[Bug tree-optimization/17724] New: [4.0 Regression] wrong dominator (eh/fold all builtins) related

[Bug tree-optimization/17724] New: [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

namespace ACE_OS {char *strcpy (char *s, const char *t);}
extern "C" char *strcpy (char *__restrict __dest, __const char *__restrict __src)  throw ();
inline char *ACE_OS::strcpy (char *s, const char *t){return ::strcpy (s, t); }
struct decon{~decon();};
int fg();
void f();
int main (int, char *[])
{
  decon cmd_opts;
  int option;
  char config_file[4096];
  ACE_OS::strcpy (config_file, "HAStatus.conf");
  while ((option = fg()) != (-1))
    if (option)
       return -1;
  return 0;
}

-- 
           Summary: [4.0 Regression] wrong dominator (eh/fold all builtins)
                    related
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code, EH
          Severity: critical
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-29 13:41 -------
This was the last preprocessed source in PR 17697, I jsut reduced it.
: Search converges between 2004-07-21-trunk (#492) and 2004-07-23-trunk (#493).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |micis at gmx dot de, jakub
                   |                            |at redhat dot com
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-09-29 13:41:56
               date|                            |
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-09-29 19:18 -------
This ICE is caused by the patch to fix http://gcc.gnu.org/bugzilla/show_bug.cgi?
id=17697

If you build a gcc without this patch you get:

src.ii: In function `int main(int, char**)':
scr.ii:7: error: Statement marked for throw, but doesn't.
D.1829_4 = D.1844_32;

src.ii:7: internal compiler error: verify_stmts failed.
Please submit a full bug report, with preprocessed source if appropriate.

With that patch you get:

src.ii: In function `int main(int, char**)':
src.ii:7: error: dominator of 4 should be 2, not 0
src.ii:7: internal compiler error: in verify_dominators, at dominance.c:860
Please submit a full bug report, with preprocessed source if appropriate.

Michael Cieslinski

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-29 21:02 -------
Actually the ICE is no caused by that patch, we just don't get passed verify_stmts before.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-29 22:31 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-09/msg02990.html>.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-09-30 08:56 -------
Many thanks to all gcc developers for your excellent work!

I applied your patch, bootstraped gcc (on ppc-linux-gnu), run make check and 
found no new regressions.
Also my application seems to compile (it takes a lot of time)

Michael Cieslinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-09-30 11:29 -------
After compiling ~2500 files I found one where this error still occures.
Below is a reduced testcase.

/usr/local/gcc40/bin/g++40 -O2 -c -o out.o src.ii -v
Reading specs from /usr/local/gcc40/lib/gcc/powerpc-unknown-linux-
gnu/4.0.0/specs
Configured with: ../gcc40/configure --prefix=/usr/local/gcc40 --program-
suffix=40 --with-cpu=G5 --enable-altivec --enable-languages=c,c++ --enable-
checking
Thread model: posix
gcc version 4.0.0 20040926 (experimental)
 /usr/local/gcc40/libexec/gcc/powerpc-unknown-linux-gnu/4.0.0/cc1plus -
fpreprocessed src.ii -quiet -dumpbase src.ii -mcpu=G5 -auxbase-strip out.o -O2 -
version -o /tmp/ccRGaRii.s
GNU C++ version 4.0.0 20040926 (experimental) (powerpc-unknown-linux-gnu)
        compiled by GNU C version 4.0.0 20040926 (experimental).
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
src.ii: In member function `int C::D()':
src.ii:11: error: dominator of 3 should be 2, not 0
src.ii:11: internal compiler error: in verify_dominators, at dominance.c:860
Please submit a full bug report, with preprocessed source if appropriate.


Michael Cieslinski


extern char *strcpy (char* d, const char* s) throw ();

class A { public: A (); ~A (); };

inline char * B (char *s, const char *t)
{ return ::strcpy (s, t); }

class C { int D (void); int E; };


int C::D (void)
{
  A aa;
  try 
    {
      char z [22];
      if (this->E)   B (z, "");
      return 0;
    }

  catch (int &f) { return -1; } 
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-09-30 11:57 -------
Jakub, I'm assigning this bug to you since you posted a patch for it.

Can you check if the new testcase is the same bug which still surfaces, or a 
new one (for which a new PR should be opened)? Thanks.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jakub at redhat dot com
                   |dot org                     |
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[jakub at gcc dot gnu dot org]

------- Additional Comments From jakub at gcc dot gnu dot org  2004-10-01 12:41 -------
Well, the testcase as is certainly doesn't fail with current CVS plus
http://gcc.gnu.org/ml/gcc-patches/2004-09/msg02990.html.
When the first line is one of:
extern char *strcpy (char* d, const char* s) throw ();
extern "C" char *strcpy (char* d, const char* s) throw ();
extern "C" char *strcpy (char* d, const char* s);

it works just fine (though probably all 4 variants should be added to the
testsuite).
But when the testcase starts with:

extern char *strcpy (char* d, const char* s);

it ICEs.  The problem is different though, particularly that
delete_unreachable_blocks nor delete_basic_block called from it updates
dominators info.

IMHO either tree-cfg.c's remove_bb needs to
recount_dominator/set_immediate_dominator if changed for all preds and succs bb's
or cleanup_tree_cfg needs to set some flag when delete_unreachable_blocks returns
true and after the loop call free_dominance_info if that flag is set.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-07 21:26 -------
Newest patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00658.html>.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-10 18:15 -------
Newest patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00853.html>.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-10-12 07:25 -------
Whith the patch http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00853.html
applied I get a new ICE when I compile ACE542:

/usr/local/gcc40/bin/g++40 -W -Wall -Wpointer-arith -fpermissive  -O2   -
D_REENTRANT -DACE_HAS_AIO_CALLS -D_GNU_SOURCE   -I/home/cie019/ace542-
gcc40 /ACE_wrappers -DACE_HAS_EXCEPTIONS -D__ACE_INLINE__ -I..  -c -
o .obj/Map_Test.o Map_Test.cpp -g
Map_Test.cpp: In function `void functionality_test(MAP&, size_t)':
Map_Test.cpp:44: internal compiler error: in cp_tree_equal, at /cp/tree.c:1586
Please submit a full bug report, with preprocessed source if appropriate.

But when I specify -save-temps, this ICE does not occur.
Also it does not occur if I omit "-g".
How should I proceed?

Michael Cieslinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-10-12 08:45 -------
If it cannot be reproduced with --save-temps, it is probably a memory 
corruption or something. I think we should do one thing at a time: once Jakub's 
patch is in, we will look into the following problem. Be patient, we will get 
to it :)

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-10-12 08:48 -------
I tried to copy all needed files in a directory and while this file set was 
still not complete I got a segmentation fault when I tried to compile.
I filed a bug report bug17944

Michael Cieslinski

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[micis at gmx dot de]

------- Additional Comments From micis at gmx dot de  2004-10-13 07:22 -------
The ICE is not related to this patch.
Michael Cieslinski

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-13 11:27 -------
Subject: Bug 17724

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	jakub@gcc.gnu.org	2004-10-13 11:27:02

Modified files:
	gcc            : ChangeLog tree-cfg.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/g++.dg/opt: pr17724-1.C pr17724-2.C pr17724-3.C 
	                          pr17724-4.C pr17724-5.C pr17724-6.C 

Log message:
	PR tree-optimization/17724
	* tree-cfg.c (tree_purge_dead_eh_edges): Free dominance info.
	
	* g++.dg/opt/pr17724-1.C: New test.
	* g++.dg/opt/pr17724-2.C: New test.
	* g++.dg/opt/pr17724-3.C: New test.
	* g++.dg/opt/pr17724-4.C: New test.
	* g++.dg/opt/pr17724-5.C: New test.
	* g++.dg/opt/pr17724-6.C: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.5856&r2=2.5857
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-cfg.c.diff?cvsroot=gcc&r1=2.74&r2=2.75
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4440&r2=1.4441
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-1.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-2.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-3.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-4.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-5.C.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/g++.dg/opt/pr17724-6.C.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724

[Bug tree-optimization/17724] [4.0 Regression] wrong dominator (eh/fold all builtins) related

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-13 12:22 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17724