Release Notes 5.1.1
for use with JBoss Enterprise Application Platform 5.1.1
Chapter 1. Introduction
1.1. About JBoss Enterprise Application Platform
1.2. About this release
1.3. What's New
1.4. Excluded, Removed, or Deprecated Items
- An item that has never featured in a product release but is otherwise part of one of the open source components of the product.
- An item that will be removed from a future release, usually the next major version.
- An item that was previously in a release of the product and is no longer included. Items will usually be deprecated before being removed.
Chapter 2. Installation Notes
2.1. Supported Configurations
2.2. Installing JBoss Enterprise Application Platform
2.3. Default Startup Profile
default, which is a base Java EE 5 server profile containing a default set of services. It contains the most frequently used services required to deploy a Java EE 5 application. It does not include the JAXR service, the IIOP service, or any of the clustering services.
defaultprofile is not intended for production use, or for running load, stress, availability or performance tests.
2.4. Source Files
2.5. Product Support
Chapter 3. New Features
HornetQ is offered as a technology preview (TP) in JBoss Enterprise Application Platform 5.1.1. You must download and install HornetQ separately to JBoss Enterprise Application Platform 5.1.1, using the standalone zip available from the Customer Portal. Refer to the HornetQ User Guide for more information.
JBoss Enterprise Application Platform 5.1.1 is now available as certified RPMs, which adds to the certified installation zip files available for Red Hat Enterprise Linux 6.
The following database and JDBC combinations are now certified to run with JBoss Enterprise Application Platform 5.1.1:
- PostgreSQL 8.4.x
- Certified to run with JDBC4 PostgreSQL Driver, Version 8.4-702
- Sybase ASE 15.5
- Certified to run with Sybase jConnect JDBC driver v7 (Build 26502/EBF17993)
- Microsoft SQL Server 2008 R2
- Certified to run with Microsoft SQL Server JDBC Driver 3.0.1301.101
Apache httpd Connectors mod_cluster, and mod_jk are certified to work with the Apache httpd version shipped in Red Hat Enterprise Linux 6. Refer to https://issues.jboss.org/browse/JBPAPP-6195 for more information.
3.2. By component
- Property substitution did not work for
<max-pool-size>in deployment descriptors. This problem has been fixed, and property substitution now works for all properties in deployment descriptors.
SchemaUpdatenow creates indexes automatically, when property variable
hibernate.hbm2ddl.autois set to either
update. The previous behavior was to only update when the variable was set to
- A new property called
ORBGracefulShutdownhas been added to the
iiop-service.xmlfile. The default value,
false, retains the old behavior. A value of
truecauses the ORB to wait for completion of outstanding calls before it shuts down.
- If multiple images were used on the same page, and
s:graphicImagewas used to do transformation on the images, page loading was sometimes slow. This has been resolved by making
- Previously, BMP images were not supported for
s:graphicImage. BMP image support is now available, as content type
- Exposing the keystore or truststore password in JaasSecurityDomain is insecure. Two methods have been added to retrieve private keys and certificates directly from the JSD, so that they can be used by external components. These methods are
getKeymethod requires you to provide a security token.
- The JBossWS configuration management layer could not reference a JaasSecurityDomain JNDI name, which also prevented the JBossWS security layer from accessing keystores and truststores normally exposed by the Security Domain. The JBossWS management and security layers were updated to permit alternate
jboss-wsse-client.xmlfiles. The alternate XML files allow a JaasSecurityDomain JNDI name to be specified. The keystores and truststores for the security domain are exposed, providing greater security for JBossWS transactions.
- Exposing the keystore or truststore password in JaasSecurityDomain is insecure. Two options have been added to retrieve private keys and certificates directly from the JSD, so that they can be used by external components. These methods are
getKeymethod requires you to provide a security token.
- JaasSecurity Domain now includes
serverAliasoptions. You can configure them in the same way as
Chapter 4. Fixed Issues
- The following unnecessary files were included in the installation of Enterprise Web Platform that only pertained to Enterprise Application Platform.
These files have been removed from the installation procedure.
- The version of Seam distributed with JBoss Enterprise Application Platform 5 included a superfluous dependency on the Hyper Structured Query Language Database (HSQLDB). This dependency has been removed.
- JBoss Native Zips did not contain the sha256sum value within the zip archive. The values were included in a separate file, which had to be downloaded separately with the native zip archive. This caused problems with the release process, and maintaining an accurate sha256sum list. The sha256sum values are now included in the zip archive.
- When a multicast address (
mcast_addr) was incorrectly configured, the warning that appeared in the log included an outdated URL. The URL has been updated.
org.jboss.system.server.profileservice.repository.clustered.local.file.AbstractFileWriteActionhas a member variable
tempFile, which is set to a newly created file in
getTempFile(). This file is called from
getOutputStream(), which is called from
writeBytesis only called if there is data in the file.If the file is empty,
tempFileis never set, and
AbstractFileWriteAction.modifyTargetpasses a null
FileUtil.localMove, where the
getTempFile(), which creates the file if needed, instead of using
encryptclass of JGroups had bugs related to threading in the Cipher routine. These have been addressed.
- A software error resulted in a situation where neither the process of serving the login page nor handling the
j_security_checkresulted in a call into
ClusteredSessionValve, which is the part of the request pipeline that triggers replication of the session by storing it in the database.
FormAuthenticatordirectly dispatched to the login page, and handled
j_security_checkitself.As a result, the session that was created before serving the login page was not persisted. When the authentication was complete and the request for the original URL came in, the session manager checked with the distributed store to see whether the local session copy was out of date. With
DataSourcePersistentManager, that check was made to class
VersionBasedOutdatedSessionCheckerreturned true because it could not find the session in the database. Because of the
truereturn value, the session manager could not use the local session. It attempted to read it from the persistent store. It did not exist in the persistent store either, so a new session was created.To address this issue,
VersionBasedOutdatedSessionCheckernow checks the session's
getLastReplicatedmethod if it cannot find a persisted session. If the value is 0, the session is new, and has never been persisted, so it cannot be outdated. In that case it returns false.
- An incorrect subsystem was set during remoting client creation, due to the order of superclass constructor operations. This created
ClassCastExceptions. Constructors have been altered to initialize the client after the correct subsystem is set.
- A timer that is shared across all channels in a group did not stop when one of the channels was closed. If OOB messages were passed around at the time a channel left a group, messages were added to the Retransmitter after the channel closed. This happened because messages which were added to the Retransmitter continued to be requested after
Retransmitter.reset()completed. This issue is resolved by fixing the condition that caused retransmission requests when a message from a node is processed at the same time the node leaves the cluster.
- In Microsoft Windows, when the network interface was disrupted, such as by removing the network cable and reinserting it, JGroups received an exception of type
NoRouteToHostExceptionevery time it attempted to send a message. This continued until the server was restarted. This problem is solved by new behavior in JGroups, which re-binds the socket to the interface, allowing JGroups to send messages again.
sym_providersecurity provider options of the ENCRYPT algorithm previously were only used by the
KeyGeneratorlookup. These options are now implemented in the
Cipherlookup as well.
- The JGroups
FD_SOCKprotocol is a failure detection protocol based on sockets. The client side of the
FD_SOCKTCP connection was not binding to a specific IP address. If you tried to open a firewall connection specific to each individual application, there was unnecessary cross-talk between applications.
bind_addrwere added to
FD_SOCK, which allows you to bind the outgoing connections to a specific IP address and/or port range.
- A race condition in the JGroups
FLUSHprotocol caused messages to be processed in the wrong order when the master node left the cluster, so the master node left before sending a new view (with a new master). Therefore, the other nodes did not detect the master's absence, or elect a new master, until the master's absence was detected by the failure detection protocol. The message ordering has been corrected, and the race condition no longer occurs.
- When a JBoss Cache instance configured for asynchronous serialization was shunned from the cluster, all future replications would fail until the cache was restarted. The cause was a naming conflict between a method and its parent class. This has been fixed by renaming method
- The JGroups configuration in the binding manager now implements the system properties added to the JGroups configurations in Enterprise Application Platform 5.1. In addition, hard-coded ports have been removed, and the default multicast addresses are set to the same values used in version 5.0 of the platform.
- If a user waits too long to log in after going to the Admin Console login page, the session times out. Previously, this caused an exception. A user-friendly timeout message is now displayed.
- If a topic or queue was created with the same name as an existing topic or queue, the new one overwrote the old one. This has been resolved by causing an attempt to create a topic or queue with a duplicate name to fail, generating an error.
ApplicationServerDiscoveryComponentclass does not use JBoss variables because the paths are hardcoded to the
jarfiles. If JBoss library locations and the JBoss variables for the libraries (
jboss.server.lib.url) were changed, the JBoss Admin Console failed to open. This issue is resolved by adding
BootstrapAction.createPluginContainerConfiguration()to the Admin Console.
- Seam Reference User Guide source code blocks did not have syntax highlighting, which made the source code difficult to read. The affected code has been updated with the correct highlighting.
- The Microcontainer User Guide contained references to some outdated Maven artifacts. Please review the document shipped with the current release for updated information.
- In the Hibernate Core User Guide, a reference was made to an incorrect POM file. This reference has been updated in the documentation for the current release.
- The configuration information in the JMS Clustering Notes section of the guide was inaccurate. The guide has been corrected with an updated configuration example, which fixes the issue.
- A change to the default value of the
max_prepared_transactionsdatasource property in PostgreSQL versions 8.2 and 8.4 resulted in XA Transactions being denied. An Important admonition was added to Appendix A.7 PostgreSQL of the Administration and Configuration Guide for the XA Datasource example, to highlight the requirement.
- The Security Guide released with earlier versions of the Enterprise Application Platform missed the instructions for encrypting the datasource password. The documentation now includes the step. The instructions to configure the
server.passwordfile by running the following command:
java -cp jboss-as/common/lib/jbosssx.jar org.jboss.security.plugins.FilePassword \ SALT COUNT MASTER_PASSWORD PASSWORD_FILE
- The HTTP Connectors Guide is a new user guide for JBoss Enterprise Application Platform 5.1.1. This guide supercedes and improves upon the mod_cluster User Guide and the HTTP Services section of the Administration and Configuration Guide. It also contains information about the ISAPI and NSAPI connectors.
- EJB3 clients did not use existing socket connections on subsequent invocations. Instead, they created a new connection for each invocation, which was destroyed when the connection was complete. A delay has been added so that connections remain open for use in subsequent invocations.
- In two cases, the
UnifiedClassLoaderwas used instead of an isolated EAR's classloader. The first case occurred when a EJB2 Entity Bean was packaged inside of an isolated EAR and an EJB client was packaged in the same isolated EAR. When the EJB Entity Bean was passivated, a
ClassNotFoundexception was thrown. This was caused by the usage of
UnifiedClassLoaderinstead of the EAR's isolated classloader. The second case occured when a deployed EJB2 Entity Bean had a local interface of the same name as an interface in the
UnifiedClassLoader, resulting in a
ClassCastException. The correct classloader is now used in both cases, and the errors no longer occur.
- A bug in
org.jboss.ejb.plugins.SecurityInterceptorcaused problems with setting the runas-identity context method invocation on stateless session EJBs that were not originally runas-deployed. Invocations used the identity of any authenticated context sent to the EJB, which resulted in invocations being executed as if the EJB was runas-deployed. The only way to stop this behavior was to restart the server.
SecurityInterceptornow looks at the run-as role of the original EJB, and ensures that runAsRole is available to any calls made by the EJB for declarative security checks.
- When an EJB Timer service was configured to use the
GeneralPurposeDatabasePersistencePlugin, and the
CachedConnectionManagerwas set to debug, the following warning was recorded in the server log:
WARN [org.jboss.resource.adapter.jdbc.WrappedConnection] Closing a result set you left open! Please close it yourself.This occurred because the
GeneralPurposeDatabasePersistencePlugindid not close an internally-used
ResultSetand correctly informed the user of the error. The
GeneralPurposeDatabasePersistencePluginhas been modified to close the
ResultSet, so the warning no longer appears.
- If you use JDK 6 update 19 or newer, and your application sends or receives
java.io.Fileobjects, a serialization exception may occur. This exception can also happen if you are using clustering with a stateful session containing a
java.io.Fileobject.To prevent this exception, a new property has been added. Set the system property
-Dorg.jboss.serial.SYNC_SERIALIZATION_BINARY_FORMATS=true, for all JBoss instances that communicate with each other.
- When Hibernate executed a cacheable query using a
ResultTransformer, it attempted to cache the results after applying the
ResultTransformer. However, the data might be modified so that Hibernate could not read it. In this case, a
ClassCastExceptionwould occur when attempting to cache the results.This issue is fixed by introducing three new classes to the API:
Review the Javadoc for Hibernate for more details.
org.hibernate.dialect.SQLServer2008Dialecthas been added to Hibernate, to address changes to the Microsoft SQL dialect introduced in Microsoft SQL Server version 2008.
- When a
refresh()method was invoked immediately prior to an
insert(), and second-level caching was enabled, the entity was inserted into the second-level cache. If the
refresh()committed successfully, however, the cached data was not automatically evicted. This occurred because
refresh()did not track entity state.
refresh()now tracks entity state, and evicts the cached data on a successful commit.
- JBoss Marshalling failed to serialize a session, due to a flaw in the serialization logic of
org.hibernate.impl.SessionImpl. This class now adheres to the Java Serialization specification listed at http://java.sun.com/javase/6/docs/platform/serialization/spec/output.html#86, and serialization works properly and as expected.
org.hibernate.type.EnumTypeused a static map to cache the enum values, because getting the value of the enum by reflecting is expensive. However, this implementation had the potential to cause memory leaks in come circumstances, especially when using Hibernate in the application server. This is because the cached objects would never be garbage-collected, so the Map acting as cache kept growing infinitely.Instead of using a global enum cache, Hibernate now uses a transient array for each
EnumTypeinstance to cache the enum values. This improves performance and eliminates the memory leak.
cascade-saveoperation was incorrectly handling auto-incremented Primary Key (PK) tables that linked to parent tables with an assigned PK. The fix now allows the
cascade-saveoperation to handle the parent-child relationship between tables.
- An issue with
QueryPlanCachewas discovered that caused a memory leak due to soft references held by the unbound
SoftLimitMRUCacheeventually filled up until a major stop-the-world garbage collection was necessary to clean the
SoftLimitMRUCachesoft references.The solution involves constraining the amount of soft and strong unbound references held in the cache. Entries are evicted using an LRU policy, or by memory pressure from the GC in the case of soft references.Two configuration options are now available, which solve the issue:
Users that want to emulate previous behavior, can set
hibernate.query.plan_cache_max_strong_referenceswhich defaults to 128
hibernate.query.plan_cache_max_soft_referenceswhich defaults to 2048
- According to the JPA persistence specification, the
ASkeyword is optional for collection declarations, such as the following:
collection_member_declaration ::= IN (collection_valued_path_expression) [AS] identification_variableA HQL/JPQA such as the one below syntax caused a parser exception:
SELECT o FROM EntityBean AS o, IN (o.items) AS l WHERE l.itemValue = '1'This occurred because Hibernate did not implement the
ASkeyword. Hibernate now follows the specification in this area, and allows the optional keyword.
ByteCodeHelper.readByteCode()was previously limited to 409600 bytes, but can now handle an entity class of any size.
ChainedPropertyAccessorwas used by
ChainedPropertyAccessorwas not serializable. Therefore, any cacheable queries using
AliasToBeanResultTransformerbroke during cache replication. For example, this query could not be replicated or cached to disk:
session.createQuery("select foo").setResultTransformer(new AliasToBeanResultTransformer(SimpleCount.class)).setCacheable(true).list();The behavior of
AliasToBeanResultTransformerhas been changed. When the
AliasToBeanResultTransformeris created/deserialized, it recreates the
ChainedPropertyAccessorinstance according to the cached
resultClass. Cacheable queries using
AliasToBeanResultTransformercontinue to work during cache replication.
propertyAccessor.hashCode(). Implementations of
PropertyAccessordo not override
equals(). Therefore, Cache lookups involving a
AliasToBeanResultTransformerresulted in a cache miss.Two
AliasToBeanResultTransformerobjects should only be equal and use the same hash code if their
resultClassis equal. Getters are determined by aliases, which are provided as an argument, so they are not needed to determine equality. Setters should be used to compute
hashCode(). However, due to the implementations of Setter, Setters do not override
equals().The fix is to cache the aliases corresponding to the Setters and use that to check for equality. Cache lookups involving a
AliasToBeanResultTransformerno longer result in a cache miss.
- When Hibernate executed a cacheable query using a
ResultTransformer, it attempted to cache the results. However, the data had the potential to be modified so that Hibernate could not read it.
PropertyAccessExceptionwas thrown when all of the following are true:
This issue is fixed by introducing three new API calls:
- The query has a
- Results are cached before being transformed
- Each result has a single value
Refer to the Hibernate Javadoc for more details.
- Hibernate Core
EntityMetamodel entityNameByInheritenceClassNameMapfield was used inconsistently. It used a class for puts to the map, but a method for gets from it. An exception was thrown when saving instances of subclasses with specific entity names. The
EntityModelclass has been updated to use class
InheritenceClassinstead of its
getName()method. Saving instances of subclasses no longer causes an exception in these cases.
org.hibernate.id.enhanced.OptimizerFactory.PooledOptimizerwas used in multiple JVMs with the same database, there was a risk of duplicate sequenced values being generated. This occurred because a method was invoked twice to initialize two different values (value and hiValue). If another JVM requested a sequence value between these invocations, the sequence related to the second invocation. This issue has been corrected by limiting the double-read to optimizer initialization, when the first read gives Hibernate the
- If an entity defined a collection with a property-ref-based key as cacheable, there was a problem loading the collection and putting it into its second level cache region. Hibernate used the incorrect key for the owning entity. It attempted to use the referenced property value instead of the corresponding ID value.Hibernate now checks whether the collection key is defined by a property-ref. If so, the key of the owner instance associated with the collection itself is used, if it exists. Hibernate resolves the owner of the collection against the owner of the Persistence Context.
- Synchronizing access to
Mapcaused performance degredation. Synchronization of
Mapis now managed by
ConcurrentHashMap, which improves performance.
- Many well-formatted and indented
Strings were built up for each element of a parsed HQL string, for each tree node. The purpose was to feed invocations of
log.trace(). The information was added to the log regardless of the log level. This caused extra output and larger log files. Hibernate now checks the log level, and only includes the formatted output when the log level is
- A customer reported an issue with the ScrollableResults JoinFetch, where ScrollableResults set child collection correctly at the first parent object, but after the second parent object only the first element of a child collection was included.The ScrollableResults logic has been improved to read sequentially. When a new parent is encountered, ScrollableResults assumes it has finished processing all rows for the previous parent.
ImportantThis fix relies on results being ordered consistently. Because some databases do not return sorted data (such as H2) you should explicitly apply "order by" for ScrollableResults result ordering.
jboss-log4j.xmlhas been updated to include the
jacorb.configlog level. The priority has been set to
ERRORto avoid printing unnecessary messages to the console during server startup.
- In the graphical installer, the directory selection dialog used the language specified in the environment where the installer was run, rather than the language requested by the user at the beginning of the installation. The installer has been updated so that the directory selection screen uses the language selected by the user at the beginning of the installation process.
- During installation of the Enterprise Application Platform on Solaris, the user was prompted for the root password. This was caused by permission issues when the installer tried to create some optional shortcuts (symbolic links). This issue has been fixed by removing the ability for Solaris platforms to create shortcuts through the installer. You can still create the symbolic links manually if needed.
- Previously, scripts for all operating systems were included when JBoss Enterprise Application Platform was installed, rather than being restricted to operating system being installed onto. The installer has been updated so that only appropriate scripts are included in the bin/ directory of the platform. Shell scripts are installed for Linux and UNIX systems, and batch scripts are included for Microsoft environments.
- The graphical installer allows you to specify a custom JAAS security domain for securing consoles and invokers. However, in the past, the Tomcat console was always secured with the JAAS security domain
jmx-console, even if it does not exist.This issue is fixed, and the Tomcat console is secured with the security domain you specify during installation.
- When a custom JAAS security domain was created during the graphical installation process, the custom JAAS security domain continued to use
jmx-console-roles.properties. This incorrect behavior has been changed, so that the custom JAAS security domain called NAME uses
- Seam examples were not correctly deployed using
ant explodedue to faulty path delimiters in the
build.propertiesfile. This caused Seam example deployment to fail. The
build.propertiesfile now uses the correct path delimiters.
- In Microsoft Windows installations of the Seam examples, the Deploy Hotel Booking Seam Demo shortcut in the Start Menu pointed to a directory one level above the JBoss Enterprise Application Platform installation. This caused an error of type
NoClassDefFoundErrorto appear in the log, and the example not to deploy. The path has been corrected, and the example now deploys from the shortcut.
- The security configuration screen of the graphical installer contained incorrect links to the documentation portal and user documentation. All links have been corrected.
- Due to a regression introduced by the fix in JBAS-5929, the JCA pool sometimes shut down without looking for checked-out connections. This issue has been fixed, and the JCA pool looks for checked-out connections before shutting down.
XAManagedConnectionFactoryclass did not allow for the
ismethodName form for getting and setting properties. This caused problems for XA DataSources when the property was a boolean type.
XAManagedConnectionFactorynow allows for the
isformat and handles boolean types without errors.
- Passing a negative timeout value to method
- The JCA code did not properly handle multi-threaded access, due to several potential causes of deadlocks involving JDBC operations and the JBossTS transaction reaper thread. The JCA code has been updated to resolve these problems, and multi-threaded access should work properly in the JCA.
ServiceMetaDataParser.parseValueFactoryParameter()method only considered the first child of the
<parameter>element. If the
<null/>element was surrounded by carriage returns, the node was treated as a text value and the parameter was not correctly substituted elsewhere. The following example failed, but now works properly.
<parameter> <null/> </parameter>
JMXOpsAccessControlFilterclass did not retain the role mappings for a user if the roles were set with the
DeploymentRolesMappingProvider. The filter class has been corrected, and the mappings are now retained as expected.
- Some debugging code was inadvertently left in
ChannelSupport.afterRecoveryEx(), and this code was visible in log files. It has been removed, making log files clearer and easier to read.
- The MessageSucker is responsible for migrating messages between different members of a cluster. The
onMessageroutine attempts to deliver messages to the local queue, among other tasks. If delivery failed, messages appeared to be lost. They were still in the database, but it was difficult to redeliver them.This was a defect in the JBoss Messaging component, and has been fixed in that component. The fix is complex, and is discussed in JBMESSAGING-1822. In short, a new layer of reliability has been added to the message delivery logic.
- A potential race condition could occur if a session was closed while a transaction was committing. In this case, an exception was thrown. A check has been added such that a committed message is not cancelled.
- Because of problems with the
pause()method of the connector, a race condition could occur when applications were undeployed. Sessions that were forwarded to the application server by mod_cluster before receiving an undeploy notification could result in an
error 503 - This application is not currently available. The method has been updated, addressing these issues.
- The status page of the mod_cluster manager was not updated upon failover, so worker nodes were listed as active and available after they had failed. The status page now updates when nodes fail.
- A server configured for failover with an application deployed in a subdirectory such as
/webapp, and another application deployed in
/, occasionally caused an error. This was because the failed-over application tried to deploy in
/webapp. This situation has been fixed, and failover now occurs as expected.
- The maximum length for cluster aliases has been raised from
64, to accommodate longer aliases.
- Clustered nodes were only retired after the worker retry timeout (set to 60 seconds). This resulted in error messages suggesting that a node was in error for 60 seconds when it was likely the node had retired correctly during this period. Nodes are now retired when the node receives a STATUS message, broadcast by the nodes approximately every 10 seconds.
mod_rewritecould not rewrite from the root (/) in httpd if there was no
ROOTcontext deployed in JBoss Enterprise Application Platform. This occurred because mod_cluster attempted to use the original URI instead of the URI provided by
mod_rewrite. The rewrite rule now works regardless of whether the
ROOTcontext is deployed to the platform.
- When AopC was in use, report generation failed with an exception of type
ArrayIndexOutOfBoundsException. This has been resolved so that a report is now generated as expected, and no exception is thrown.
- In isolated deployments, a
ClassNotFoundExceptionwas thrown when the application server tried to deserialize an info object associated with the Timer. This occurred because the wrong classloader (
threadContextClassLoader) was used to deserialize the object. This bug has been fixed and the correct classloader is now used.
- Deployment file names were not previously checked for legal paths. Poorly-constructed file names could cause unexpected file deletions or alterations. The deployment files are now checked for legal paths, and an exception is thrown if an illegal path is used.
- Due to a casting problem where a long was cast to an int, using
org.jboss.varia.scheduler.ScheduleManagerwith a short period, such as 1 ms, could cause negative repeat values. This has been remedied by using a long instead of an int.
- A bug in the Profile service caused the status of deployments to be reported incorrectly if they are started and stopped very quickly. This bug is still present. However, modifications have been made to the console to work around this by updating the status after a short delay, or after the page is refreshed. In normal usage, the status is shown as expected.
- Sending multiple requests from a WS client and using
fastinfosetcaused an increase of sockets in a
CLOSE_WAITstate and incorrect shutdown. This issue was fixed by introducing the
org.jboss.ws.client.remoting.disconnect.after.useJVM property, which causes the client remote to disconnect immediately. This property is enabled by default. If you disable it,
spring-hibernate-contactsexample failed with the error:
java.lang.IllegalArgumentException: object is not an instance of declaring classThis occurred because the
getContactById(@PathParam("id") Long id)method in the
ContactServiceImplclass was missing the @GET annotation. This error has been corrected, and the example code now works as expected.
Scripts and commands
- When starting the server using
service.bat, no options were inherited from
run.conf.bat. This occurred because
JAVA_OPTSto a placeholder value, which was not overridden by values set in
run.conf.batdue to a faulty check. This issue has been fixed, and options set in
run.conf.batare now used when starting the server with
JpdlParserin JBPM included some hard-coded file names of XSDs, which referred to internet locations and ignored XSDs in the classpath. This caused failures when those internet locations were unavailable. The hard-coded paths have been removed and the classpath is now searched for the XSDs.
NullPointerExceptionoccured when undeploying or uninstalling Seam examples. It did not cause any application to fail, but the error was incorrect. The issue occurred in the nestedbooking, dvdstore, itext, and excel examples, but not in the ui example. The cause was a bug in the EJB3 deployer, and an upgrade to that component resolved the issue. The
NullPointerExceptionno longer occurs in the examples listed above.
- When uploading files with
<h:dataTable>, errors occurred which led to identical filenames and incorrect content. This occurred because the local value was not returned by the
getLocalValue()method. This problem has been fixed by correcting the method, and the problems with the filenames and content no longer occur.
- The IBM JVM v1.6 was not handling annotations that were UNKNOWN during runtime. This caused problems with the Chatroom example. The underlying issue is resolved on IBM JVM version 1.6.0 (SR9 FP1).
NoteTo make the chatroom example work with IBM JVM versions prior to 1.6.0 (SR9 FP1), copy http://repository.jboss.org/maven2/net/jcip/jcip-annotations/1.0/jcip-annotations-1.0.jar into JBOSS_HOME/server/PROFILE/lib.Restart the server after adding the specified file.
- The Seampay example had an issue with the duration between the first payment and the second payment in the example. Users with a combination of fast hardware, Java 6 runtime, and Windows Server 2003 found the set delay between payments to be too short. The example has been fixed to extend the duration between payments, which fixes the issue.
jta.jarwas being included as a third-party dependency of Eclipse by the seam-gen process. The
jboss-transaction-api.jarwas included in the seam-gen process, but the two files caused a
.classpathconflict. All Hibernate dependencies have been updated to exclude the
jta.jarfile, which fixes the dependency issue.
- The SeamBay example shipped as part of the Enterprise Platform contained an incompatibility with Internet Explorer 8. This has been resolved, and the example once again works with Internet Explorer 8.
- There were two identical
ant.jarlibraries in the Seam distribution within the Enterprise Platform:
This issue has been fixed by removing the
- An asynchronous invocation of JBPM which required transactional capabilities could cause the asynchronous action not to execute, if a transaction was not started. An exception was thrown and processing did not continue. This is fixed by checking whether transactions are required, and starting a new one if it is not already active, before initiating the asynchronous JBPM processing.
- During transactions where application exceptions were thrown, the transaction was neither committed nor rolled back. Instead it remained in memory. Transactions now roll back properly when exceptions are thrown, and are cleared from memory.
- The configuration for persistence service in
jbpm.cfg.xmlhas been updated in the Todo and DvdStore examples, to use the following value:
<service name="persistence" factory="org.jbpm.persistence.jta.JtaDbPersistenceServiceFactory" />
- If any exception occurred while a Seam EL expression was being evaluated during the
jbpm-jpdlprocess definition, JBPM's transaction handling sometimes committed transactions that should have been rolled back. Rollback only occurred on the Seam side, because Seam tried to close the JBPM Context while the JBPM process was committing. Seam now closes the JPBM context immediately after rolling back the transaction, and the dirty commits no longer occur.
- Invoking EntityQuery.resultCount() resulted in a query such as
select count( entity ).... However, if the persistence provider is Hibernate and the Entity has a composite key, the exception
[ERROR JDBCExceptionReporter] Operand should contain 1 column(s)is triggered, because Hibernate does not support composite keys in
countqueries.This constraint prevented you from using seam-gen and JBoss Developer Studio to generate CRUDs if you had Entities with composite keys and you use Hibernate. Hibernate extensively uses
EntityQuery.resultCount()in the generated code, and this class triggered the exception.Seam has been updated, so that if the persistence provider is Hibernate, Seam builds the query using the notation
select count( * )instead of
select count ( entity ). This method works, even with composite keys. With this fix in place,
EntityQuery.resultCount()works for all entities and all persistence providers, as do generated CRUDs.
- Seam-RESTEasy integration module allowed anemic session requests to remain open when an exception occured during the JAX-RS request invocation. Accessing previously authenticated sessions was possible even if incorrect credentials were passed in a request. The code responsible for invalidating the session is now contained in a Java
finallyblock. This fix prevents anemic session requests from remaining open.
s:linkwas not working if pages containing the
s:linkwere not previously rendered on the JBoss Enterprise Platform instance. This caused an IllegalStateException: Unable to read <page>, because the method binding for the Seam UI page actions were not added to the SafeActions on other servers.
viewIdnow includes the path to the web context root in each searching path, which allows pages to remain active when workload is moved between nodes.
- In seam-gen, Entity fields that were annotated by
@Versionwere set as an Integer data type. The Update operation was not compatible with the Integer data type, and terminated with a
@Versiondata type has been changed from Integer to int. Entity Update operations now complete successfully.
- It was found that JBoss Seam did not properly block access to JBoss Expression Language (EL) constructs in page exception handling, allowing arbitrary Java methods to be executed. A remote attacker could use this flaw to execute arbitrary code through a specially-crafted URL provided to certain applications based on the JBoss Seam 2 framework.
NoteA properly configured and enabled Java Security Manager would prevent exploitation of this flaw. CVE-2011-1484)
- It was found that the fix for CVE-2011-1484 was incomplete: JBoss Seam 2 did not block access to all malicious JBoss Expression Language (EL) constructs in page exception handling, allowing arbitrary Java methods to be executed. A remote attacker could use this flaw to execute arbitrary code via a specially-crafted URL provided to certain applications based on the JBoss Seam 2 framework.
NoteA properly configured and enabled Java Security Manager would prevent exploitation of this flaw. (CVE-2011-2196)
- Applying the workaround for JBAS-7049 causes servers using OpenJDK 6 and running the security manager to fail to start the Enterprise Application Platform. This was due to upstream issues in OpenJDK. These issues have been addressed in the OpenJDK software, which is a third-party product. Therefore, the Enterprise Application Platform now starts as expected on affected systems with upgraded OpenJDK.
- Due to an error in the order of services starting up,
JNDIContextEstablishmentcould potentially start later than expected, causing
NamingExceptionsbecause the context
java:/jaaswas not bound in JNDI. This binding is now handled in
JaasSecurityManagerService, which starts before any service in the
unauthenticatedIdentityThe option did not work with the
LdapExtLoginModuleclass, even when the method invoked did not require a specific role. The security libraries have been updated to fix this issue.
org.jboss.security.plugins.FilePasswordfile previously required write permission even for decoding the password. The file could not be changed to read-only for regular users because it was inaccessible. This issue has been fixed by changing the file permissions from to read-only.
LdapExtLoginModulequeried LDAP for the
roleNameAttributeIDinstead of checking the returned distinguished name. This was slow. To improve performance, a
parseRoleNameFromDNoption has been added to the module.
QueuedPessimisticEJBLockin JBoss Transactions did not always detect that a waiting transaction had timed out, so the lock would remain active when it should have been released. This eventually resulted in a threadpool-wide lock, with all transactions waiting for the lock to release, or the locked transaction to be registered as timed out. The
isTxExpired()method has been modified to ensure that timeouts are registered, so that transactions are informed of locks.
- The JBossTS TransactionReaper contained a bug which caused it to execute continuously when it was running in dynamic mode, rather than pausing between runs. This caused performance degradation. To fix this issue, JBossTS was updated. As a result, the Reaper now pauses between runs, leading to better performance.
- The Transactions Recovery Manager searched for XIDs in need of recovery in such a way that sometimes recoverable XIDs were missed. This happened when consecutive recovery scans for a given resource manager returned non-equal lists of in-doubt XIDs. The matching algorithm has been enhanced to handle recoverable XIDs.When using TX Resource managers, which are JDBC drivers that support XA, it is an anti-pattern to rely on the order of the XIDs the Resource Manager returns to you, because you risk losing XIDs and causing heuristic exceptions, in which you will need to manually commit or roll back the transaction.
- When nodes unexpectedly die during active transaction processing, zero length transaction journal files may be created. Upon node restart, the processing of these files caused logged warning messages, even though they do not contain any information that can be used to recover the transactions in question. This fix suppresses the log for this specific situation.
- A bug in class
org.jboss.mail.SessionObjectFactorycaused only one mail service to ever be resolved, even if more than one service was configured and present in the JNDI tree. The class now resolves more than one mail service if one is configured.
sendNotificationwas present in
org.jboss.web.tomcat.service.deployers.TomcatServices.java, but was disabled, so the
jboss.tomcat.connectors.started notificationwas missing. The Barrier Controller depended on that notification and was unavailable to manage subscriptions. This problem has been fixed by reinstating
- Customers migrating from Oracle WebLogic need the ability to rewrite URLs in a different context from the webapp. This feature is now implemented. To override the context, set
- The HTTPS connector used
addHandshakeCompletedListener()to verify a SSL handshake was successfully negotiated between the connector and listener.
addHandshakeCompletedListener()added the listener after the notification thread started running, which required SSL handshake renegotiation.
setEnabledCipherSuites(new String)is now used to verify SSL Handshake negotiations, which fixes the issue. This issue was introduced by a fix for CVE-2009-3555.
- When using the
org.apache.tomcat.util.http.ServerCookie.VERSION_SWITCH=falseproperty, cookies containing characters
:were truncated to the first occurrence of one of the listed characters. Changes have been made to the way these characters are handled in the code which prevents cookies containing these characters being incorrectly truncated.
- Fault signing and encryption have been added to JBoss WS. These features can be enabled by specifying
includeFaults="true"on a sign or encrypt element. For the client, specify
includeFaults="true"in the encryption and signature elements that are within a
<requires>tag to enforce signed/encrypted faults.
- When services were deployed to JBoss Web Services Native, any XML Schema Declaration (XSD) imports with relative URLs were rewritten as absolute URLs relative to the WSDL service contract. This worked fine, unless an imported XSD imported a second XSD using a relative path. In this case, the path of the second XSD was incorrectly rewritten to be relative to the first XSD instead of the WSDL service contract. This problem has been resolved, and the paths are now rewritten relative to the WSDL service contract.
- You can now disable chunked encoding using the SAAJ API, by calling
- Generating WSDL for JAX-WS resulted in an incorrect namespace prefix for the message part in the WSDL. The wrong namespace prefix was referenced if the exception was mapped to a different namespace that used the @WebFault annotation. This issue has been fixed by correcting how WSDL is generated, and the correct namespaces prefixes are now used.
- Due to an error in the SAAJ implementation in the JBossWS libraries, SOAP nodes previously had the same prefix as the
<Envelope>node. This has been corrected, so that SOAP nodes now have their own prefix.
- DescriptorDeploymentAspect loaded the
/cxf.xmlin the classpath and put it in the deployment attachment as the CXF endpoint deployment descriptor. This prevented the
WEB-INF/jbossws-cxf.xmlfrom being loaded. These are intended to co-exist with
/cxf.xml, which creates the CXF bus. The code has been updated so that these files are now loaded properly.
- Appending an
SOAPBodyresulted in a
java.lang.IllegalArgumentException. The code has been updated so that adding an
Elementnow results in the
Elementbeing converted to a
SOAPElement, and no exception is thrown.
- The logging level for
MessageContextJAXWSwas too high when the conversion of a WSDL URL to a URI failed. This caused excessive log messages. The logging level for these messages has been lowered to reflect the true severity.
- When a request with an
application/fastinfosetcontent-type header was sent to the
FastInfosetweb service, the response was returned with an incorrect header of
text/xml. This issue has been corrected with an upgrade of the JBoss Web Services component.
- JBossWS does not reuse an established SSL connection when sending multiple requests over SSL to JBoss Enterprise Application Platform. This means an SSL handshake occurs on every request. This issue has been fixed and the SSL handshake occurs once and is reused for every subsequent request, if a local WSDL is used.
Chapter 5. Known Issues
- If you use the partitionName property on the
MCBean:ServerConfigProfile Service managed component, it returns a
nullvalue. Instead, use the partionName from the
MCBean:HAPartitionmanaged component, which is the correct property.
- Using the Hibernate SecondLevel Cache at the
InitialStatephase causes a
NullPointerException. To work around this issue, disable the query cache. To disable the query cache, edit
persistence.xmlto either explicitly set the
false, as follows, or remove the line altogether, since the query cache is disabled by default.
<property name="hibernate.cache.use_query_cache" value="false"/>
- If you change the database properties via the Admin Console, the changes are persistent. However, if you undo the changes, the old values still persist when you restart the Application Server. This problem is not yet resolved, but you can work around it by editing the associated attachments file in
server/PROFILE/data/attachments, to force the datasource configuration to use the new settings.
jboss-ejb3-core-1.3.5with default setting
LANG=Ccauses an unmappable character for encoding ASCII error. Use
LANG=en_US.UTF-8to work around this issue.
- The deployment of a persistence unit into an EAR will fail when the unit is outside of the EAR and the bean attempting to inject the persistence unit is within the EAR. The injection fails because the persistence unit cannot be found. This is expected behavior of the EJB3 spec. For strict EJB3-spec compliance, the persistence unit should be packaged within the EAR. A JBoss-specific behavior allows persistence units to exist outside EARs. This is configured in the file
deployers/ejb3.deployer/META-INF/jpa-deployer-jboss-beans.xmlunder a JBoss AS server profile. The relevant section is:
<bean name="PersistenceUnitDependencyResolver" class="org.jboss.jpa.resolvers.DynamicPersistenceUnitDependencyResolver"/>
- When using the Criteria API with LEFT OUTER JOIN to add criteria to children, the child collections will only contain those children matching the criteria. This behavior also applies when not using any Filters. For example:
criteria.createCriteria("children", JoinFragment.LEFT_OUTER_JOIN)There is no workaround for this issue.
- org.hibernate.PropertyAccessException may be thrown if an Entity contains the following conditions:
- Uses @EmbeddedId
- Uses @JoinTable on a collection or association property/field, which references another property/field of the entity.
@EmbeddedId private MyPk id; private Long name; @CollectionOfElements @JoinTable( name="GLOBAL_NOTES", joinColumns=@JoinColumn(name="text_id", referencedColumnName="name")) private Set<String> globalNotes = new HashSet<String>();There is no workaround for this issue, other than not using @EmbeddedId.
- Oracle JDBC driver version 22.214.171.124.0 causes the JBoss Messaging Transaction Service to fail with a
SQLException("Bigger type length than Maximum") on Oracle 11g R1.This is caused by a regression in Oracle JDBC driver 126.96.36.199.0.We recommend Oracle JDBC driver version 188.8.131.52.0 for use with Oracle 11g R1, Oracle 11g R2, Oracle RAC 11g R1 and Oracle RAC 11g R2.
- Two JBoss Messaging Test Suite tests fail on Oracle 11g R1, R2 and RAC with the latest JDBC driver, version 184.108.40.206.0:
These tests use a large value for the fullSize queue configuration parameter, which is passed to the
setFetchSizemethod on the
java.sql.PreparedStatement. A problem with the JDBC driver means that more than the usual amount of memory is consumed when
executeQuery()is called, which results in a
java.lang.OutOfMemoryError, causing the test to fail.
- When using Sybase database with the JDBC driver Sybase jConnect JDBC driver v7 (Build 26502), a
MaxParamsattribute in the
sybase-persistence-service.xmlconfiguration file cannot be set higher than
481due to a limitation of the
PreparedStatementclass in this driver. Unexpected failures can occur when the
MaxParamsattribute is set to a value higher than
481. To avoid this issue, set
MaxParamsto a smaller value than
- If a message-driven bean is deployed with the default settings
useDLQ=true, DLQMaxResent=5and message redelivery is prompted, a message remains in the queue in a
deliveringstate even after the message is delivered to the dead letter queue. The workaround for this issue is to disable the dead letter queue handling on the MDB side by setting
- IPv6 is not supported in version 5 of the Enterprise Platform.
- Resteasy-guice applications fail to deploy because of a
java.lang.SecurityException. An error message similar to the following is displayed:
java.lang.SecurityException: class "org.jboss.resteasy.examples.guice.hello.DefaultGreeter$$FastClassByGuice$$70fd68d0"'s signer information does not match signer information of other classes in the same packageThis occurs because the
cglib.jarin JBoss Enterprise Application Platform is signed, and the
cglib-instrumentedproxy uses the
cglib.jarsigner information instead of the signer information of the application target class.
- The TwitterClient example is deprecated in RESTEasy 1.2.x due to Twitter deprecating the Basic Authentication method in August 2010. All applications must now use OAuth. RESTEasy 2.x contains a reworked TwitterClient example that includes OAuth. Download the example from RESTEasy 2.x for testing purposes.
jettison.jarfile is not included in the
jboss-eap-noauth-5.1.0.CR3.zip. To work around this issue, use the
jettison.jarfile included in the Seam distribution,
- The script
jboss_init_redhat.shhas been removed from distribution. Instead, use your own script to customize start-up of the Enterprise Platform.
- An incompatibility with Microsoft Internet Explorer 8 and the Seam Tasks Example resulted in text appended to a task being represented as a blank line in the task list. There is no work-around to this issue.
- Some examples included in the
jboss-seam2-examplesdistribution are designed to showcase functionality of the JBoss Enterprise Application Platform server. Those examples that require a full Application Platform server will not run on a JBoss Enterprise Web Platform server.
java.sql.Date.valueOfattempts to parse dates of the format yyyy-mm-dd, the TCK test threw a java.lang.IllegalArgumentException. This was due to a regression in the latest Sun JVM, Sun JDK 1.6.0_24 (see http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6898593 for more information). The workaround for this issue is to downgrade to Sun JDK 1.6.0_17.
- A NullPointerException was thrown when a datasource depended upon by a webapp was restarted. The error has been changed to an IllegalStateException, which returns "WebModules cannot be restarted, and must be redeployed".
Appendix A. Revision History
|Revision 5.1.1-105||Mon Jul 18 2011|