[Bug middle-end/17902] New: [4.0] ICE from tree_verify_flow_info

[Bug middle-end/17902] New: [4.0] ICE from tree_verify_flow_info

[gcc-bugzilla at gcc dot gnu dot org]

g++ crashes compiling the code below, when i use 
 -O3 -mtune=pentiumpro:

$ /usr/local/gcc/libexec/gcc/i686-pc-linux-gnu/4.0.0/cc1plus -quiet  x.cc -quiet  -mtune=pentiumpro  -O3
x.cc: In function `void fill_ntuples()':
x.cc:16: internal compiler error: tree check: expected class 'declaration', have 'statement' (label_expr) in tree_verify_flow_info, at /tree-cfg.c:3504
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
$


At this point, all my attempts to simplify the example further make the
crash go away.

Environment:
System: Linux karma 2.6.8.1 #20 Mon Sep 13 23:48:47 EDT 2004 i686 i686 i386 GNU/Linux
Architecture: i686

	<machine, os, target, libraries (multiple lines)>
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: i686-pc-linux-gnu
configured with: /home/sss/gcc/gcc/configure --prefix=/usr/local/gcc --enable-threads=posix --enable-long-long --enable-languages=c,c++,f95

How-To-Repeat:

Compile the code below with  -O3 -mtune=pentiumpro.

------------------------------------------------
extern int foo();

bool a;
bool b;
bool f() { return a || b; }

struct A { ~A(){} };

struct vector
{
  ~vector() { if (f()) foo();}
  A _M_impl;
};

 
void fill_ntuples()
{      
  vector smt_states;

  int nbar4=0;
  bool log3 = false;
  for (int j=0;j<foo();j++){
    int jdeter = foo();
    if (0!=j&&jdeter==1) foo();
    
    if (jdeter==1&&0!=j)
    {
      if (foo()==4){
        nbar4++;
      }
    }
    if (foo()==3) log3=true;
  }
       
  int jjhi=foo();
  for (int jj=0;jj<jjhi;jj++)
  {
    if (
        (jj==0)||
        (jj==1&&log3&&nbar4>=3)
        )
    {
      vector substates;
    }
  }
}
------------------------------------------------
------- Additional Comments From snyder at fnal dot gov  2004-10-08 21:54 -------
Fix:
	<how to correct or work around the problem, if known (multiple lines)>

-- 
           Summary: [4.0] ICE from tree_verify_flow_info
           Product: gcc
           Version: 0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: snyder at fnal dot gov
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug middle-end/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 22:08 -------
Confirmed but not reduced all the way.  This is caused by -funswitch-loops on the tree level.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rakdver at gcc dot gnu dot
                   |                            |org
           Keywords|                            |ice-on-valid-code
            Summary|[4.0] ICE from              |[4.0 Regression] ICE from
                   |tree_verify_flow_info       |tree_verify_flow_info
   Target Milestone|---                         |4.0.0
            Version|0.0                         |2.95


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug middle-end/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 22:10 -------
This is also exception handling related as -fno-exceptions also does not ICE.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |EH


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |tree-optimization


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|2.95                        |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 22:24 -------
Here is something which produces the least amount of BBs (only one exception handling try/catch 
block):
extern int foo();
extern int foo1() throw();


bool a;
bool b;
static inline bool f()throw() { return a || b; }

struct A { ~A()throw(){} };

struct vector
{
  ~vector() { if (f()) foo1();}
  A _M_impl;
};

 
void fill_ntuples()throw()
{      
  vector smt_states;

  int nbar4=0;
  bool log3 = false;
  int t = foo1();
  for (int j=0;j<t;j++)
  {
    int jdeter = foo1();
    if (0!=j)if(jdeter==1) foo();
    if (0!=j)if(jdeter==1)
    {
      if (foo1()==4){
        nbar4++;
      }
    }
    if (foo1()==3) log3=true;
  }
  int jjhi=foo1();
  for (int jj=0;jj<jjhi;jj++)
  {
    if (jj==1&&log3&&nbar4>=3)
    {
      vector substates;
    }
  }
}


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-10-08 22:24:17
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[rakdver at gcc dot gnu dot org]

------- Additional Comments From rakdver at gcc dot gnu dot org  2004-10-08 22:38 -------
This seems to be a bug in phi opts.  Concretely they must somewhere assume that
there is exactly one label at the start of the basic block, or something like 
that.  Before phi opts:

;; prev block 11, next block 13
;; pred:       10 [32.6%]  (false,exec) 11 [100.0%]  (fallthru)
;; succ:       13 [100.0%]  (fallthru,dfs_back)
# jj_116 = PHI <1(10), jj_58(11)>;
<L29>:;
<L30>:;

After phi opts:

;; basic block 12, loop depth 1, count 0
;; prev block 10, next block 13
;; pred:       10 [32.6%]  (fallthru,exec)
;; succ:       13 [100.0%]  (fallthru,dfs_back)
<L29>:;
jj_116 = jj_58;
<L30>:;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 22:39 -------
I will look at it then.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 22:41 -------
Note this does not effect the tcb branch as phi-opt is differently.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[reichelt at gcc dot gnu dot org]

------- Additional Comments From reichelt at gcc dot gnu dot org  2004-10-08 22:52 -------
Here's something shorter. Maybe related to PR 17766 which is also
triggered by -funswitch-loops.

==============================================
void foo();

struct A { ~A(){ foo(); } };

struct B { A a; };

void bar()
{
    A a;
    bool b = false;
    int i, j;

    for (j=0; j<i; j++)
    {
        if (i) b=true;
        if (j && i) foo();
        if (j && i) i++;
    }

    for (j=0; j<i; j++)
        if ( !j || (j==1 && b && i) )
            B x;
}
==============================================


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 23:03 -------
I have a patch, the problem is that PHI-OPT should be using bsi_after_labels instead of bsi_start.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 23:10 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00792.html>.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-08 23:24 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-08 23:26 -------
Subject: Bug 17902

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	pinskia@gcc.gnu.org	2004-10-08 23:25:51

Modified files:
	gcc            : ChangeLog tree-ssa-phiopt.c 

Log message:
	2004-10-08  Andrew Pinski  <pinskia@physics.uc.edu>
	
	PR tree-opt/17902
	* tree-ssa-phiopt.c (conditional_replacement): Use bsi_after_labels
	instead of bsi_start.
	(value_replacement): Likewise.
	(abs_replacement): Likewise

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.5818&r2=2.5819
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-ssa-phiopt.c.diff?cvsroot=gcc&r1=2.16&r2=2.17



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902

[Bug tree-optimization/17902] [4.0 Regression] ICE from tree_verify_flow_info

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-08 23:29 -------
Subject: Bug 17902

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	pinskia@gcc.gnu.org	2004-10-08 23:29:36

Modified files:
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/g++.dg/opt: pr17902.C 

Log message:
	2004-10-08  Andrew Pinski  <pinskia@physics.uc.edu>
	
	PR tree-opt/17902
	* g++.dg/opt/pr17902.C: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4417&r2=1.4418
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17902.C.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17902