[Bug middle-end/43951] New: [4.6 Regression] 483.xalancbmk in SPEC CPU 2006 is miscompiled

[Bug middle-end/43951] New: [4.6 Regression] 483.xalancbmk in SPEC CPU 2006 is miscompiled

[hjl dot tools at gmail dot com]

On Linux/x86-64, revision 158930 miscompiled 483.xalancbmk in SPEC CPU 2006
with -O2 -ffast-math:

  Running 483.xalancbmk ref base o2 default

483.xalancbmk: copy #0 non-zero return code (rc=0, signal=11)

Error with
'/export/gnu/import/svn/gcc-test/spec/2006/x86_64/spec/bin/specinvoke -E -d
/export/gnu/import/svn/gcc-test/spec/2006/x86_64/spec/benchspec/CPU2006/483.xalancbmk/run/run_base_ref_o2.0000
-c 1 -e compare.err -o compare.stdout -f compare.cmd': check file
'/export/gnu/import/svn/gcc-test/spec/2006/x86_64/spec/benchspec/CPU2006/483.xalancbmk/run/run_base_ref_o2.0000/.err'
Error: 1x483.xalancbmk

Revision 158912 is OK.


-- 
           Summary: [4.6 Regression] 483.xalancbmk in SPEC CPU 2006 is
                    miscompiled
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hjl dot tools at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] 483.xalancbmk in SPEC CPU 2006 is miscompiled

[hjl dot tools at gmail dot com]

-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.6.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #1 from hjl dot tools at gmail dot com  2010-04-30 22:00 -------
Revision 158918:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|hubicka at gcc dot gnu dot  |
                   |org                         |
            Summary|[4.6 Regression]            |[4.6 Regression] Revsion
                   |483.xalancbmk in SPEC CPU   |158918 miscompiled
                   |2006 is miscompiled         |483.xalancbmk in SPEC CPU
                   |                            |2006


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #2 from hjl dot tools at gmail dot com  2010-05-01 00:13 -------
StylesheetExecutionContextDefault.cpp is miscompiled.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #3 from hjl dot tools at gmail dot com  2010-05-01 00:15 -------
Revision 158918:

http://gcc.gnu.org/ml/gcc-cvs/2010-04/msg01025.html

is the cause.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2010-05-01 00:24 -------
(In reply to comment #3)
> is the cause.

What happens if you revert the build_new_1 part?  That is not part of the
changelog.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #4 from hjl dot tools at gmail dot com  2010-05-01 00:23 -------
This patch:

---
Index: cp/init.c
===================================================================
--- cp/init.c   (revision 158918)
+++ cp/init.c   (working copy)
@@ -1911,7 +1911,7 @@ build_new_1 (VEC(tree,gc) **placement, t

   is_initialized = (TYPE_NEEDS_CONSTRUCTING (elt_type) || *init != NULL);

-  if (*init == NULL)
+  if (*init == NULL && !type_has_user_provided_constructor (elt_type))
     {
       bool maybe_uninitialized_error = false;
       /* A program that calls for default-initialization [...] of an
---

fixed the problem. Jason, does it make any senses?


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at redhat dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revsion 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #6 from hjl dot tools at gmail dot com  2010-05-01 01:15 -------
(In reply to comment #5)
> (In reply to comment #3)
> > is the cause.
> 
> What happens if you revert the build_new_1 part?  That is not part of the
> changelog.
> 

It works since my patch in comment #4 is the minimal change to revert
the build_new_1 change.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #7 from hjl dot tools at gmail dot com  2010-05-01 03:10 -------
Here is a testcase:

[hjl@gnu-34 483.xalancbmk]$ cat x.cc
class Foo 
{
public:
  Foo (void);
  const int xxx;
};
Foo *
bar ()
{
  return new Foo;
}
[hjl@gnu-34 483.xalancbmk]$ /export/gnu/import/rrs/158918/usr/bin/g++ -S -O2
x.cc
[hjl@gnu-34 483.xalancbmk]$ cat x.s
        .file   "x.cc"
        .text
        .p2align 4,,15
.globl _Z3barv
        .type   _Z3barv, @function
_Z3barv:
.LFB0:
        .cfi_startproc
        rep
        ret
        .cfi_endproc
.LFE0:
        .size   _Z3barv, .-_Z3barv


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug middle-end/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #8 from hjl dot tools at gmail dot com  2010-05-01 03:13 -------
A run-time testcase:

[hjl@gnu-34 483.xalancbmk]$ cat x.cc
extern "C" void abort (void);

class Foo 
{
public:
  Foo (void) : xxx (1) {};
  const int xxx;
};
Foo *
bar ()
{
  return new Foo;
}

int
main ()
{
  Foo *p = bar ();

  if (p->xxx != 1)
    abort ();
  return 0;
}
[hjl@gnu-34 483.xalancbmk]$ /export/gnu/import/rrs/158918/usr/bin/g++  -O2 x.cc
[hjl@gnu-34 483.xalancbmk]$ ./a.out 
Segmentation fault
[hjl@gnu-34 483.xalancbmk]$ 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|middle-end                  |c++
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-05-01 03:17:24
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #9 from hjl at gcc dot gnu dot org  2010-05-01 13:53 -------
Subject: Bug 43951

Author: hjl
Date: Sat May  1 13:52:52 2010
New Revision: 158959

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158959
Log:
Revert the accidental checkin in revision 158918.

gcc/cp/

2010-05-01  H.J. Lu  <hongjiu.lu@intel.com>

        PR c++/43951
        * init.c (build_new_1): Revert the accidental checkin in
        revision 158918.

gcc/testsuite/

2010-05-01  H.J. Lu  <hongjiu.lu@intel.com>

        PR c++/43951
        * g++.dg/init/new28.C: New.

Added:
    trunk/gcc/testsuite/g++.dg/init/new28.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/init.c
    trunk/gcc/testsuite/ChangeLog


------- Comment #10 from hjl dot tools at gmail dot com  2010-05-02 01:27 -------
Fixed.


-- 

hjl dot tools at gmail dot com changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[fabien dot chene at gmail dot com]

------- Comment #11 from fabien dot chene at gmail dot com  2010-05-02 16:36 -------
Hi,

Sorry for introducing this bug, thanks for the analysis, but unfortunately,
this is the wrong fix.
The change reverted wasn't accidental as it is redundant with the check done in
diagnose_uninitialized_cst_or_ref_member_1.
The problem here is that we return error_mark_node whereas the code can be
valid.
I'll fix that soon.

Could someone please reopen it, then assign me ?
Thanks.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #12 from rguenth at gcc dot gnu dot org  2010-05-02 18:13 -------
Reopened per request.


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |fabien dot chene at gmail
                   |dot org                     |dot com
             Status|REOPENED                    |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[jason at gcc dot gnu dot org]

------- Comment #13 from jason at gcc dot gnu dot org  2010-05-07 15:52 -------
Subject: Bug 43951

Author: jason
Date: Fri May  7 15:52:06 2010
New Revision: 159158

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159158
Log:
        PR c++/43951
        * init.c (diagnose_uninitialized_cst_or_ref_member_1): Returns the
        error count. Emit errors only if compain is true.
        (build_new_1): Do not return error_mark_node if
        diagnose_uninitialized_cst_or_ref_member_1 does not diagnose any
        errors. Delay the check for user-provided constructor.
        (perform_member_init): Adjust.
        * cp-tree.h (diagnose_uninitialized_cst_or_ref_member): Change the
        prototype.

Added:
    trunk/gcc/testsuite/g++.dg/init/new29.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/cp-tree.h
    trunk/gcc/cp/init.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951

[Bug c++/43951] [4.6 Regression] Revision 158918 miscompiled 483.xalancbmk in SPEC CPU 2006

[fabien dot chene at gmail dot com]

------- Comment #14 from fabien dot chene at gmail dot com  2010-05-11 16:43 -------
Fixed (committed by Jason).


-- 

fabien dot chene at gmail dot com changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43951