[Bug lto/54728] New: [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[Bug lto/54728] New: [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[dimhen at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

             Bug #: 54728
           Summary: [4.8 regression] ICE in input_gimple_stmt, at
                    gimple-streamer-in.c:254
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: lto
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dimhen@gmail.com


gcc version 4.8.0 20120927 (experimental) [trunk revision 191799] (GCC)

$ cat fun1.cpp
#include <sstream>
#include <string>

void fun1(std::string &s)
{
    std::ostringstream os;
    s = os.str();
}
$ sed 's/fun1/fun2/' < fun1.cpp > fun2.cpp
$ g++ -flto -c fun1.cpp
$ g++ -flto -c fun2.cpp
$ g++ fun1.o fun2.o -flto -flto-partition=none -o libx.so
In member function 'str':
lto1: internal compiler error: in input_gimple_stmt, at
gimple-streamer-in.c:254
0xbf6cfb input_gimple_stmt
    /home/dimhen/src/gcc-current/gcc/gimple-streamer-in.c:253
0xbf6cfb input_bb(lto_input_block*, LTO_tags, data_in*, function*, int)
    /home/dimhen/src/gcc-current/gcc/gimple-streamer-in.c:310
0x70d47c input_function
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:870
0x70d47c lto_read_body
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:994
0x70d47c lto_input_function_body(lto_file_decl_data*, tree_node*, char const*)
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:1039
0x4cc9ca lto_materialize_function
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:226
0x4cc9ca materialize_cgraph
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:3045
0x4d137a lto_main()
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:3307
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.
lto-wrapper: g++ returned 1 exit status
/usr/bin/ld: fatal error: lto-wrapper failed
collect2: error: ld returned 1 exit status

$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/local/gcc_current/libexec/gcc/x86_64-unknown-linux-gnu/4.8.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: /home/dimhen/src/gcc-current/configure
--prefix=/usr/local/gcc_current --with-multilib-list=m64 --enable-__cxa_atexit
--enable-shared --enable-checking=df,fold,rtl,tree,yes
--enable-gnu-unique-object --enable-linker-build-id
--enable-languages=c,c++,lto --enable-plugin
--enable-version-specific-runtime-libs
Thread model: posix
gcc version 4.8.0 20120927 (experimental) [trunk revision 191799] (GCC)

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-09-27
   Target Milestone|---                         |4.8.0
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-09-27 12:17:22 UTC ---
Confirmed.

252       else if (code == GIMPLE_LABEL)
253         gcc_assert (emit_label_in_global_context_p (gimple_label_label
(stmt))
254                     || DECL_CONTEXT (gimple_label_label (stmt)) ==
fn->decl);

DECL_CONTEXT seems to be off.  Not sure if we ever assert this during regular
verification though (so might be a bug without LTO as well).

(gdb) call debug_cgraph_node (node)
_ZNKSt15basic_stringbufIcSt11char_traitsIcESaIcEE3strEv/818 (str)
@0x7ffff69d5138
  Type: function
  Visibility: external public visibility_specified
  previous sharing asm name: 253
  References: __gxx_personality_v0/445 (addr)
  Referring: 
  Read from file: t2.o
  Availability: available
  Function flags: analyzed body finalized

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[dimhen at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #2 from Dmitry G. Dyachenko <dimhen at gmail dot com> 2012-09-27 12:27:33 UTC ---
for reduced test case:
$ cp fun1.ii fun2.ii
$ g++...


$ cat fun1.ii
class G;
template < typename, typename = G > class A;
class B
{
};
class G:B
{
};
template < typename _CharT, typename _Alloc > class A
{
    struct C
    {
        static C & _S_empty_rep ();
        _CharT *_M_refdata ();
    };
    struct H:_Alloc
    {
        H (_CharT *, _Alloc):_Alloc ()
        {
        }
    };
    H _M_dataplus;
    C & _S_empty_rep ()
    {
        return C::_S_empty_rep ();
    }
public:
    A ():_M_dataplus (_S_empty_rep ()._M_refdata (), _Alloc ())
    {
    }
};

extern template class A < char >;
class D
{
};
class F
{
};
template < typename _CharT > class I:D
{
    typedef _CharT char_type;
    typedef A < char_type > __string_type;
    __string_type _M_string;
};

template < typename _CharT > class J:F
{
    typedef I < _CharT > __stringbuf_type;
    __stringbuf_type _M_stringbuf;
};

J < char >a;

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[dimhen at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #3 from Dmitry G. Dyachenko <dimhen at gmail dot com> 2012-10-01 10:16:50 UTC ---
different (less) testcase

this code must be compiled with '-O1'.
w/o '-O1' or with '-O0' err not triggered

$ cat f1.ii
namespace std
{
template < typename _CharT > class A
{
public:
    A (_CharT *);
};
class B
{
};
}

class D:
    std::B
{
public:
    D (const D &):
        std::B (),
        wfile (0)
    {
    } D ():
    std::B (),
        wfile (0)
    {
    } std::A < int >
    wfile;
};

class
    C
{
    virtual bool
    get_current_pin ()
    {
        throw
        D ();
    }
};
$ g++ -O -flto -c -o f1.o f1.ii
$ g++ -O -flto -c -o f2.o f1.ii
$ g++ f1.o f2.o -flto -flto-partition=none -o .libs/libp.so
In member function 'get_current_pin':
lto1: internal compiler error: in input_gimple_stmt, at
gimple-streamer-in.c:254
0xbf78ab input_gimple_stmt
    /home/dimhen/src/gcc-current/gcc/gimple-streamer-in.c:253
0xbf78ab input_bb(lto_input_block*, LTO_tags, data_in*, function*, int)
    /home/dimhen/src/gcc-current/gcc/gimple-streamer-in.c:310
0x70f3e4 input_function
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:853
0x70f3e4 lto_read_body
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:977
0x70f3e4 lto_input_function_body(lto_file_decl_data*, tree_node*, char const*)
    /home/dimhen/src/gcc-current/gcc/lto-streamer-in.c:1022
0x4ccd8a lto_materialize_function
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:226
0x4ccd8a materialize_cgraph
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:3045
0x4d173a lto_main()
    /home/dimhen/src/gcc-current/gcc/lto/lto.c:3307
Please submit a full bug report,

gcc version 4.8.0 20121001 (experimental) [trunk revision 191884] (GCC)

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #4 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> 2012-10-11 07:06:58 UTC ---
Created attachment 28417
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28417
Testcase

Compile with `g++ -g -O1 -flto -flto-partition=none 1.cc 2.cc'.

Another (perhaps related) issue. GCC 4.8.0 20121007 (experimental) report
"internal compiler error: SSA corruption" with `-O1' and "internal compiler
error: in input_gimple_stmt, at gimple-streamer-in.c:254" with `-O0'. Older 4.8
builds fail with "fatal error: LTO_tags out of range: Range is 0 to 348, value
is 15872".

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #6 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> 2012-10-11 07:08:29 UTC ---
Created attachment 28419
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28419
Backtrace with -O1

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #5 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> 2012-10-11 07:07:51 UTC ---
Created attachment 28418
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28418
Backtrace with -O0

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-11-16 03:36:35 UTC ---
here is the most reduced testcase.
cat > foo1.cc << EOF
template<typename _CharT>     struct basic_string
{
  struct _Rep   {    _CharT*  _M_refdata() throw()  {  return 0; }        };
  _CharT * dst;
  static _Rep&       _S_empty_rep();
  basic_string()       : dst(_S_empty_rep()._M_refdata())  {  }
};
extern template class basic_string<char>;
struct basic_ostringstream    {
  basic_string<char>  _M_stringbuf;
};
basic_ostringstream foo1;
EOF
sed s/foo1/foo2/ < foo1.cc > foo2.cc
g++ -Wfatal-errors   t1.cc t2.cc  -fpic  -flto -c
g++ -Wfatal-errors   t1.o t2.o -shared -flto -o t.so -flto-partition=none
In member function ‘_M_refdata’:
lto1: internal compiler error: in input_gimple_stmt, at
gimple-streamer-in.c:255
0xc565b6 input_gimple_stmt
....

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
                 CC|                            |hubicka at gcc dot gnu.org

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> 2012-12-07 11:24:52 UTC ---
fn->decl is 0x7ffff6929300 which is _not_ the prevailing one!  DECL_CONTEXT
of the label points to the prevailing one instead.

I wonder why we materialize a non-prevailing function ... Honza?

This is from

static void
materialize_cgraph (void)
{
...
  FOR_EACH_FUNCTION (node)
    {
      if (node->symbol.lto_file_data)
        {
          lto_materialize_function (node);
          lto_stats.num_input_cgraph_nodes++;
        }
    }

at mixed WPA/LTRANS stage (-flto-partition=none).

We arrive in lto_symtab_merge_cgraph_nodes_1 for both and have
!symtab_real_symbol_p (e) for the non-prevailing decl:

 <function_decl 0x7ffff6929300 _M_refdata
    type <method_type 0x7ffff691fb28
        type <pointer_type 0x7ffff691fbd0 type <integer_type 0x7ffff691fc78
char>
            unsigned DI
            size <integer_cst 0x7ffff67e2d20 constant 64>
            unit size <integer_cst 0x7ffff67e2d40 constant 8>
            align 64 symtab 0 alias set -1 canonical type 0x7ffff67fb2a0>
        QI
        size <integer_cst 0x7ffff67e2ee0 constant 8>
        unit size <integer_cst 0x7ffff67e2f00 constant 1>
        align 8 symtab 0 alias set -1 canonical type 0x7ffff691fd20 method
basetype <record_type 0x7ffff691fe70 _Rep>
        arg-types <tree_list 0x7ffff691aed8 value <pointer_type 0x7ffff691fdc8>
            chain <tree_list 0x7ffff67ed6b8 value <void_type 0x7ffff67f3bd0
void>>>
        pointer_to_this <pointer_type 0x7ffff6924b28>>
    addressable used nothrow public static external autoinline QI defer-output
file t2.ii line 3 col 31 align 16 context <record_type 0x7ffff691fe70 _Rep>
    arguments <parm_decl 0x7ffff6926400 this
        type <pointer_type 0x7ffff6924150 type <record_type 0x7ffff691fe70
_Rep>
            readonly unsigned DI size <integer_cst 0x7ffff67e2d20 64> unit size
<integer_cst 0x7ffff67e2d40 8>
            align 64 symtab 0 alias set -1 canonical type 0x7ffff6924150>
        readonly unsigned DI file t2.ii line 3 col 50 size <integer_cst
0x7ffff67e2d20 64> unit size <integer_cst 0x7ffff67e2d40 8>
        align 64 context <function_decl 0x7ffff6920900 _M_refdata> arg-type
<pointer_type 0x7ffff6924150>>
    result <result_decl 0x7ffff6927618 D.2421 type <pointer_type
0x7ffff691fbd0>
        unsigned ignored DI file t2.ii line 3 col 31 size <integer_cst
0x7ffff67e2d20 64> unit size <integer_cst 0x7ffff67e2d40 8>
        align 64 context <function_decl 0x7ffff6920900 _M_refdata>>>

for which then the merging keeps the cgraph node around ...

The node is analyzed (it has a body!) but it's DECL_EXTERNAL, it doesn't
have callers and is not refered to.

So - is the above decl broken (public static external - huh), or should
lto_symtab_merge_cgraph_nodes_1 simply drop !symtab_real_symbol_p nodes
that have a body?

Should the above body materialization use

  FOR_EACH_FUNCTION (node)
    {
      if (symtab_real_symbol_p (&node->symbol))
        {
          lto_materialize_function (node);
          lto_stats.num_input_cgraph_nodes++;
        }
    }

instead?

Honza, please have a look here.

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #9 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-12-21 14:16:09 UTC ---
When writting symtab_real_symbol_p I made external nodes to not be real symbols
unless they are explicitelly called or refereed. 

The reason is that we keep around virtual extenral functions for possible
devirtualization and I did not want those symbol appear to refereed in
lto-plugin symbol table. While this works well with lto-plugin symbol table
output it does not fare well with the use in lto-symtab.

I am testing patch to separate both predicates.

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #10 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-12-21 15:01:29 UTC ---
Created attachment 29021
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=29021
Patch I am testing.

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #11 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-12-22 23:27:54 UTC ---
Author: hubicka
Date: Sat Dec 22 23:27:49 2012
New Revision: 194700

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194700
Log:
    PR lto/54728
    * cgraph.h (symtab_real_symbol_p): Drop code looking for external
functions.
    * lto-streamer-out.c (output_symbol_p): New function.
    (produce_symtab) Use it.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cgraph.h
    trunk/gcc/lto-streamer-out.c

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

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

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

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> 2013-01-02 10:41:32 UTC ---
Fixed.

[Bug lto/54728] [4.8 regression] ICE in input_gimple_stmt, at gimple-streamer-in.c:254

[dimhen at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54728

--- Comment #13 from Dmitry G. Dyachenko <dimhen at gmail dot com> 2013-01-02 13:47:43 UTC ---
fix original problem for me too
Thank You