-
Bug
-
Resolution: Not an Issue
-
P1
-
None
-
1.4.0
-
generic
-
generic
Name: poR10007 Date: 06/19/2001
Java VM from jdk1.4.0beta fails to pass JCK-runtime tests in singleJVM mode.
During the build of NSK part of JCK testsuite starting from some point tests
fail with java.lang.OutOfMemoryError. Sometimes the Javatest agent crashes
dumping core and producing the following error message:
Unexpected Signal : 11 occurred at PC=0xFE5941F4
Function=JVM_NewInstanceFromConstructor+0x47E4
Library=/export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/client/libjvm.so
Dynamic libraries:
0x10000 /export/ld25/java/dest/jdk1.4.0beta-b65/solsparc/bin/../bin/sparc/native_threads/java
0xff360000 /usr/lib/libthread.so.1
0xff3a0000 /usr/lib/libdl.so.1
0xff280000 /usr/lib/libc.so.1
0xff260000 /usr/platform/SUNW,Ultra-Enterprise/lib/libc_psr.so.1
0xfe400000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/client/libjvm.so
0xff210000 /usr/lib/libCrun.so.1
0xff1f0000 /usr/lib/libsocket.so.1
0xff100000 /usr/lib/libnsl.so.1
0xff1c0000 /opt/SUNWspro/lib/libm.so.1
0xff240000 /usr/lib/libw.so.1
0xff0e0000 /usr/lib/libmp.so.2
0xff0a0000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/native_threads/libhpi.so
0xff070000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libverify.so
0xff030000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libjava.so
0xfe7e0000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libzip.so
0xfc960000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libnio.so
0xfc940000 /usr/lib/libposix4.so.1
0xfc910000 /export/ld25/java/dest/jdk1.4.0beta-b65/solsparc/bin/../jre/lib/sparc/libnet.so
0xfc860000 /usr/lib/libaio.so.1
0xfc840000 /usr/lib/nss_nisplus.so.1
0xfc810000 /usr/lib/libdoor.so.1
0xf3a00000 /export/ld51/java/pas/build/JCK-runtime-b65-cf-2/unzip/JCK-runtime-14/lib/libjckjni.so
0xfa7b0000 /export/ld51/java/pas/build/JCK-runtime-b65-cf-2/unzip/JCK-runtime-14/lib/libjckatr.so
0xfa480000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libawt.so
0xf3880000 /export/ld25/java/dest/jdk1.4.0beta-b65/solsparc/bin/../jre/lib/sparc/libmlib_image.so
0xfa410000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/motif21/libmawt.so
0xf2880000 /usr/lib/libXm.so.4
0xf3290000 /usr/openwin/lib/libXt.so.4
0xfa590000 /usr/openwin/lib/libXext.so.0
0xf43e0000 /usr/openwin/lib/libXtst.so.1
0xf3000000 /usr/openwin/lib/libX11.so.4
0xf42a0000 /usr/openwin/lib/libdps.so.5
0xf43c0000 /usr/openwin/lib/libSM.so.6
0xf4390000 /usr/openwin/lib/libICE.so.6
0xf3fb0000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libcmm.so
0xf3eb0000 /export/ld54/java/dest/jdk1.4.0beta-b65/solsparc/jre/lib/sparc/libjpeg.so
Local Time = Fri Jun 15 04:56:53 2001
Elapsed Time = 3560
#
# HotSpot Virtual Machine Error : 11
# Error ID : 4F530E43505002D7 01
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# Java VM: Java HotSpot(TM) Client VM (1.4.0-beta-b65 mixed mode)
#
OutOfMemory failure is quite reproducible, crash of Hotspot is intermittent.
The failure was reproduced on jdk1.4.0beta-b58, b64, b65, b68, on both
Solaris 7 and Solaris 8.
We tested the following modes attempting to find a cause of the
failure (all these configurations didn't pass testing either):
1. java -ms32m -mx64m -Xfuture (default build mode)
2. java -ms32m -mx512m -Xfuture
3. java -ms32m -mx64m -Xfuture -Xconcgc
4. using -loaddir parameter in Javatest environment to force
the agent to load classes for each test with its own
new classloader
5. using logging of gc activity (-Xloggc option). The last lines
of the log before the agent has died were the following:
3257.27: [Full GC 20541K->14155K(34736K), 1.2465899 secs]
3258.52: [Full GC 14155K->12914K(34736K), 1.1989310 secs]
3262.3: [Full GC 14166K->13055K(34736K), 1.1658022 secs]
3263.57: [Full GC 13098K->13065K(34736K), 1.1147656 secs]
3264.75: [Full GC
6. using nonconcurrent mode of the agent for testing (default
concurrency factor in the build is 2, i.e. the agent
runs two tests in a time)
Our test suite consists of about 7300 tests (lang, vm and some
of API tests). We also tried to divide the whole testsuite
into parts and run them separately. Lang and vm tests (5570 tests)
pass when runnning without API tests. API tests pass running
separately too.
To reproduce the bug do the following:
1. cd into directory where the report and work directories will be created
2. run script /net/sqeel.eng/export/disk04/dev02/users/NSKrotator/JCK-merlin-runtime/start
You can contact our JCK rotator (###@###.###) if you have
any questions on reproducing this bug.
======================================================================
###@###.### 07.17.2001
We created a little program to run tests from a big testlist within
the same JVM in a loop without using a JavaTest agent. We reproduced
this bug with the program on jdk1.4.0beta-b70 - VM hangs after
OutOfMemoryError. We also tried to run it with -Xss2m and using java_g
- VM crashes producing core dump or hangs after OutOfMemoryError too.
You can use the program by running the following script:
/net/sqeel.eng/export/disk04/dev02/users/NSKrotator/JCK-merlin-runtime/sameJVM.run
If you meet a program failure due to a security exception please
create .java.policy file in your home directory with the following contents:
grant{
permission java.security.AllPermission ;
};
==============================================================================
Pavel Ozhdikhin ###@###.### 2001-07-24
Now the problem is reduced to OutOfMemory errors only. The crash
we observed before was caused by absence of needed patch on our
Solaris 5.7 machine. Sorry for confusion. The words of evaluation
of 2001-06-27 saying "Do not pay attention to OutOfMemory" are
incorrect. Now the only problem preventing JCK to be passed in
singleJVM mode is JVM hang immediately after OutOfMemory error.