diff options
| author | heather <heather@ae88bc3d-4319-0410-8dbf-d08b4c9d3795> | 2003-02-17 15:14:14 +0000 |
|---|---|---|
| committer | heather <heather@ae88bc3d-4319-0410-8dbf-d08b4c9d3795> | 2003-02-17 15:14:14 +0000 |
| commit | a23c86d62742366374a854af733cf48487c4f48e (patch) | |
| tree | 1f85cc89d2ce2e385d784cef6079ed4e7e7c035e | |
| parent | 868d283cf2fc990d4f500f216895c3eb55bcf46b (diff) | |
| download | ATCD-a23c86d62742366374a854af733cf48487c4f48e.tar.gz | |
ChangeLogTag: Mon Feb 17 09:09:54 2003 Heather Drury <drury_h@ociweb.com>
| -rw-r--r-- | TAO/ChangeLog | 1040 | ||||
| -rw-r--r-- | TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp | 2 | ||||
| -rw-r--r-- | TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h | 4 |
3 files changed, 528 insertions, 518 deletions
diff --git a/TAO/ChangeLog b/TAO/ChangeLog index c138f88373f..c488933e0b5 100644 --- a/TAO/ChangeLog +++ b/TAO/ChangeLog @@ -1,82 +1,90 @@ +Mon Feb 17 09:09:54 2003 Heather Drury <drury_h@ociweb.com> + + * orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h: + * orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp: Fixed compile error + that was happening because the base class signature for + establish_components() did not match the derived class method + signature. + Sat Feb 15 10:14:08 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * examples/RTCORBA/Activity/Activity.h: Fixed compile errors in - builds where exceptions have been turned off. + * examples/RTCORBA/Activity/Activity.h: Fixed compile errors in + builds where exceptions have been turned off. Fri Feb 14 21:12:22 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/ast/ast_module.cpp: - Changed a misapplied static_cast to const_cast. + Changed a misapplied static_cast to const_cast. Fri Feb 14 18:27:09 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/ast/ast_decl.cpp: - * TAO_IDL/ast/ast_module.cpp: - * TAO_IDL/include/ast_module.h: - - Fixes for an error in IDL_Test, where recent changes to the - IDL compiler induced an error in typeprefix handling, and - a bug reported by Will Christhof <Will.Christhof@krones.de> - in #pragma prefix handling. + * TAO_IDL/ast/ast_module.cpp: + * TAO_IDL/include/ast_module.h: - * TAO_IDL/util/utl_scope.cpp: + Fixes for an error in IDL_Test, where recent changes to the + IDL compiler induced an error in typeprefix handling, and + a bug reported by Will Christhof <Will.Christhof@krones.de> + in #pragma prefix handling. - Cosmetic fix. + * TAO_IDL/util/utl_scope.cpp: - * tests/IDL_Test/main.cpp: + Cosmetic fix. - Changed error reporting and handling to allow multiple - errors to be reported in one run, and to allow the ORB - and POA to clean up at exit whether there have been errors - or not. + * tests/IDL_Test/main.cpp: + + Changed error reporting and handling to allow multiple + errors to be reported in one run, and to allow the ORB + and POA to clean up at exit whether there have been errors + or not. Fri Feb 14 14:37:55 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/be/be_predefined_type.cpp (compute_repoID): - Fixed function to catch a reference to CORBA::Object - in an IDL file. The repository ID must be handled as - a special case, since this node type is not looked up - in the AST. + Fixed function to catch a reference to CORBA::Object + in an IDL file. The repository ID must be handled as + a special case, since this node type is not looked up + in the AST. - * TAO_IDL/be/be_visitor_typecode/typecode_defn.cpp: + * TAO_IDL/be/be_visitor_typecode/typecode_defn.cpp: - Fixed incorrect typecode generation for members of type - CORBA::Object. Thanks to Ferran Boladeres Salvadó <bds@iis.fhg.de> - for sending in the example code that uncovered the bug. + Fixed incorrect typecode generation for members of type + CORBA::Object. Thanks to Ferran Boladeres Salvadó <bds@iis.fhg.de> + for sending in the example code that uncovered the bug. Fri Feb 14 18:39:01 UTC 2003 Craig Rodrigues <crodrigu@bbn.com> - * tao/RTCORBA/RT_ORBInitializer.cpp: - If you do: - ACE_DEBUG((LM_DEBUG, "%p" )); - to print errno, then that is wrong and will crash. + * tao/RTCORBA/RT_ORBInitializer.cpp: + If you do: + ACE_DEBUG((LM_DEBUG, "%p" )); + to print errno, then that is wrong and will crash. - You need to do: - ACE_DEBUG((LM_DEBUG, "%p", "somestring )); - I've changed things to be consistent: - ACE_DEBUG((LM_DEBUG, "errno: %d %m", errno)); + You need to do: + ACE_DEBUG((LM_DEBUG, "%p", "somestring )); + I've changed things to be consistent: + ACE_DEBUG((LM_DEBUG, "errno: %d %m", errno)); Fri Feb 14 09:12:12 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/RTCORBA/MT_Client_Protocol_Priority/client.conf: New file - with the scheduling policy information for the client. - - * tests/RTCORBA/MT_Client_Protocol_Priority/server.conf: Added the - scheduling policy information for the server. - - * tests/RTCORBA/MT_Client_Protocol_Priority/run_test.pl: Sript now - use the above configuration files when running the test. + * tests/RTCORBA/MT_Client_Protocol_Priority/client.conf: New file + with the scheduling policy information for the client. + + * tests/RTCORBA/MT_Client_Protocol_Priority/server.conf: Added the + scheduling policy information for the server. - * tests/RTCORBA/MT_Client_Protocol_Priority/svc.conf: Removed from - the repository. + * tests/RTCORBA/MT_Client_Protocol_Priority/run_test.pl: Sript now + use the above configuration files when running the test. + + * tests/RTCORBA/MT_Client_Protocol_Priority/svc.conf: Removed from + the repository. Fri Feb 14 08:35:14 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * examples/RTCORBA/Activity/Activity.cpp: Fixed compile errors in - builds where exceptions have turned off. Thanks to Pradeep Gore - for sending the patches. + * examples/RTCORBA/Activity/Activity.cpp: Fixed compile errors in + builds where exceptions have turned off. Thanks to Pradeep Gore + for sending the patches. Fri Feb 14 12:02:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -85,197 +93,197 @@ Fri Feb 14 12:02:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> Thu Feb 13 19:30:55 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/RTCORBA/Persistent_IOR/Makefile: Linked the client with - RTCORBA library. + * tests/RTCORBA/Persistent_IOR/Makefile: Linked the client with + RTCORBA library. Thu Feb 13 15:15:29 2003 Priyanka Gontla <pgontla@ociweb.com> - * tao/MCAST_Parser.cpp (multicast_query): + * tao/MCAST_Parser.cpp (multicast_query): Modified the diagnostic message that Craig added with the - changelog entry + changelog entry - Wed Feb 12 00:45:04 UTC 2003 Craig Rodrigues <crodrigu@bbn.com> + Wed Feb 12 00:45:04 UTC 2003 Craig Rodrigues <crodrigu@bbn.com> - to reflect the actual service name instead of being NameService - always. Added a service_name check to print out the NameService - related debugging message if it is indeed NameService. + to reflect the actual service name instead of being NameService + always. Added a service_name check to print out the NameService + related debugging message if it is indeed NameService. - * orbsvcs/ImplRepo_Service/tao_imr_i.cpp: - * orbsvcs/ImplRepo_Service/ImR_Activator_i.cpp: - A change which I did a while ago seemed to be missing. Nanbor - applied the relevant fixes for Windows. Removing the ifdefs - around the same code so that even for non-windows, the whole - hostname including the domain is used. + * orbsvcs/ImplRepo_Service/tao_imr_i.cpp: + * orbsvcs/ImplRepo_Service/ImR_Activator_i.cpp: + A change which I did a while ago seemed to be missing. Nanbor + applied the relevant fixes for Windows. Removing the ifdefs + around the same code so that even for non-windows, the whole + hostname including the domain is used. Thu Feb 13 17:45:35 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * utils/catior/Makefile: Do not build if RTCORBA is turned off. + * utils/catior/Makefile: Do not build if RTCORBA is turned off. Thu Feb 13 17:41:37 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/RTCORBA/Profile_And_Endpoint_Selection/svc.conf: Added the - scheduling policy to the configuration file. + * tests/RTCORBA/Profile_And_Endpoint_Selection/svc.conf: Added the + scheduling policy to the configuration file. Thu Feb 13 16:35:24 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/Exposed_Policies/svc.conf: Configuration file for - scheduling policy. + * tests/Exposed_Policies/svc.conf: Configuration file for + scheduling policy. Thu Feb 13 13:32:50 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * utils/catior/catior.dsp: Fixed the debug and release configurations. + * utils/catior/catior.dsp: Fixed the debug and release configurations. Thu Feb 13 14:21:43 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * utils/catior/catior.cpp: Fixed warnings and errors on various - platforms. + * utils/catior/catior.cpp: Fixed warnings and errors on various + platforms. Thu Feb 13 12:43:27 2003 Pradeep Gore <pradeep@oomworks.com> - * examples/RTCORBA/Activity/Periodic_Task.h: - * examples/RTCORBA/Activity/Thread_Task.cpp: - * examples/RTCORBA/Activity/Thread_Task.h: - Added missing code to convert CORBA priority to native priority - when activating the thread task. - - Made the following changes to get the example running in the - autobuilds: - * examples/RTCORBA/Activity/run_test.pl: - This runs a simple test of the activity example. This test is - meant to be run in non-root mode. - * examples/RTCORBA/Activity/Activity.cpp: - * examples/RTCORBA/Activity/Activity.h: - - Added code to obtain the Priority Mapping Manager. - - Added a "-Started_Flag <file_name> " option. When an activity instance - completes its setup, it generates a file to signal the perl - script that it has finished initialization. - * examples/RTCORBA/Activity/README: - Added a new section on the new command-line option. - * examples/RTCORBA/Activity/server.conf: - * examples/RTCORBA/Activity/client.conf: - These conf files are read by run_test.pl to make a sanity check - on the activity program. + * examples/RTCORBA/Activity/Periodic_Task.h: + * examples/RTCORBA/Activity/Thread_Task.cpp: + * examples/RTCORBA/Activity/Thread_Task.h: + Added missing code to convert CORBA priority to native priority + when activating the thread task. + + Made the following changes to get the example running in the + autobuilds: + * examples/RTCORBA/Activity/run_test.pl: + This runs a simple test of the activity example. This test is + meant to be run in non-root mode. + * examples/RTCORBA/Activity/Activity.cpp: + * examples/RTCORBA/Activity/Activity.h: + - Added code to obtain the Priority Mapping Manager. + - Added a "-Started_Flag <file_name> " option. When an activity instance + completes its setup, it generates a file to signal the perl + script that it has finished initialization. + * examples/RTCORBA/Activity/README: + Added a new section on the new command-line option. + * examples/RTCORBA/Activity/server.conf: + * examples/RTCORBA/Activity/client.conf: + These conf files are read by run_test.pl to make a sanity check + on the activity program. Thu Feb 13 10:30:56 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/ast/ast_module.cpp: - Fixed an error in comparing the prefixes of a module and a previous - opening of that module. The error appears only when the previous - module opening is in an included IDL file, and the IDL compiler - is executed from a subdirectory of the one where the IDL files - are located. Thanks to Will Christof <Will.Christof@krones.de> - for reporting the bug and for providing the example IDL files and - build conditions. This closes [BUGID:1442]. + Fixed an error in comparing the prefixes of a module and a previous + opening of that module. The error appears only when the previous + module opening is in an included IDL file, and the IDL compiler + is executed from a subdirectory of the one where the IDL files + are located. Thanks to Will Christof <Will.Christof@krones.de> + for reporting the bug and for providing the example IDL files and + build conditions. This closes [BUGID:1442]. Wed Feb 12 14:58:46 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/IIOP_Transport.cpp: Methods tear_listen_point_list () and - get_listen_point () accessed the connection_handler without any - protection and violated the protocol that was set in pre-1.2 - timeframe. + * tao/IIOP_Transport.cpp: Methods tear_listen_point_list () and + get_listen_point () accessed the connection_handler without any + protection and violated the protocol that was set in pre-1.2 + timeframe. - * orbsvcs/orbsvcs/SSLIOP/SSLIOP_Transport.cpp: Same changes as - above. This code needs refactoring. + * orbsvcs/orbsvcs/SSLIOP/SSLIOP_Transport.cpp: Same changes as + above. This code needs refactoring. - * tao/Transport.h: Cosmetic fix. + * tao/Transport.h: Cosmetic fix. Wed Feb 12 12:32:57 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/RTCORBA/Client_Protocol/svc.conf: - * tests/RTCORBA/Server_Protocol/server_iiop.conf: - * tests/RTCORBA/Server_Protocol/server_reverse.conf: - * tests/RTCORBA/Server_Protocol/server_reverse_nt.conf: - * tests/RTCORBA/Server_Protocol/server_shmiop.conf: - * tests/RTCORBA/Server_Protocol/server_uiop.conf: Added - scheduling policy information to the above configuration - files. This should fix the problems that shows up in our daily - builds. + * tests/RTCORBA/Client_Protocol/svc.conf: + * tests/RTCORBA/Server_Protocol/server_iiop.conf: + * tests/RTCORBA/Server_Protocol/server_reverse.conf: + * tests/RTCORBA/Server_Protocol/server_reverse_nt.conf: + * tests/RTCORBA/Server_Protocol/server_shmiop.conf: + * tests/RTCORBA/Server_Protocol/server_uiop.conf: Added + scheduling policy information to the above configuration + files. This should fix the problems that shows up in our daily + builds. Wed Feb 12 12:14:13 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * utils/catior/catior.cpp: Fixed a warning in g++ builds. + * utils/catior/catior.cpp: Fixed a warning in g++ builds. Wed Feb 12 00:45:04 UTC 2003 Craig Rodrigues <crodrigu@bbn.com> - * tao/MCAST_Paser.cpp: Add improved diagnostic message when - locating NameService via multicast fails. + * tao/MCAST_Paser.cpp: Add improved diagnostic message when + locating NameService via multicast fails. Wed Feb 12 00:19:58 UTC 2003 Craig Rodrigues <crodrigu@bbn.com> - * tao/RTCORBA/Linear_Priority_Mapping.cpp (to_CORBA): - Unconditionally display error message upon priority mapping failure. - * tao/RTCORBA/RT_Protocol_Hooks.cpp: Display errno value and message - if setting the native thread priority fails. - * tao/RTCORBA/RT_ORBInitiializer.h: - Create a new class member variable sched_policy_flags_. - The value of sched_policy_flags_ is a bitmask value that can be one of - THR_SCHED_RR, THR_SCHED_FIFO, THR_SCHED_DEFAULT, etc. - sched_policy_flags_ value is stored in the - TAO_ORB_Core TAO_ORB_Parameters structure. - In TAO_Thread_Lane::create_dynamic_threads(), the sched_policy_flags_ - is passed to the ACE_Task_Base::activate() call for creating new - threads for a RTCORBA thread pool. - - sched_policy_ is an integer value that can be one of - ACE_SCHED_RR, ACE_SCHED_FIFO, ACE_SCHED_OTHER, etc. This value - can be passed directly to the newly modified ACE_OS::thr_setpriority() - call. For POSIX pthreads, this value maps directly to the - policy parameter of the pthread_setschedparam() call. - ACE_SCHED_* values can also be passed directly to ACE_Sched_Params - for determining the minimum and maximum priority value for a particular - policy. - - * tao/RTCORBA/RT_ORBInitiializer.cpp: - Change all assignments of sched_policy_ variable to use - ACE_SCHED_* values and not THR_SCHED_* values. - In the TAO_RT_ORBInitializer constructor, set sched_policy_flags_ - to the appropriate THR_SCHED_* values depending on what ACE_SCHED_* - value is specified. - - In TAO_RT_ORBInitializer::pre_init(), pass sched_policy_flags_ - to orb_core()->orb_params()->sched_policy(), since this value - will be later used in the activate() call in - TAO_Thread_Lane::create_dynamic_threads(). - - At the end of TAO_RT_ORBInitializer::pre_init(), set the policy - of the ACE_Thread::self() to the policy specified by -ORBSchedPolicy, and - set the priority of ACE_Thread::self() to the lowest priority - of the specified policy. If this is not done, then errors will - occur on Linux, because the default policy in ACE_OS::thr_setprio() - is 0 (which maps to SCHED_OTHER), which will cause strange errors - when the scheduling policy is set to something else, like SCHED_RR. - - Since the return value of ACE_OS::thr_getprio() can be invalid for - a given policy, ie. 0 is valid for SCHED_OTHER, but invalid for - SCHED_RR, I then decided to set the priority to the minimum - priority of the policy/scheduling class. - - In addition, in RTPortableServer's - TAO_RT_Servant_Dispatcher::post_invoke() function, the priority - of the thread will be reset to what it was before the invocation. - This means that on platforms which use pthread_setschedparam(), - the initial priority and policy of the thread need to be - set to acceptable initial values, otherwise CORBA::DATA_CONVERSION - errors will be thrown in this TAO_RT_Servant_Dispatcher::post_invoke(). + * tao/RTCORBA/Linear_Priority_Mapping.cpp (to_CORBA): + Unconditionally display error message upon priority mapping failure. + * tao/RTCORBA/RT_Protocol_Hooks.cpp: Display errno value and message + if setting the native thread priority fails. + * tao/RTCORBA/RT_ORBInitiializer.h: + Create a new class member variable sched_policy_flags_. + The value of sched_policy_flags_ is a bitmask value that can be one of + THR_SCHED_RR, THR_SCHED_FIFO, THR_SCHED_DEFAULT, etc. + sched_policy_flags_ value is stored in the + TAO_ORB_Core TAO_ORB_Parameters structure. + In TAO_Thread_Lane::create_dynamic_threads(), the sched_policy_flags_ + is passed to the ACE_Task_Base::activate() call for creating new + threads for a RTCORBA thread pool. + + sched_policy_ is an integer value that can be one of + ACE_SCHED_RR, ACE_SCHED_FIFO, ACE_SCHED_OTHER, etc. This value + can be passed directly to the newly modified ACE_OS::thr_setpriority() + call. For POSIX pthreads, this value maps directly to the + policy parameter of the pthread_setschedparam() call. + ACE_SCHED_* values can also be passed directly to ACE_Sched_Params + for determining the minimum and maximum priority value for a particular + policy. + + * tao/RTCORBA/RT_ORBInitiializer.cpp: + Change all assignments of sched_policy_ variable to use + ACE_SCHED_* values and not THR_SCHED_* values. + In the TAO_RT_ORBInitializer constructor, set sched_policy_flags_ + to the appropriate THR_SCHED_* values depending on what ACE_SCHED_* + value is specified. + + In TAO_RT_ORBInitializer::pre_init(), pass sched_policy_flags_ + to orb_core()->orb_params()->sched_policy(), since this value + will be later used in the activate() call in + TAO_Thread_Lane::create_dynamic_threads(). + + At the end of TAO_RT_ORBInitializer::pre_init(), set the policy + of the ACE_Thread::self() to the policy specified by -ORBSchedPolicy, and + set the priority of ACE_Thread::self() to the lowest priority + of the specified policy. If this is not done, then errors will + occur on Linux, because the default policy in ACE_OS::thr_setprio() + is 0 (which maps to SCHED_OTHER), which will cause strange errors + when the scheduling policy is set to something else, like SCHED_RR. + + Since the return value of ACE_OS::thr_getprio() can be invalid for + a given policy, ie. 0 is valid for SCHED_OTHER, but invalid for + SCHED_RR, I then decided to set the priority to the minimum + priority of the policy/scheduling class. + + In addition, in RTPortableServer's + TAO_RT_Servant_Dispatcher::post_invoke() function, the priority + of the thread will be reset to what it was before the invocation. + This means that on platforms which use pthread_setschedparam(), + the initial priority and policy of the thread need to be + set to acceptable initial values, otherwise CORBA::DATA_CONVERSION + errors will be thrown in this TAO_RT_Servant_Dispatcher::post_invoke(). Tue Feb 11 13:58:23 2003 Nanbor Wang <nanbor@cs.wustl.edu> - * orbsvcs/vc7_orbsvcs.mwc: Added this temporary file that - specifies nightly VC7 build targets under orbsvcs. + * orbsvcs/vc7_orbsvcs.mwc: Added this temporary file that + specifies nightly VC7 build targets under orbsvcs. - * tao/tao.mpc: Added LF_Invocation_Event.cpp and LF_CH_Event.cpp. + * tao/tao.mpc: Added LF_Invocation_Event.cpp and LF_CH_Event.cpp. Tue Feb 11 12:48:12 UTC 2003 Gautam Thaker <gthaker@atl.lmco.com> * utils/catior/catior.cpp: * utils/catior/Makefile: Modified catior to better print Priority Model Policy. - Required changes to Makefile to now link wiht RTCORBA which - should be ok since RT CORBA is built by default. Did not - wish to have two versions of catior pgm, or some conditional - compiles. + Required changes to Makefile to now link wiht RTCORBA which + should be ok since RT CORBA is built by default. Did not + wish to have two versions of catior pgm, or some conditional + compiles. Tue Feb 11 07:11:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -285,67 +293,67 @@ Tue Feb 11 07:11:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> Mon Feb 10 17:12:30 2003 Otis Nyandoro <o.nyandoro@ee.wits.ac.za> - * tests/RTCORBA/check_supported_priorities.cpp: Changed the - function to take in a ORB pointer. The function now queries the - ORB to get the scheduling policy in place before checking the - priority levels. Using SCHED_OTHER as it was done before doesnt - work well on Linux and it was broken. - - * tests/RTCORBA/Banded_Connections/client.cpp: - * tests/RTCORBA/Banded_Connections/server.cpp: - * tests/RTCORBA/Client_Propagated/client.cpp: - * tests/RTCORBA/Client_Propagated/server.cpp: - * tests/RTCORBA/Collocation/Collocation.cpp: - * tests/RTCORBA/Destroy_Thread_Pool/Destroy_Thread_Pool.cpp: - * tests/RTCORBA/Explicit_Binding/client.cpp: - * tests/RTCORBA/Explicit_Binding/server.cpp: - * tests/RTCORBA/Linear_Priority/client.cpp: - * tests/RTCORBA/Linear_Priority/server.cpp: - * tests/RTCORBA/MT_Client_Protocol_Priority/client.cpp: - * tests/RTCORBA/MT_Client_Protocol_Priority/server.cpp: - * tests/RTCORBA/ORB_init/ORB_init.cpp: - * tests/RTCORBA/Persistent_IOR/client.cpp: - * tests/RTCORBA/Persistent_IOR/server.cpp: - * tests/RTCORBA/Policy_Combinations/client.cpp: - * tests/RTCORBA/Policy_Combinations/server.cpp: - * tests/RTCORBA/Private_Connection/client.cpp: - * tests/RTCORBA/Private_Connection/server.cpp: - * tests/RTCORBA/Profile_And_Endpoint_Selection/client.cpp: - * tests/RTCORBA/Profile_And_Endpoint_Selection/server.cpp: - * tests/RTCORBA/RTMutex/server.cpp: - * tests/RTCORBA/Server_Declared/client.cpp: - * tests/RTCORBA/Server_Declared/server.cpp: - * tests/RTCORBA/Server_Protocol/client.cpp: - * tests/RTCORBA/Server_Protocol/server.cpp: - * tests/RTCORBA/Thread_Pool/client.cpp: - * tests/RTCORBA/Thread_Pool/server.cpp: We now do a check on - supported priorities after the ORB is initialized. - - * tests/RTCORBA/Banded_Connections/svc.conf: - * tests/RTCORBA/Explicit_Binding/svc.conf: - * tests/RTCORBA/Linear_Priority/svc.conf: - * tests/RTCORBA/Policy_Combinations/svc.conf: Added scheduling - policy to all the configurations. - - * tests/RTCORBA/Client_Propagated/svc.conf: - * tests/RTCORBA/Collocation/svc.conf: - * tests/RTCORBA/Destroy_Thread_Pool/svc.conf: - * tests/RTCORBA/MT_Client_Protocol_Priority/svc.conf: - * tests/RTCORBA/ORB_init/svc.conf: - * tests/RTCORBA/Persistent_IOR/svc.conf: - * tests/RTCORBA/Policies/svc.conf: - * tests/RTCORBA/Private_Connection/svc.conf: - * tests/RTCORBA/RTMutex/svc.conf: - * tests/RTCORBA/Server_Declared/svc.conf: - * tests/RTCORBA/Server_Protocol/svc.conf: - * tests/RTCORBA/Thread_Pool/svc.conf: New files added with - scheduling information. This is not perfect for all cases, but a - good starting point for the bug reports that showed up about - RTCORBA on Linux. - - We need to relook into stressing these tests after 1.3.1 using - multiple scheduling policies. We have very less time to within - the short timeframe available. + * tests/RTCORBA/check_supported_priorities.cpp: Changed the + function to take in a ORB pointer. The function now queries the + ORB to get the scheduling policy in place before checking the + priority levels. Using SCHED_OTHER as it was done before doesnt + work well on Linux and it was broken. + + * tests/RTCORBA/Banded_Connections/client.cpp: + * tests/RTCORBA/Banded_Connections/server.cpp: + * tests/RTCORBA/Client_Propagated/client.cpp: + * tests/RTCORBA/Client_Propagated/server.cpp: + * tests/RTCORBA/Collocation/Collocation.cpp: + * tests/RTCORBA/Destroy_Thread_Pool/Destroy_Thread_Pool.cpp: + * tests/RTCORBA/Explicit_Binding/client.cpp: + * tests/RTCORBA/Explicit_Binding/server.cpp: + * tests/RTCORBA/Linear_Priority/client.cpp: + * tests/RTCORBA/Linear_Priority/server.cpp: + * tests/RTCORBA/MT_Client_Protocol_Priority/client.cpp: + * tests/RTCORBA/MT_Client_Protocol_Priority/server.cpp: + * tests/RTCORBA/ORB_init/ORB_init.cpp: + * tests/RTCORBA/Persistent_IOR/client.cpp: + * tests/RTCORBA/Persistent_IOR/server.cpp: + * tests/RTCORBA/Policy_Combinations/client.cpp: + * tests/RTCORBA/Policy_Combinations/server.cpp: + * tests/RTCORBA/Private_Connection/client.cpp: + * tests/RTCORBA/Private_Connection/server.cpp: + * tests/RTCORBA/Profile_And_Endpoint_Selection/client.cpp: + * tests/RTCORBA/Profile_And_Endpoint_Selection/server.cpp: + * tests/RTCORBA/RTMutex/server.cpp: + * tests/RTCORBA/Server_Declared/client.cpp: + * tests/RTCORBA/Server_Declared/server.cpp: + * tests/RTCORBA/Server_Protocol/client.cpp: + * tests/RTCORBA/Server_Protocol/server.cpp: + * tests/RTCORBA/Thread_Pool/client.cpp: + * tests/RTCORBA/Thread_Pool/server.cpp: We now do a check on + supported priorities after the ORB is initialized. + + * tests/RTCORBA/Banded_Connections/svc.conf: + * tests/RTCORBA/Explicit_Binding/svc.conf: + * tests/RTCORBA/Linear_Priority/svc.conf: + * tests/RTCORBA/Policy_Combinations/svc.conf: Added scheduling + policy to all the configurations. + + * tests/RTCORBA/Client_Propagated/svc.conf: + * tests/RTCORBA/Collocation/svc.conf: + * tests/RTCORBA/Destroy_Thread_Pool/svc.conf: + * tests/RTCORBA/MT_Client_Protocol_Priority/svc.conf: + * tests/RTCORBA/ORB_init/svc.conf: + * tests/RTCORBA/Persistent_IOR/svc.conf: + * tests/RTCORBA/Policies/svc.conf: + * tests/RTCORBA/Private_Connection/svc.conf: + * tests/RTCORBA/RTMutex/svc.conf: + * tests/RTCORBA/Server_Declared/svc.conf: + * tests/RTCORBA/Server_Protocol/svc.conf: + * tests/RTCORBA/Thread_Pool/svc.conf: New files added with + scheduling information. This is not perfect for all cases, but a + good starting point for the bug reports that showed up about + RTCORBA on Linux. + + We need to relook into stressing these tests after 1.3.1 using + multiple scheduling policies. We have very less time to within + the short timeframe available. Mon Feb 10 19:28:43 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -367,51 +375,51 @@ Fri Feb 7 14:04:59 2003 Jeff Parsons <j.parsons@vanderbilt.edu> Thu Feb 6 16:47:02 2003 Mayur Deshpande <mayur@ics.uci.edu> - * tests/AMH_Oneway - * tests/AMH_Oneway/AMH_Oneway.dsw - * tests/AMH_Oneway/Makefile - * tests/AMH_Oneway/Makefile.bor - * tests/AMH_Oneway/Test.idl - * tests/AMH_Oneway/client.bor - * tests/AMH_Oneway/client.cpp - * tests/AMH_Oneway/client.dsp - * tests/AMH_Oneway/run_test.pl - * tests/AMH_Oneway/server.bor - * tests/AMH_Oneway/server.cpp - * tests/AMH_Oneway/server.dsp - * tao/PortableServer/AMH_Response_Handler.cpp: - - Test to test Oneways and AMH. Client make 10 calls. Servers - counts the calls. Before thr fix, the server hung and this - generated an error mesg when the perl script (run_test.pl) ran. + * tests/AMH_Oneway + * tests/AMH_Oneway/AMH_Oneway.dsw + * tests/AMH_Oneway/Makefile + * tests/AMH_Oneway/Makefile.bor + * tests/AMH_Oneway/Test.idl + * tests/AMH_Oneway/client.bor + * tests/AMH_Oneway/client.cpp + * tests/AMH_Oneway/client.dsp + * tests/AMH_Oneway/run_test.pl + * tests/AMH_Oneway/server.bor + * tests/AMH_Oneway/server.cpp + * tests/AMH_Oneway/server.dsp + * tao/PortableServer/AMH_Response_Handler.cpp: + + Test to test Oneways and AMH. Client make 10 calls. Servers + counts the calls. Before thr fix, the server hung and this + generated an error mesg when the perl script (run_test.pl) ran. Thu Feb 06 18:35:05 2003 Pradeep Gore <pradeep@oomworks.com> - * orbsvcs/orbsvcs/Notify/Proxy_T.h: - * orbsvcs/orbsvcs/Notify/Any/CosEC_ProxyPushSupplier.h: - * orbsvcs/orbsvcs/Notify/Any/ProxyPushSupplier.h: - * orbsvcs/orbsvcs/Notify/Sequence/SequenceProxyPushSupplier.h: - * orbsvcs/orbsvcs/Notify/Structured/StructuredProxyPushSupplier.h: + * orbsvcs/orbsvcs/Notify/Proxy_T.h: + * orbsvcs/orbsvcs/Notify/Any/CosEC_ProxyPushSupplier.h: + * orbsvcs/orbsvcs/Notify/Any/ProxyPushSupplier.h: + * orbsvcs/orbsvcs/Notify/Sequence/SequenceProxyPushSupplier.h: + * orbsvcs/orbsvcs/Notify/Structured/StructuredProxyPushSupplier.h: - Added the patch from Bruce MacDonald <brucemac@netcomuk.co.uk> to get - Notification compiling on VC7/.NET + Added the patch from Bruce MacDonald <brucemac@netcomuk.co.uk> to get + Notification compiling on VC7/.NET Wed Feb 5 12:10:08 2003 Mayur Deshpande <mayur@ics.uci.edu> - * tao/PortableServer/AMH_Response_Handler.cpp: - - Added fix so that oneways work with AMH. Because of semantics of - oneways, the client doesn't expect any reply. The RH doesn't - know this and thinking that a RH is being destroyed before being - sent, it tries to send a exception to the client. This hangs the - server. The current fix handles oneways separately and doesn't - try and be 'intelligent' in the RH destruction process (and - doesn't try and send an exception to the cleint). Thanks to - Andrey Nechypurenko@mchp <Andrey.Nechypurenko@mchp.siemens.de> - for reporting a problem that led to uncovering of this - problem. Thanks to Balachandran Natarajan - <bala@isis-server.isis.vanderbilt.edu> for the suggestion to - retrieve the state from Server_Request to differentiate oneways. + * tao/PortableServer/AMH_Response_Handler.cpp: + + Added fix so that oneways work with AMH. Because of semantics of + oneways, the client doesn't expect any reply. The RH doesn't + know this and thinking that a RH is being destroyed before being + sent, it tries to send a exception to the client. This hangs the + server. The current fix handles oneways separately and doesn't + try and be 'intelligent' in the RH destruction process (and + doesn't try and send an exception to the cleint). Thanks to + Andrey Nechypurenko@mchp <Andrey.Nechypurenko@mchp.siemens.de> + for reporting a problem that led to uncovering of this + problem. Thanks to Balachandran Natarajan + <bala@isis-server.isis.vanderbilt.edu> for the suggestion to + retrieve the state from Server_Request to differentiate oneways. Wed Feb 5 13:34:44 2003 Jeff Parsons <j.parsons@vanderbilt.edu> @@ -430,20 +438,20 @@ Wed Feb 5 13:34:44 2003 Jeff Parsons <j.parsons@vanderbilt.edu> Wed Feb 5 10:02:10 2003 Ossama Othman <ossama@uci.edu> - * orbsvcs/orbsvcs/LoadBalancing/LB_CPU_Monitor.cpp (loads): + * orbsvcs/orbsvcs/LoadBalancing/LB_CPU_Monitor.cpp (loads): - Removed "ACE_OS::" scope modifier from fscanf() call. fscanf() - doesn't exist in that scope. Fixes compile-time problems - revealed by some compilers. Thanks to Matthias Blankenhaus - <matthias@ingrian.com> for reporting the problem. + Removed "ACE_OS::" scope modifier from fscanf() call. fscanf() + doesn't exist in that scope. Fixes compile-time problems + revealed by some compilers. Thanks to Matthias Blankenhaus + <matthias@ingrian.com> for reporting the problem. Wed Feb 5 10:59:16 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/be/be_visitor_enum/enum_cs.cpp: - Added a missing check for typecode support before generating - the typecode for an enum. Thanks to Mario Hofmann - <Mario.Hofmann@SystemaGmbH.de> for reporting the bug. + Added a missing check for typecode support before generating + the typecode for an enum. Thanks to Mario Hofmann + <Mario.Hofmann@SystemaGmbH.de> for reporting the bug. Wed Feb 5 09:44:33 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -467,11 +475,11 @@ Tue Feb 4 13:23:51 2003 Krishnakumar B <kitty@cs.wustl.edu> Sun Feb 02 21:01:12 2003 Frank Hunleth <frank@hunleth.com> - * orbsvcs/tests/Miop/McastHello/run_test.pl: Fixed a race condition - in the test script that could sometime cause the client to send - its requests before the second server starts. This occurs under - WinXP when using the Microsoft Loopback Adapter. Thanks to - Wolfgang Schroeder <cpp_kurs@arcor.de> for the configuration. + * orbsvcs/tests/Miop/McastHello/run_test.pl: Fixed a race condition + in the test script that could sometime cause the client to send + its requests before the second server starts. This occurs under + WinXP when using the Microsoft Loopback Adapter. Thanks to + Wolfgang Schroeder <cpp_kurs@arcor.de> for the configuration. Sun Feb 2 13:05:12 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -482,38 +490,38 @@ Sun Feb 2 13:05:12 2003 Johnny Willemsen <jwillemsen@remedy.nl> Sat Feb 1 09:03:55 2003 Ossama Othman <ossama@uci.edu> - * orbsvcs/orbsvcs/LoadBalancing/LB_CPU_Monitor.cpp (loads): + * orbsvcs/orbsvcs/LoadBalancing/LB_CPU_Monitor.cpp (loads): - Added missing completion status argument to CORBA:NO_IMPLEMENT - constructor call. Fixes compile-time problems for some - compilers (e.g. egcs 1.1.2). Thanks to Matthias Blankenhaus - <matthias@ingrian.com> for reporting the problem. + Added missing completion status argument to CORBA:NO_IMPLEMENT + constructor call. Fixes compile-time problems for some + compilers (e.g. egcs 1.1.2). Thanks to Matthias Blankenhaus + <matthias@ingrian.com> for reporting the problem. Fri Jan 31 17:07:20 2003 Pradeep Gore <pradeep@oomworks.com> - Fix for crash in Notify due to race condition in the Worker Task + Fix for crash in Notify due to race condition in the Worker Task shutdown. This crash showed up on the SunOS_Forte build. - * orbsvcs/orbsvcs/Notify/Worker_Task.h: - The is a Refcountable object now. + * orbsvcs/orbsvcs/Notify/Worker_Task.h: + The is a Refcountable object now. - * orbsvcs/orbsvcs/Notify/Reactive_Task.cpp: - * orbsvcs/orbsvcs/Notify/Reactive_Task.h: - Modifications for Refcounting support. + * orbsvcs/orbsvcs/Notify/Reactive_Task.cpp: + * orbsvcs/orbsvcs/Notify/Reactive_Task.h: + Modifications for Refcounting support. - * orbsvcs/orbsvcs/Notify/ThreadPool_Task.cpp: - * orbsvcs/orbsvcs/Notify/ThreadPool_Task.h: - Instead of relying on the thread count, the <close> method - decrements the count on the Task Object. + * orbsvcs/orbsvcs/Notify/ThreadPool_Task.cpp: + * orbsvcs/orbsvcs/Notify/ThreadPool_Task.h: + Instead of relying on the thread count, the <close> method + decrements the count on the Task Object. - * orbsvcs/orbsvcs/Notify/Object.cpp: - Apply refcounting to Task objects. + * orbsvcs/orbsvcs/Notify/Object.cpp: + Apply refcounting to Task objects. Fri Jan 31 14:25:07 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * orbsvcs/orbsvcs/SSLIOP/IIOP_SSL_Connection_Handler.cpp: Fixed a - unused warning in the daily builds. Thanks to Heather Drury for - letting me know. + * orbsvcs/orbsvcs/SSLIOP/IIOP_SSL_Connection_Handler.cpp: Fixed a + unused warning in the daily builds. Thanks to Heather Drury for + letting me know. Fri Jan 31 07:47:32 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -530,28 +538,28 @@ Fri Jan 31 07:47:32 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> Thu Jan 30 17:30:04 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/IIOP_Connection_Handler.cpp: - * tao/IIOP_Connection_Handler.h: Added a new protected constructor - that can be used by the derived classes. This constructor just - initializes itself and its base classes without initializing the - transport. This is needed since the derived classes may want to - initialize its own version of the transport. + * tao/IIOP_Connection_Handler.cpp: + * tao/IIOP_Connection_Handler.h: Added a new protected constructor + that can be used by the derived classes. This constructor just + initializes itself and its base classes without initializing the + transport. This is needed since the derived classes may want to + initialize its own version of the transport. - * orbsvcs/orbsvcs/SSLIOP/IIOP_SSL_Connection_Handler.cpp: Use the - new constructor to initialize the base class. + * orbsvcs/orbsvcs/SSLIOP/IIOP_SSL_Connection_Handler.cpp: Use the + new constructor to initialize the base class. - This fixes [BUG 1421]. Thanks to Jon Reis and Wayne Erchak for - reporting the problem. + This fixes [BUG 1421]. Thanks to Jon Reis and Wayne Erchak for + reporting the problem. Thu Jan 30 10:40:16 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/be_include/be_visitor_context.h: - If there is a system macro defined called 'interface', which - there is on some versions of MingW, redefine - it around the declaration of class be_visitor_context, which - contains a method by that name. Thanks to Johnny Willemsem - <jwillemsem@remedy.nl> for sending in the patch. + If there is a system macro defined called 'interface', which + there is on some versions of MingW, redefine + it around the declaration of class be_visitor_context, which + contains a method by that name. Thanks to Johnny Willemsem + <jwillemsem@remedy.nl> for sending in the patch. Thu Jan 30 07:03:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -564,34 +572,34 @@ Thu Jan 30 07:03:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> Wed Jan 29 12:43:06 2003 Frank Hunleth <frank@hunleth.com> - * tao/ORB_Core.{cpp,h}: Encapsulated all static members of - TAO_ORB_Core into a TAO_ORB_Core_Static_Resources singleton. By - accessing all of these variables through the singleton, this - alleviates the constructor call ordering problem inwhich - constructors in other compliation units (a.k.a some TAO shared - libraries) are called before those in TAO_ORB_Core and modify - the TAO_ORB_Core static variables before they are constructed. - This caused segfaults when it happened, and seemed to affect - OpenBSD the most. + * tao/ORB_Core.{cpp,h}: Encapsulated all static members of + TAO_ORB_Core into a TAO_ORB_Core_Static_Resources singleton. By + accessing all of these variables through the singleton, this + alleviates the constructor call ordering problem inwhich + constructors in other compliation units (a.k.a some TAO shared + libraries) are called before those in TAO_ORB_Core and modify + the TAO_ORB_Core static variables before they are constructed. + This caused segfaults when it happened, and seemed to affect + OpenBSD the most. - Note that this change is selfcontained and does not affect any - other classes. + Note that this change is selfcontained and does not affect any + other classes. Wed Jan 29 12:15:31 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tests/Bug_1361_Regression/run_test.pl: Reduced the number of - iterations that this test runs. Removed -ORBdebuglevel from teh - scripts. This was used for debugging purposes long back. + * tests/Bug_1361_Regression/run_test.pl: Reduced the number of + iterations that this test runs. Removed -ORBdebuglevel from teh + scripts. This was used for debugging purposes long back. - * tests/Bug_1361_Regression/Server_Thread_Pool.cpp: Removed a - debug statement where we print exceptions. This is expected in - this test. + * tests/Bug_1361_Regression/Server_Thread_Pool.cpp: Removed a + debug statement where we print exceptions. This is expected in + this test. Tue Jan 28 15:59:45 2003 Priyanka Gontla <pgontla@ece.uci.edu> - * docs/Options.html: - Added documentation on support to ORBListenEndpoints, ORBId and - ORBServerId ORB options. + * docs/Options.html: + Added documentation on support to ORBListenEndpoints, ORBId and + ORBServerId ORB options. Tue Jan 28 16:07:21 2003 Jeff Parsons <j.parsons@vanderbilt.edu> @@ -610,125 +618,125 @@ Tue Jan 28 16:07:21 2003 Jeff Parsons <j.parsons@vanderbilt.edu> Tue Jan 28 14:27:01 2003 Pradeep Gore <pradeep@oomworks.com> - * orbsvcs/orbsvcs/Notify/Properties.cpp: - In case no conf. file is specified, the EC will default to - reactive concurrency. This fixes the problem of the Notify - crashing without any conf. file. Thanks to Dennis Sporcic - <dennis@hsa.com.au> and Frederic Langlet - <Frederic.Langlet@alcatel.com> for reporting this problem. + * orbsvcs/orbsvcs/Notify/Properties.cpp: + In case no conf. file is specified, the EC will default to + reactive concurrency. This fixes the problem of the Notify + crashing without any conf. file. Thanks to Dennis Sporcic + <dennis@hsa.com.au> and Frederic Langlet + <Frederic.Langlet@alcatel.com> for reporting this problem. Mon Jan 27 23:01:43 2003 Frank Hunleth <frank@hunleth.com> - * orbsvcs/orbsvcs/PortableGroup/UIPMC_Transport.cpp: Fixed code - that was supposed to fake out a good error return when the call - to send() fails. This fixes an exception break on WinXP when - the network cable is unplugged and a program makes a MIOP - request. The justification behide ignoring the error is - explained further in the code, but basically, it takes advantage - of the fact that MIOP is unreliable by specification, and hopes - that the source of the error will be corrected. Thanks to - Antonio Leonforte <a.leonforte@mclink.it> for pointing out the - exception break. + * orbsvcs/orbsvcs/PortableGroup/UIPMC_Transport.cpp: Fixed code + that was supposed to fake out a good error return when the call + to send() fails. This fixes an exception break on WinXP when + the network cable is unplugged and a program makes a MIOP + request. The justification behide ignoring the error is + explained further in the code, but basically, it takes advantage + of the fact that MIOP is unreliable by specification, and hopes + that the source of the error will be corrected. Thanks to + Antonio Leonforte <a.leonforte@mclink.it> for pointing out the + exception break. Mon Jan 27 19:16:33 2003 Pradeep Gore <pradeep@oomworks.com> - Made the following changes to fix the race related crashes of - Notify seen in orbsvcs/tests/Notify/performance-tests/RedGreen - - - * orbsvcsy/orbsvcs/Notify/Builder.cpp: - Added a refcount on each parent object per child object. - This way the parent is always alive when the child is destroyed. - When the child is destroyed, the parent count is decremented. - The ensures a bottom-up destruction of the Notify object tree. - - * orbsvcs/orbsvcs/Notify/Object.cpp: - * orbsvcs/orbsvcs/Notify/Object.h: - Made the TAO_NS_Object class a TAO_NS_Refcountable. Moved the - <has_shutdown> method to this class from TAO_NS_Object_T. - The <shutdown> method sets the <shutdown_> flag atomically. - It returns 1 if this object has been shutdown already. All - derived classes should consult the TAO_NS_Object::shutdown - method's return value to determine if the shutdown should - proceed in case of multiple calls to shutdown. - - * orbsvcs/orbsvcs/Notify/Object_T.cpp: - * orbsvcs/orbsvcs/Notify/Object_T.h: - Decrement the count on this Objects parent in the destructor. - Changed the signature of destroy similar to <shutdown> to avoid - multiple destoy's. - - * orbsvcs/orbsvcs/Notify/Container_T.cpp: - * orbsvcs/orbsvcs/Notify/Container_T.h: - The <collection_> owned by the Container is destroyed in the - <shutdown> method now. The releases the reference count on child - objects. - - * orbsvcs/orbsvcs/Notify/EventChannel.cpp: - * orbsvcs/orbsvcs/Notify/EventChannelFactory.cpp: - The <destroy> method checks the return value of base destroy - method before proceeding. - - * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.cpp: - * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.h: - * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.inl: - Made the Event_Map_Entry a refcounted object. - The <usage_count_> member is incremented when the <find> method - returns a suitable entry. - The caller is responsible for decrementing the count when it no - longer needs the entry. - - * orbsvcs/orbsvcs/Notify/Event_Map_T.cpp: - * orbsvcs/orbsvcs/Notify/Event_Map_T.h: - * orbsvcs/orbsvcs/Notify/Event_Map_T.inl: - <find> method now returns the entry (from which the collection - can be obtained). - The <remove> method decrements the count on the entry instead of - deleteting it. - - * orbsvcs/orbsvcs/Notify/Method_Request_Lookup.cpp: - Obtains the correct map entry and uses the collection in that - entry. The map is released when it is done using it. - Note that the map pointer is guaranteed to be valid because the - Lookup object holds a reference count to the proxy that holds a - count on its parent and so on up to the EventChannel. The - EventChannel only destroys the Event_manager in its destructor. - - * orbsvcs/orbsvcs/Notify/Notify_Service.cpp: - * orbsvcs/orbsvcs/Notify/Notify_Service.h: - Added a fake <_decr_refcnt> to satisfy the template requirements - from TAO_NS_Object_T. This will be rectified in future revisions. - - * orbsvcs/orbsvcs/Notify/ProxyConsumer.cpp: - * orbsvcs/orbsvcs/Notify/ProxyConsumer.h: - * orbsvcs/orbsvcs/Notify/ProxySupplier.cpp: - * orbsvcs/orbsvcs/Notify/ProxySupplier.h: - Modified the <shutdown> implementation to follow the checks as - described above. - - * orbsvcs/orbsvcs/Notify/Timer.h: - * orbsvcs/orbsvcs/Notify/Timer_Reactor.cpp: - * orbsvcs/orbsvcs/Notify/Timer_Reactor.h: - Made the Timer a refcounted object. - - * orbsvcs/orbsvcs/Notify/Timer_Queue.cpp: - * orbsvcs/orbsvcs/Notify/Timer_Queue.h: - Added an <impl> method to obtain the timer implementation. - - * orbsvcs/orbsvcs/Notify/Worker_Task.h: - Removed the <init> method which is not required here. - - * orbsvcs/orbsvcs/Notify/Reactive_Task.cpp: - * orbsvcs/orbsvcs/Notify/Reactive_Task.h: - Added an <init> method in which the Reactor based timer is - allocated. - - * orbsvcs/orbsvcs/Notify/ThreadPool_Task.cpp: - * orbsvcs/orbsvcs/Notify/ThreadPool_Task.h: - The Timer Queue is allocated in the <init> method. - - * orbsvcs/orbsvcs/Notify/Sequence/SequencePushConsumer.cpp: - Because the timer is now refcounted, we must release it on - shutdown. + Made the following changes to fix the race related crashes of + Notify seen in orbsvcs/tests/Notify/performance-tests/RedGreen - + + * orbsvcsy/orbsvcs/Notify/Builder.cpp: + Added a refcount on each parent object per child object. + This way the parent is always alive when the child is destroyed. + When the child is destroyed, the parent count is decremented. + The ensures a bottom-up destruction of the Notify object tree. + + * orbsvcs/orbsvcs/Notify/Object.cpp: + * orbsvcs/orbsvcs/Notify/Object.h: + Made the TAO_NS_Object class a TAO_NS_Refcountable. Moved the + <has_shutdown> method to this class from TAO_NS_Object_T. + The <shutdown> method sets the <shutdown_> flag atomically. + It returns 1 if this object has been shutdown already. All + derived classes should consult the TAO_NS_Object::shutdown + method's return value to determine if the shutdown should + proceed in case of multiple calls to shutdown. + + * orbsvcs/orbsvcs/Notify/Object_T.cpp: + * orbsvcs/orbsvcs/Notify/Object_T.h: + Decrement the count on this Objects parent in the destructor. + Changed the signature of destroy similar to <shutdown> to avoid + multiple destoy's. + + * orbsvcs/orbsvcs/Notify/Container_T.cpp: + * orbsvcs/orbsvcs/Notify/Container_T.h: + The <collection_> owned by the Container is destroyed in the + <shutdown> method now. The releases the reference count on child + objects. + + * orbsvcs/orbsvcs/Notify/EventChannel.cpp: + * orbsvcs/orbsvcs/Notify/EventChannelFactory.cpp: + The <destroy> method checks the return value of base destroy + method before proceeding. + + * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.cpp: + * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.h: + * orbsvcs/orbsvcs/Notify/Event_Map_Entry_T.inl: + Made the Event_Map_Entry a refcounted object. + The <usage_count_> member is incremented when the <find> method + returns a suitable entry. + The caller is responsible for decrementing the count when it no + longer needs the entry. + + * orbsvcs/orbsvcs/Notify/Event_Map_T.cpp: + * orbsvcs/orbsvcs/Notify/Event_Map_T.h: + * orbsvcs/orbsvcs/Notify/Event_Map_T.inl: + <find> method now returns the entry (from which the collection + can be obtained). + The <remove> method decrements the count on the entry instead of + deleteting it. + + * orbsvcs/orbsvcs/Notify/Method_Request_Lookup.cpp: + Obtains the correct map entry and uses the collection in that + entry. The map is released when it is done using it. + Note that the map pointer is guaranteed to be valid because the + Lookup object holds a reference count to the proxy that holds a + count on its parent and so on up to the EventChannel. The + EventChannel only destroys the Event_manager in its destructor. + + * orbsvcs/orbsvcs/Notify/Notify_Service.cpp: + * orbsvcs/orbsvcs/Notify/Notify_Service.h: + Added a fake <_decr_refcnt> to satisfy the template requirements + from TAO_NS_Object_T. This will be rectified in future revisions. + + * orbsvcs/orbsvcs/Notify/ProxyConsumer.cpp: + * orbsvcs/orbsvcs/Notify/ProxyConsumer.h: + * orbsvcs/orbsvcs/Notify/ProxySupplier.cpp: + * orbsvcs/orbsvcs/Notify/ProxySupplier.h: + Modified the <shutdown> implementation to follow the checks as + described above. + + * orbsvcs/orbsvcs/Notify/Timer.h: + * orbsvcs/orbsvcs/Notify/Timer_Reactor.cpp: + * orbsvcs/orbsvcs/Notify/Timer_Reactor.h: + Made the Timer a refcounted object. + + * orbsvcs/orbsvcs/Notify/Timer_Queue.cpp: + * orbsvcs/orbsvcs/Notify/Timer_Queue.h: + Added an <impl> method to obtain the timer implementation. + + * orbsvcs/orbsvcs/Notify/Worker_Task.h: + Removed the <init> method which is not required here. + + * orbsvcs/orbsvcs/Notify/Reactive_Task.cpp: + * orbsvcs/orbsvcs/Notify/Reactive_Task.h: + Added an <init> method in which the Reactor based timer is + allocated. + + * orbsvcs/orbsvcs/Notify/ThreadPool_Task.cpp: + * orbsvcs/orbsvcs/Notify/ThreadPool_Task.h: + The Timer Queue is allocated in the <init> method. + + * orbsvcs/orbsvcs/Notify/Sequence/SequencePushConsumer.cpp: + Because the timer is now refcounted, we must release it on + shutdown. Mon Jan 27 17:36:51 2003 Jeff Parsons <j.parsons@vanderbilt.edu> @@ -742,9 +750,9 @@ Mon Jan 27 17:36:51 2003 Jeff Parsons <j.parsons@vanderbilt.edu> Mon Jan 27 16:47:31 2003 Frank Hunleth <frank@hunleth.com> - * orbsvcs/tests/Miop/McastHello/run_test.pl: Fixed incorrect - reference to $SV in error handling code. Thanks to Craig - Rodrigues for pointing this out. + * orbsvcs/tests/Miop/McastHello/run_test.pl: Fixed incorrect + reference to $SV in error handling code. Thanks to Craig + Rodrigues for pointing this out. Mon Jan 27 09:48:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> @@ -757,90 +765,90 @@ Mon Jan 27 09:48:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> Sun Jan 26 17:29:34 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/Leader_Follower.cpp (wait_for_event): Set LFS_TIMEOUT on the - LF_Event only if the follower timesout. The follower could - return with a -1 from wait () even if errors other than timeout - occurs, like connection closure and such. + * tao/Leader_Follower.cpp (wait_for_event): Set LFS_TIMEOUT on the + LF_Event only if the follower timesout. The follower could + return with a -1 from wait () even if errors other than timeout + occurs, like connection closure and such. - * tao/LF_CH_Event.cpp: Added a comment + * tao/LF_CH_Event.cpp: Added a comment Sun Jan 26 17:23:22 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/Notify_Handler.cpp (TAO_Notify_Handler): - * tao/Notify_Handler.h: + * tao/Notify_Handler.cpp (TAO_Notify_Handler): + * tao/Notify_Handler.h: - - Added a new constructor and a create_handler () method which - takes in a transport and the ACE_HANDLE that this handler would - be used for. + - Added a new constructor and a create_handler () method which + takes in a transport and the ACE_HANDLE that this handler would + be used for. - - Deprecating the old create_handler () and the constructor. - They have been left there for backward compatibility. + - Deprecating the old create_handler () and the constructor. + They have been left there for backward compatibility. - The above changes were made to get around a problem that Yakov - Becker and Johnny Willemsen observed in one of their - applications. We havent been able to reproduce the problem here - at all. The gist of the problem is + The above changes were made to get around a problem that Yakov + Becker and Johnny Willemsen observed in one of their + applications. We havent been able to reproduce the problem here + at all. The gist of the problem is - - A server thread after buffering lots of incoming messages - starts processing one of the messages. Keeping the handle - suspended it wakes up another thread in the Reactor. + - A server thread after buffering lots of incoming messages + starts processing one of the messages. Keeping the handle + suspended it wakes up another thread in the Reactor. - - The first thread after processing the request tries to write - a reply only to find the socket closed. It starts the - cleanup and disassociates the transport and the handler. So - far so good. + - The first thread after processing the request tries to write + a reply only to find the socket closed. It starts the + cleanup and disassociates the transport and the handler. So + far so good. - - The second thread that is woken up now tries to get to the - handler and then to the transport to process the queued - messages. But the transport is already disassociated from the - handler and the assert fails. + - The second thread that is woken up now tries to get to the + handler and then to the transport to process the queued + messages. But the transport is already disassociated from the + handler and the assert fails. - We address this problem by using the transport as the mechanism - to process messages instead of the handler. The above changes - were made to accomodate the change in philosophy. + We address this problem by using the transport as the mechanism + to process messages instead of the handler. The above changes + were made to accomodate the change in philosophy. - * tao/Transport.cpp: Used the new method to create the - Notify_Handler. + * tao/Transport.cpp: Used the new method to create the + Notify_Handler. Sat Jan 25 18:10:10 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * orbsvcs/tests/Security/MT_IIOP_SSL/Makefile: Fixed dependencies. - * orbsvcs/tests/Security/MT_IIOP_SSL/test_i.cpp: Cosmetic fixes. - * orbsvcs/tests/Security/Makefile: Added the new file to the DIRS - section. + * orbsvcs/tests/Security/MT_IIOP_SSL/Makefile: Fixed dependencies. + * orbsvcs/tests/Security/MT_IIOP_SSL/test_i.cpp: Cosmetic fixes. + * orbsvcs/tests/Security/Makefile: Added the new file to the DIRS + section. Sat Jan 25 17:55:10 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * orbsvcs/tests/Security/MT_IIOP_SSL/client.dsp: - * orbsvcs/tests/Security/MT_IIOP_SSL/server.dsp: Added necessary - libraries for linking. + * orbsvcs/tests/Security/MT_IIOP_SSL/client.dsp: + * orbsvcs/tests/Security/MT_IIOP_SSL/server.dsp: Added necessary + libraries for linking. Sat Jan 25 18:55:47 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * orbsvcs/tests/Security/MT_IIOP_SSL/test_i.h: Fixed a compilation error. + * orbsvcs/tests/Security/MT_IIOP_SSL/test_i.h: Fixed a compilation error. Sat Jan 25 18:50:32 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/Transport_Cache_Manager.cpp (close_i): Avoid totally - unaesthetic temporaries in the for loop. + * tao/Transport_Cache_Manager.cpp (close_i): Avoid totally + unaesthetic temporaries in the for loop. Sat Jan 25 18:30:27 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * orbsvcs/tests/Security/MT_IIOP_SSL: A multi-threaded stress test - for IIOP_SSL which basically provides a insecure path for - communication incase an SSL connection cannot be setup. This - will also serve as a stress test for BUG 1421. + * orbsvcs/tests/Security/MT_IIOP_SSL: A multi-threaded stress test + for IIOP_SSL which basically provides a insecure path for + communication incase an SSL connection cannot be setup. This + will also serve as a stress test for BUG 1421. Sat Jan 25 18:06:51 2003 Balachandran Natarajan <bala@isis-server.isis.vanderbilt.edu> - * tao/Transport_Cache_Manager.h: - * tao/Transport_Cache_Manager.inl: Added an accessor for the hash - map used inside the Cache_Manager. Useful for writing regression - tests as we will demonstrate shortly in a test. + * tao/Transport_Cache_Manager.h: + * tao/Transport_Cache_Manager.inl: Added an accessor for the hash + map used inside the Cache_Manager. Useful for writing regression + tests as we will demonstrate shortly in a test. Fri Jan 24 11:42:57 2003 Boris S Kolpackov <boris@isis-server.isis.vanderbilt.edu> - * orbsvcs/tests/Security/Crash_Test/run_test.pl: fixed yet another + * orbsvcs/tests/Security/Crash_Test/run_test.pl: fixed yet another possible race conditions. Fri Jan 24 10:59:49 2003 Jeff Parsons <j.parsons@vanderbilt.edu> @@ -858,9 +866,9 @@ Fri Jan 24 09:04:28 2003 Jeff Parsons <j.parsons@vanderbilt.edu> * TAO_IDL/fe/fe_interface_header.cpp: - Fixed a bug in error-checking the inheritance of abstract - interfaces. Thanks to Michael Soden <soden@ikv.de> for - reporting the IDL compiler's incorrect error message. + Fixed a bug in error-checking the inheritance of abstract + interfaces. Thanks to Michael Soden <soden@ikv.de> for + reporting the IDL compiler's incorrect error message. Fri Jan 24 08:23:12 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> diff --git a/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp b/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp index 59415c1d927..abefc409f72 100644 --- a/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp +++ b/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.cpp @@ -48,7 +48,7 @@ void TAO_LB_IORInterceptor::establish_components ( PortableInterceptor::IORInfo_ptr ACE_ENV_ARG_DECL_NOT_USED) - ACE_THROW_SPEC (()) + ACE_THROW_SPEC ((CORBA::SystemException)) { } diff --git a/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h b/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h index 609d0ec2cf6..8b69b2b9b32 100644 --- a/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h +++ b/TAO/orbsvcs/orbsvcs/LoadBalancing/LB_IORInterceptor.h @@ -81,7 +81,9 @@ public: virtual void establish_components ( PortableInterceptor::IORInfo_ptr info ACE_ENV_ARG_DECL_WITH_DEFAULTS) - ACE_THROW_SPEC (()); + ACE_THROW_SPEC (( + CORBA::SystemException + )); virtual void components_established ( PortableInterceptor::IORInfo_ptr info |