[Bug java/18699] New: [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[Bug java/18699] New: [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[ovidr at users dot sourceforge dot net]

I have upgraded to gcc version 4.0.0 20041127 (experimental) and my app now
receives this:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 1103858608 (LWP 4735)]
GC_local_gcj_malloc (bytes=20, ptr_to_struct_containing_descr=0x86d1848)
    at /datal/gcc/gcc/boehm-gc/pthread_support.c:364
364             ptr_t my_entry = *my_fl;
(gdb) bt
#0  GC_local_gcj_malloc (bytes=20, ptr_to_struct_containing_descr=0x86d1848)
    at /datal/gcc/gcc/boehm-gc/pthread_support.c:364
#1  0x405ea18b in _Jv_AllocObjectNoFinalizer (klass=0x86d18a0) at java-gc.h:46
(rest is in my code)
...
(gdb) info local
index = 3
my_fl = (ptr_t *) 0x234
my_entry = Variable "my_entry" is not available.


There were recent changes to the THREAD_LOCAL_ALLOC define I believe.  Whether
related or not I'm not sure.

<builddir>/i686-pc-linux-gnu/libjava/include/java-gc.h (is a symlink to
/libjava/include/boehm-gc.h)

#ifdef THREAD_LOCAL_ALLOC
  return GC_local_gcj_malloc (size, klass->vtable);   // this is line 46
#else
  return GC_gcj_malloc (size, klass->vtable);
#endif

-- 
           Summary: [4.0 Regression] SIGSEGV in GC_local_gcj_malloc
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: java
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ovidr at users dot sourceforge dot net
                CC: gcc-bugs at gcc dot gnu dot org,java-prs at gcc dot gnu
                    dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|java                        |libgcj
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[ovidr at users dot sourceforge dot net]

------- Additional Comments From ovidr at users dot sourceforge dot net  2004-11-28 04:20 -------
And this..

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 1113430960 (LWP 4763)]
0x405e8f93 in really_start (x=0x8771a40) at
/datal/gcc/gcc/libjava/posix-threads.cc:396
396       if (! (info->data->flags & FLAG_DAEMON))
Current language:  auto; currently c++
(gdb) bt
#0  0x405e8f93 in really_start (x=0x8771a40) at
/datal/gcc/gcc/libjava/posix-threads.cc:396
#1  0x435f979c in start_thread () from /lib/tls/libpthread.so.0
#2  0x433daf2a in clone () from /lib/tls/libc.so.6



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[mark at gcc dot gnu dot org]

------- Additional Comments From mark at gcc dot gnu dot org  2004-11-28 14:50 -------
Also seeing this with some simple AWT programs (see bug #18697) and a recent
gjdoc build (as described at http://gcc.gnu.org/ml/java/2004-11/msg00169.html).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-11-28 14:50:40
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[mark at gcc dot gnu dot org]

------- Additional Comments From mark at gcc dot gnu dot org  2004-11-28 15:28 -------
Reverting this patch seems to make things work again:

Index: ChangeLog
===================================================================
RCS file: /cvs/gcc/gcc/boehm-gc/ChangeLog,v
retrieving revision 1.212
retrieving revision 1.213
diff -u -r1.212 -r1.213
--- ChangeLog   26 Nov 2004 05:31:47 -0000      1.212
+++ ChangeLog   27 Nov 2004 17:44:30 -0000      1.213
@@ -1,3 +1,7 @@
+2004-11-27  Bryce McKinlay  <mckinlay@redhat.com>
+
+       * include/gc_config.h.in: Add comment. Define only THREAD_LOCAL_ALLOC.
+
 2004-11-26  Bryce McKinlay  <mckinlay@redhat.com>

        Fix "PACKAGE_NAME redefined" errors.
Index: include/gc_config.h.in
===================================================================
RCS file: /cvs/gcc/gcc/boehm-gc/include/gc_config.h.in,v
retrieving revision 1.3
retrieving revision 1.4
diff -u -r1.3 -r1.4
--- include/gc_config.h.in      26 Nov 2004 05:31:48 -0000      1.3
+++ include/gc_config.h.in      27 Nov 2004 17:44:31 -0000      1.4
@@ -1,3 +1,7 @@
-#undef ALL_INTERIOR_POINTERS
-#undef GC_GCJ_SUPPORT
+/* gc_config.h. This autoconf header contains macro definitions that are
+used by GC clients, such as libjava.

+This file is not generated by autoheader. We only want to define what we
+really need in order to avoid namespace pollution. */
+
+#undef THREAD_LOCAL_ALLOC


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[ovidr at users dot sourceforge dot net]

------- Additional Comments From ovidr at users dot sourceforge dot net  2004-11-28 16:41 -------
I reverted that patch and 'make clean' in build's boehm-gc directory but now I
get the following.  Do I need a complete rebuild? 

Collecting from unknown thread.

Program received signal SIGABRT, Aborted.
[Switching to Thread 1112734640 (LWP 10238)]
0x432e8c32 in _dl_sysinfo_int80 () from /lib/ld-linux.so.2
(gdb) bt
#0  0x432e8c32 in _dl_sysinfo_int80 () from /lib/ld-linux.so.2
#1  0x4332798f in raise () from /lib/tls/libc.so.6
#2  0x43329302 in abort () from /lib/tls/libc.so.6
#3  0x407f0d95 in GC_abort (msg=0x0) at /datal/gcc/gcc/boehm-gc/misc.c:1072
#4  0x407f75e7 in GC_push_all_stacks () at
/datal/gcc/gcc/boehm-gc/pthread_stop_world.c:270
#5  0x407f28d7 in GC_default_push_other_roots () at
/datal/gcc/gcc/boehm-gc/os_dep.c:2045
#6  0x407f0589 in GC_push_roots (all=1, cold_gc_frame=0x0) at
/datal/gcc/gcc/boehm-gc/mark_rts.c:643
#7  0x407ef9d5 in GC_mark_some (cold_gc_frame=0x4252f6d4 "") at
/datal/gcc/gcc/boehm-gc/mark.c:326
#8  0x407e67bb in GC_stopped_mark (stop_func=0x407e6410 <GC_never_stop_func>)
    at /datal/gcc/gcc/boehm-gc/alloc.c:518
#9  0x407e716a in GC_try_to_collect_inner (stop_func=0x407e6410
<GC_never_stop_func>)
    at /datal/gcc/gcc/boehm-gc/alloc.c:365
#10 0x407e7b16 in GC_collect_or_expand (needed_blocks=1, ignore_off_page=0)
    at /datal/gcc/gcc/boehm-gc/alloc.c:1020
#11 0x407e7c8b in GC_allocobj (sz=2, kind=4) at /datal/gcc/gcc/boehm-gc/alloc.c:1071
#12 0x407ec55a in GC_generic_malloc_inner (lb=8, k=4) at
/datal/gcc/gcc/boehm-gc/malloc.c:136
#13 0x407eb6a4 in GC_gcj_malloc (lb=8, ptr_to_struct_containing_descr=0x40b36948)
    at /datal/gcc/gcc/boehm-gc/gcj_mlc.c:157
#14 0x405ea18b in _Jv_AllocObjectNoFinalizer (klass=0x40b36880) at java-gc.h:48
#15 0x405ea1f5 in _Jv_AllocObject (klass=0x40b36880) at
/datal/gcc/gcc/libjava/prims.cc:435
#16 0x4061286d in java::lang::Object::clone (this=0x8a399c0) at
/datal/gcc/gcc/libjava/java/lang/natObject.cc:101
#17 0x4067c2e7 in java.util.HashSet.clone() (this=0x8a399c0) at
/datal/gcc/gcc/libjava/java/util/HashSet.java:173
#18 0x4068463b in java.util.Observable.notifyObservers(java.lang.Object)
(this=0x8ac7d40, obj=0x8a399d8)
    at /datal/gcc/gcc/libjava/java/util/Observable.java:164

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[ovidr at users dot sourceforge dot net]

------- Additional Comments From ovidr at users dot sourceforge dot net  2004-11-28 21:24 -------
Recompiled libgcj and I still get the above stacktrace.  Here is a testcase.

import java.util.Observable;
import java.util.Observer;

class TT extends Observable implements Runnable, Observer {

  public static void main(String[] args) {
    TT tt1 = new TT();
    TT tt2 = new TT();
    tt1.addObserver(tt2);
    tt2.addObserver(tt1);
    new Thread(tt1).start();
    new Thread(tt2).start();
  }

  public void run() {
    int c = 0;
    String s = "";
    while (++c < 50) {
      System.out.println("Running: " + Thread.currentThread());
      this.setChanged();
      s = "";
      for (int i = 0; i < 200; i++)
        s += String.valueOf(i);
      this.notifyObservers(s);
    }
  }

  public void update(Observable o, Object obj) {
    System.out.println(o);
  }

}

gcj -g --main=TT TT.java -o TT

Collecting from unknown thread.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[mark at gcc dot gnu dot org]

------- Additional Comments From mark at gcc dot gnu dot org  2004-11-28 21:57 -------
That TT.java example program also fails for me.
The other testcases I had work again when reverting the patch in comment #3.
The TT testcase fails even when I revert that patch.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[andreast at gcc dot gnu dot org]

------- Additional Comments From andreast at gcc dot gnu dot org  2004-11-29 09:42 -------
Reverting those two patches makes TT work on darwin.

2004-11-27  Bryce McKinlay  <mckinlay@redhat.com>

        * include/gc_config.h.in: Add comment. Define only THREAD_LOCAL_ALLOC.

2004-11-26  Bryce McKinlay  <mckinlay@redhat.com>

        Fix "PACKAGE_NAME redefined" errors.
        * configure.ac (AC_CONFIG_HEADERS): Build both include/config.h
        and include/gc_config.h.
        * configure: Regenerated.
        * include/gc_config.h.in: No longer auto-generated. Define only the
        macros needed by libjava.
        * include/config.h.in: Generated by autoheader.
        * include/gc.h: Include config.h, not gc_config.h.

http://gcc.gnu.org/ml/gcc-cvs/2004-11/msg01333.html
http://gcc.gnu.org/ml/gcc-cvs/2004-11/msg01278.html


These patches are also responsible for the many failure we see in libgcj testsuite.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[mckinlay at redhat dot com]

------- Additional Comments From mckinlay at redhat dot com  2004-11-29 15:48 -------
I'm investigating...

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[Hans dot Boehm at hp dot com]

------- Additional Comments From Hans dot Boehm at hp dot com  2004-11-29 18:18 -------
I have not tried to reproduce this.  This is purely conjecture:

All failures could be explained if the collector is no longer intercepting 
pthread_create.  Threads should be created with GC_pthread_create.

The original failure could also be explained by GC_init_parallel() not being 
called, e.g. due to inconsistent THREAD_LOCAL_ALLOC definitions within the GC.

THREAD_LOCAL_ALLOC should be defined in libgcj only if it's defined in the GC.  
But I don't think that's the original problem here.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[tromey at gcc dot gnu dot org]

------- Additional Comments From tromey at gcc dot gnu dot org  2004-11-29 20:17 -------
If the GC headers that we include rely on the defines in
the GC's config.h, then that would explain the problem.

We don't actually need PACKAGE_NAME and friends, I think.
One fix would be to simply undef them in boehm-gc.h before
including gc_config.h.

BTW, one reason to prefer using results of the GC's configure,
rather than hard-coding stuff in libgcj, is that it lets people
play with changing the GC; then they don't need parallel changes.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[mckinlay at redhat dot com]

------- Additional Comments From mckinlay at redhat dot com  2004-11-30 00:39 -------
Something strange is going on here. I have reverted the boehm-gc config patches
for now.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-11-30 00:39 -------
Subject: Bug 18699

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	bryce@gcc.gnu.org	2004-11-30 00:38:57

Modified files:
	boehm-gc       : ChangeLog configure.ac configure 
	boehm-gc/include: gc.h gc_config.h.in 
Removed files:
	boehm-gc/include: config.h.in 

Log message:
	2004-11-29  Bryce McKinlay  <mckinlay@redhat.com>
	
	PR libgcj/18699
	Revert patches of 2004-11-26 and 2004-11-27.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/ChangeLog.diff?cvsroot=gcc&r1=1.214&r2=1.215
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/configure.ac.diff?cvsroot=gcc&r1=1.9&r2=1.10
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/configure.diff?cvsroot=gcc&r1=1.91&r2=1.92
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/include/gc.h.diff?cvsroot=gcc&r1=1.13&r2=1.14
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/include/gc_config.h.in.diff?cvsroot=gcc&r1=1.4&r2=1.5
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/boehm-gc/include/config.h.in.diff?cvsroot=gcc&r1=1.1&r2=NONE



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-12-01 18:35 -------
Subject: Bug 18699

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	bryce@gcc.gnu.org	2004-12-01 18:34:56

Modified files:
	libjava        : ChangeLog 
Added files:
	libjava/testsuite/libjava.lang: PR18699.java PR18699.out 

Log message:
	2004-12-01  Bryce McKinlay  <mckinlay@redhat.com>
	
	PR libgcj/18699
	* testsuite/libjava.lang/PR18699.java,
	testsuite/libjava.lang/PR18699.out: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libjava/ChangeLog.diff?cvsroot=gcc&r1=1.3254&r2=1.3255
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libjava/testsuite/libjava.lang/PR18699.java.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libjava/testsuite/libjava.lang/PR18699.out.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699

[Bug libgcj/18699] [4.0 Regression] SIGSEGV in GC_local_gcj_malloc

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-04 04:04 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18699