diff options
Diffstat (limited to 'cpp')
| -rw-r--r-- | cpp/CHANGES | 312 |
1 files changed, 158 insertions, 154 deletions
diff --git a/cpp/CHANGES b/cpp/CHANGES index d358fa701b9..e7499144b71 100644 --- a/cpp/CHANGES +++ b/cpp/CHANGES @@ -1,9 +1,10 @@ Changes since version 3.1.1 --------------------------- -- createObjectAdapter() and propertyToProxy() print warnings if unknown - properties are found during creation of the adapter/proxy. These - warnings can be disabled by setting Ice.Warn.UnknownProperties=0. +- createObjectAdapter() and propertyToProxy() print warnings if + unknown properties are found during creation of the adapter or + proxy. These warnings can be disabled by setting the property + Ice.Warn.UnknownProperties=0. - Fixed a race condition that could cause an assertion failure when using thread-per-connection. @@ -18,13 +19,14 @@ Changes since version 3.1.1 by default. If this is the desired behavior set the property IceBox.InheritProperties to 1. -- Fixed a bug where new connections are be accepted but not serviced - correctly when using the thread pool concurrency model if the number - of file descriptors registered with the thread pool exceeds FD_SETSIZE. +- Fixed a bug where new connections are being accepted but not + serviced correctly when using the thread pool concurrency model if + the number of file descriptors registered with the thread pool + exceeds FD_SETSIZE. -- Added Communicator::isShutdown() and ObjectAdapter::isDeactivated() to - allow applications to determine, respectively, whether communicator has - been shutdown or object adapter has been deactivated. +- Added Communicator::isShutdown() and ObjectAdapter::isDeactivated() + to allow applications to determine whether a communicator has been + shutdown or an object adapter has been deactivated, respectively. - The Glacier2 SessionControl object is now registered with the Server adapter rather than the Admin adapter. @@ -35,16 +37,17 @@ Changes since version 3.1.1 Glacier2.Filter.Category.AcceptUser. The old property name has been deprecated and will be removed in a future release. -- Added destroy method to the ObjectAdapter interface. It is necessary - to call destroy to clean up the adapter resources and be able to - recreate a new adapter with the same name. Communicator destroy - automatically calls destroy on all its adapters. +- Added destroy() operation to the ObjectAdapter interface. This + operation must be called in order to reclaim the adapter's resources + and be able to recreate a new adapter with the same name. Destroying + a communicator automatically calls destroy on all of its adapters. - Added a per-process logger and two methods to get/set the logger, - Ice::getProcessLogger() and Ice::setProcessLogger(). Get will create a - default process logger when called if set has not been called. The - process logger will also be used as the default logger for all - communicators that do not have a logger explicitly set. + Ice::getProcessLogger() and Ice::setProcessLogger(). Calling + getProcessLogger() creates a default logger if setProcessLogger() + has not been called. The process logger is also used as the default + logger for all communicators that do not have a logger explicitly + set. - Added the property Glacier2.ReturnClientProxy to control whether Glacier2's implementation of the Router::getClientProxy() operation @@ -52,36 +55,36 @@ Changes since version 3.1.1 For backward compatibility with clients that use Ice versions older than 3.2, this property must be set to 1. -- Added support for a user callback when SIGINT like signals are - sent to applications that use the Application interface. See - Application::interruptCallback and Application::callbackOnInterrupt - for details. +- Added support for a user-defined signal handler in Ice::Application. + A subclass must override the interruptCallback method and invoke + callbackOnInterrupt to activate it. - Message batches are now automatically flushed when the total size of - the batch reaches the message size maximum. This feature can be controlled - using the Ice.BatchAutoFlush property, which is on by default. + the batch reaches the message size maximum. This feature can be + controlled using the Ice.BatchAutoFlush property, which is on by + default. -- It's now possible to specify logs file for IceGrid servers - with the `log' XML element of a server descriptor. Server logs can - be viewed remotely with the IceGrid GUI or the icegridadmin command - line tool. +- It is now possible to identify log files for IceGrid servers using + the `log' XML element of a server descriptor. Server logs can be + viewed remotely by the IceGrid GUI or the icegridadmin command- + line utility. -- Added support for viewing the node, registry and server standard - error and output files from the IceGrid GUI and the icegridadmin - command line tool. +- Added support to the IceGrid GUI and icegridadmin command-line + utility for viewing the standard error and standard output files + of an IceGrid node, registry, and server. - Added new `property' attribute to IceGrid object descriptors. If - this attribute is specified, the registry will generate a property - named with the value of the attribute, for example: + this attribute is specified, the registry generates a property + that supplies the object's identity. For example: <object id="hello" property="Identity"/> - will generate the property "Identity=hello". + generates the property "Identity=hello". - Patching a distribution with IceGrid should no longer result in an Ice::TimeoutException exception if the patching takes a long time. -- Fixed IceGrid to no longer throw Ice::NoEndpointException to a +- Fixed IceGrid to no longer return Ice::NoEndpointException to a client if the object adapter of a server managed by an IceGrid node is being deactivated. IceGrid will now queue the locate request from the client, reactivate the server and send the response of the @@ -96,11 +99,11 @@ Changes since version 3.1.1 at least one such adapter is deactivated. - Added findAllReplicas() method to the IceGrid::Query interface. This - operation allows to retrieve all the replicas associated to a given - proxy. + operation allows you to retrieve all the replicas associated with a + given proxy. - Added support for service property sets in server instances. These - property sets allow to specify additional properties for IceBox + property sets allow you to specify additional properties for IceBox services hosted by IceBox server template instances. For example: <server-instance template="MyIceBoxTemplate"> @@ -109,36 +112,37 @@ Changes since version 3.1.1 </properties> </server-instance> -- Added new IceGrid `Ordered' load balancing policy. The endpoint of a - replica group configured with this load balancing will be ordered by - priority of each replica. The priority can be set with the new - object adapter `priority' attribute. +- Added new IceGrid `Ordered' load balancing policy. The endpoints of + a replica group configured with this load balancing policy are + ordered by the priority of each replica. A replica's priority is set + with the new object adapter `priority' attribute. -- Removed the IceGrid `None' load balancing policy. This load - balancing is equivalent to the `Random' load balancing configured to - return the endpoints of all the replicas (i.e.: with the load - balancing `n-replicas' attribute set to 0). +- Removed the `None' load balancing policy from IceGrid. This load + balancing policy is equivalent to the `Random' policy configured to + return the endpoints of all the replicas (i.e., with the + `n-replicas' attribute set to 0). - Setting the n-replicas attribute of an IceGrid replica group load - balancing policy to 0 is no longer equivalent to 1. It now specifies - to return the endpoints of all the replicas which are members of the - replica group. + balancing policy to 0 is no longer equivalent to 1. It now causes + the locator to return the endpoints of all the replicas that are + members of the replica group. - The IceGrid registry now supports replication in a master/slave scheme. See the Ice manual for information on the configuration of replicated IceGrid registries. -- The communcator object adapter now throw an InitializationException - if the adapter has no configuration. It is possible to explicitly - create a adapter with no configuration, useful for use with - bi-directional connections, by calling createObjectAdapter(""). +- During the creation of an object adapter, the communicator now + raises InitializationException if the adapter has no configuration. + It is possible to create an adapter with no configuration, which + is useful for bidirectional connections, by passing an empty string + as the argument to createObjectAdapter. - It is now possible to start and stop individual IceBox services using the IceBox admin utility. - Added Communicator::propertyToProxy() which creates a proxy from - a set of properties. This function allows you to set various local - proxy settings, such as the Locator cache timeout, which cannot be + a set of properties. This operation allows you to set various local + proxy settings, such as the locator cache timeout, which cannot be set using stringToProxy(). - Added ["cpp:class"] C++ metadata to map a slice struct to a @@ -150,54 +154,53 @@ Changes since version 3.1.1 - The Ice.Logger.Timestamp property has been removed. Timestamps are now always printed. -- Added new property Ice.Default.PreferSecure, which if set to 1 - causes secure connections to be prefered over non-secure connections. - By default this property is set to 0. Also added functions to control - this behavior on a per-proxy rather global basis, ice_isPreferSecure() - and ice_preferSecure(). +- Added the property Ice.Default.PreferSecure, which if set to 1 + causes secure endpoints to be preferred over non-secure endpoints. + By default this property is set to 0. Also added functions to + control this behavior on a per-proxy rather than global basis: + ice_isPreferSecure() and ice_preferSecure(). -- Added a demo to illustrate use of AMI/AMD. +- Added a demo to illustrate the use of AMI/AMD. - IceBox services are now stopped in the reverse order from which they were started. -- If client endpoint configuration does not contain a hostname setting - (-h paramater), the client will try to connect using only the loopback - interface (127.0.0.1). +- If a proxy does not specify a hostname (via -h HOST), the client + only attempts to connect to the loopback interface (127.0.0.1). - All Ice exceptions now derive from std::exception. -- Attempts to change the properties of a fixed proxy now result in a +- Attempts to change the attributes of a fixed proxy now result in a FixedProxyException rather than just ignoring the change request and - returning the same proxy. Also attempts to marshal a fixed proxy now - result in a FixedProxyException rather than a MarshalException. + returning the same proxy. Also, attempts to marshal a fixed proxy + now result in a FixedProxyException rather than a MarshalException. -- Fixed a bug in the IceGrid locator implementation which could cause +- Fixed a bug in the IceGrid locator implementation that could cause a memory corruption when using replica groups. -- New feature: implicit context - When you invoke an operation on a proxy and don't pass an explicit - context, Ice uses the "implicit context" combined with the context - associated with the proxy (if there is one) - You can retrieve and set this ImplicitContext using - Communicator::getImplicitContext(). Three ImplicitContext implementations - are available. You select an implementation by setting the - Ice.ImplicitContext property to one of the following values: - * None: (the default) No implicit context at all - * PerThread: The implementation maintains a Context per thread - * Shared: The implementation maintains a single Context shared - by all threads, and serializes access to this Context. - -- Removed defaultContext from InitializationData - -- Communicator::setDefaultContext(), Communicator::getDefaultContext() +- Added support for implicit request contexts. When you invoke an + operation on a proxy and don't pass an explicit context, Ice uses + the "implicit context" combined with the per-proxy context (if any). + You can obtain the implicit context using the new communicator + operation getImplicitContext(). Three implementations of implicit + context are available; you select one by setting the property + Ice.ImplicitContext to one of the following values: + + * None: No implicit context at all (default). + * PerThread: The implementation maintains a context per thread. + * Shared: The implementation maintains a single thread-safe context + that is shared by all threads. + +- Removed defaultContext from InitializationData. + +- Communicator::setDefaultContext(), Communicator::getDefaultContext() and Proxy::ice_defaultContext() have been deprecated. - Ice now uses the epoll() system call on Linux and the poll() system call on other Unix platforms instead of select(). -- Setting the IceGrid.Node.Name property is now required, the hostname - isn't the default value anymore. +- Setting the IceGrid.Node.Name property is now required; this + property no longer uses the hostname as its default value. - For object adapters without endpoints, Ice.Trace.Network >= 2 now prints a trace message. @@ -208,79 +211,83 @@ Changes since version 3.1.1 - IceStorm changes: Changed IceStorm federation such that IceStorm now detects when - a downstream topic is unavailable. Once a link is diagnosed as + a downstream topic is unavailable. Once a link is diagnosed as non-functional, events are thrown out until a timer expires (specified using the property IceStorm.Discard.Interval in seconds, - default value is 1 minute). At that point, delivery of the link + default value is 1 minute). At that point, delivery of the link is attempted once again. - Added a new property IceStorm.Send.Timeout. This specifies in - milliseconds a timeout value is that is set on all link and - subscriber proxies. The default value is 1 minute. + Added a new property IceStorm.Send.Timeout. This specifies a timeout + in milliseconds that is used for all link and subscriber proxies. + The default value is 1 minute. All message sending is now faster. In particular, oneway QoS is now significantly faster. All message propagation is now fully buffered through a dynamic - thread pool. Tracing for which can be enabled through the property - IceStorm.Trace.SubscriberPool. The thread pool size is controlled - through the following properties. - - * IceStorm.SubscriberPool.Size. The initial size of the thread - pool, the default being 1. - * IceStorm.SubscriberPool.SizeMax. The maximum size of the thread - pool, the default being 0, which means unlimited. - * IceStorm.SubscriberPool.SizeWarn. If the size of the thread pool - exceeds this value a warning is issued. The default is 0, which + thread pool. Tracing for the thread pool can be enabled using the + property IceStorm.Trace.SubscriberPool. The thread pool size is + controlled by the following properties. + + * IceStorm.SubscriberPool.Size: The initial size of the thread + pool. The default value is 1. + * IceStorm.SubscriberPool.SizeMax: The maximum size of the thread + pool. The default value is 0, which means unlimited. + * IceStorm.SubscriberPool.SizeWarn: A warning is issued if the size + of the thread pool exceeds this value. The default is 0, which means never warn. - * IceStorm.SubscriberPool.Timeout. Default value is 1s. If a - stall is detected for more than this timeout value (in - milliseconds) a new thread is spawned up to the maximum size. - - It is now possible to call unlink with a topic proxy which refers - to an IceStorm topic that is now currently reachable. - - Added per-subscriber publisher objects to IceStorm. - IceStorm::Topic::subscribeAndGetPublisher now returns a proxy which - is used to send an event to only that particular subscriber. See - demo/IceStorm/counter for a demo of this feature. - - The possible values of the reliability QoS passed to - subscribeAndGetPublisher are now different. The only allowable value - is either "ordered" or the empty string. The delivery reliability - quality of service used by IceStorm is now derived from the the - proxy mode. For example, if oneway delivery is required then a - oneway proxy must be passed. The "ordered" QoS requires a twoway - proxy otherwise a BadQoS exception is raised. Topic::subscribe has - been deprecated and retains the old subscription QoS semantics. - - The property IceStorm.InstanceName is now used to produce unique identities - for each IceStorm topic. The identities used are now: + * IceStorm.SubscriberPool.Timeout: Specifies a timeout in + milliseconds. IceStorm spawns a new thread, up to the maximum + size of the thread pool, if a stall is detected for longer than + this timeout value. The default value is one second. + + It is now possible to call unlink with a topic proxy that refers + to an IceStorm topic that is not currently reachable. + + Added per-subscriber publisher objects to IceStorm. The new + operation Topic::subscribeAndGetPublisher returns a proxy that is + used to send an event to only that particular subscriber. See + demo/IceStorm/counter for an example of this feature. + + The legal values for the reliability QoS passed to the operation + subscribeAndGetPublisher() are different than that of subscribe(). + The new operation accepts only "ordered" or an empty string and + derives the desired QoS from the mode of the subscriber's proxy. + For example, passing a subscriber proxy configured for oneway + invocations enables oneway delivery in IceStorm. The "ordered" QoS + requires a twoway proxy otherwise a BadQoS exception is raised. + Topic::subscribe has been deprecated and retains the old QoS + semantics. + + The property IceStorm.InstanceName is now used to produce unique + identities for each IceStorm topic. The identities used are now: <instance-name>/topic.<topic name> <instance-name>/publisher.<topic name> <instance-name>/link.<topic name> - IceStorm is still fully compatible with pre 3.2 IceStorm. + IceStorm is still fully compatible with prior releases. - Due to identity change the IceStorm database format was changed. The - script config/upgradeicestorm.py can be used to upgrade old IceStorm - databases to the new format. + The IceStorm database format was changed to accommodate the new + identity semantics. The script config/upgradeicestorm.py can be + used to upgrade existing IceStorm databases to the new format. -- icestorm admin changes +- Changes in icestormadmin: - Removed support for the "graph" command from icestormadmin. + Removed support for the "graph" command. The various icestormadmin commands take the topic manager as an - optional part of the arguments. For example, create MyTopicManager/foo - creates the topic "foo" on the topic manager using the InstanceName - MyTopicManager. See the Ice manual for more information. - - These proxies are configured using the property prefix - IceStormAdmin.TopicManager.<any>. IceStormAdmin.TopicManager.Default - is the default proxy (used by the commands if no topic provided). - IceStorm.TopicManager.Proxy also sets the default topic manager - for backwards compatibility reasons. + optional part of the arguments. For example, the command + "create MyTopicManager/foo" creates the topic "foo" on the topic + manager using the instance name MyTopicManager. See the Ice manual + for more information. + + The program obtains the proxy for a topic manager using a property + of the form IceStormAdmin.TopicManager.NAME. If an icestormadmin + command does not specify a topic manager, the program uses the + proxy in the property IceStormAdmin.TopicManager.Default. For + backward compatibility, the property IceStorm.TopicManager.Proxy + also sets the default topic manager. Added a new command "current" to set the default topic manager. @@ -290,12 +297,9 @@ Changes since version 3.1.1 - Added Ice.Override.Secure which forces a client to only use secure connections. -- If a proxy contains no -h parameter, an Ice client will now attempt - to connect using loopback as well as all other local interfaces. - -- IceGrid registry now cleans up any created admin and regular client - sessions that do not correctly time out. This could occur if the - Glacier2 router terminates unexpectedly. +- IceGrid registry now cleans up any existing administrative and + regular client sessions that do not correctly time out. This could + occur if the Glacier2 router terminates unexpectedly. - Added Glacier2::SessionControl::getSessionTimeout. @@ -310,9 +314,9 @@ Changes since version 3.1.1 - The IceGrid registry Admin endpoints have been removed. A new set of endpoints has been added for the Glacier2 session manager. These - should only be enabled through the - IceGrid.Registry.SessionManager.Endpoints property if Glacier2 - integration is desired since it presents a security threat. + should only be enabled if Glacier2 integration is desired since it + presents a security threat. The endpoints are defined by the + property IceGrid.Registry.SessionManager.Endpoints. - The ice_name() member function of exceptions derived from Ice::Exception (and, hence, IceUtil::Exception) now returns a @@ -347,8 +351,8 @@ Changes since version 3.1.1 if(opts.isSet("v")) // show version... - Either test works correctly, regardless of of whether the actual - option passed to the executable was -v or --version. + Either test works correctly, regardless of whether the actual option + passed to the executable was -v or --version. Note that, for this to work, you must add both long and short option in a single call to addOpt(): @@ -381,10 +385,10 @@ Changes since version 3.1.1 - Added the NullSSLPermissionsVerifier to the IceGrid registry. -- Added more tracing to the IceSSL plug-in to help with debugging - the TrustOnly properties. +- Added more tracing to the IceSSL plug-in to aid in using the + TrustOnly properties. -- Added --ip and --dns command line options to the iceca sign script. +- Added --ip and --dns command line options to the iceca sign command. This adds support for subjectAltName in certificates. - Merged the simple CA scripts into a single iceca script. @@ -429,7 +433,7 @@ Changes since version 3.1.1 corresponding servant base class. - ["freeze:read"] or ["freeze:write"] marks the operation as "read" - or "write" for the Freeze evictor. The default is the freeze mode + or "write" for the Freeze evictor. The default is the Freeze mode of the enclosing interface/class. - ["nonmutating"] provides backward compatibility for idempotent @@ -440,7 +444,7 @@ Changes since version 3.1.1 - New Slice metadata for interfaces/classes: - - ["freeze:read"] or ["freeze:write"] defines the default freeze + - ["freeze:read"] or ["freeze:write"] defines the default Freeze mode for operations on the interface/class. The default is read. This metadata has no effect on derived types. |