--- Comment #41 from andreast at gcc dot gnu dot org 2010-02-02 08:19
---
Subject: Bug 41991
Author: andreast
Date: Tue Feb 2 08:19:26 2010
New Revision: 156446
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=156446
Log:
2010-02-02 Jack Howarth
PR java/41991
--- Comment #40 from andreast at gcc dot gnu dot org 2010-02-02 08:19
---
Subject: Bug 41991
Author: andreast
Date: Tue Feb 2 08:18:48 2010
New Revision: 156445
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=156445
Log:
2010-02-02 Jack Howarth
PR java/41991
--- Comment #39 from andreast at gcc dot gnu dot org 2010-02-02 08:18
---
Subject: Bug 41991
Author: andreast
Date: Tue Feb 2 08:18:08 2010
New Revision: 156444
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=156444
Log:
2010-02-02 Jack Howarth
PR java/41991
--- Comment #38 from howarth at nitro dot med dot uc dot edu 2009-12-19
00:35 ---
I've confirmed that both the WalkerTest failures and the gcj compiler crashes
on java code are eliminated on darwin10 if I duplicate the code for
_Unwind_FindEnclosingFunction() as _darwin10_Unwind_FindEnc
--- Comment #37 from howarth at nitro dot med dot uc dot edu 2009-12-18
14:54 ---
I noticed that in libjava/sysdep/arm/backtrace.h, arm has its own definition of
_Unwind_FindEnclosingFunction. Could we use the same approach for darwin10 to
override the default _Unwind_FindEnclosingFunct
--- Comment #36 from howarth at nitro dot med dot uc dot edu 2009-12-17
23:30 ---
(In reply to comment #31)
> Interestingly gcc-4.4.2 with the proposed patch,
> http://gcc.gnu.org/ml/java/2009-12/msg00027.html, shows gcj crashing the same
> way as gcc trunk with the same patch
>
>
--- Comment #35 from howarth at nitro dot med dot uc dot edu 2009-12-10
14:32 ---
This might be a red herring but after what I observed for PR42333
(http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42333#c29) I went back and looked
at how the libjava libraries and executables are linked
--- Comment #34 from bryce at gcc dot gnu dot org 2009-12-09 18:46 ---
Subject: Bug 41991
Author: bryce
Date: Wed Dec 9 18:46:11 2009
New Revision: 155114
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155114
Log:
2009-12-09 Bryce McKinlay
PR java/41991
*
--- Comment #33 from bryce at gcc dot gnu dot org 2009-12-09 18:45 ---
Subject: Bug 41991
Author: bryce
Date: Wed Dec 9 18:45:13 2009
New Revision: 155113
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155113
Log:
2009-12-09 Bryce McKinlay
PR java/41991
*
--- Comment #32 from howarth at nitro dot med dot uc dot edu 2009-12-05
09:02 ---
Disassembling the crash on gcc-4.4.2 with the proposed patch on
x86_64-apple-darwin10 shows...
(gdb) x/10i 0x0001fcbc
0x1fcbc
<_ZN3gnu9classpath13VMStackWalker21getCallingClassLoaderEJPN4java4
--- Comment #31 from howarth at nitro dot med dot uc dot edu 2009-12-05
08:54 ---
Interestingly gcc-4.4.2 with the proposed patch,
http://gcc.gnu.org/ml/java/2009-12/msg00027.html, shows gcj crashing the same
way as gcc trunk with the same patch
(gdb) r testme.java
-fbootclasspath=
--- Comment #30 from howarth at nitro dot med dot uc dot edu 2009-12-05
08:21 ---
Confirmed that, as expected, the proposed patch,
http://gcc.gnu.org/ml/java/2009-12/msg00027.html, eliminates the crashes in gcj
in gcc 4.4.2 on x86_64-apple-darwin9 without any other changes.
--
http
--- Comment #29 from howarth at nitro dot med dot uc dot edu 2009-12-05
07:03 ---
The proposed patch to properly implement passing -allow_stack_execute on
SYSTEMSPEC...
http://gcc.gnu.org/ml/java/2009-12/msg00027.html
eliminates the crashes in gcj when compiling java code provided tha
--- Comment #28 from howarth at nitro dot med dot uc dot edu 2009-12-03
15:33 ---
Possible fix which is untested...
Index: libjava/configure.ac
===
--- libjava/configure.ac(revision 154950)
+++ libjava/configure.ac
--- Comment #27 from howarth at nitro dot med dot uc dot edu 2009-12-01
04:56 ---
Created an attachment (id=19192)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19192&action=view)
walk from last _Unwind_RaiseException breakpoint until crash
--
http://gcc.gnu.org/bugzilla/sho
--- Comment #26 from howarth at nitro dot med dot uc dot edu 2009-11-29
07:48 ---
I finally managed to sort out the darwin build to get access to the unwinder
debug symbols.
Walking through the testme.java test case using the _Unwind_RaiseException 39
times, I
then used a uw_frame_state
--- Comment #25 from howarth at nitro dot med dot uc dot edu 2009-11-29
07:41 ---
Created an attachment (id=19174)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19174&action=view)
gdb walk through w_frame_state_for calls in unwinder
--
http://gcc.gnu.org/bugzilla/show_bug.cgi
--- Comment #24 from howarth at nitro dot med dot uc dot edu 2009-11-28
23:01 ---
I figured out for darwin9 that the dual linkage to the system libgcc and the
FSF libgcc from...
http://gcc.gnu.org/viewcvs?view=revision&revision=154283
http://gcc.gnu.org/viewcvs?view=revision&revision=1
--- Comment #23 from howarth at nitro dot med dot uc dot edu 2009-11-28
20:32 ---
No luck here with setting...
setenv DYLD_LIBRARY_PATH /sw/lib/gcc4.5/lib
I suspect this can randomly pass in some cases
as you have seen before.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41991
--- Comment #22 from andreast at gcc dot gnu dot org 2009-11-28 20:27
---
I follow this one too, that is why I ask!
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41991
--- Comment #21 from howarth at nitro dot med dot uc dot edu 2009-11-28
20:25 ---
Andreas,
Actually we have another issue at the moment. See [Bug target/42208]. It
appears that at some point in late Sept or earlier Oct, darwin starting linking
both the system and the gcc built libgc
--- Comment #20 from andreast at gcc dot gnu dot org 2009-11-28 20:14
---
Jack, can you point DYLD_LIBRARY_PATH to your installed libgcc_s.dylib and try
to run a gcj -C testme.java?
I did so with todays trunk and the patch from lxo (Alex), the one you sent me
as well as mine from comme
--- Comment #19 from howarth at nitro dot med dot uc dot edu 2009-11-27
21:38 ---
In case it helps, a break point on _Unwind_RaiseException () is hit 39 times.
When you continue from the 39th instance of that break point having hit, when
compiling the testme.java test code, you get...
--- Comment #18 from howarth at nitro dot med dot uc dot edu 2009-11-27
21:28 ---
Created an attachment (id=19164)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19164&action=view)
complete walk from _Jv_Throw as bz2 archive
Used proposed patch from http://gcc.gnu.org/bugzilla/sho
--- Comment #17 from andreast at gcc dot gnu dot org 2009-11-19 21:15
---
Unfortunately I'm now in the same boat with you:
[deuterium:gcc/head/objdir-x86_64] andreast% ./gcc/xgcc -v
Using built-in specs.
COLLECT_GCC=./gcc/xgcc
Target: x86_64-apple-darwin9
Configured with: /Volumes/deve
--- Comment #16 from howarth at nitro dot med dot uc dot edu 2009-11-18
14:02 ---
The dSYM issues looks like part of PR41473. I'll revert to debugging in gcc
4.4.2 for now which shouldn't have the issue.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41991
--- Comment #15 from howarth at nitro dot med dot uc dot edu 2009-11-18
01:39 ---
Argh, the dSYM issue looks like another variation of the current dsymutil
issues...
libtool: link:
/sw/src/fink.build/gcc45-4.4.999-20091116/darwin_objdir/./gcc/xgcc
-shared-libgcc -B/sw/src/fink.build/g
--- Comment #14 from howarth at nitro dot med dot uc dot edu 2009-11-18
01:19 ---
The missing .dSYM for libgcj is a blocker for debugging this issue on darwin...
(gdb) break _Jv_Throw
Breakpoint 1 at 0x20c49ba3dcddf8: file
../../../gcc-4.5-20091116/libjava/exception.cc, line 100.
(gdb)
--- Comment #13 from howarth at nitro dot med dot uc dot edu 2009-11-18
01:04 ---
One oddity in the libgcj.dylib build on darwin is that this only shared library
in libjava which doesn't get a .dSYM directory built for it. All the other
shared libs have one...
ls -R libjava | grep dSYM
--- Comment #12 from howarth at nitro dot med dot uc dot edu 2009-11-17
00:39 ---
The attached unwinder_walk.txt is the log of walk of ecj1 when compiling the
testme,java test code. This uses r154217 with the patch from comment 3 and
with the installed libgcc replaced with a copy compi
--- Comment #11 from howarth at nitro dot med dot uc dot edu 2009-11-17
00:37 ---
Created an attachment (id=19027)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19027&action=view)
gdb walk from _Jv_Throw breakpoint
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41991
--- Comment #10 from howarth at nitro dot med dot uc dot edu 2009-11-16
01:26 ---
This doesn't solve the abort in ejc1 on x86_64-apple-darwin10, but shouldn't we
have...
Index: libjava/Makefile.in
===
--- libjava/Makefile.
--- Comment #9 from howarth at nitro dot med dot uc dot edu 2009-11-15
22:52 ---
Comparing the linkages of ecjx and gij on x86_64-apple-darwin10 with your
patch, one thing I did notice is that there is a slight difference in the link
flags. The ecjx linkage is passed -findirect-dispatch
--- Comment #8 from howarth at nitro dot med dot uc dot edu 2009-11-15
19:20 ---
I have the same problem with current gcc trunk and the proposed patch on a
Core2Duo under x86_64-apple-darwin10...
gcj --main=testme -O testme.java
gcj: Internal error: Abort trap (program ecj1)
Please sub
--- Comment #7 from howarth at nitro dot med dot uc dot edu 2009-11-15
01:57 ---
Yes, I have...
[MacPro:darwin_objdir/x86_64-apple-darwin10.2.0/libjava] howarth% grep
extra_gij_ldflags *
Makefile:extra_gij_ldflags = -Wl,-allow_stack_execute
I have been testing on a late 2008 MacPro w
--- Comment #6 from andreast at gcc dot gnu dot org 2009-11-14 20:57
---
I tried on another core2duo (x86_64-apple-darwin9), same issue. W/o
extra_gij_ldflags=-Wl,-allow_stack_execute for ecjx_LINK I fail with the same
entry in Library/Logs/CrashReporter:
ecj1_2009-11-14-202538_imac-ca
--- Comment #5 from howarth at nitro dot med dot uc dot edu 2009-11-13
14:26 ---
I tried the patch from comment 3 on x86_64-apple-darwin9.8.0 and the gcj
crashes still occur. I wonder if this change only makes the bug potentially go
latent on certain hardware for intel darwin. It is unc
--- Comment #4 from howarth at nitro dot med dot uc dot edu 2009-11-12
02:43 ---
Unfortunately, the proposed patch doesn't solve the crashes in gcj on intel
darwin. I would note that if I use javac to convert testme.java into
testme.class that...
gcj --main=testme -O testme.class
comp
--- Comment #3 from andreast at gcc dot gnu dot org 2009-11-11 22:40
---
Does this help:
[deuterium:head/gcc/libjava] andreast% svn diff Makefile.am Makefile.in
Index: Makefile.am
===
--- Makefile.am (revision 154097)
+++
--- Comment #2 from howarth at nitro dot med dot uc dot edu 2009-11-09
00:45 ---
(In reply to comment #1)
> A backtrace of the segfault under i686-apple-darwin10 for gcc trunk is...
>
This should have read...
A backtrace of the segfault under i686-apple-darwin9 for gcc trunk is...
--- Comment #1 from howarth at nitro dot med dot uc dot edu 2009-11-09
00:44 ---
A backtrace of the segfault under i686-apple-darwin10 for gcc trunk is...
gdb /sw/lib/gcc4.5/libexec/gcc/i686-apple-darwin9/4.5.0/ecj1
...
(gdb) r testme.java
-fbootclasspath=./:/sw/lib/gcc4.5/share/java/l
41 matches
Mail list logo
