side)
{
    printf("move: side is %d\n", (int)side);
    adjust(side, false);
    adjust(side, true);
}

int main()
{
    A t;
    t.move(Side::Left);
}

$ g++49 -v
Using built-in specs.
COLLECT_GCC=g++49
COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc49/gcc/x86_64-unknown-linux-gnu/4.10.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-trunk/configure --prefix=/usr/local --disable-nls
--libdir=/usr/local/lib/gcc49 --libexecdir=/usr/local/libexec/gcc49
--program-suffix=49 --with-gmp=/usr/local
--with-gxx-include-dir=/usr/local/lib/gcc49/include/c++/
--with-libiconv-prefix=/usr/local --with-system-zlib --disable-rpath
--enable-languages=c,c++ --mandir=/usr/local/man
--infodir=/usr/local/info/gcc49 --disable-initfini-array
--disable-install-libiberty
Thread model: posix
gcc version 4.10.0 20140804 (experimental) (GCC)

$ g++49 -Wall -Wextra -o t t.cc -std=c++11 -O3
$ ./t
move: side is 0
adjust: side is 2, final 0
adjust: side is 0, final 1

The output should be 'side is 0' on all three lines.

[Bug ipa/62015] [4.8/4.9/4.10 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
                 CC|                            |jamborm at gcc dot gnu.org
      Known to work|                            |4.7.2
   Target Milestone|---                         |4.8.4
            Summary|ipa-cp-clone uses a clone   |[4.8/4.9/4.10 Regression]
                   |that is too specialized for |ipa-cp-clone uses a clone
                   |the call context            |that is too specialized for
                   |                            |the call context

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[mikpelinux at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

Mikael Pettersson <mikpelinux at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpelinux at gmail dot com

--- Comment #1 from Mikael Pettersson <mikpelinux at gmail dot com> ---
I can reproduce this wrong-code on x86_64-linux with gcc 5-20140824,
4.9-20140827, and 4.8-20140828.  4.7-20140607 works.

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2014-09-02
           Assignee|unassigned at gcc dot gnu.org      |jamborm at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Mine

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> ---
I have proposed a fix on the mailing list:

https://gcc.gnu.org/ml/gcc-patches/2014-09/msg00210.html

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

--- Comment #4 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Author: jamborm
Date: Wed Sep  3 14:26:38 2014
New Revision: 214878

URL: https://gcc.gnu.org/viewcvs?rev=214878&root=gcc&view=rev
Log:
2014-09-03  Martin Jambor  <mjambor@suse.cz>

    PR ipa/62015
    * ipa-cp.c (intersect_aggregates_with_edge): Handle impermissible
    pass-trough jump functions correctly.

testsuite/
    * g++.dg/ipa/pr62015.C: New test.


Added:
    trunk/gcc/testsuite/g++.dg/ipa/pr62015.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-cp.c
    trunk/gcc/testsuite/ChangeLog

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Author: jamborm
Date: Wed Sep  3 16:16:28 2014
New Revision: 214883

URL: https://gcc.gnu.org/viewcvs?rev=214883&root=gcc&view=rev
Log:
2014-09-03  Martin Jambor  <mjambor@suse.cz>

    PR ipa/62015
    * ipa-cp.c (intersect_aggregates_with_edge): Handle impermissible
    pass-trough jump functions correctly.

testsuite/
    * g++.dg/ipa/pr62015.C: New test.


Added:
    branches/gcc-4_9-branch/gcc/testsuite/g++.dg/ipa/pr62015.C
Modified:
    branches/gcc-4_9-branch/gcc/ChangeLog
    branches/gcc-4_9-branch/gcc/ipa-cp.c
    branches/gcc-4_9-branch/gcc/testsuite/ChangeLog

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

--- Comment #6 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Author: jamborm
Date: Wed Sep  3 16:36:06 2014
New Revision: 214885

URL: https://gcc.gnu.org/viewcvs?rev=214885&root=gcc&view=rev
Log:
2014-09-03  Martin Jambor  <mjambor@suse.cz>

    PR ipa/62015
    * ipa-cp.c (intersect_aggregates_with_edge): Handle impermissible
    pass-trough jump functions correctly.

testsuite/
    * g++.dg/ipa/pr62015.C: New test.


Added:
    branches/gcc-4_8-branch/gcc/testsuite/g++.dg/ipa/pr62015.C
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/ipa-cp.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog

[Bug ipa/62015] [4.8/4.9/5 Regression] ipa-cp-clone uses a clone that is too specialized for the call context

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62015

Martin Jambor <jamborm at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Fixed.