[Bug tree-optimization/46130] New: [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[Bug tree-optimization/46130] New: [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

           Summary: [4.6 Regression] ICE: SIGSEGV in
                    walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2
                    -fno-tree-dce
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz


Created attachment 22117
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22117
reduced testcase

Compiler output:
$ gcc -O2 -fno-tree-dce pr46130.c
pr46130.c: In function 'foo':
pr46130.c:15:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Related valgrind output:
==22348== Invalid read of size 2
==22348==    at 0x6EA541: walk_stmt_load_store_addr_ops (gimple.c:4894)
==22348==    by 0xAAB87A: rebuild_cgraph_edges (cgraphbuild.c:471)
==22348==    by 0x796C5E: execute_one_pass (passes.c:1560)
==22348==    by 0x796EF4: execute_pass_list (passes.c:1615)
==22348==    by 0x79616B: do_per_function_toporder (passes.c:1153)
==22348==    by 0x7973A5: execute_ipa_pass_list (passes.c:1932)
==22348==    by 0xAB09D7: cgraph_optimize (cgraphunit.c:1696)
==22348==    by 0xAB0B99: cgraph_finalize_compilation_unit (cgraphunit.c:1016)
==22348==    by 0x4ABB8B: c_write_global_declarations (c-decl.c:9828)
==22348==    by 0x88A553: toplev_main (toplev.c:931)
==22348==    by 0x658ABBC: (below main) (in /lib64/libc-2.11.2.so)
==22348==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==22348== 
testcase.c: In function 'foo':
testcase.c:15:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.


Tested revisions:
r165822 - crash
r161659 - crash
r159696 - OK
4.5 r165781 - OK

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #1 from Zdenek Sojka <zsojka at seznam dot cz> 2010-10-22 13:48:55 UTC ---
Created attachment 22118
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22118
another testcase

This testcase is shorter and doesn't take address of an function, but it needs
-flto

$ gcc -O2 -flto -fno-tree-dce pr46130-2.c
==30450== Invalid read of size 2
==30450==    at 0x6EA541: walk_stmt_load_store_addr_ops (gimple.c:4894)
==30450==    by 0xAAB87A: rebuild_cgraph_edges (cgraphbuild.c:471)
==30450==    by 0x796C5E: execute_one_pass (passes.c:1560)
==30450==    by 0x796EF4: execute_pass_list (passes.c:1615)
==30450==    by 0x79616B: do_per_function_toporder (passes.c:1153)
==30450==    by 0x7973A5: execute_ipa_pass_list (passes.c:1932)
==30450==    by 0xAB09D7: cgraph_optimize (cgraphunit.c:1696)
==30450==    by 0xAB0B99: cgraph_finalize_compilation_unit (cgraphunit.c:1016)
==30450==    by 0x4ABB8B: c_write_global_declarations (c-decl.c:9828)
==30450==    by 0x88A553: toplev_main (toplev.c:931)
==30450==    by 0x658ABBC: (below main) (in /lib64/libc-2.11.2.so)
==30450==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==30450== 
pr46130-2.c: In function 'foo':
pr46130-2.c:8:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.10.22 13:58:34
                 CC|                            |hjl.tools at gmail dot com,
                   |                            |rguenth at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> 2010-10-22 13:58:34 UTC ---
It is caused by revision 161655:

http://gcc.gnu.org/ml/gcc-cvs/2010-07/msg00006.html

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #3 from Zdenek Sojka <zsojka at seznam dot cz> 2010-10-22 14:05:10 UTC ---
Created attachment 22119
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22119
different crash

Very similiar testcase, but it crashes with different message:

$ gcc -O2 -fno-tree-dce pr46130-3.c 
pr46130-3.c: In function 'foo':
pr46130-3.c:16:1: internal compiler error: in split_function, at
ipa-split.c:1014
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Tested revisions:
r165822 - crash
r163636 - crash
r161659 - segfault (the same as in other testcases)

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
   Target Milestone|---                         |4.6.0

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-10-22 14:08:18 UTC ---
Mine.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|rguenth at gcc dot gnu.org  |hubicka at gcc dot gnu.org

--- Comment #5 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-11-02 12:49:59 UTC ---
First testcase:

Program received signal SIGSEGV, Segmentation fault.
0x000000000080e5fd in walk_stmt_load_store_addr_ops (stmt=0x7ffff5b47900, 
    data=0x7ffff5b01000, visit_load=0xd0293a <mark_load>, 
    visit_store=0xd02a5e <mark_store>, visit_addr=0xd027ca <mark_address>)
    at /space/rguenther/src/svn/trunk/gcc/gimple.c:5153
5153              if (TREE_CODE (op) == ADDR_EXPR)
(gdb) p op
$1 = (tree) 0x0
(gdb) call debug_gimple_stmt (stmt)
a_1 = PHI <0(2), (3)>

triggered by function splitting.  Re-assinging to Honza.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #6 from Zdenek Sojka <zsojka at seznam dot cz> 2010-11-16 00:52:52 UTC ---
Created attachment 22414
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22414
another simple testcase

This one needs -flto too:

$ gcc -O2 -flto -fno-tree-dce testcase.c
==9353== Invalid read of size 2
==9353==    at 0x710735: walk_stmt_load_store_addr_ops (gimple.c:5024)
==9353==    by 0xAC5C12: rebuild_cgraph_edges (cgraphbuild.c:474)
==9353==    by 0x7BCA0E: execute_one_pass (passes.c:1560)
==9353==    by 0x7BCCD4: execute_pass_list (passes.c:1615)
==9353==    by 0x7BBEFB: do_per_function_toporder (passes.c:1153)
==9353==    by 0x7BD185: execute_ipa_pass_list (passes.c:1932)
==9353==    by 0xACAD67: cgraph_optimize (cgraphunit.c:1694)
==9353==    by 0xACAF29: cgraph_finalize_compilation_unit (cgraphunit.c:1015)
==9353==    by 0x4D7D8B: c_write_global_declarations (c-decl.c:9837)
==9353==    by 0x8A9405: toplev_main (toplev.c:885)
==9353==    by 0x6376BBC: (below main) (in /lib64/libc-2.11.2.so)
==9353==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==9353== 
testcase.c: In function 'foo':
testcase.c:4:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

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

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

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-19 16:29:06 UTC ---
Testcase that needs just -O2 -fno-tree-dce:
static void
foo (__builtin_va_list ap)
{
  __builtin_va_arg (ap, char *)[0];
}

int
bar (__builtin_va_list ap)
{
  foo (ap);
  foo (ap);
}

The issue is that I see nothing that checks the SSA_NAMEs in return_bb PHIs
which originate from the split function are just in phi setting retval and
nothing else.
In this case because of -fno-tree-dce there is a dead PHI which the *.part.0
function doesn't pass through to the caller.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-19 16:59:59 UTC ---
Yet another testcase:
extern int bar (int);

static int foo (int x)
{
  int z, w;
  if (x <= 1024)
    {
      z = 16;
      w = 17;
    }
  else
    {
      bar (bar (bar (bar (bar (bar (bar (bar (bar (16)))))))));
      if (x > 131072)
    w = 19;
      else
    w = 21;
      z = 32;
    }
  w = w + 121;
  return z;
}

int
baz (int x)
{
  return foo (x + 6) + foo (x + 15) + foo (x + 24);
}

This one shows one PHI node for retval we handle correctly and one we don't.
What we perhaps could handle is if the phi argument in the other PHI is the
same in all the edges from split_bbs and originates outside of the split bbs,
but probably it isn't worth handling it.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-19 17:38:25 UTC ---
Created attachment 23031
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23031
gcc46-pr46130.patch

Untested fix.

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-20 12:02:38 UTC ---
Author: jakub
Date: Thu Jan 20 12:02:33 2011
New Revision: 169052

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=169052
Log:
    PR tree-optimization/46130
    * ipa-split.c (consider_split): If return_bb contains non-virtual
    PHIs other than for retval or if split_function would not adjust it,
    refuse to split.

    * gcc.dg/pr46130-1.c: New test.
    * gcc.dg/pr46130-2.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr46130-1.c
    trunk/gcc/testsuite/gcc.dg/pr46130-2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-split.c
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/46130] [4.6 Regression] ICE: SIGSEGV in walk_stmt_load_store_addr_ops (gimple.c:4894) with -O2 -fno-tree-dce

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46130

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

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

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-01-20 12:03:56 UTC ---
Fixed.