[Bug rtl-optimization/44012] New: ICE: SIGSEGV in ira_merge_allocno_live_ranges

[Bug rtl-optimization/44012] New: ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

Tested revisions:
r159062 - crash
r159045 - crash
r158683 - OK

Output with valgrind:
$ valgrind --trace-children=yes -q
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/bin/g++ -O2 -fno-strict-aliasing
sqcompiler-noempty.ii
==28996== Invalid read of size 4
==28996==    at 0x801DD0: ira_merge_allocno_live_ranges (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x802FFE: remove_unnecessary_regions (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x80478C: ira_build (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x7FEE2A: rest_of_handle_ira (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x8592BB: execute_one_pass (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x859544: execute_pass_list (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x859556: execute_pass_list (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x94E5B5: tree_rest_of_compilation (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0xAD6CB5: cgraph_expand_function (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0xAD9569: cgraph_optimize (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0xAD9CE4: cgraph_finalize_compilation_unit (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==    by 0x5515EA: cp_write_global_declarations (in
/mnt/svn/gcc-trunk/binary-159062-lto-fortran/libexec/gcc/x86_64-unknown-linux-gnu/4.6.0/cc1plus)
==28996==  Address 0xafafafafafafafb7 is not stack'd, malloc'd or (recently)
free'd
==28996==
sqcompiler-noempty.ii: In member function 'void SQCompiler::BitwiseOrExp()':
sqcompiler-noempty.ii:2877:2: 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.

Testcase will be attached soon.


-- 
           Summary: ICE: SIGSEGV in ira_merge_allocno_live_ranges
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: zsojka at seznam dot cz
  GCC host triplet: x86_64-pc-linux-gnu
GCC target triplet: x86_64-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #1 from zsojka at seznam dot cz  2010-05-06 13:50 -------
Created an attachment (id=20585)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20585&action=view)
partially reduced testcase

Compiler flags:
g++ -O1 -fgcse -findirect-inlining -fipa-sra
or
g++ -O2 -fno-strict-aliasing

Valgind output:
$ valgrind --trace-children=yes -q
/mnt/svn/gcc-trunk/binary-159045-lto-fortran/bin/g++ -O1 -fgcse
-findirect-inlining -fipa-sra testcase-min2.ii
==18710== Invalid read of size 4
==18710==    at 0x801A40: ira_merge_allocno_live_ranges (ira-build.c:863)
==18710==    by 0x802C7E: remove_unnecessary_regions (ira-build.c:1977)
==18710==    by 0x804408: ira_build (ira-build.c:2778)
==18710==    by 0x7FEA6E: rest_of_handle_ira (ira.c:3215)
==18710==    by 0x8588AD: execute_one_pass (passes.c:1590)
==18710==    by 0x858B34: execute_pass_list (passes.c:1645)
==18710==    by 0x858B46: execute_pass_list (passes.c:1646)
==18710==    by 0x94CDB5: tree_rest_of_compilation (tree-optimize.c:413)
==18710==    by 0xAD49F5: cgraph_expand_function (cgraphunit.c:1627)
==18710==    by 0xAD7289: cgraph_optimize (cgraphunit.c:1706)
==18710==    by 0xAD7A04: cgraph_finalize_compilation_unit (cgraphunit.c:1166)
==18710==    by 0x550FCA: cp_write_global_declarations (decl2.c:3837)
==18710==  Address 0xafafafafafafafb7 is not stack'd, malloc'd or (recently)
free'd
==18710==
testcase-min2.ii: In member function 'void SQCompiler::BitwiseXorExp()':
testcase-min2.ii:371:2: internal compiler error: Segmentation fault

I will post more reduced testcase, delta is quite slow with this code.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #2 from zsojka at seznam dot cz  2010-05-06 14:38 -------
When reducing this testcase, PR43791 was triggered. Upgrading to r159108 to
verify this still crashes...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[hjl dot tools at gmail dot com]

------- Comment #3 from hjl dot tools at gmail dot com  2010-05-06 14:45 -------
It was caused by revision 158827:

http://gcc.gnu.org/ml/gcc-cvs/2010-04/msg00934.html


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #4 from zsojka at seznam dot cz  2010-05-06 15:04 -------
Created an attachment (id=20589)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20589&action=view)
reduced testcase

Command line:
gcc -O1 -fgcse -findirect-inlining pr44012.c

not tested with r159108 yet


-- 

zsojka at seznam dot cz changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #20585|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #5 from zsojka at seznam dot cz  2010-05-06 15:54 -------
Still crashes in r159109, so not a dup of PR43791


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[jamborm at gcc dot gnu dot org]

------- Comment #6 from jamborm at gcc dot gnu dot org  2010-05-06 16:38 -------
(In reply to comment #4)
> Created an attachment (id=20589)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20589&action=view) [edit]
> reduced testcase
> 
> Command line:
> gcc -O1 -fgcse -findirect-inlining pr44012.c
> 

It crashes for me on both i686 and x86_64 even without
-findirect-inlining.  On the other hand, -fno-inline does prevent the
ICE.

(In reply to comment #3)
> It was caused by revision 158827:
> 
> http://gcc.gnu.org/ml/gcc-cvs/2010-04/msg00934.html
> 

The testcase from comment #4 crashes for me even on revision 158826
(on x86-64) and so I don't think that my commit is relevant.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #7 from zsojka at seznam dot cz  2010-05-06 16:47 -------
You are right, the reduced testcase from comment #4 crashes with:

Command line:
$ gcc -O1 -fgcse pr44012.c

Tested revisions:
r159109 - crash
r158683 - crash
r157702 - crash
r153685 - crash
4.5 r158978 - crash
4.4 r158133 - OK
4.4 r157895 - OK


-- 

zsojka at seznam dot cz changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|ICE: SIGSEGV in             |[4.5/4.6 Regression] ICE:
                   |ira_merge_allocno_live_range|SIGSEGV in
                   |s                           |ira_merge_allocno_live_range
                   |                            |s


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[hjl dot tools at gmail dot com]

------- Comment #8 from hjl dot tools at gmail dot com  2010-05-06 18:57 -------
(In reply to comment #4)
> Created an attachment (id=20589)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20589&action=view) [edit]
> reduced testcase
> 
> Command line:
> gcc -O1 -fgcse -findirect-inlining pr44012.c
> 
> not tested with r159108 yet
> 

This is caused by revision 147852:

http://gcc.gnu.org/ml/gcc-cvs/2009-05/msg00829.html

There may be 2 bugs.


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-05-06 18:57:42
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[hjl dot tools at gmail dot com]

------- Comment #9 from hjl dot tools at gmail dot com  2010-05-06 21:01 -------
Created an attachment (id=20590)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20590&action=view)
A patch

This patch avoids ICE.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[zsojka at seznam dot cz]

------- Comment #10 from zsojka at seznam dot cz  2010-05-07 00:32 -------
(In reply to comment #9)

That patch fixes all testcases and flag combinations I tried (eg. gcc doesn't
crash)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[rguenth at gcc dot gnu dot org]

------- Comment #11 from rguenth at gcc dot gnu dot org  2010-05-07 09:46 -------
There may be a dup of this bug (can't find it right now).


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.5.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[vmakarov at redhat dot com]

------- Comment #12 from vmakarov at redhat dot com  2010-05-07 17:49 -------
  When allocno is finished, its some info is propagated into upper allocno. 
When several allocnos with same regno are finished, info can be propagated
directly to survived upper allocno or through one allocno will be finished.  It
depends on region configuration and order of allocnos with the same regno in
the corresponding list.  The sigsegv occurs in the second case when we remove
allocno and propagates this info through allready removed allocno.  It happens
because regno_allocno_map which is used to find allocno into which the info to
propagate is not nullified after removing allocno.

H.J.'s patch idea is right but the patch is complicated.  I'll send a simplier
patch soon.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[hjl at gcc dot gnu dot org]

------- Comment #13 from hjl at gcc dot gnu dot org  2010-05-10 14:02 -------
Subject: Bug 44012

Author: hjl
Date: Mon May 10 14:02:43 2010
New Revision: 159223

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159223
Log:
Nullify regno_allocno_map of the removed allocno.

gcc/

2010-05-10  H.J. Lu<hongjiu.lu@intel.com>
            Vladimir Makarov<vmakarov@redhat.com>

        PR rtl-optimization/44012
        * ira-build.c (remove_unnecessary_allocnos): Nullify
        regno_allocno_map of the removed allocno.

gcc/testsuite/

2010-05-10  H.J. Lu  <hongjiu.lu@intel.com>

        PR rtl-optimization/44012
        * gcc.dg/pr44012.c: New.

Added:
    trunk/gcc/testsuite/gcc.dg/pr44012.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ira-build.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[hjl at gcc dot gnu dot org]

------- Comment #14 from hjl at gcc dot gnu dot org  2010-05-10 14:05 -------
Subject: Bug 44012

Author: hjl
Date: Mon May 10 14:05:16 2010
New Revision: 159224

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159224
Log:
Nullify regno_allocno_map of the removed allocno.

gcc/

2010-05-10  H.J. Lu  <hongjiu.lu@intel.com>

        Backport from mainline
        2010-05-10  H.J. Lu<hongjiu.lu@intel.com>
                    Vladimir Makarov<vmakarov@redhat.com>

        PR rtl-optimization/44012
        * ira-build.c (remove_unnecessary_allocnos): Nullify
        regno_allocno_map of the removed allocno.

gcc/testsuite/

2010-05-10  H.J. Lu  <hongjiu.lu@intel.com>

        Backport from mainline
        2010-05-10  H.J. Lu  <hongjiu.lu@intel.com>

        PR rtl-optimization/44012
        * gcc.dg/pr44012.c: New.

Added:
    branches/gcc-4_5-branch/gcc/testsuite/gcc.dg/pr44012.c
Modified:
    branches/gcc-4_5-branch/gcc/ChangeLog
    branches/gcc-4_5-branch/gcc/ira-build.c
    branches/gcc-4_5-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012

[Bug rtl-optimization/44012] [4.5/4.6 Regression] ICE: SIGSEGV in ira_merge_allocno_live_ranges

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2010-05-19 12:54 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
      Known to fail|                            |4.5.0
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44012