CR233415

Looking at allocation stacktraces in the Memory Leak Detector could show stacktraces where the selected type was not allocated, as well as stacktraces where it was allocated. This has been fixed.

CR224247

An issue that prevented the Java Plugin from running applets in Internet Explorer is solved in this version. JRockit 5.0 will now run applets in Internet Explorer.

CR214904

Under high network load, JRockit could sometimes incorrectly throw exceptions of the type "java.net.SocketException: Socket closed." This has now been fixed.

CR224896

Options that are not valid JRockit options, have sometimes been parsed as a different, valid option. This happened if the valid option began with the same characters as the invalid one. For example, -Xcheck could be parsed as -XcheckedStacks. This problem is now fixed.

CR204924

FileChannel.read now returns -1 at end of file instead of throwing an exception.

CR224377

Calling Object.wait() in a system with high lock contention could cause JRockit to throw a NullPointException. This has now been fixed.

CR225425

The Memory Leak Detector now lists the top 10 items (if applicable) when listing the largest arrays.

CR226240

The JRockit Analyzer (JRA) does not crash the JVM anymore when the JVM runs with the gencon garbage collector.

CR227186

The garbage collector MXBean method getNurserySize does not throw an internal RuntimeException anymore when a nursery is not available.

CR206727

JRockit will no longer crash when installed as a web browser plugin and you press "v" in the Java Console.

CR215677

Hour-glass does not stay on the screen in Management Console anymore.

CR221097

JRockit issues events on multiple threads during garbage collection. New option, -Xjvmpi:singlegc=true, has been added to force events being issued via one thread only.

CR221238

JRockit memory usage has been improved in cases where there are many classloaders with few classes per classloader.

CR221635

User defined actions in the Management Console are now working.

CR221996

JRockit startup time has been improved.

CR222428

Stack overflow issues are now handled more safely.

CR222443

Previously JRockit displayed incorrect nursery size information when using -Xverbose:memory with an optimizing garbage collector. This information is no longer displayed.

CR222539

The optimizing compiler could generate inaccurate code for StringReader.read. This problem is now fixed.

CR222737

A race condition that could lead to a null pointer exception when generating the detail panes in the Management Console is now eliminated.

CR222973

Prevent applications from changing the private field parent in java.lang.ClassLoader.

CR222990

Can now create arrays larger than 256*1024*1024 elements.

CR223407

The MemoryMBean attribute ObjectPendingFinalizationCount now returns the correct value.

CR223426

On 64-bit platforms, JRockit could crash if the Memory Leak Detector tried to retrieve the largest arrays for a type. This has now been fixed.

CR223640

When using IterateOverReachableObjects in JVMTI, JRockit reported bad referrer_index for JVMTI_REFERENCE_FIELD. This has now been fixed.

CR225144

Reverted the java.vendor* system properties back to BEA specific property values.

CR225152

Eliminated sporadic deadlocks on platforms with weak memory guarantees.

CR226873

Previously JRockit could crash during shutdown, possibly without generating either a crash dump or a core file. This was an uncommon problem, but has now been fixed.

CR209583

CGLIB does not throw IllegalAccessError.

CR211573

JRockit on x64 may crash due to a floating point code generation bug. This is now fixed.

CR211938

Overriding package local final Methods now works.

CR211961

java.vm.info property now exists.

CR212193

When iterating over stack frames, getStackAccessControlContext used the wrong iterator (frame instead of call), which could result in JRockit missing optimized frames.

CR212198

Throwing ZipException instead of IOException when encountering empty or invalid zipfile

CR212423

System.arraycopy had an off-by-one error with short-arrays when doing a reverse copy. This is now fixed.

CR214656

lphello does not dump in cgGetCallChainOnAddr.

CR215089

Large page for IA32 Linux with 2.6 kernels are now working.

CR215143

Problems with method resolving in JNI now fixed.

CR215444

WebLogic Enterprise Platform arraycopy might read outside of heap on X86. This issue is now fixed.

CR217645

JRockit is now able to throw ClassCircularityError.

CR218157

Optimizing problems causing a thread dump, sometimes occurring with optimizeIt, now fixed.

CR219032

Heap/object allocation scaling bottleneck solved by modifying -XXaggressive to set the cachelistpercentage to 10% and by adding jrockit.heap.cachelistpercentage property. This sets the percentage of free heap memory that will be kept in the freelist caches that are used for speeding up object allocation (for example set: -Djrockit.heap.cachelistpercentage=15 on the command line to set the percentage to 15%).

CR220374

JRockit dump in zipReleaseEntry during adminServer (WLS) startup, R25.1.0-96. This is a known issue in a third-party product used by BEA JRockit.

CR220655

Rare "timeout" with gencon showed out to be an internal signal error. This is now fixed.

CR220819

In the BEA Installer the Java Control Panel is now installed.

CR220825

If JRockit is uninstalled and thereafter reinstalled, the installer fails to set JRockit as a plugin to any browser. This problem has now been fixed.

CR221005

BEA Installer: Multiple instances of JRockit installation problem has been fixed.

CR221009

If user installs JRockit in another location than the default location, the Registry keys still point to the default location (which do not exist). This problem is now fixed.

CR221016

JRockit does not fail to install Java Web Start anymore.

CR221029

Java Application Cache Viewer can now be launched from bin as well as jre/bin.

CR222056

Internet Explorer is now shown as a browser alternative.

CR194599

Support for profiling of VM internal locks are now implemented.

CR195900

External system properties handling is now cleaned up.

CR205334

sun.reflect.ReflectionFactory.newConstructorForSerialization does not work.

Note: JRockit does not support that API.

CR205501

IPF largepages MAP_PRIVATE is now supported.

CR206811

Invocation compiler newInstance now generates call to registerfinalizer.

CR206862

Performance degradation over time with -Xgcprio:throughput due to incorrect strategy choice. Heuristics updated and no degradation can be observed.

CR207085

JRockit crashed when starting a server with OptimizeIt due to differences between the handling of startEvents in JVMTI and JVMPI. This is now fixed.

CR207247

Optimistically enable HT optimizations if processor supports it but OS cannot confirm.

To enable the optimistic approach, the jrockit.cpu.ia32.ht property has been added.

CR209071

JRockit now returns the class with the most specific return type from Class#getDeclaredMethod when several matching methods are found.

CR209777

JRockit VM seems to hang when attaching optimizeIt and trying to deploy a client. OptimizeIt uses getthreadstate frequently. The implementation of that method used to suspend native threads to retrieve information. Now JRockit does not need to suspend native threads when retrieving their status.

CR210093

ClassFormatError with resin 3.0.10. Now allows multiple LocalVar*Table:s in any order.

CR210268

StackOverFlowError while viewing JRA profile has been fixed.

CR211066

Bytecode verification is now less strict to be compliant with the reference VM.

CR211177

Potential deadlock in JVMTI/rawmonitors when using gcevents has been eliminated.

CR211205

During shut-down, JRockit crashed when failing to acquire shared memory. This is now fixed.

CR211322

GetLineNumberTable now returns JVMTI_ERROR_NATIVE_METHOD in case of native methods.

CR212065

Upgrade to 1.5.0_02: Instances of java.lang.Enum are now never finalized, Sun bug 5098065.

CR212317

Fixed race condition in add_JVMPI_entryexit_hooks.

CR212612

JRockit can now bootstrap when launched from UNC path.

CR214838

Fixed ObjectMonitor memory leak.

CR214982

JRA now samples even if started with -Xnoopt.

CR215448

Out of memory error even though heap could still be expanded. Now heap expansion has been corrected.

CR216068

Missing JNI support, like JAWT, is now added.

CR216076

File handle leak in FileNativeIO.open on directories is now fixed.

CR216086

JRockit max heap size on win_ia32 can now increase beyond 1574304kb with /3GB parameter in boot entry.

CR216620

Fixed rapid increase of heap usage after IllegalMonitorStateException.

CR216626

Starting JRockit with the option -Xcleartype has no effect. The option is now deprecated and may be removed in a future release.

CR217056

The jvmtiGetTag now works.

CR217060

JVMTI now reports JVMTI_HEAP_ROOT_THREAD to the correct callback in iterateover objects.

CR217076

Can now add static initializer to Object.

CR217100

Remote JMX using authentication, file + password, now works.

CR217466

Eclipse dumps when running on JRockit. This problem is now solved.

CR217647

JRockit does not hang anymore when trying to throw StackOverflowError.

CR218061

jrcmd now resolves path to JRockit executable correctly.

CR218251

JRockit now supports the -Xrs option.

CR219600

Large unused nursery and still getting an out of memory error. This problem has now been fixed.

CR199232

Eternal garbage collection if an application stores too many finished threads. This has now been fixed and JRockit will not collect garbage eternally.

CR200091

JGCW does no longer expose problems with finalizers.

CR206447

The name of the -XXhpc option has changed to -XXhpm due to confusion with industry acronym for "high performance computing."

CR208302

No more npe when adding StringBuffer.append(char[]) to method profiling.

CR208310

Management Console does not deadlock anymore when reconnecting to a new VM.

CR208532

Warn if the host computer swaps during collection. If verbose:memory is turned on, JRockit now warns if the total number of page faults during garbage collection is bigger than 5% of the number of pages in the heap.

CR209826

fix_javahome, realpath, and symlinks cause java.home to be incorrect.

In this release, JRockit handles longer realpaths from symlink to java.home, which corrects this problem.

CR210867

Previously JRockit missed to report some compile_method_events. This is no longer the case.

CR212649

JRockit can only allocate 1500m heap on Windows 2003 IA32. The JVM performance counters are now initialized after the heap is allocated, which corrects this problem.

CR214354

Occasional JRockit crash when starting a JRA recording has now been fixed.

CR214516

The initialization behavior and value of sun.boot.class.path differs between JRockit and Sun. The initialization behavior has now been rewritten to mirror the behavior of Sun.

CR214537

Getting threads by name from MAPI now works.

CR215113

Cannot build update-site in Eclipse.

CR216646

Counter for encountered deadlocks added.

CR217666

The Management Console now shows deadlock detection data in thread stack dumps.

CR218492

Parts of JVMTI system properties interface implementation now follow the specification.

CR208906

A table is added for native lock profiling on the JRA's Lock Profiling page.

Australian Daylight Savings Time change for 2006 not supported.

This release does not contain a fix for the Australian Daylight Savings Time change for 2006. Please contact BEA Support for a patch.

JRockit crashes when using a heap larger than 5375 MB on Red Hat 4.0 for EM64T/AMD64.

You can workaround this problem by using a heap smaller than 5375 MB.

This will be fixed in the next service pack (CR247853).

A signal being sent to wrong thread on Linux 2.6 kernels on Itanium (2.6.11 and previous Itanium kernels) might cause random hanging or crashing JVM.

BEA JRockit is known to hang or crash on 2.6 kernels on Itanium, due to a bug in the Linux 2.6.11 (and previous) kernels. The bug is in the kernel sigprocmask() syscall and will be fixed in the 2.6.12 kernel.

We cannot estimate how often the bug actually occurs, but under different circumstances it has been seen as often as once per hour or as seldom as once every two days.

BEA believes that this issue affects SMP systems more than single CPU systems.

Affected Linux versions are: SuSE Linux Enterprise Server 9.0 and RedHat Enterprise Linux 4.0 on Itanium running on 2.6.11 (or previous) kernels. (CR230226 and CR218035).

SuSE has confirmed that this bug will be fixed in their upcoming SLES 9.0 SP2 release. We are working with RedHat to get this fix into RHEL 4.0 Update 2.

You can see the committed patch at:

http://kernel.org/git/?p=linux/kernel/git/
torvalds/linux-2.6.git;a=commit;
h=a2a64769d0d3cc0380b4b6ecdcb781a2f790a69e

Customers running SMP systems must patch their kernel to include the kernel fix or upgrade to a newer kernel. At the time of our release (June 22, 2005) no offically updated kernels are available. If you are running on SLES 9.0 Itanium, you should upgrade to SP2 once it is made available.

The RedHat internal reference number is 74397. For SuSE the internal reference number is 78084.

BEA JRockit Memory Leak Detector hangs when displaying aquired instances.

The BEA JRockit Memory Leak Detector could in some instances cause JRockit to freeze or crash. (CR228592)

JRockit deadlock on SUSE Linux Enterprise Server 9.0 running on x64/SMP

There is a kernel bug in SUSE Linux Enterprise Server 9.0 x64 version, which causes signals to be lost on rare occations on multi-CPU systems. This is not noticable for most applications, but JRockit relies heavily on signalling, and is therefore affected by this bug. The consequence of the bug is that JRockit deadlocks. For this reason, it is *not recommended* to use JRockit for production on this platform, until this problem has been fixed by SUSE. BEA is working with SUSE to get the fix into the kernel.

Affected Linux version is SUSE Linux Enterprise Server 9.0 SP1. The problem is also present in RedHat Enterprise Linux 4.0 GA on x64, but it will be fixed in the upcoming RHEL 4.0 QU1. (CR230225 CR220658)

Crash in garbage collector under rare circumstances

JRockit may crash with an Illegal memory access and point to a problem in mmNurseryAddTLAsToCache. This is a timing dependent problem that might occur when run with a generational garbage collector. Low probability (CR229720).

Crashes in cbGetCodeInfoOnAddr

Under certain circumstances, JRockit can crash when unloading classes (CR229723).

JRockit may deadlock

If JRockit has been installed as a java plugin for a browser, pressing "v" (dump threads) in the Java Console has a high chance of causing JRockit to deadlock (CR225442)

demo\jpda trace example fails

This is because line numbers are not available for the classes in the jrockit.jar file (CR221291).

Corrupt or missing fonts or icons

When running AWT/java2d/java3d applications with JRockit, on certain configurations, the AWT DirectX acceleration might fail to initialize. This is due to virtual memory space conflicts.

Workaround: set an explicit -mx flag to approximately half the amount of the physical RAM or less.

Static field values are reset to zero/null.

Using dynamic code replace from a debugger will cause static field values to be reset to zero/null, and can also cause spurious crashes at a later time due to un-updated references to the old field values.

Workaround: do not redefine class definitions containing static fields

JRockit is crashing due to a signal handling conflict.

For Linux users only.

If you are using JRockit in conjunction with a native library that relies on OS signals you may experience crashes due to a signal handling conflict between JRockit and the native library.

Workaround: set the environment variable LD_PRELOAD as follows:

export LD_PRELOAD=$JROCKIT_HOME/jre/lib/i386/libjsig.so

BEA Engineering found this conflict using IBM's MQSeries native drivers, and it may be present in other libraries that rely on native code.

For more information, see http://java.sun.com/j2se/1.5.0/docs/guide/vm/signal-chaining.html