- This will be the final release of OpenDDS that will build with the current patch level of TAO 1.4a. This release is compatible with (the current patch levels of) TAO 1.4a, 1.5a, and 1.6a, as well as the current DOC Group beta/micro release. Future releases will be compatible with TAO 1.5a, 1.6a and the DOC group -- see the README file for details on TAO versions.
-
Implemented the Communication Status entity structures and behaviors to conform with the version 1.2 specification (formal/07-01-01). This includes member name changes and additions to LivelinessChangedStatus; modifying and extending the SampleRejectedStatusKind enumeration; and, adding new members to the PublicationMatchStatus and SubscriptionMatchStatus structures.
-
Implemented the TRANSPORT_PRIORITY Quality of service policy using the policy value to set thread priorities and DiffServ codepoint values. The current implementation does not honor the Changeable behavior of the policy. Only the intially established policy value is honored for any given DataWriter Entity. Thread priorities are not currently set on Windows systems.
-
Implemented the LATENCY_BUDGET Quality of service policy as a monitoring feature with an extended listener callback for exceeded budget values. Statistics for latency are made available via an OpenDDS-specific extended API as well.
-
Implemented the ReadCondition interface and associated operations on the DataReader (such as read_w_condition).
-
Implemented the QueryCondition interface, but only queries of the form "ORDER BY a[, b, c...]" have any effect. In other words, it's a mechanism that controls the order of the data-set received from any given read or take operation. The ORDER BY fields can be either primitive types (including strings) or primitive types in nested structs (using a.b).
-
Corrected a bug in the LIFESPAN QoS implementation where the Topic QoS was used. Now the DataWriter QoS takes effect.
-
The PARTITION QoS policy now accepts character class wildcards using [].
-
Corrected a bug in the PARTITION QoS implementation where the default QoS would not match against the '*' wildcard.
-
The DURABILITY QoS implementation was corrected so that readers requesting the VOLATLIE kind will not receive historic samples from writers. Since there were some use-cases for the old behavior, it is available through an OpenDDS-specific interface -- see $DDS_ROOT/dds/DdsDcpsSubscriptionExt.idl.
-
When the writer's DURABILITY QoS is PERSISTENT, the format of the data stored on the filesystem has changed. Instead of a single file in the current working directory, a directory is now used. This is configurable using the option -DCPSPersistentDataDir.
-
A Java Message Service 1.1 (JMS) provider built on top of OpenDDS is now available in $DDS_ROOT/java/jms. JMS Topics are supported (not queuing). See the README in that directory for more info.
-
The $DDS_ROOT/bin/federation utility has been renamed to opendds_repo_ctl and now supports shutting down a remote non-federated DCPSInfoRepo process.
-
Deleting a DataWriter from a Publisher now blocks until pending samples have been delivered by the underlying transport. By default, the DataWriter will block indefinitely. To change the default blocking behavior, you may specify a number of seconds to block with the -DCPSPendingTimeout option.
-
Instance handles are now compliant with version 1.2 of the DDS specification. Additionally, the contains_entity and get_instance_handle operations have been implemented.
-
Implemented DataWriter::wait_for_acknowledgments() from formal/07-01-01.
-
Added support for Windows Mobile 6 (WinCE) Standard SDK, and ACE_USES_WCHAR builds in general. See docs/design/wchar.txt.
-
Implemented the WaitSet, GuardCondition, and StatusCondition interfaces from the OMG DDS specification.
-
Implemented DomainParticipant::get_current_time() from formal/07-01-01.
-
Added the ability for using multiple repositories to provide redundant service metadata management.
The ability to federate repositories and redirect applications to attach to a new repository in the event a repository fails or becomes unreachable. This is an experimental feature.
-
Removed requirement for repository domain specification files.
Repositories will manage DomainParticipants and their contained Entities in any domain without requiring the domain to be prespecified in a configuration file. This removes the '-d' command line argument from the repositories as well.
-
Internal service Entity identifiers have been expanded to 128 bits.
These internal identifiers were previously 32 bits. The 128 format follows the format of OMG Specification formal/08-04-09 (The Real-time Publish-Subscribe Wire Protocol DDS Interoperability Wire Protocol Specification, Version 2.0) GUID_t unique identifier type.
OpenDDS places the identifier value for the DataWriter Entity of a publication on the wire to identify the source of data samples. This means that this version of OpenDDS WILL NOT INTEROPERATE with previous versions since the 128 values are not compatible with the previous 32 bit values.
-
Java Bindings for OpenDDS
See $DDS_ROOT/java/README, $DDS_ROOT/java/INSTALL, $DDS_ROOT/java/FAQ
-
dcps_ts.pl changes
Modified this script to create a single idl, header, and source file no matter how many types are contained within the input idl file. This allows the output file names to be based upon the input file name (instead of the types found in the idl file). Users of prior OpenDDS versions can use the script $DDS_ROOT/bin/update_dcps.pl to assist with this transition.
Added a command line argument to allow an output directory to be specified. All output files will be placed into this directory if present. They will continue to be placed in the current directory if this argument is not present.
-
DCPSInfoRepo changes
A new library has been introduced (libDCPSInfoRepoServ) which provides the neccessary hooks for running the DCPSInfoRepo within another process. For example usage, see: $DDS_ROOT/dds/InfoRepo/DCPSInfoRepo.cpp.
-
Fixed bug in DCPSInfoRepo that did not re-evaluate the compatibility and associations when QoS change is made.
-
Fixed bugs that affected connection establishment.
Fixed DataWriterImpl to handle the case that remove_association is called before fully_associated is called.
Fixed a race condition that connection delay deletion removes new assocaitions.
Corrected fully association ack verification to fix issues when multiple associations of a datalink are added simultaneously.
-
Added datalink_release_delay and removed keep_link configuration.
The datalink_release_delay is the delay for datalink release after no associations. It defaults to 10 seconds. This is to fix disconnect problem happened when removing and adding association related to same datalink during short period.
Removed keep_link which can be replaced with a big value of datalink_release_delay.
-
Fixed a potential deadlock in reliable multicast transport.
-
Added support for PARTITION QoS policy.
This QoS policy allows publishers and subscribers to belong to one or more logical partitions. Partition names may contain wildcards. However, only wildcards containing "*" or "?" are currently supported.
-
Added support for DEADLINE QoS policy.
The DEADLINE QoS policy allows the application to detect when data is not written or read within a specified amount of time. The {Offered,Requested}DeadlineMissesStatus::last_instance_handle is currently not updated.
-
Added support for LIFESPAN QoS policy.
This policy allows the application to specify when data expires, meaning invalid. Expired data will not be delivered to subscribers.
-
Added support for the TRANSIENT kind of DURABILITY QoS policy.
Data will outlive DataWriters when TRANSIENT DURABILITY is configured, and will be sent once a new DataWriter is created within the same domain for the same topic and type. The cached data will only last as long as the process.
-
Added support for the PERSISTENT kind of DURABILITY QoS policy.
PERSISTENT DURABILITY provides basically the same functionality as TRANSIENT DURABILITY, except that cached data will survive process destruction.
-
Added support for the DURABILITY_SERVICE QoS policy.
This QoS policy provides a way to specify resource limits for domain/topic/type-specific data in a TRANSIENT or PERSISTENT DURABILITY cache. The resource limits in question are analogous to, although independent of, those found in the HISTORY and RESOURCE_LIMITS policies.
-
Added a new command-line option, "-DCPSTransportDebugLevel".
This can be used to specify a debug level for the OpenDDS transport layer at run-time. OpenDDS debug statements range from 0 to 5 in increasing order of verbosity.
-
Added the ability to attach to multiple DCPSInfoRepo processes.
Applications which communicate using DDS are no longer restricted to communicating using domains controlled by a single DCPSInfoRepo process. Individual domain publications and subscriptions within a process can be attached to a designated repository. Each domain is controlled by a single repository and this association may not be changed for the lifetime of the repository or the application.
-
Added QoS propagation.
The client applications can change the qos of DCPS entity via set_qos() call. The qos updates are propagated to DCPSInfoRepo which publishes the updated qos to corresponding BIT datawriter and also save it to persistent repository.
-
Fixed a few bugs in persistent DCPSInfoRepo support.
Fixed incorrect actor type and qos type for subscriber/datareader problem in UpdateManager::add(const UA& actor). This bug causes the reloading from persistent file maps the subscriber/datareader as publisher/datawriter.
Fixed a bug in qos serialization. The TAO_OutputCDR stream was not fully copied to buffer when the qos contains sequence data. The TAO_OutputCDR stream needs be consolidated before memcpy.
-
Fixed compliance with qos RxO.
Fixed compliance with the DDS specification that qos parameters that Durability kind and Deadline.period be compatible between the subscriber and the publisher (RxO - Received x Offered)
-
Improved Memory Mangament.
Fixed issues with reference counting for classes derived from LocalObject. If a user defined application is crashing on exit or cleanup, see the OpenDDS FAQ.
-
Added support for IPv6.
The requirement for using IPv6 is the fully qualified domain name (FQDN) needs be resolved as IPv6 address as it is used by default configuration to support BuiltInTopic.
The local_address in the configuration file should use IPv6 decimal address or hostname that can be resolved as IPv6 address.
-
Used the address string instead of 4-byte IP address and port number for pub/sub endpoint address advertisement.
The address string in the configuration can use decimal IPv4/IPv6 address or hostname that can be resolved on both pub and sub sides.
The default address string is "host:port". The "host" is the fully qualified domain name and the "port" is picked randomly by OS.
-
Fixed a problem that the default local_address does not work when hostname is resolved as the loopback ip address.
With the default local_address, the acceptor now is listening on IPv4/IPv6 ANY, but use the fully qualified hostname for address advertisement.
-
Added default configuration for the reliable multicast transport type.
The multicast_group_address defaults to the ACE default multicast address (224.9.9.2:20001 for IPv4 and ff05:0::ff01:1:20001 for IPv6) and the local_address defaults to ANY with a random port picked by the OS.
-
Fixed a bug in DataReaderImpl::remove_associations() and DataWriterImpl::remove_associations() that should not attempt to remove the already removed writer/reader.
-
Fixed a bug in SimpleTCP that an active disconnect due to the peer graceful disconnect does not terminate send. This bug made the re-established connection fail on send strategy state check.
-
Added dispose() and unregister() notifications. When dispose() is explicitly called on the DataWriter side, the DataReader listener on_data_available() callback should receive a sample with the valid_data flag == 0 and the instance state == NOT_ALIVE_DISPOSED_INSTANCE_STATE.
When unregister() is explicitly called on the DataWriter side, the DataReader listener on_data_available() callback should receive a sample with the valid_data flag == 0 and the instance state == NOT_ALIVE_NO_WRITERS_INSTANCE_STATE.
-
Moved TRANSPORTTYPE_QOS_POLICY_NAME and TRANSPORTTYPE_QOS_POLICY_ID from DDS namespace to OpenDDS namespace since they are specific to OpenDDS.
-
This product has been renamed to OpenDDS from "TAO DDS"
Namespace and module names have been changed accordingly, except for a few cases where we are counting on certain names being generated by the TAO_IDL compiler.
-
The default participant setting for BITs is now "on"
This change was actually part of 0.12 but the release notes were not updated to include it. Please see the entry with the same title under the "Version 0.12" section below for more details.
-
Simplified .mpc and .mpb files
Projects that use DDS no longer need to explicitly list "portableserver" as a base project. It will be included by the "dcps*" base projects such as "dcpsexe" and "dcpsexe_with_tcp". Also, an .mpb file is provided for each of the transport libraries (dcps_tcp.mpb, dcps_unreliable_dgram.mpb, dcps_reliable_multicast.mpb).
-
Supported built-in sequences (the CORBA::*Seq sequences).
-
DDS keeps copies of the $TAO_ROOT/tao/*Seq.pidl in $DDS_ROOT/dds/CorbaSeq directory and renamed to *.idl files. These idl files are compiled with a new IDL option -Gdcpsonly to just generate the DDS specific code for built in sequences. The DDS CORBA::*Seq generated code will include the TAO CORBA::*Seq generated code so both TAO and DDS functions will be available.
-
When a DDS application contains an idl file that includes $TAO_ROOT/tao/Seq.pidl, the generated code will include the DDS specific built in sequences code(under $DDS_ROOT/dds/ directory) instead of TAO specific built in sequences code(under $TAO_ROOT/tao/ directory).
-
The idl files in $DDS_ROOT/dds/CorbaSeq directory are internally used by DDS and CAN NOT be included in your idl files.
-
Currently the CORBA::AnySeq is not supported since DDS does not support serialization of Any type data.
-
The DDS entities (e.g. DomainParticipant, FooDataReader) have changed from remote IDL types to local IDL types and the data sequence and info sequence types have changed from normal IDL sequences to types with extra functionality required to support zero-copy reads.
This change requires changes to DDS user code.
-
The ZCSeq and ::TAO::DCPS::SampleInfoZCSeq no longer exist. The Seq and DDS::SampleInfoSeq have been changed to support both single-copy reads (as before) and zero-copy reads (like ZCSeq should have).
Change ::TAO::DCPS::SampleInfoZCSeq to ::DDS::SampleInfoSeq. Change ZCSeq to [Modlue::]Seq.
The default constructor for Seq enables zero-copy reads. Constructing a Seq(num) with num > 0 enables single-copy reads.
-
Listeners will now inherit from DDS::Listner instead of POA_DDS:: or POA_TAO::DCPS::DataReader/WriterListener and should inherit as a local object (not a servant). DDS provides a helper template that adds reference counting. Use of this template (as shown below) is highly recommended.
For example:
class DataReaderListenerImpl : public virtual POA\_DDS::DataReaderListener, public virtual PortableServer::RefCountServantBase {...};Will become:
class DataReaderListenerImpl : public virtual TAO::DCPS::LocalObject<DDS::DataReaderListener> {...};But the header and implementation should require no other changes.
NOTE: if you are implementing the TAO specific reconnection callbacks, on*disconnected, on_*_reconnected, on_*_lost, on_connection_deleted then you should change inheritance from POA_TAO::DCPS::DataReaderListener to TAO::DCPS::LocalObject<TAO::DCPS::DataReaderListener> and remove inheritance from PortableServer::RefCountServantBase
-
Implementation reference counting cleanup code is no longer used
For example:
SimpleTypeSupportImpl sts_servant = new SimpleTypeSupportImpl; PortableServer::ServantBase_var safe_servant = sts_servant; SimpleTypeSupport_var fts = TAO::DCPS::servant_to_reference (sts_servant); if (::DDS::RETCODE_OK != fts->register_type(dp.in (), MY_TYPE))becomes:
SimpleTypeSupport_var fts = new SimpleTypeSupportImpl; if (::DDS::RETCODE_OK != fts->register_type(dp.in (), MY_TYPE))Note: Although the old style is discouraged it will still work.
-
Since the DCPS interfaces are now local, there is very little performance gain to convert from _var to a servant pointer (using reference_to_servant).
For example:
::DDS::DataWriter_var dw = pub->create_datawriter( topic.in (), dw_qos, ::DDS::DataWriterListener::_nil()); Test::SimpleDataWriter_var foo_dw = Test::SimpleDataWriter::_narrow(dw.in ()); // This is unnecessary but will still work. // Previously fast_dw was used to increase the // performance of writing samples. Test::SimpleDataWriterImpl* fast_dw = TAO::DCPS::reference_to_servant<Test::SimpleDataWriterImpl>(foo_dw.in()); -
Also, a pointer to the servant is no longer needed for making calls on zero-copy read supporting overloaded methods.
-
If the user defined DDS type is in a module then the generated types will also be in that same module.
Given a "Foo" DDS type defined in the module "Test": old new -------- ------------ type Test::Foo Test::Foo sequence FooSeq Test::FooSeq reader FooDataReader Test::FooDataReader writer FooDataWriter Test::FooDataWriter -
The --module option to dcps_ts.pl is no longer supported. The module is set as described in point #5 above.
-
What did not change. You may use the following: _var, _ptr ::_narrrow() // might use this for a listener ::_duplicate() TAO::DCPS::servant_to_reference() TAO::DCPS::reference_to_servant() TAO::DCPS::deactivate() // now a no-op
- if you used:
- TAO::DCPS::servant_to_reference()
- TAO::DCPS::reference_to_servant()
- TAO::DCPS::deactivate_object
- for non-DDS interfaces then you may change to:
- remote_reference_to_servant
- servant_to_remote_reference
- deactivate_remote_object
- Made the sub/pub repo id generated by DCPSInfoRepo to be unique per DCPSInfoRepo instance instead of being unique per domain. This would allow multiple domains in the same process(connect to the same DCPSInfoRepo instance) share the same transport.