path: root/TAO/ChangeLogs/ChangeLog-02a

Sun Jan 30 19:59:30 2000  Carlos O'Ryan  <coryan@uci.edu>

        * Changed the basic collection classes that keep both the consumer
          and supplier proxies.  The new scheme uses the strategy pattern:
          the concrete collection (such as a list or RB-tree) is the
          strategy to store/retrieve objects, while different classes
          implement the serialization protocol (such as using
          copy-on-read, or delayed changes).
          This new implementation uses templates, because the code is
          generic for both the consumer and supplier proxies.  It is
          possible that the same implementation can be used for other
          Event Services (such as Notification or CosEvents)

        * orbsvcs/orbsvcs/Event/EC_Proxy_Collection.h:
        * orbsvcs/orbsvcs/Event/EC_Proxy_Collection.i:
        * orbsvcs/orbsvcs/Event/EC_Proxy_Collection.cpp:
          Implement the different serialization protocols for the proxy
          collections.
          In this new implementation there are no iterators exposed to the
          rest of the EC, instead a for_each() method is implemented, each
          component that needs to iterate over a consumer or supplier
          collection implements a different Worker object for the task.

        * orbsvcs/orbsvcs/Event/EC_Concrete_Proxy_Set.h:
        * orbsvcs/orbsvcs/Event/EC_Concrete_Proxy_Set.i:
        * orbsvcs/orbsvcs/Event/EC_Concrete_Proxy_Set.cpp:
          Implement the concrete collections, currently we have two of
          them, one based on RB-trees and another based on an ordered
          list.

        * orbsvcs/orbsvcs/Event/EC_Worker.h:
        * orbsvcs/orbsvcs/Event/EC_Worker.i:
        * orbsvcs/orbsvcs/Event/EC_Worker.cpp:
          Defines the interfaces for the Worker objects.

        * orbsvcs/orbsvcs/Event/EC_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Event_Channel.h:
        * orbsvcs/orbsvcs/Event/EC_Event_Channel.i:
          Modified to create the new collection classes.
          Removed the methods to create consumer and supplier admin locks,
          because they have been superseeded by the collection classes.

        * orbsvcs/orbsvcs/Event/EC_Defaults.h:
          Added support for the new collection scheme, and removed old
          defaults for the previous style.

        * orbsvcs/orbsvcs/Event/EC_Default_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.i:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.cpp:
          Added support for the new proxy collections.
          The factory can create both list-based or RB-tree-based
          collections, decorated with any of the three basic algorithms
          (i.e. simple locking, aka immediate changes, delayed changes or
           copy-on-read)
          The file contains most of the template instantiations, must fix
          that.

        * docs/ec_options.html:
          The options in the default resource factory have changed, to
          handle the new collection types.

        * orbsvcs/orbsvcs/Event/EC_ConsumerAdmin.h:
        * orbsvcs/orbsvcs/Event/EC_ConsumerAdmin.i:
        * orbsvcs/orbsvcs/Event/EC_ConsumerAdmin.cpp:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.i:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.cpp:
        * orbsvcs/orbsvcs/Event/EC_ObserverStrategy.h:
        * orbsvcs/orbsvcs/Event/EC_ObserverStrategy.i:
        * orbsvcs/orbsvcs/Event/EC_ObserverStrategy.cpp:
        * orbsvcs/orbsvcs/Event/EC_SupplierControl.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.h:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.i:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.h:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.i:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.i:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.cpp:
        * orbsvcs/orbsvcs/Event/EC_Trivial_Supplier_Filter.cpp:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.i:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.cpp:
          Changed all the iterations in the Event Service implementation
          to use Workers instead of iterators.

        * orbsvcs/orbsvcs/Event/EC_Command.h:
        * orbsvcs/orbsvcs/Event/EC_Command.i:
        * orbsvcs/orbsvcs/Event/EC_Command.cpp:
          Added new Reconnected command for delayed reconnections.
          Normalized the name of the Shutdown command.

        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.cpp:
          Create the equivalent proxy collections corresponding to the old
          strategies.

        * docs/tutorials/Quoter/Event_Service/ec.conf:
        * orbsvcs/CosEvent_Service/svc.conf:
        * orbsvcs/Event_Service/svc.conf:
        * orbsvcs/examples/RtEC/Schedule/svc.conf:
        * orbsvcs/examples/RtEC/Simple/ec.conf:
        * orbsvcs/tests/EC_Mcast/svc.conf:
        * orbsvcs/tests/EC_Throughput/ec.conf:
        * orbsvcs/tests/EC_Throughput/ec.mt.conf:
        * orbsvcs/tests/Event/Basic/mt.svc.conf:
        * orbsvcs/tests/Event/Basic/observer.conf:
        * orbsvcs/tests/Event/Basic/sched.conf:
        * orbsvcs/tests/Event/Basic/svc.conf:
        * orbsvcs/tests/Event/Performance/ec.mt.conf:
        * orbsvcs/tests/Event/Performance/ec.st.conf:
        * orbsvcs/tests/Event/Performance/ec.rb_tree.conf:
        * orbsvcs/tests/Event/Performance/ec.list.conf:
          All the configuration files were updated to avoid any obsolete
          flags.

        * orbsvcs/orbsvcs/Event/EC_Dispatching_Task.h:
        * orbsvcs/orbsvcs/Event/EC_Dispatching_Task.i:
        * orbsvcs/orbsvcs/Event/EC_Dispatching_Task.cpp:
        * orbsvcs/orbsvcs/Event/EC_MT_Dispatching.cpp:
        * orbsvcs/orbsvcs/Event/EC_Priority_Dispatching.cpp:
          Normalized the name of the Shutdown_Command, the one to shutdown
          a task is now called Shutdown_Task.

        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.i:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set_T.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set_T.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set_T.i:
          Removed, the EC_Proxy_Collection and EC_Concreate_Proxy_Set
          classes have superseeded this implementation.

        * orbsvcs/tests/Event/lib/Consumer.h:
        * orbsvcs/tests/Event/lib/Consumer.cpp:
        * orbsvcs/tests/Event/lib/Supplier.h:
        * orbsvcs/tests/Event/lib/Supplier.cpp:
          Optimized activation and de-activation of consumers and
          suppliers.

Sun Jan 30 18:52:45 2000  Darrell Brunsch <brunsch@uci.edu>

        * orbsvcs/ImplRepo_Service/ImplRepo_i.cpp:
        * orbsvcs/ImplRepo_Service/ImplRepo_i.h:
        * orbsvcs/ImplRepo_Service/Options.cpp:
        * orbsvcs/ImplRepo_Service/Options.h:
          Added a timeout to the startup code so if the startup
          server does not respond within the timeout, then kill
          it and return an exception.  The timeout period can
          be controlled by the -t option.

          Also replaced the ACE_Process code with the ACE_Process_Manager.
          This should automatically clean up the child processes
          when they exit.

        * orbsvcs/ImplRepo_Service/tao_imr_i.cpp:
          Prints out more information when a CannotActivate exception
          occurs.

Sat Jan 29 12:57:53 2000  bala  <bala@cs.wustl.edu>

        * TAO version 1.0.13 released.

Fri Jan 28 20:11:23 2000  Carlos O'Ryan  <coryan@uci.edu>

        * tao/ORB_Core.cpp:
          When the last ORB is detroyed it must reset the first_orb_
          variable in TAO_ORB_Table to 0.  Thanks to Alex for reporting
          this bug and providing a fix.

Fri Jan 28 18:35:15 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/Request.cpp:
          Removed the check on work_pending() in method
          get_repsonse(), used in deferred synchronous requests.
          Recent changes to perform_work() and/or work_pending()
          have made it possible for work_pending() to return 0
          even if the DII request has not completed (received its
          response). This was causing get_response() to return
          TRUE erroneously, and the resulting null message
          block pointer would eventually cause a client crash.
          Now get_response() returns TRUE only if the
          repsonse has been received.

Fri Jan 28 17:27:19 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_sequence/cdr_op_cs.cpp:
          Fix sent in by Alex Arulanthu <Alex.Arulanthu@sylantro.com> that
          plugs a memory leak in generated code when decoding a sequence
          of length 0.  Thanks to Jianfei Xu <jxu@yahoo.com> for reporting
          this.

Fri Jan 28 14:05:32 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * performance-tests/Latency/st_client.cpp:
          Added 'x' to the string passed to get_opts(). It was in
          the switch statement and in the usage message, as it is
          in the other client types in this test suite.

Fri Jan 28 12:51:26 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tests/Leader_Followers: Minor changes to make it work with
          different compilers.

Thu Jan 27 20:25:37 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/Leader_Follower:

          wait_for_client_leader_to_complete: If a client thread (a thread
          making a remote request) is currently running the event loop, an
          event loop thread (a thread running ORB->run or
          ORB->perform_work) must wait for that client thread to complete.

          Additional state: <client_thread_is_leader_> keep track of
          whether a client thread the current leader.
          <server_threads_waiting_> keeps track of whether server threads
          waiting for the client leader to complete.
          <server_threads_condition_> is the condition variable used by
          event loop threads waiting for the client leader to complete.

          TAO_LF_Server_Thread_Helper: Auto-pointer like class similar to
          TAO_LF_Client_Thread_Helper and TAO_LF_Leader_Thread_Helper that
          helps manage the leader follower object.  This class helps with
          grabbing the lock and calling set_server_thread and grabbing the
          lock, calling reset_server_thread, and electing a new leader.

          set_server_thread: This method now calls
          wait_for_client_leader_to_complete if needed.  Also it now has a
          relevant return value for timeouts and errors.

          set_leader_thread and reset_leader_thread: Added the setting and
          resetting <client_thread_is_leader_>.

          elect_new_leader: Election of new leader must first consider
          event loop threads before considering waiting clients.  If
          server threads are waiting, we wake them all up.  Note that it
          is necessary to give precedence to server threads over client
          threads since only one client runs at any given time; with a tp
          reactor, multiple server threads can execute simultaneously.

        * tao/ORB_Core.cpp (run): Use of TAO_LF_Server_Thread_Helper
          helped with abstracting and simplifying some of the leader
          follower code.  In additions, timeouts are determined by
          examining the <timeout> variable.  Therefore, we don't need
          <break_on_timeouts> anymore.

        * tao/ORB (run and perform_work): The new ORB_Core::run simplified
          ORB::run and ORB::perform_work.

        * tao/Wait_Strategy.cpp (TAO_Wait_On_Reactor::wait,
          TAO_Exclusive_Wait_On_Leader_Follower::wait, and
          TAO_Muxed_Wait_On_Leader_Follower::wait): The while loop in
          TAO_Wait_On_Reactor::wait and
          TAO_Exclusive_Wait_On_Leader_Follower::wait was incorrect.

          This:

          *max_wait_time != ACE_Time_Value::zero

          should have been:

          *max_wait_time == ACE_Time_Value::zero

          While fixing this, I simplified the while loop making it easier
          to read and understand.

          Also fixed TAO_Muxed_Wait_On_Leader_Follower::wait which did not
          handle timeouts correctly.

        * tests/Leader_Followers: This is a test for server applications
          that have client threads (threads making remote calls) starting
          before server threads (threads running the event loop).

          The event loop threads should wait for the client threads that
          become leaders to give up leadership before continuing.

        * version_tests.dsw:
        * tests/Makefile (DIRS):
        * run_all_list.pm:

          Added new test Leader_Followers.

Thu Jan 27 08:36:17 2000  Darrell Brunsch <brunsch@uci.edu>

        * orbsvcs/ImplRepo_Service/tao_imr_i.cpp: The help messages
          were still displaying "tao_ir" instead of "tao_imr".  Also,
          if an Implementation Repository wasn't running, tao_imr
          wouldn't get a chance to parse the command line arguments -
          so it wouldn't even display the help messages unless an
          ImplRepo was running.  Thanks to Dave Zumbro
          <zumbro_d@ociweb.com> for reporting these.  [Bugs 412, 413]

Wed Jan 26 16:17:24 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_scope.cpp:
        * TAO_IDL/be/be_visitor_interface_fwd/interface_fwd_ch.cpp:
        * TAO_IDL/be/be_visitor_interface_fwd/interface_fwd_ci.cpp:
        * TAO_IDL/be/be_visitor_interface_fwd/cdr_op_ch.cpp:
          A fix for what I broke yesterday. At least we should now
          again be able to compile files where a forward declared
          interface is defined later in the file.

        * TAO_IDL/be/be_visitor_interface/cdr_op_ch.cpp:
          Cosmetic changes.

        * tests/IDL_Test/interface.idl:
          Commented out the forward declared interface that's not
          defined in the same file for now, until it is working on
          more platforms.

Tue Jan 25 19:42:54 2000  Balachandran Natarajan  <bala@cs.wustl.edu>

        * tao/UIOP_Profile.cpp (decode):
        * tao/IIOP_Profile.cpp : A problem when TAO clients were
          interacting with GIOP 1.2 servers. The profile received by the
          TAO client is parsed in decode (). In the code for IIOP_Profile,
          a check was made for a higher minor version but no action was
          taken. In the process, the version information that the
          IIOP_Profile class was holding had the minor version information
          of the received profile and  not what the TAO client can
          send. So, changes have been made to check for the minor version
          number. If the minor version is less than or equal to the minor
          version that TAO supports, we save that information. If it is
          greater than the one we support, the {IIOP, UIOP}_Profile
          classes would highest minor version that we support. This
          information gets passed on through the invocation classes to the
          transport and then to the messaging classes. Thanks to Klemen
          Zagar <klemen.zagar@irj.si> for reporting this problem.

Tue Jan 25 19:27:24 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * docs/releasenotes/index.html:
          Added entry for forward declared interface change to
          IDL compiler.

Tue Jan 25 19:01:33 2000  Carlos O'Ryan  <coryan@cs.wustl.edu>

        * orbsvcs/orbsvcs/Event/EC_Event_Channel.cpp:
          Fixed shutdown problem for the Timer Module.

        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.cpp:
          Fixed memory leak, if the supplier disconnects while some
          consumers are still attached it has to shutdown() its filter to
          clear the extra reference counts.

        * orbsvcs/tests/Event/Basic/Atomic_Reconnect.cpp:
        * orbsvcs/tests/Event/Basic/MT_Disconnect.cpp:
        * orbsvcs/tests/Event/Basic/Negation.cpp:
        * orbsvcs/tests/Event/Basic/Observer.cpp:
        * orbsvcs/tests/Event/lib/Driver.cpp:
          Fixed several memory leaks in the tests, to verify that the
          service is ok.

        * orbsvcs/tests/Event/Basic/run_test.pl:
          Increased allowed execution time for one of the tests.

        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.cpp:
          Cosmetic fixes

Tue Jan 25 18:51:42 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tests/IDL_Test/interface.idl:
        * tests/IDL_Test/generic_object.idl:
          Added tests for new forward declared interface rule
          mentioned below and for generic objects used as
          parameters in a request.

Tue Jan 25 18:19:49 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/tao_idl.cpp:
        * TAO_IDL/ast/ast_interface_fwd.cpp:
        * TAO_IDL/be/be_visitor_scope.cpp:
        * TAO_IDL/be/be_visitor_factory.cpp:
        * TAO_IDL/be/be_visitor_root/root.cpp:
        * TAO_IDL/be/be_visitor_interface_fwd/interface_fwd_ch.cpp:
        * TAO_IDL/be/be_visitor_interface_fwd.cpp:
        * TAO_IDL/be_include/be_visitor_interface_fwd.h:
          Modified files

        * TAO_IDL/be_include/be_visitor_interface_fwd/cdr_op_ch.h:
        * TAO_IDL/be/be_visitor_interface_fwd/cdr_op_ch.cpp:
          New files.

          Changes to support code generation and C++ compilation
          for IDL files containing a forward declared interface
          that is not defined in the same file. The spec had
          required that the definition be present, but the rule
          was recently relaxed in CORBA 2.3.1 (99-10-07). If
          the full definition is not in some generated .cpp file
          that is included in the build, the C++ compiler will
          produce a link error. This closes Bugzilla entry #401.

Tue Jan 25 14:23:04 2000  Ossama Othman  <ossama@uci.edu>

        * docs/pluggable_protocols/index.html:

          Updated/corrected some URLs.

Tue Jan 25 14:05:42 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/Leader_Follower.h (TAO_LF_Client_Thread_Helper and
          TAO_LF_Leader_Thread_Helper): Auto-pointer like classes that
          help manage the leader follower object.
          TAO_LF_Client_Thread_Helper helps with set_client_thread and
          reset_client_thread while TAO_LF_Leader_Thread_Helper helps with
          set_leader_thread and reset_leader_thread.

        * tao/Wait_Strategy.cpp (TAO_Muxed_Wait_On_Leader_Follower::wait
          and TAO_Exclusive_Wait_On_Leader_Follower::wait):

          We now use guards to make sure that the leader follower object
          has consistent information even when errors occur.  Previously,
          there were many return and exceptional conditions where the
          leader follower object was left in an inconsistent state.

Tue Jan 25 12:17:35 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_operation/rettype_vardecl_cs.cpp:
          Missed some cases in the previous checkin of this file.

Tue Jan 25 10:38:31 2000  Pradeep Gore  <pradeep@flamenco.cs.wustl.edu>

        * orbsvcs/orbsvcs/Notify/Notify_SupplierAdmin_i.cpp:
        * orbsvcs/orbsvcs/Notify/Notify_ConsumerAdmin_i.cpp:
          Fixed "unreachable statement" warnings on kai.

Tue Jan 25 02:07:58 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * examples/Simple/time/README: Added description on getting
          JACE.Misc.GetOpt.

        * examples/Simple/time/Time_Client.java: Modified the Java client
          so that it, like C++ client, can use command line args to
          determine ways of acquiring Time objref.  This test now uses
          JACE.Misc.GetOpt (a complete JACE installation is not required.)

Mon Jan 24 18:47:22 2000  Ossama Othman  <ossama@uci.edu>

        * docs/Options.html:

          Clarified what UID and GID in the documentation for the
          -ORBSetUID and -ORBSetGID options.

Mon Jan 24 13:17:31 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_operation/rettype_vardecl_cs.cpp:
          Russell L. Carter <rcarter@pinyon.org> reported a bug when
          the return type of an operation is CORBA::Object. It turns
          out that CORBA::TypeCode, interface, forward declared interface,
          valuetype, and forward declared valuetype needed the same fix.

Sun Jan 23 18:15:32 2000  Krishnakumar Elakkara Pathayapura  <krish@polka.cs.wustl.edu>

        * docs/releasenotes/index.html:
          Updated the release notes for the Logging Service

Sun Jan 23 14:09:30 2000  Ossama Othman  <ossama@uci.edu>

        * docs/Options.html:

          Added descriptions of the newly added "-ORBSetUID" and
          "-ORBSetGID" ORB options.

        * tao/ORB_Core.cpp (init):

          Added two new ORB options "-ORBSetUID" and "-ORBSetGID."  These
          options can be used to the set the effective user and group IDs,
          respectively.  This is useful when starting the ORB under the
          super-user but it is desired to run the ORB under different user
          and/or group IDs.  A typical use would be to start a TAO daemon,
          under the 'nobody' user via System V style initialization
          scripts.

          Note that these options are only supported if the platform
          implements the setuid() and setgid() POSIX system calls.

          In general, it is a good idea to only run the ORB with
          super-user privileges when necessary, e.g. when performing
          benchmarking experiments, and when access to the real-time
          scheduling class is needed.  If the ORB must be started via the
          super-user account, and the real-time scheduling class won't be
          used, than the ORB's effective user ID should at least be
          changed to an unprivileged one, such as 'nobody'.

Sat Jan 22 21:08:07 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * orbsvcs/orbsvcs/orbsvcs.dsp:
        * orbsvcs/orbsvcs/orbsvcs_static.dsp: Renamed
          EC_SupplierFiltering.* to EC_Supplier_Filter.*.

Sat Jan 22 18:25:51 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.cpp:
          Fixed typo in the inline include file, only a problem in builds
          that disable inlining.

Sat Jan 22 17:04:14 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_ConsumerAdmin.h:
        * orbsvcs/orbsvcs/Event/EC_ConsumerAdmin.cpp:
        * orbsvcs/orbsvcs/Event/EC_Event_Channel.h:
        * orbsvcs/orbsvcs/Event/EC_Event_Channel.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.cpp:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.cpp:
        * orbsvcs/orbsvcs/Event/EC_Trivial_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Trivial_Supplier_Filter.cpp:
          Added support for atomic reconnections.  Before this change the
          event channel could loose messages when a consumer reconnected,
          because the implementation basically disconnected the consumer
          and then connected it again.  There was an interval where the
          suppliers could push events, while the consumer was temporarily
          disconnected.
          The new implementation pushes a reconnected() message through
          the event channel, the consumer and supplier admins and finally
          to each proxy.
          There were no changes on the critical path, or regular
          connect/disconnect calls, so this only affect the performance of
          reconnections.
          This fixes bug #271

        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.cpp:
          Another attempt at fixing this code to compile without CORBA
          Messaging support. It happens that ACE_HAS_CORBA_MESSAGING is
          always defined, but sometimes to 0 and other times to 1.

        * orbsvcs/tests/Event/Event.dsw:
        * orbsvcs/tests/Event/Basic/Atomic_Reconnect.cpp:
        * orbsvcs/tests/Event/Basic/Atomic_Reconnect.dsp:
        * orbsvcs/tests/Event/Basic/Atomic_Reconnect.h:
        * orbsvcs/tests/Event/Basic/Basic.dsw:
        * orbsvcs/tests/Event/Basic/Makefile:
        * orbsvcs/tests/Event/Basic/run_test.pl:
          New test to check that Atomic reconnects actually work.

        * orbsvcs/tests/Event/Basic/mt_disconnect.conf:
        * orbsvcs/tests/Event/Basic/mt.svc.conf:
          Renamed because it is also used in the Atomic_Reconnect test.

Sat Jan 22 17:11:30 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_codegen.cpp:
          Brian Harris <harris_b@ociweb.com> reported that when using
          the -GI option with the IDL compiler to generate implementation
          files for an IDL file 'foo', an included IDL file 'bar' caused
          the generation of #include "barI.h" in the implementation header
          file, when it should be #include "barC.h". However, since barC.h
          is indirectly included by the line #include "fooS.h", it does
          not need to be there at all. The block that generates this line
          in be_codegen.cpp is commented out for now, to see if some use
          case arises where it is needed. This closes out bug 410.

Fri Jan 21 21:19:26 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Makefile:
          Updated dependencies again, i forget LifeCycle last time.

Fri Jan 21 16:46:36 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event_Utilities.cpp:
        * orbsvcs/orbsvcs/Event_Utilities.h:
        * orbsvcs/orbsvcs/Event_Utilities.i:
          The QoS factories now take an (optional) initializer function
          from the user to help with customized Event Channels that
          transmit a union as part of the EventPayload.

        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.cpp:
          Fixed some very obscure concurrency problems while disconnecting
          suppliers: if a supplier and a consumer where disconnecting at
          the same time there was a very small window for a crash.  The
          race condition has been eliminated.

        * orbsvcs/tests/Event/Event.dsw:
        * orbsvcs/tests/Event/Basic/Basic.dsw:
        * orbsvcs/tests/Event/Basic/MT_Disconnect.cpp:
        * orbsvcs/tests/Event/Basic/MT_Disconnect.dsp:
        * orbsvcs/tests/Event/Basic/MT_Disconnect.h:
        * orbsvcs/tests/Event/Basic/Makefile:
        * orbsvcs/tests/Event/Basic/mt_disconnect.conf:
        * orbsvcs/tests/Event/Basic/run_test.pl:
          New test that performs concurrent supplier and consumer
          connect/disconnect calls.

        * orbsvcs/orbsvcs/Makefile:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.i:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter.cpp:
        * orbsvcs/orbsvcs/Event/EC_SupplierFiltering.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierFiltering.i:
        * orbsvcs/orbsvcs/Event/EC_SupplierFiltering.cpp:
          Renamed the EC_SupplierFiltering class to EC_Supplier_Filter, it
          is more consitent that way (next step: rename EC_Filter to
          EC_Consumer_Filter)

        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Per_Supplier_Filter.cpp:
        * orbsvcs/orbsvcs/Event/EC_Supplier_Filter_Builder.h:
        * orbsvcs/orbsvcs/Event/EC_Trivial_Supplier_Filter.h:
        * orbsvcs/orbsvcs/Event/EC_Trivial_Supplier_Filter.cpp:
          Propagate the EC_SupplierFiltering name change.

        * orbsvcs/tests/Event_Latency/Event_Latency.cpp:
          The ORB initialized was not the default ORB used by the resource
          factory for the event service.

Fri Jan 21 08:39:10 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.cpp:
          More #ifdefs to compile when CORBA messaging is disabled.

Thu Jan 20 09:07:47 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.cpp:
          Minimum CORBA does not provide the _non_existent() method, we
          disable the test in that case.

Thu Jan 20 07:11:16 2000  David L. Levine  <levine@cs.wustl.edu>

        * orbsvcs/orbsvcs/Makefile: updated dependencies, in
          a workspace that had Notify (and all of the other
          orbsvcs) built.

Wed Jan 19 17:51:39 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/ast/ast_check.cpp:
          Added update of AST_Decl's filename pointer before
          outputting a 'forward declaration not defined' error.
          In certain cases, #xxxx statements in the IDL file
          cause the preprocessor to add #<line number> "<filename>"
          statements where <filename> is a repeat of the main
          file name. This would invalidate
          the filename pointer held deep down in a forward
          declared interface node. If a forward declared interface
          is found not to have a full definition, the error
          function that is called tries to use this filename
          pointer, causing a crash. Thanks to
          Hessel Idzenga <idzenga@lucent.com> for sending in
          the example file that uncovered this bug.

Wed Jan 19 15:44:23 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_union/union_ci.cpp:
        * TAO_IDL/be/be_visitor_union/union_cs.cpp:
          Uninlined union destructor. In certain cases
          where the union in nested inside other structs
          and/or unions, Dec Unix picks up on the fact that
          the inner union's destructor is called before
          it is defined inline. Although he didn't report
          this bug, thanks to Hugh Arnold <harnold@itginc.com>
          for sending in the example IDL file that rooted it
          out.

Wed Jan 19 12:07:20 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_union_branch/cdr_op_ci.cpp:
          Fixed demarshaling for unions having members with
          multiple case labels. Thanks to
          Christopher Kohlhoff <chris@kohlhoff.com> for reporting
          the bug.

Tue Jan 18 20:45:36 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tests/Timed_Buffered_Oneways/run_test.pl:
          TAO/tests/Timed_Buffered_Oneways is timing out on our HPUX_aCC
          auto compile.  Added extra timing information to narrow down
          this problem.

Tue Jan 18 20:26:11 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/Transport_Mux_Strategy.cpp
          (TAO_Exclusive_TMS::dispatch_reply): The following change to
          this method was incorrect:

          Thu Jan 06 22:26:42 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

          This method is not the correct place mark the transport as being
          available for reuse.  If we do so here, another thread can start
          using the transport and reset the message state before we have
          had a chance of reading the reply.  However, we still need to
          find a place to safely mark the transport as idle otherwise we
          will not end up reusing network connections.  Note that this was
          not a problem for Oneway, Twoway, and Locate Request invocations
          since their destructors were setting the transport to idle
          anyway.  The problem was for DII and Asynch invocations since
          their destructors were (correctly) not setting the transport to
          idle.

          Solution: First we revived
          <Transport_Mux_Strategy::idle_after_reply> which was there but
          commented out. <TAO_Exclusive_TMS::idle_after_reply> idles the
          transport while <TAO_Muxed_TMS::idle_after_reply> is a no-op.
          Note that this is the opposite of how <idle_after_send> works.

          Second we make <~TAO_Asynch_Reply_Dispatcher> and
          <~TAO_DII_Deferred_Reply_Dispatcher> call
          <Transport_Mux_Strategy::idle_after_reply>.  These reply
          dispatchers are told about their transports when
          <Invocation::invoke_i> has succeeded.

          Third we change Oneway, Twoway, and Locate Request invocations
          to call <idle_after_reply> instead of <idle> in their
          destructors.  This will prevent double idling of the transports
          when used with the TAO_Muxed_TMS strategy.

        * examples/Buffered_AMI/client.cpp (main): Checking
          <message_count == -1> was not a effective trick since the
          <(iterations % message_count) != 0> failed.  Therefore,
          introduced a new variable <setup_buffering> to simplify this
          madness.

Tue Jan 18 18:45:13 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_union.cpp:
          Increased the signed long default discriminator
          value by 1. It was being set to ACE_INT32_MIN,
          which is -2147483648. The literal constant is
          generated in the _default() method, and when
          MSVC sees the line
                this->disc_ = -2147483648;
          it outputs the warning "unary minus operator applied to
          unsigned type, result still unsigned". If the line is
                this->disc_ = -2147483647;
          we get no warning.

Tue Jan 18 17:26:18 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/deep_free.cpp:
          Fixed a memory leak in recursive unions. The fix also
          removes leaks in unions that contain anonymous
          sequences. Thanks to
          Lothar Werzinger <werzinger.lothar@krones.de> for
          reporting this leak and sending in an example IDL
          file.
        * tao/DynStruct_i.cpp:
          Fixed some mismatched forms of ACE_CHECK that I added
          earlier today.

Tue Jan 18 13:04:24 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.cpp:
          Fixed warnings on platforms without native C++ exception
          support.

Tue Jan 18 12:08:00 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/DynStruct_i.cpp:
          Added patches to from_any() and to_any() sent in by
          Philippe Merle <Philippe.Merle@lifl.fr>.These
          patches (de)marshal the repository ID if the Any
          contains an exception.

Tue Jan 18 11:37:31 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tests/MT_Client/client.cpp (parse_args): Only shutdown server
          when specified by the user.

        * tests/MT_Client/run_test.pl: Added explicit shutdown of the
          server.

Tue Jan 18 08:56:20 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.i:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.h:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.i:
          Added a protected modified for the supplier (or consumer in the
          ProxySupplier case), this is useful for users that inherit from
          the ProxySupplier or ProxyConsumer and change the policies used
          to communicate with the peer.  Notice that this changes should
          only be made once the connection is complete, i.e. after
          connect_push_XXX() has completed.  Thanks to Dave Meyer
          <dmeyer@std.saic.com> for suggesting this.

Mon Jan 17 22:13:05 2000  Darrell Brunsch <brunsch@uci.edu>

        * orbsvcs/ImplRepo_Service/tao_imr_i.cpp:
        * orbsvcs/ImplRepo_Service/tao_imr_i.h:
          Added a new command, ior to the tao_imr program.  This
          command creates a simplified IOR by attaching a given
          name with the IMR's endpoint.  This will only work with
          servers that register the simplified IORs with the
          IOR lookup table (Interoperable Naming Service stuff).

        * orbsvcs/tests/ImplRepo/run_test.pl:
          Change the nestea_ir test to use this scheme.

        * docs/implrepo/index.html:
        * docs/implrepo/tao_imr.html
        * docs/implrepo/usersguide.html
          Updated the documentation.

Mon Jan 17 22:53:53 2000  Carlos O'Ryan  <coryan@cs.wustl.edu>

        * orbsvcs/orbsvcs/orbsvcs.dsp:
        * orbsvcs/orbsvcs/orbsvcs_static.dsp:
          Updated the NT project files.

Tue Jan 18 20:44:02 2000  Carlos O'Ryan  <coryan@uci.edu>

        * tao/Object.cpp:
          The _non_existent() method must catch the
          CORBA::OBJECT_NOT_EXIST exception and return the appropriate
          value.

Tue Jan 18 20:39:50 2000  Carlos O'Ryan  <coryan@uci.edu>

        * orbsvcs/orbsvcs/Makefile:
        * orbsvcs/orbsvcs/Event/EC_SupplierControl.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierControl.i:
        * orbsvcs/orbsvcs/Event/EC_SupplierControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_ConsumerControl.h:
        * orbsvcs/orbsvcs/Event/EC_ConsumerControl.i:
        * orbsvcs/orbsvcs/Event/EC_ConsumerControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.h:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.i:
        * orbsvcs/orbsvcs/Event/EC_Reactive_ConsumerControl.cpp:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.h:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.i:
        * orbsvcs/orbsvcs/Event/EC_Reactive_SupplierControl.cpp:
          New components to the event channel that deal with stale
          consumer and/or supplier proxies.
          The first implementations simply poll the consumers and
          suppliers periodically, using the _non_existent() method to
          check if the object is there, to avoid dead-locks they use the
          timeout policies in the ORB to bound the time the spend
          polling.

        * orbsvcs/orbsvcs/Event/EC_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Null_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Basic_Factory.cpp:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.h:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.i:
        * orbsvcs/orbsvcs/Event/EC_Default_Factory.cpp:
          The factory must create the new components now.
          Use CORBA::ORB_init() instead of TAO_ORB_Core_instance() to get
          to the ORB.  This is standard and it can be easily configured
          using the ORBid.

        * docs/ec_options.html:
          Documented the new options in the TAO_EC_Default_Factory.

        * orbsvcs/orbsvcs/Event/EC_Defaults.h:
          New defaults for the consumer and supplier control policies.

        * orbsvcs/orbsvcs/Event/EC_Event_Channel.h:
        * orbsvcs/orbsvcs/Event/EC_Event_Channel.cpp:
          The event channel creates the consumer and supplier control
          policies (using the factory), and then activates them.

        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.h:
        * orbsvcs/orbsvcs/Event/EC_ProxyConsumer.cpp:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.h:
        * orbsvcs/orbsvcs/Event/EC_ProxySupplier.cpp:
          Addded methods to invoke _non_existent() while holding the state
          lock, to avoid race conditions (as in one thread disconnecting
          a consumer while another thread tries to check if the consumer
          still exists).

        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.h:
        * orbsvcs/orbsvcs/Event/EC_SupplierAdmin.i:
          New accessor to use the collection lock.

        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set.h:
          Fixed nasty bug in the definition of the Disconnect command
          object, the typedef was wrong.

        * orbsvcs/orbsvcs/Event/EC_ProxyPushSupplier_Set_T.cpp:
          Removed old debug message.

        * orbsvcs/orbsvcs/Event/EC_Reactive_Timeout_Generator.cpp:
          Removed references to TAO_ORB_Core_instance().

Mon Jan 17 19:54:34 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/GIOP.cpp (write_request_header_lite): Made
          write_request_header_lite similar to write_request_header_std -
          both for checking the <response_flags> correctly and for dealing
          with the SYNC_WITH_SERVER option.  Thanks to Bala for pointing
          this out.

Mon Jan 17 15:49:25 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_field/field_ch.cpp:
          Fix to scoping of field name generation in header
          file when structs and/or unions are nested. Thanks to
          Hugh Arnold <harnold@itginc.com> for reporting this
          bug and sending in the example file.

        * TAO_IDL/be/be_visitor_scope.cpp:
          Cosmetic changes.

        * tests/IDL_Test/union.idl:
          Added example IDL file mentioned above to test suite.

        * tests/IDL_Test/idl_test.dsp:
          Updated project settings to make custom build settings
          identical for all the IDL files in the test suite.

Tue Jan 18 10:54:49 2000  Carlos O'Ryan  <coryan@uci.edu>

        * docs/Options.html:
          Documented the -ORBReactorRegistry option in the resource
          factory.

Sun Jan 16 15:54:39 2000  Ossama Othman  <ossama@uci.edu>

        * tao/default_resource.cpp:

          Moved template instantiations related to ACE_LOCAL_MEMORY_POOL
          to ace/ACE.cpp.  The idea is to place this set of template
          instantiations in a "common" area, since ace/Configuration.cpp
          also needs these templates instantations.

Fri Jan 14 08:00:00 2000  Michael Kircher  <Michael.Kircher@mchp.siemens.de>

        * tests/AMI/README,
          docs/releasenotes/index.html,
          examples/AMI/FL_Callback/README:
          Realized that IDL_HAS_VALUETYPE is defined by default in the TAO
          IDL compiler. Updated the documentation about AMI accordingly.

Fri Jan 14 06:01:00 2000  Michael Kircher  <Michael.Kircher@mchp.siemens.de>

        * TAO_IDL/be/be_interface.cpp,
          TAO_IDL/be_include/be_interface.h:
          Added a method named 'replacement' giving access to an
          interface node, which can serve as an replacement for the original.

        * TAO_IDL/be/be_interface_strategy.cpp,
          TAO_IDL/be/be_visitor_ami_pre_proc.cpp,
          TAO_IDL/be/be_visitor_interface/ami_interface_ch.cpp,
          TAO_IDL/be_include/be_interface_strategy.h:
          Enhanced the AMI preprocessor and the associated interface
          strategies to handle replacement nodes.
          This was necessary to pass information around for AMI code generation.
          The bug triggering this was a multiply defined typedef. Thanks
          to Bala for figuring this out.

Thu Jan 13 20:11:55 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/ORB_Core.cpp (run): Generalized to work with <perform_work>.
          In <perform_work>, we only run the loop once.

        * tao/ORB.cpp (perform_work): The method now calls
          orb_core->run(), specifying that it is <perform_work> which
          prevents looping of the event loop.

        * tao/ORB.h (CORBA_ORB): There are now three versions of
          perform_work() and are similar to the run() methods.

Tue Jan 11 21:27:51 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao/Messaging.pidl (Messaging): Added two TAO specific SYNC
          options that control how buffering takes place in the ORB.

          (a) SYNC_EAGER_BUFFERING is the same as SYNC_NONE.  These
          options will first queue the message in the buffering queue and
          then check the buffering constraints.  If the constraints have
          been reached, the buffered messages will be delivered to the
          server.  If the delivery of the messages to the server times
          out, the queued messages will remain queued until the buffering
          constraints are reached again.

          (b) SYNC_DELAYED_BUFFERING will first check the buffering queue.
          If the queue already has buffered messages, then this option
          will behave the same as the above option, i.e., buffer the
          message for later delivery.  However, if the queue is empty, it
          will try to deliver the message immediately. If the message
          delivery to the server times out, the message will be queued for
          later delivery when the buffering constraints are reached.

        * tao: Added code to handle the new options and strategies related
          to the above change.  The following files were effected:

          - Invocation.cpp
          - GIOP.cpp
          - MessagingC.cpp
          - MessagingC.h
          - ORB_Core.cpp
          - ORB_Core.h
          - ORB_Core.i
          - Stub.cpp
          - Sync_Strategies.cpp
          - Sync_Strategies.h
          - TAOC.cpp
          - TAOC.h

        * tao/GIOP.cpp (write_request_header_std): Comparison with majic
          numbers is evil ;-) Changed it to compare against the constants.

        * tao/Pluggable.h (TAO_Transport): Added friendship between the
          transport class and the sync classes.

        * tests/Timed_Buffered_Oneways/client.cpp
          (setup_buffering_constraints): Added the ability to either
          choose eager or delayed buffering.

Tue Jan 11 17:25:16 2000  bala  <bala@cs.wustl.edu>

        * TAO version 1.0.12 released.

Tue Jan 11 16:46:40 2000  Balachandran Natarajan  <bala@cs.wustl.edu>

        * tao/PollableS.cpp:
        * tao/PollableC.cpp:
        * tao/MessagingS.h: Fixed compile errors in AMI code.

Tue Jan 11 11:20:27 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * performance-tests/RTCorba/Oneways/Reliable/server.cpp (set_rt_mode):
        * performance-tests/RTCorba/Oneways/Reliable/client.cpp (set_rt_mode):

          If errors occur while setting the thread priority, print a
          message but don't abandon the test.

Tue Jan 11 00:27:51 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * orbsvcs/Notify_Service/Notify_Service.dsp
        * orbsvcs/Notify_Service/Notify_Service.dsw
        * orbsvcs/orbsvcs/orbsvcs.dsp:
        * orbsvcs/orbsvcs/orbsvcs_static.dsp:
          Added Notify Service.

        * orbsvcs/orbsvcs/Notify/NotifyPublish_i.h:
        * orbsvcs/orbsvcs/Notify/NotifySubscribe_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_Constraint_Interpreter.h:
        * orbsvcs/orbsvcs/Notify/Notify_Constraint_Visitors.h:
        * orbsvcs/orbsvcs/Notify/Notify_ConsumerAdmin_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_Dispatcher.h:
        * orbsvcs/orbsvcs/Notify/Notify_EventChannelFactory_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_EventChannel_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_FilterAdmin_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_FilterFactory_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_Filter_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_ProxyConsumer_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_ProxyPushConsumer_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_ProxyPushSupplier_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_ProxySupplier_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_QoSAdmin_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_StructuredProxyPushConsumer_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_StructuredProxyPushSupplier_i.h:
        * orbsvcs/orbsvcs/Notify/Notify_StructuredPushConsumer.h:
        * orbsvcs/orbsvcs/Notify/Notify_StructuredPushSupplier.h:
        * orbsvcs/orbsvcs/Notify/Notify_SupplierAdmin_i.h: Added
          TAO_ORBSVCS_Export.  Thanks to Stephane Chatre
          <schatre@oresis.com> for noticing these.

Mon Jan 10 18:12:55 2000  Balachandran Natarajan  <bala@cs.wustl.edu>

        * tao/PollableC.cpp: Fixed two instances of wrong usage of
          ACE_NEW_RETURN.

Mon Jan 10 16:43:09 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_union_branch.cpp:
        * TAO_IDL/util/utl_scope.cpp:
          Lookup and code generation of case labels for unions
          in IDL did not correctly handle cases where the
          label names are scoped (for example, if the enum
          discriminator is defined in a different module than
          the union, the case label names may or may not be
          scoped). Thanks to Hessel Idzenga <idzenga@lucent.com>
          for reporting this bug.

Mon Jan 10 12:53:08 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/GIOP.cpp:
        * tao/GIOP_Server_Request.cpp:
          Since Reliable Oneways depend on a feature of GIOP 1.2,
          and GIOP 1.2 has not yet been fully implemented in TAO,
          a temporary workaround was added so that a TAO server
          will not confuse a twoway request from a GIOP 1.1
          client with a oneway SYNC_WITH_SERVER request. When
          GIOP 1.2 is fully implemented in TAO, the TAO server
          can then check the version in the message header and
          behave accordingly. [bug 397]

Mon Jan 10 11:51:33 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_exception/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_structure/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_sequence/any_op_cs.cpp:
          Turned out to be the same change as below, but these
          were warnings from SunCC.

Mon Jan 10 11:40:20 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_union/any_op_cs.cpp:
          Fix to generated code for >>=. We were getting a
          compile error from HPUX.

Sat Jan  8 12:13:43 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_union_branch/cdr_op_ci.cpp:
          Fixed some uninitialized declarations. Reported by
          g++ David Levine.

Sat Jan  8 03:30:00 2000  Michael Kircher  <Michael.Kircher@mchp.siemens.de>

        * TAO/docs/releasenotes/index.html
          TAO/examples/AMI/FL_Callback/README,
          TAO/examples/Buffered_AMI/README,
          TAO/tests/AMI/README:
          Made the documentation of 'defines' necessary to compile
          AMI consistent across various documentations.
          Currently, you need to set TAO_HAS_CORBA_MESSAGING and
          TAO_HAS_AMI_CALLBACK to activate AMI code in the TAO library.
          (TAO_HAS_VALUETYPE is needed as well, but is defined by default)
          IDL_HAS_VALUETYPE needs to be defined for the TAO IDL compiler,
          this is due to the fact that AMI uses valuetypes for exception
          handling, though this might change in the future, because the
          spec might change in this area. Use -GC on the IDL compiler
          to generate AMI stubs.
          Thanks to "Russell L. Carter" <rcarter@consys.com> for pointing
          this out.

Fri Jan  7 15:22:38 2000  Ossama Othman  <othman@cs.wustl.edu>

        * tao/ORB.cpp (destroy):

          Corrected behavior where ORB::destroy() would throw a
          BAD_INV_ORDER exception if the ORB was already shutdown, which
          does not agree with the CORBA specification.  It is now possible
          to call ORB::destroy() after ORB::shutdown().  Thanks to Mogens
          Hansen <mogens_h@dk-online.dk> for reporting this problem.

Fri Jan  7 11:52:17 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * TAO_IDL/be/be_visitor_union_branch/cdr_op_ci.h:
        * TAO_IDL/be/be_visitor_union_branch/cdr_op_ci.cpp:
          Fixed bug where discriminator in a union with an
          explicit default case was not getting set after
          transmission. Thanks to Hugh Arnold <harnold@itginc.com>
          for reporting this bug.

Thu Jan 06 22:26:42 2000  Irfan Pyarali  <irfan@cs.wustl.edu>

        * tao: Minimum CORBA in TAO meant a minimum POA.  This
          relationship was changed such that minimum CORBA does not have
          to mean a minimum POA.  This will allow the user to use a
          minimum CORBA build of TAO with the full functionality of the
          POA.  The default, however, remains the same, i.e., minimum
          CORBA enables a minimum POA.

          Added a new variable TAS_HAS_MINIMUM_POA to orbconf.h.
          TAS_HAS_MINIMUM_POA support is disabled by default if TAO is not
          configured for minimum CORBA.  If TAO is configured for minimum
          CORBA, then TAS_HAS_MINIMUM_POA will be enabled by default.
          Also, TAO_HAS_MINIMUM_POA_MAPS is now influenced by the value of
          TAO_HAS_MINIMUM_POA rather than that of TAO_HAS_MINIMUM_CORBA if
          not explicitly set by the user.

          The following files were involved in this change:

          - POA.cpp
          - POA.h
          - POA.i
          - POA.pidl
          - POAC.cpp
          - POAC.h
          - POAC.i
          - POAManager.cpp
          - POAManager.h
          - POAManager.i
          - POAS.cpp
          - POAS.h
          - POAS.i
          - Object_Adapter.cpp
          - Object_Adapter.h
          - Object_Adapter.i

          Thanks to Erik Johannes <ejohannes@oresis.com> for motivating
          this change.

        * tests/Timed_Buffered_Oneways: New test for oneways with
          buffering and timing constraints.  The client sends is setup to
          send large requests to the server.  The server is setup to take
          a long time to process these requests.  The combination will
          cause flow control for the client.  The timing constraints on
          the client ORB will prevent the client from blocking because of
          flow control.  The request is queued up for later delivery once
          the flow control subsides.

        * tao/IIOP_Connect.cpp (handle_timeout):
        * tao/UIOP_Connect.cpp (handle_timeout):

          Added code to access the thread or ORB roundtrip timeout
          policies.  This allows us to perform a timed send operation
          instead of a blocking one. We do not consider the object
          roundtrip timeout policy since we do not have access to it.

        * tao/ORB_Core.h (stubless_relative_roundtrip_timeout):  Added a
          new method to access the RoundtripTimeoutPolicy policy set on
          the thread or on the ORB.  In this method, we do not consider
          the stub since we do not have access to it.

        * tao/Sync_Strategies.cpp: Moved the TAO-specific SYNCH_FLUSH
          option from the Messaging::SyncScope policy and changed it to
          BUFFER_FLUSH as one of the buffering constraint modes in the
          TAO-specific BufferingConstraint policy.  Since we now have
          BUFFER_FLUSH, we don't need BUFFER_NONE anymore since they both
          do the same thing.  Also, we didn't need the
          TAO_Flush_Sync_Strategy anymore and the Buffered_Oneway example
          was changed to accommodate this change.

        * tao/Pluggable.cpp

          (send_buffered_messages): The return value of 0 is considered
          EOF.  Therefore, changed the return values in the case of
          timeouts and empty queues to be 1.  Also, fixed how timeouts are
          handled.

          (reset_message): The resetting of the queued messages was not
          correct.  It was deleting excessively. This was fixed.  Also,
          the resetting was decoupled from the queue so that it can be
          used with independent message blocks.

        * tao/Wait_Strategy.cpp:

          TAO_Exclusive_Wait_On_Leader_Follower::handle_input: The check
          for <expecting_response_> was conflicting with the ability to
          buffer asynchronous calls.  If we mark the asynchronous call as
          a twoway call, then buffering cannot take place.  If we mark it
          as a oneway call, then the check for <expecting_response_>
          fails.  For now I have selected to disable the check.  The long
          term fix is to separate out the two concerns (a) can the call be
          buffered and (b) are we expecting a response.

        * tao/Transport_Mux_Strategy.cpp
          (TAO_Exclusive_TMS::dispatch_reply): Once we receive our reply,
          we need to mark the transport as being available for reuse
          again.

        * tao: Renamed TAO_RelativeRoundtripTimeoutPolicy_i to
          TAO_RelativeRoundtripTimeoutPolicy.

        * tao/Messaging_Policy_i.cpp (set_time_value): Factored out common
          code in Invocation.cpp and IIOP_Connect.cpp.

        * tao/MessagingS.cpp: Added an overloaded relative_expiry(), which
          is a hacky TAO extension to reduce a call to
          CORBA::Environment::default_environment () since this method
          will never raise exceptions.

Thu Jan 06 21:26:17 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * tests/Param_Test/param_test_i.cpp
          (test_unbounded_struct_sequence): Changed the explicit cast to
          [(unsigned int) 0] to unsigned literal [0u] to avoid MSVC fro
          whining.

Wed Jan  5 15:54:38 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/Any.cpp:
          Fix to new Any extractor to pointer to const Any,
          prompted by a SunCC 4.2 compile error.

        * TAO_IDL/be/be_visitor_union/any_op_cs.cpp:
          Removed a couple of unnecessary 'ACE_reinterpret_cast's
          in generated code for the new Any extractor to pointer
          to const Any. They were causing compile errors with
          Kai, SunCC and other compilers.

        * performance-tests/Latency/deferred_synch_client.dsp:
          Added missing paths for library linking.

        * performance-tests/RTCorba/Oneways/Reliable/client.dsp:
        * performance-tests/RTCorba/Oneways/Reliable/server.dsp:
          Put in the complete relative path to tao_idl.exe, as
          it is in all other projects with IDL files.

Tue Jan  4 11:46:42 2000  Jeff Parsons  <parsons@cs.wustl.edu>

        * tao/Any.h:
        * tao/Any.cpp:
          Added >>= operators to extract to const CORBA::Any *&,
          as well as to const char *& and const wchar *&, as outlined in
          the CORBA 2.3.1 C++ mapping.

        * TAO_IDL/be/be_visitor_exception/any_op_ch.cpp:
        * TAO_IDL/be/be_visitor_exception/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_sequence/any_op_ch.cpp:
        * TAO_IDL/be/be_visitor_sequence/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_structure/any_op_ch.cpp:
        * TAO_IDL/be/be_visitor_structure/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_union/any_op_ch.cpp:
        * TAO_IDL/be/be_visitor_union/any_op_cs.cpp:
        * TAO_IDL/be/be_visitor_union/any_op_cs.h:
          Added code generation for Any extraction operator to
          pointer to const whenever extraction to a pointer is called
          for. Left the former generated operators (same but without
          the 'const') for backward compatibility, but marked them
          as 'deprecated' in the generated header file. Also modified
          code for both operators to return a null pointer if the
          extraction fails. All this is in accordance with CORBA 2.3.1.

        * TAO_IDL/be/be_visitor_interface/any_op_cs.cpp:
          Modified Any extraction operator to return a null pointer
          if the extraction fails, for the reason given above.

Tue Jan 04 07:30:04 2000  David L. Levine  <levine@cs.wustl.edu>

        * TAO version 1.0.11 released.

Tue Jan  4 00:34:57 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * orbsvcs/tests/Trading/Service_Type_Exporter.cpp:
        * orbsvcs/tests/Simple_Naming/client.cpp:
        * orbsvcs/LifeCycle_Service/Factory_Trader.cpp: Fixes for KAI that
          missed the last round and fixes that were needed after the last
          round.

Mon Jan  3 20:33:11 2000  Balachandran Natarajan  <bala@cs.wustl.edu>

        * orbsvcs/Trading_Service/Trading_Service.cpp:
        * orbsvcs/tests/Simple_Naming/client.cpp:
        * orbsvcs/tests/Trading/Service_Type_Exporter.cpp:
        * orbsvcs/tests/Trading/Offer_Exporter.cpp: Fixes for KAI that
          missed the last round and fixes that were needed after the last
          round.

Mon Jan  3 15:08:47 2000  Balachandran Natarajan  <bala@cs.wustl.edu>

        * orbsvcs/Trading_Service/Trading_Service.cpp:
        * orbsvcs/tests/Trading/Offer_Exporter.cpp:
        * orbsvcs/tests/Trading/Offer_Importer.cpp:
        * orbsvcs/tests/Trading/Service_Type_Exporter.cpp:
        * orbsvcs/tests/Simple_Naming/client.cpp:
        * orbsvcs//LifeCycle_Service/Factory_Trader.cpp:
        * orbsvcs/tests/Property/client.cpp: KAI needed fixes in these
          files to compile clean. There was a need to explicitly use an
          unsigned integer to avoid overloading amibiguity. Essentially
          the same fix as 'Mon Jan 03 10:55:36 2000  David L. Levine'.

Mon Jan  3 14:45:22 2000  Nanbor Wang  <nanbor@cs.wustl.edu>

        * tao/ORB.h:
        * tao/corbafwd.h: Removed compiler-specific alignment adjusting
          preprocessor directives for MSVC and BCB completely.  Thanks to
          Christopher Kohlhoff <chris@kohlhoff.com> for reporting this.

Mon Jan 03 10:55:36 2000  David L. Levine  <levine@cs.wustl.edu>

        * orbsvcs/Time_Service/Clerk_i.cpp (get_first_IOR):
          explicitly use unsigned 0 index for bindings_list
          array element to avoid overloading ambiguity on Irix
          and HP/UX.

Mon Jan 03 10:46:31 2000  David L. Levine  <levine@cs.wustl.edu>

        * orbsvcs/orbsvcs/Trader/Offer_Iterators.cpp (next_n):
          modified look index so that its comparison is > 0, not
          >= 0.  Now that it's unsigned, it was causing a compiler
          warning.

Sun Jan 02 10:48:59 2000  Michael Kircher  <Michael.Kircher@mchp.siemens.de>

        * orbsvcs/orbsvcs/Trader/Offer_Iterators.cpp:
          Fixed a MSVC warning, complaining about an array
          operator.

Sat Jan 01 09:18:59 2000  David L. Levine  <levine@cs.wustl.edu>

        * orbsvcs/orbsvcs/Makefile: updated dependencies, including
          Notification Service files.

Sat Jan 01 09:17:40 2000  David L. Levine  <levine@cs.wustl.edu>

        * ChangeLog: moved to ChangeLog-99c.