[Bug c++/60600] New: ICE in ipa_get_indirect_edge_target_1

[Bug c++/60600] New: ICE in ipa_get_indirect_edge_target_1

[lucdanton at free dot fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

            Bug ID: 60600
           Summary: ICE in ipa_get_indirect_edge_target_1
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: lucdanton at free dot fr

Created attachment 32404
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32404&action=edit
Testcase

$ g++-trunk --version
g++-trunk (GCC) 4.9.0 20140319 (experimental)

$ cat main.cpp 
struct data {
    data(int);
};

struct top {
    virtual int topf();
};

struct intermediate: top {
    int topf() /* override */ { return 0; }
};

struct child1: top {
    void childf()
    {
        data d(topf());
    }
};

struct child2: intermediate {};

void test(top& t)
{
    child1& c = static_cast<child1&>(t);
    c.childf();
    child2 d;
    test(d);
}

$ g++-trunk main.cpp -O3
main.cpp:28:1: internal compiler error: in ipa_get_indirect_edge_target_1, at
ipa-cp.c:1645
 }
 ^
0x10ac52e ipa_get_indirect_edge_target_1
    ../../gcc/gcc/ipa-cp.c:1644
0xa1570c estimate_edge_devirt_benefit
    ../../gcc/gcc/ipa-inline-analysis.c:2955
0xa1570c estimate_edge_size_and_time
    ../../gcc/gcc/ipa-inline-analysis.c:2987
0xa1570c estimate_calls_size_and_time
    ../../gcc/gcc/ipa-inline-analysis.c:3040
0xa1602c estimate_node_size_and_time
    ../../gcc/gcc/ipa-inline-analysis.c:3127
0xa18016 estimate_ipcp_clone_size_and_time(cgraph_node*, vec<tree_node*,
va_heap, vl_ptr>, vec<tree_node*, va_heap, vl_ptr>, vec<ipa_agg_jump_function*,
va_heap, vl_ptr>, int*, int*, int*)
    ../../gcc/gcc/ipa-inline-analysis.c:3163
0x10b17fc estimate_local_effects
    ../../gcc/gcc/ipa-cp.c:2004
0x10b17fc propagate_constants_topo
    ../../gcc/gcc/ipa-cp.c:2229
0x10b17fc ipcp_propagate_stage
    ../../gcc/gcc/ipa-cp.c:2329
0x10b17fc ipcp_driver
    ../../gcc/gcc/ipa-cp.c:3707
0x10b17fc execute
    ../../gcc/gcc/ipa-cp.c:3806
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.

-------------------------------------

This only happens with -O3, in all C++ modes. This may be a regression – the
testcase was built from the preprocessed source of a third-party library, with
stock settings. Presumably it built at some point before shipping.
>From gcc-bugs-return-446985-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Thu Mar 20 13:40:57 2014
Return-Path: <gcc-bugs-return-446985-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 23577 invoked by alias); 20 Mar 2014 13:40:56 -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 23522 invoked by uid 48); 20 Mar 2014 13:40:53 -0000
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/60580] aarch64 generates wrong code for __attribute__ ((optimize("no-omit-frame-pointer")))
Date: Thu, 20 Mar 2014 13:40:00 -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: 4.8.3
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: jakub at gcc dot gnu.org
X-Bugzilla-Status: UNCONFIRMED
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:
Message-ID: <bug-60580-4-MNG19wjA7J@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-60580-4@http.gcc.gnu.org/bugzilla/>
References: <bug-60580-4@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: 2014-03/txt/msg01854.txt.bz2
Content-length: 233

http://gcc.gnu.org/bugzilla/show_bug.cgi?id`580

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I'm not going to hack on this any time soon, so if aarch64 maintainers could
take it over, it would be appreciated.

[Bug c++/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-03-20
                 CC|                            |hubicka at gcc dot gnu.org,
                   |                            |jakub at gcc dot gnu.org,
                   |                            |jamborm at gcc dot gnu.org
   Target Milestone|---                         |4.9.0
            Summary|ICE in                      |[4.9 Regression] ICE in
                   |ipa_get_indirect_edge_targe |ipa_get_indirect_edge_targe
                   |t_1                         |t_1
     Ever confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r205019.

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
          Component|c++                         |ipa

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

--- Comment #2 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Well, this is ICE on code with undefined behavior.  Function test
calls itself with a parameter which is a reference to an object of
type child2 and then static_casts it to a reference to child1.  These
are non-PODs, neither of these types is an ancestor of the other and
thus the use of static_cast is not allowed by the C++ standard.

ipa-prop based devirtualization does not see the cast, it thinks the
object is of type child2, and therefore the virtual method it looks up
is intermediate::topf.  Then it does a consistency check to see
whether type hierarchy based devirtualization has that function among
its possible targets.  However, this devirtualization starts with
child1 as its outer type and thus it concludes the only possible
target is top::topf.  The consistency check fails and we get an ICE on
assert.

The assert has to go but I wonder whether we want to give up when
possible_polymorphic_call_target_p returns false so that we don't try
to optimize such undefined cases.  We might even give a warning,
although making the warning useful to the user might require some
effort.

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[lucdanton at free dot fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

--- Comment #5 from lucdanton at free dot fr ---
(In reply to Martin Jambor from comment #2)
> Well, this is ICE on code with undefined behavior.  Function test
> calls itself with a parameter which is a reference to an object of
> type child2 and then static_casts it to a reference to child1.  These
> are non-PODs, neither of these types is an ancestor of the other and
> thus the use of static_cast is not allowed by the C++ standard.
> […]

I must apologise for the peculiar testcase I produced. I normally try to write
a sensible program or fragment, but in this case I couldn’t figure out where
the problematic behaviour was. In the end I resigned myself to (almost
mechanically) discard bits and bits of the original TU until I was left with
those things that triggered the ICE and nothing else.

(I can’t help but notice that if the testcase is made a complete program by
e.g. adding an empty main then strictly speaking there is no UB as control flow
never reaches the invalid cast — with the ICE still happening.)

For the record, and as best as I can tell, the original code was carefully
performing the downcasts inside a (type)case, with no overlap or fallthrough.
And if I do apply the patch you supplied, GCC does compile the library without
issue.
>From gcc-bugs-return-447157-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sat Mar 22 12:52:53 2014
Return-Path: <gcc-bugs-return-447157-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 13778 invoked by alias); 22 Mar 2014 12:52:53 -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 13747 invoked by uid 48); 22 Mar 2014 12:52:47 -0000
From: "gnugcc at marino dot st" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug ada/60620] New: missing gnattools dependency causes highly parallel build failure with --disable-bootstrap
Date: Sat, 22 Mar 2014 12:52:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: new
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: ada
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: gnugcc at marino dot st
X-Bugzilla-Status: UNCONFIRMED
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_id short_desc product version bug_status bug_severity priority component assigned_to reporter
Message-ID: <bug-60620-4@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: 2014-03/txt/msg02026.txt.bz2
Content-length: 3966

http://gcc.gnu.org/bugzilla/show_bug.cgi?id`620

            Bug ID: 60620
           Summary: missing gnattools dependency causes highly parallel
                    build failure with --disable-bootstrap
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ada
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gnugcc at marino dot st

GCC 4.9.0 has been brought into FreeBSD ports (lang/gcc-aux) which supports 4
platforms: i386-FreeBSD, i386-DragonFly, x86_64-FreeBSD, and x86-64-DragonFly.
On machines with low -j inputs (say -j4) it would build fine, but on powerful
machines with like -j8 or greater, we'd seen a clean jail (no base compiler lib
in path) build break on gnattools such as this:

checking for scalbnl... ../../xg++ -B../../
-B../../../x86_64-aux-freebsd10.0/libstdc++-v3/src/.libs
-B../../../x86_64-aux-freebsd10.0/libstdc++-v3/libsupc++/.libs
-L../../../x86_64-aux-freebsd10.0/libstdc++-v3/src/.libs
-L../../../x86_64-aux-freebsd10.0/libstdc++-v3/libsupc++/.libs
-static-libstdc++ -static-libgcc -I- -I../rts -I.
-I/work/a/ports/lang/gcc-aux/work/gcc-4.9-20140316/gcc/ada -static-libstdc++
-static-libgcc  -DIN_GCC  -g -O2 -W -Wall -o ../../gnatlink b_gnatl.o
gnatlink.o a-except.o ali.o alloc.o butil.o casing.o csets.o debug.o fmap.o
fname.o gnatvsn.o hostparm.o indepsw.o interfac.o i-c.o i-cstrin.o namet.o
opt.o osint.o output.o rident.o s-exctab.o s-secsta.o s-stalib.o s-stoele.o
sdefault.o snames.o stylesw.o switch.o system.o table.o targparm.o tree_io.o
types.o validsw.o widechar.o ../link.o ../targext.o ../../ggc-none.o
../../libcommon-target.a ../../libcommon.a ../../../libcpp/libcpp.a
../rts/libgnat.a ../../../intl/libintl.a
../../../libbacktrace/.libs/libbacktrace.a ../../../libiberty/libiberty.a
yes
checking for sinf... /usr/bin/ld: cannot find -lstdc++
collect2: error: ld returned 1 exit status
gmake[5]: *** [../../gnatlink] Error 1
gmake[5]: *** Waiting for unfinished jobs....


It seems this doesn't happen with a full bootstrap.  For reference, this is a
typical configuration:

/work/a/ports/lang/gcc-aux/work/gcc-4.9-20140316/configure
--enable-languages=c\ c++\ ada\ fortran\ objc --build=x86_64-aux-freebsd10.0
--prefix=/usr/local/gcc-aux --with-system-zlib --with-gmp=/usr/local
--with-mpfr=/usr/local --with-mpc=/usr/local  --enable-shared
--enable-threads=posix --disable-libmudflap --disable-libgomp --disable-libssp
--enable-libquadmath --enable-nls --disable-bootstrap

It appears the linking using g++ is intentional as -static-libstdc++ is passed,
so that means that libstdc++ is required to build gnattools but that dependency
is not documented.  The following patches solved the parallel build problems
with --disable-bootstrap that we were seeing:

--- Makefile.def.orig    2013-10-29 13:37:47.000000000 -0500
+++ Makefile.def
@@ -336,6 +336,7 @@ dependencies = { module=all-libcpp; on=a
 dependencies = { module=all-fixincludes; on=all-libiberty; };

 dependencies = { module=all-gnattools; on=all-target-libada; };
+dependencies = { module=all-gnattools; on=all-target-libstdc++-v3; };

 dependencies = { module=all-lto-plugin; on=all-libiberty; };


--- Makefile.in.orig    2014-03-07 07:58:27.000000000 -0500
+++ Makefile.in
@@ -46730,6 +46730,7 @@ all-stageprofile-libcpp: maybe-all-stage
 all-stagefeedback-libcpp: maybe-all-stagefeedback-intl
 all-fixincludes: maybe-all-libiberty
 all-gnattools: maybe-all-target-libada
+all-gnattools: maybe-all-target-libstdc++-v3
 all-lto-plugin: maybe-all-libiberty

 all-stage1-lto-plugin: maybe-all-stage1-libiberty


It should be a simple matter to confirm that libstdc++ is indeed required for
gnattools and that it's not listed as a dependency.  It is also possible that
gcc 4.8.x is affected by this, but I cannot say as I have never tried to build
that version, nor do I know if libstdc++ is required for gnattools there.

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
The patch looks fine to me with an testcase added checking for the warning.
I sort of hoped that the type based devirt code in ipa-cp won't get into
completely contradicting answers, but that was at a time when
possible_polymorphic_call_target_p did not consider outer types, so we
definitely can construct testcases for that now.

The patch is OK if it passes testing.
Honza

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-p
                   |                            |atches/2014-03/msg01363.htm
                   |                            |l

--- Comment #7 from Martin Jambor <jamborm at gcc dot gnu.org> ---
(In reply to lucdanton from comment #5)
> I must apologise for the peculiar testcase I produced. 

No need to apologize, I should have thought the downcast was
originally performed after a check.  The testcase is in fact very
nice.  And of course, gcc should not ICE even if you feed it undefined
or even invalid input.

The patch I am going to commit is at:
http://gcc.gnu.org/ml/gcc-patches/2014-03/msg01363.html

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

--- Comment #8 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Author: jamborm
Date: Tue Mar 25 18:22:41 2014
New Revision: 208818

URL: http://gcc.gnu.org/viewcvs?rev=208818&root=gcc&view=rev
Log:
2014-03-25  Martin Jambor  <mjambor@suse.cz>

    PR ipa/60600
    * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
    inconsistent devirtualizations to __builtin_unreachable.

testsuite/
    * g++.dg/ipa/pr60600.C: New test.


Added:
    trunk/gcc/testsuite/g++.dg/ipa/pr60600.C
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-cp.c
    trunk/gcc/testsuite/ChangeLog

[Bug ipa/60600] [4.9 Regression] ICE in ipa_get_indirect_edge_target_1

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60600

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

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

--- Comment #9 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Fixed.