[Bug tree-optimization/24231] New: ICE: SSA corruption

[Bug tree-optimization/24231] New: ICE: SSA corruption

[rguenth at gcc dot gnu dot org]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2289 bytes --]

/usr/lib/gcc/powerpc64-suse-linux/4.1.0/cc1plus -fpreprocessed
AttributeListImpl.ii -quiet -dumpbase AttributeListImpl.cpp -auxbase-strip
../../obj/AttributeListImpl.o -O2 -Wall -Wall -version -fmessage-length=0
-fno-elide-constructors -fPIC -o AttributeListImpl.s
 Conflict tempVector$m_memoryManager_1215(ab) and
tempVector$m_memoryManager_1153(ab) across an abnormal edge from BB39->BB42
/usr/src/packages/BUILD/xml-xalan/c/src/xalanc/PlatformSupport/AttributeListImpl.cpp:
In member function ‘xalanc_1_9::AttributeListImpl&
xalanc_1_9::AttributeListImpl::operator=(const
xalanc_1_9::AttributeListImpl&)’:
/usr/src/packages/BUILD/xml-xalan/c/src/xalanc/PlatformSupport/AttributeListImpl.cpp:102:
internal compiler error: SSA corruption
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.suse.de/feedback> for instructions.

#0  internal_error (gmsgid=0x10bbc18c "SSA corruption") at diagnostic.c:534
#1  0x108e3140 in coalesce_abnormal_edges (map=0x10d41878, graph=0x10d41e48,
    rv=0x10d222f8) at tree-outof-ssa.c:643
#2  0x108e4548 in coalesce_ssa_name (map=0x10d41878, flags=1)
    at tree-outof-ssa.c:912
#3  0x108e89d8 in remove_ssa_form (dump=0x0, map=0x10d41878, flags=1)
    at tree-outof-ssa.c:2402
#4  0x108e94c8 in rewrite_out_of_ssa () at tree-outof-ssa.c:2593
#5  0x10817060 in execute_one_pass (pass=0x10be3d80) at passes.c:827

(gdb) up 5
#5  0x10817060 in execute_one_pass (pass=0x10be3d80) at passes.c:827
827         pass->execute ();
(gdb) print *pass
$1 = {name = 0x10bbc7a4 "optimized", gate = 0,
  execute = 0x108e93ac <rewrite_out_of_ssa>, sub = 0x0, next = 0x10be3e50,
  static_pass_number = 97, tv_id = 69, properties_required = 732,
  properties_provided = 668, properties_destroyed = 64, todo_flags_start = 28,
  todo_flags_finish = 3, letter = 0 '\0'}


-- 
           Summary: ICE: SSA corruption
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: ice-on-invalid-code
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at gcc dot gnu dot org
GCC target triplet: powerpc-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] ICE: SSA corruption

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2005-10-06 10:06 -------
Created an attachment (id=9906)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9906&action=view)
testcase

Preprocessed testcase.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] ICE: SSA corruption

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2005-10-06 11:11 -------
Reducing.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2005-10-06 13:01 -------
-O1 -ftree-vrp -ftree-pre is only needed to reproduce this.

By the way, *pass here is useless as we don't check for this until we are doing
out of SSA.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
            Summary|ICE: SSA corruption         |[4.1 Regression] ICE: SSA
                   |                            |corruption
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2005-10-06 16:11 -------
Reduced testcase at -O1 -ftree-vrp -ftree-pre:
struct MemoryManager {
  virtual void deallocate() = 0;
};
struct XalanVector {
  ~XalanVector()   {
    m_memoryManager->deallocate();
  }
  void swap(XalanVector& theOther)   {
    MemoryManager* const theTempManager = m_memoryManager;
    m_memoryManager = theOther.m_memoryManager;
    theOther.m_memoryManager = theTempManager;
    theOther.m_size = 0;
  }
  void push_back()   {
    XalanVector theTemp(*this);
    theTemp.push_back();
    swap(theTemp);
  }
  MemoryManager* m_memoryManager;
  int m_size;
};
void f(void) {
  XalanVector tempVector;
  tempVector.push_back();
}


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-10-06 16:11:31
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[rguenth at gcc dot gnu dot org]

------- Comment #5 from rguenth at gcc dot gnu dot org  2005-10-06 16:15 -------
Reproduced on i686-pc-linux-gnu.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|powerpc-*-*                 |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2005-10-06 16:17 -------
This is a PRE bug as far as I can tell.
The diff between reassoc and PRE:
 <bb 1>:
   theTempManager_86 = theTemp$m_memoryManager_265(ab);
   #   VUSE <SFT.15_91>;
   D.1851_87 = theTemp.m_memoryManager;
   theTemp$m_memoryManager_268(ab) = D.1851_87;
   #   SFT.15_213 = V_MUST_DEF <SFT.15_91>;
   theTemp.m_memoryManager = theTempManager_86;
   #   SFT.14_214 = V_MUST_DEF <SFT.14_94>;
   theTemp.m_size = 0;
-  #   VUSE <SFT.15_213>;
-  D.1852_90 = theTemp.m_memoryManager;
+  D.1852_90 = theTemp$m_memoryManager_265(ab);
   #   VUSE <SFT.14_214>;
   #   VUSE <SFT.15_213>;
   D.1853_92 = D.1852_90->_vptr$MemoryManager;
   #   VUSE <SFT.14_214>;
   #   VUSE <SFT.15_213>;
   D.1854_93 = *D.1853_92;
   #   SFT.14_215(ab) = V_MAY_DEF <SFT.14_214>;
   #   SFT.15_216(ab) = V_MAY_DEF <SFT.15_213>;
   OBJ_TYPE_REF(D.1854_93;D.1852_90->0) (D.1852_90);
   goto <bb 3> (<L47>);

Notice how both theTemp$m_memoryManager_265 and theTemp$m_memoryManager_268 are
alive.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dberlin at gcc dot gnu dot
                   |                            |org
 GCC target triplet|                            |powerpc-*-*
      Known to fail|4.1.0                       |
      Known to work|3.3.2 4.0.2                 |3.3.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[janis187 at us dot ibm dot com]

------- Comment #7 from janis187 at us dot ibm dot com  2005-10-06 19:25 -------
A regression hunt using the testcase from comment #4 identified this patch
from pinskia:

  http://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00277.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] ICE: SSA corruption

[pinskia at gcc dot gnu dot org]

------- Comment #8 from pinskia at gcc dot gnu dot org  2005-10-06 19:58 -------
(In reply to comment #7)
> A regression hunt using the testcase from comment #4 identified this patch
> from pinskia:
And this is a latent bug because the dumps show that my patch just caused more
constant prop of &theTemp into *this_6 which causes to be simplified to
theTemp.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2005-10-13 19:01 -------
hmm, we need some more checking to catch this ICE earlier than out of SSA and
then people will believe me when I say this is a PRE bug.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions

[pinskia at gcc dot gnu dot org]

------- Comment #10 from pinskia at gcc dot gnu dot org  2005-10-13 22:56 -------
I think I know why PRE is doing somthing wrong.

we have before PRE:

  theTempManager_86 = theTemp$m_memoryManager_265(ab);
...
  theTemp$m_memoryManager_268(ab) = D.1851_87;
....
  theTemp.m_memoryManager = theTempManager_86;
...
  D.1852_90 = theTemp.m_memoryManager;

Daniel could you look at this further?

There is recursive inlining, constant propagation and SRA getting involved
which is gets to the state we have by the time we get to PRE.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[pinskia at gcc dot gnu dot org]

------- Comment #11 from pinskia at gcc dot gnu dot org  2005-10-14 01:30 -------
Here is a testcase for FRE which can reproduced without -ftree-vrp -ftree-pre
at -O1:
struct f
{
  int i;
};
struct h{h();};
int g(void);
int g1(void) throw();
int h2222(f*);
void ghh(int);

int main(void)
{
  int i;
  f t;
  try
  {
    i = g1();
    try
    {
      i = g();
    }catch(...)
    {}
    int j = i;
    try
    {
      i = g();
    }catch(...)
    {}
    t.i = j;
    i = 2;
    int h = t.i;
    ghh (h);

    g();
  }catch(...)
  {}
  return i;
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[pinskia at gcc dot gnu dot org]

------- Comment #12 from pinskia at gcc dot gnu dot org  2005-10-14 01:42 -------
Thanks to Daniel, this is caused by try_look_through_load in tree-ssa-pre.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[pinskia at gcc dot gnu dot org]

------- Comment #13 from pinskia at gcc dot gnu dot org  2005-10-14 02:16 -------
Here is another testcase which is a related bug but is fixed easier by a check
for SSA_NAME_OCCURS_IN_ABNORMAL_PHI in try_look_through_load:
struct f
{
  int i;
};
struct h{h();};
int g(void);
int g1(void) throw();
int h2222(f*);
void ghh(int);

int main(void)
{
  int i;
  f t;
  try
  {
    i = g1();
    try
    {
      i = g();
    }catch(...)
    {}
    int j = i;
    try
    { t.i = i;
      i = g();
    }catch(...)
   {}
    i = 2;
    int h = t.i;
    ghh (h);

    g();
  }catch(...)
  {}
  return i;
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[pinskia at gcc dot gnu dot org]

------- Comment #14 from pinskia at gcc dot gnu dot org  2005-10-17 15:03 -------
Assigning to D. Berlin per his request.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |dberlin at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[cvs-commit at gcc dot gnu dot org]

------- Comment #15 from cvs-commit at gcc dot gnu dot org  2005-10-19 03:34 -------
Subject: Bug 24231

CVSROOT:        /cvs/gcc
Module name:    gcc
Changes by:     dberlin@gcc.gnu.org     2005-10-19 03:34:50

Modified files:
        gcc            : ChangeLog tree-ssa-pre.c 
Added files:
        gcc/testsuite/g++.dg/tree-ssa: pr24231-1.C pr24231-2.C 
                                       pr24231-3.C 

Log message:
        2005-10-18  Daniel Berlin <dberlin@dberlin.org>

        Fix PR tree-optimization/24231

        * tree-ssa-pre.c (try_look_through_load): Skip abnormal phi names
        (compute_avail): Ditto.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.10181&r2=2.10182
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-ssa-pre.c.diff?cvsroot=gcc&r1=2.99&r2=2.100
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/tree-ssa/pr24231-1.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/tree-ssa/pr24231-2.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/tree-ssa/pr24231-3.C.diff?cvsroot=gcc&r1=NONE&r2=1.1


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231

[Bug tree-optimization/24231] [4.1 Regression] SSA corruption with C++ code and exceptions and loads

[dberlin at gcc dot gnu dot org]

------- Comment #16 from dberlin at gcc dot gnu dot org  2005-10-19 03:35 -------
Fixed


-- 

dberlin at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24231