[Bug tree-optimization/50326] New: [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[Bug tree-optimization/50326] New: [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

             Bug #: 50326
           Summary: [4.7 regression] ICE in set_lattice_value, at
                    tree-ssa-ccp.c:456
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Keywords: build
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: schwab@linux-m68k.org
            Target: ia64-*-*


configure: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
make[1]: *** [configure-stage2-target-libgcc] Error 1

configure:3562: /home/schwab/src/gcc/test/./gcc/xgcc
-B/home/schwab/src/gcc/test/./gcc/ -B/usr/local/ia64-unknown-linux-gnu/bin/
-B/usr/local/ia64-unknown-linux-gnu/lib/ -isystem
/usr/local/ia64-unknown-linux-gnu/include -isystem
/usr/local/ia64-unknown-linux-gnu/sys-include    -c -g -O2  conftest.c >&5
conftest.c: In function 'main':
conftest.c:16:1: internal compiler error: in set_lattice_value, at
tree-ssa-ccp.c:456

[Bug tree-optimization/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

[Bug rtl-optimization/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bernds at gcc dot gnu.org
          Component|tree-optimization           |rtl-optimization

--- Comment #1 from Andreas Schwab <schwab@linux-m68k.org> 2011-09-08 17:29:29 UTC ---
31adcf5623e6b926531c3a2973eb1b0c22d3c4c8 is the first bad commit
commit 31adcf5623e6b926531c3a2973eb1b0c22d3c4c8
Author: bernds <bernds@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Wed Sep 7 15:57:45 2011 +0000

        * regrename.c (struct du_head): Make nregs signed.
        (closed_chains): Remove.
        (create_new_chain): Return the new chain.
        (chain_from_id): New static function.
        (dump_def_use_chain): Change argument to be an int, indicating
        the first ID to print.  All callers changed.
        (merge_overlapping_regs): Use chain_from_id.  Assert that
        chains don't conflict with themselves.
        (rename_chains): Take no argument.  Iterate over id_to_chain
        rather to find chains to rename.  Clear tick before the main
        loop.
        (struct incoming_reg_info): New struct.
        (struct bb_rename_info): New struct.
        (init_rename_info, set_incoming_from_chain, merge_chains): New
        static functions.
        (regrename_analyze): New static function, broken out of
        regrename_optimize.  Record and make use of open chain information
        at basic block boundaries, and merge chains where possible.
        (scan_rtx_reg): Make this_nregs signed.  Don't update
        closed_chains.
        (build_def_use): Return a bool to indicate success.  All callers
        changed.  Don't initialize global data here.
        (regrename_optimize): Move most code out of here into
        regrename_analyze.
        * regs.h (add_range_to_hard_reg_set, remove_range_from_hard_reg_set,
        range_overlaps_hard_reg_set_p, range_in_hard_reg_set_p): New
        static inline functions.
        * vec.h (FOR_EACH_VEC_ELT_FROM): New macro.


    git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@178645
138bc75d-0d04-0410-961f-82ee72b054a4

[Bug rtl-optimization/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[bernds at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #2 from Bernd Schmidt <bernds at gcc dot gnu.org> 2011-09-09 11:22:00 UTC ---
Hmm, are you sure you bisected that correctly? I did a build on gcc60 with the
patch reverted, and it failed with the same error.

[Bug rtl-optimization/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-09-09 11:35:28 UTC ---
It's indeed a very unlikely candidate (unless cc1[plus] is miscompiled of
course).

[Bug tree-optimization/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mjambor at suse dot cz
          Component|rtl-optimization            |tree-optimization

--- Comment #4 from Andreas Schwab <schwab@linux-m68k.org> 2011-09-10 15:21:49 UTC ---
c746e7c154c7956a558958496ff6a0a04274dcb3 is the first bad commit
commit c746e7c154c7956a558958496ff6a0a04274dcb3
Author: jamborm <jamborm@138bc75d-0d04-0410-961f-82ee72b054a4>
Date:   Wed Sep 7 14:25:39 2011 +0000

    2011-09-07  Martin Jambor  <mjambor@suse.cz>

        PR tree-optimization/49911
        * tree-sra.c (analyze_access_subtree): Change type of to-be-replaced
        enumerations to the corresponding plain integer type.

        * testsuite/g++.dg/tree-ssa/pr49911.C: New test.



    git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@178639
138bc75d-0d04-0410-961f-82ee72b054a4

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu.org
          Component|tree-optimization           |bootstrap

--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> 2011-09-13 13:14:12 UTC ---
I can confirm it is caused by the aforementioned commit.  The stage 2
compiler ICEs on 

/* confdefs.h */
#define PACKAGE_NAME "GNU C Runtime Library"
#define PACKAGE_TARNAME "libgcc"
#define PACKAGE_VERSION "1.0"
#define PACKAGE_STRING "GNU C Runtime Library 1.0"
#define PACKAGE_BUGREPORT ""
#define PACKAGE_URL "http://www.gnu.org/software/libgcc/"
/* end confdefs.h.  */

int
main ()
{

  ;
  return 0;
}

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-09-16
     Ever Confirmed|0                           |1

--- Comment #6 from Martin Jambor <jamborm at gcc dot gnu.org> 2011-09-16 12:59:12 UTC ---
The miscompiled function seems to be evaluate_stmt or possibly the
functions it is inlined to because the transformation happens in early
SRA.

Nevertheless, when we revert to the old behavior (enum type scalar
replacements) when creating a replacement for val.lattice_val in this
function the bootstrap finishes.

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #7 from Martin Jambor <jamborm at gcc dot gnu.org> 2011-09-19 18:21:25 UTC ---
The compilation before and after the patch seems to diverge at expand
time and only in one instruction when processing this particular
gimple statement:

  MEM[(struct prop_value_d *)&D.39146].lattice_val = val$lattice_val_443;

When val$lattice_val is an enum type, it is expanded to:

(insn 4927 4926 4928 593 (set (mem/s/c:SI (reg/f:DI 2698) [10 MEM[(struct
prop_value_d *)&D.39146].lattice_val+0 S4 A128])
        (reg:SI 433 [ val$lattice_val ]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1685 -1
     (nil))

whereas when it is <unnamed-unsigned:32>, it is expanded to:

(insn 4927 4926 4928 593 (set (mem/s/c:SI (reg/f:DI 2698) [4 MEM[(struct
prop_value_d *)&D.39146].lattice_val+0 S4 A128])
        (reg:SI 433 [ val$lattice_val ]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1685 -1
     (nil))

The difference is 4 instead of 10.  Now I am entering an entirely
unchartered territory for me on quite a few levels so don't hold your
breath for any progress.  I will keep looking into this, though.

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #8 from Martin Jambor <jamborm at gcc dot gnu.org> 2011-09-21 12:18:46 UTC ---
The different alias set (4 instead of 10) is then just carried along
in the RTL dumps not causing any different behavior until
tree-ssa-ccp.c.192r.postreload where we get the following diff:

--- /tmp/zzz/1    2011-09-21 14:08:26.000000000 +0200
+++ /tmp/zzz/2    2011-09-21 14:10:18.000000000 +0200
@@ -48,6 +48,8 @@
 verify found no changes in insn with uid = 4249.
 verify found no changes in insn with uid = 4435.
 verify found no changes in insn with uid = 4531.
+rescanning insn with uid = 5957.
+deleting insn with uid = 5957.
 verify found no changes in insn with uid = 5017.
 verify found no changes in insn with uid = 5065.
 verify found no changes in insn with uid = 5096.
@@ -16908,12 +16910,12 @@
         (reg:DI 15 r15 [orig:2693 val ] [2693]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1685 5 {movdi_internal}
      (nil))

-(insn 4927 4910 5957 500 (set (mem/s/c:SI (reg/f:DI 14 r14 [2688]) [10
MEM[(struct prop_value_d *)&D.39146].lattice_val+0 S4 A128])
+(insn 4927 4910 5957 500 (set (mem/s/c:SI (reg/f:DI 14 r14 [2688]) [4
MEM[(struct prop_value_d *)&D.39146].lattice_val+0 S4 A128])
         (reg:SI 16 r16 [orig:433 val$lattice_val ] [433]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1685 4 {movsi_internal}
      (nil))

 (insn 5957 4927 5963 500 (set (reg:DI 8 r8)
-        (mem/s/c:DI (reg/f:DI 14 r14 [2688]) [0+0 S8 A128]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1686 5 {movdi_internal}
+        (reg:DI 15 r15 [orig:2693 val ] [2693]))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1686 5 {movdi_internal}
      (nil))

 (insn 5963 5957 4970 500 (use (reg/i:OI 8 r8))
/abuild/mjambor/trunk/src/gcc/tree-ssa-ccp.c:1686 -1

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[rsandifo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #9 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> 2011-09-27 17:18:32 UTC ---
Author: rsandifo
Date: Tue Sep 27 17:18:22 2011
New Revision: 179285

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=179285
Log:
gcc/
    PR middle-end/50386
    PR middle-end/50326
    * tree-sra.c (build_ref_for_model): Use the type of the field as
    the type of the COMPONENT_REF.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-sra.c

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jamborm at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

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

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

--- Comment #10 from Martin Jambor <jamborm at gcc dot gnu.org> 2011-09-28 11:51:35 UTC ---
This is now fixed.

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[sje at cup dot hp.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Steve Ellcey <sje at cup dot hp.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
             Blocks|                            |50350
         Resolution|FIXED                       |

--- Comment #11 from Steve Ellcey <sje at cup dot hp.com> 2011-10-13 16:15:40 UTC ---
The bad commit listed in comment #4 was back ported to the 4.6 branch so the
check in from comment #9 should be ported over to the 4.6 branch as well.  I am
reopening the bug for that reason.

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Mikael Pettersson <mikpe at it dot uu.se> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpe at it dot uu.se

--- Comment #12 from Mikael Pettersson <mikpe at it dot uu.se> 2011-10-18 13:04:09 UTC ---
(In reply to comment #11)
> The bad commit listed in comment #4 was back ported to the 4.6 branch so the
> check in from comment #9 should be ported over to the 4.6 branch as well.  I am
> reopening the bug for that reason.

The "bad" PR49911 commit was backported to 4.5 branch too, in r178701.  I
haven't checked yet if it caused any regressions there or not.

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-10-19 08:46:12 UTC ---
Author: jakub
Date: Wed Oct 19 08:45:45 2011
New Revision: 180179

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=180179
Log:
    PR target/50350
    Backport from mainline
    2011-09-27  Richard Sandiford  <rdsandiford@googlemail.com>

    PR middle-end/50386
    PR middle-end/50326
    * tree-sra.c (build_ref_for_model): Use the type of the field as
    the type of the COMPONENT_REF.

Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/tree-sra.c

[Bug bootstrap/50326] [4.7 regression] ICE in set_lattice_value, at tree-ssa-ccp.c:456

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50326

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org
         Resolution|                            |FIXED

--- Comment #14 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-10-21 13:35:17 UTC ---
> The "bad" PR49911 commit was backported to 4.5 branch too, in r178701.  I
> haven't checked yet if it caused any regressions there or not.

The 4.5 branch doesn't have build_ref_for_model and the only call to build3 on
a COMPONENT_REF already uses the type of the field, so it shouldn't be
affected.