diff options
author | Ossama Othman <ossama-othman@users.noreply.github.com> | 2003-12-10 22:40:27 +0000 |
---|---|---|
committer | Ossama Othman <ossama-othman@users.noreply.github.com> | 2003-12-10 22:40:27 +0000 |
commit | fb333a425750570124bd2003ea85d5d7ecba73c8 (patch) | |
tree | f4f2e85ff84824fdc8f0d0bed6d177998510122c /TAO/ChangeLog | |
parent | a3b293790f31913fe1732c6a2fb5ace5a3dd5a0f (diff) | |
download | ATCD-fb333a425750570124bd2003ea85d5d7ecba73c8.tar.gz |
ChangeLogTag:Wed Dec 10 16:39:25 2003 Ossama Othman <ossama@dre.vanderbilt.edu>
Diffstat (limited to 'TAO/ChangeLog')
-rw-r--r-- | TAO/ChangeLog | 111 |
1 files changed, 89 insertions, 22 deletions
diff --git a/TAO/ChangeLog b/TAO/ChangeLog index d226057a353..4895e7bd2a8 100644 --- a/TAO/ChangeLog +++ b/TAO/ChangeLog @@ -1,3 +1,63 @@ +Wed Dec 10 16:39:25 2003 Ossama Othman <ossama@dre.vanderbilt.edu> + + * tao/Synch_Invocation.cpp (handle_system_exception): + + Corrected repository IDs for non-fatal system exceptions being + detected (TRANSIENT, OBJ_ADAPTER, NO_RESPONSE). The + "IDL:omg.org/CORBA/" prefix and ":1.0" suffix were missing, thus + preventing code that should have been run in the presence of + these exceptions from being run. + + Attempt to try another profile in the non-fatal exception case. + This fixes a problem where valid profiles were not retried when + they should have been. + + * orbsvcs/orbsvcs/LoadBalancing/LB_LeastLoaded.h + (LL_DEFAULT_LOAD_PERCENT_DIFF_CUTOFF): + + New constant used in thundering herd fix described below. + + * orbsvcs/orbsvcs/LoadBalancing/LB_LeastLoaded.cpp (get_location): + + Fixed problem where least loaded object group member was not + always selected, thus reducing scalability. + + If two locations have loads that differ by a "small" amount + (i.e. LL_DEFAULT_LOAD_PERCENT_DIFF_CUTOFF), choose one of them + at random. This helps alleviate a thundering herd phenomenon + that could occur when an object group member becomes available + for request handling. Thanks to Carlos, Marina and Jody at ATD + for coming up with this solution to this form of the thundering + herd problem. + + * orbsvcs/orbsvcs/LoadBalancing/LB_ServerRequestInterceptor.cpp + (receive_request_service_contexts): + + Always allow LoadMonitor and LoadAlert operations to proceed + regardless of the alert condition. Addresses a problem where it + was not possible to disable alert states and query loads from a + LoadMonitor collocated with a server once a load alert was + enabled. + + * orbsvcs/performance-tests/LoadBalancing/ORBInitializer.h: + + Improved documentation of some class members. + + * orbsvcs/performance-tests/LoadBalancing/ORBInitializer.cpp: + + Cosmetic change. + + * orbsvcs/performance-tests/LoadBalancing/RPS_Monitor.cpp (loads): + + If the elapsed-time is zero, set the load to zero to prevent a + division-by-zero error. + + * orbsvcs/performance-tests/LoadBalancing/Roundtrip.cpp + (test_method): + + Commented out a bunch of code that should not be used when + performing benchmarks. + Wed Dec 10 13:58:41 2003 Chris Cleeland <cleeland_c@ociweb.com> * tao/default_client.cpp (parse_args): Corrected erroneous option @@ -7,26 +67,26 @@ Wed Dec 10 13:58:41 2003 Chris Cleeland <cleeland_c@ociweb.com> * tao/IIOP_Transport.cpp (recv): Corrected method name printed in debug message. - * tao/GIOP_Message_Base.cpp: - * tao/GIOP_Message_Base.h: - * tao/GIOP_Message_Base.i: - * tao/GIOP_Message_Generator_Parser_Impl.inl: - * tao/GIOP_Message_Lite.cpp: - * tao/GIOP_Message_Lite.h: - * tao/GIOP_Message_State.cpp: - * tao/GIOP_Message_State.h: - * tao/GIOP_Message_State.inl: - * tao/Incoming_Message_Queue.cpp: - * tao/Incoming_Message_Queue.h: - * tao/Incoming_Message_Queue.inl: - * tao/Pluggable_Messaging.h: - * tao/Transport.cpp: - * tao/Transport.h: + * tao/GIOP_Message_Base.cpp: + * tao/GIOP_Message_Base.h: + * tao/GIOP_Message_Base.i: + * tao/GIOP_Message_Generator_Parser_Impl.inl: + * tao/GIOP_Message_Lite.cpp: + * tao/GIOP_Message_Lite.h: + * tao/GIOP_Message_State.cpp: + * tao/GIOP_Message_State.h: + * tao/GIOP_Message_State.inl: + * tao/Incoming_Message_Queue.cpp: + * tao/Incoming_Message_Queue.h: + * tao/Incoming_Message_Queue.inl: + * tao/Pluggable_Messaging.h: + * tao/Transport.cpp: + * tao/Transport.h: * tao/orbconf.h: * tao/Strategies/DIOP_Transport.cpp * tao/Strategies/SHMIOP_Transport.cpp * tao/Strategies/SHMIOP_Transport.h - + Integrated aggregate fix to infamous Parse Magic Bytes (PMB) problem. The fix was originally developed in OCI's repository and migrated here. An overall description of the problem and @@ -108,7 +168,7 @@ Tue Mar 18 14:57:07 2003 Chris Cleeland <cleeland_c@ociweb.com> * tao/RTCORBA/RT_ORBInitializer.h: * tao/RTCORBA/RT_ORB_Loader.cpp: * tao/RTCORBA/RT_Protocols_Hooks.cpp: - + Merged in changes from TAO 1.3.1. The actual tag for the merge was pmb_branch_mainline_mergeout_1. The command was @@ -191,7 +251,7 @@ Mon Mar 17 11:09:00 2003 Chris Cleeland <cleeland_c@ociweb.com> the queue. Also, removed old code leftover from when the incoming queue could have partial GIOP messages in it. Finally, insured that the return value matched previous requirements. - + Thu Mar 6 15:27:14 2003 Chris Cleeland <cleeland_c@ociweb.com> @@ -264,7 +324,7 @@ Wed Feb 26 21:44:34 2003 Chris Cleeland <cleeland_c@ociweb.com> Make message_type() static. * TAO/tao/GIOP_Message_Generator_Parser_Impl.inl: - + Reimplement check_revision() to be exact about which GIOP revisions are okay. This is different from before where it loosely assumed that everything was valid unless it want higher @@ -459,9 +519,9 @@ Wed Feb 26 21:44:34 2003 Chris Cleeland <cleeland_c@ociweb.com> * TAO/tests/InterOp-Naming/INS_test_client.cpp: Changes that don't appear to be related in any way to PMB... - ----------------------------------------- - + +---------------------------------------- + Wed Dec 10 15:12:34 UTC 2003 Johnny Willemsen <jwillemsen@remedy.nl> * tests/Smart_Proxies/Collocation/Smart_Proxy_Impl.cpp: @@ -703,10 +763,17 @@ Mon Dec 8 19:33:03 2003 Balachandran Natarajan <bala@dre.vanderbilt.edu> * tao/Synch_Invocation.cpp: +<<<<<<< ChangeLog + Fixed a minor problem with reinvocation. We were not handling + errors properly, more precisely the restart errors if they show + up during wait_for_reply (). Thanks to Milan Cvetkovic for + motivating this change. +======= Fixed a minor problem with reinvocation. We were not handling errors properly, more precisely the retsrat errors if they show up during wait_for_reply (). Thanks to Milan Cvetkovic for motivating this change. +>>>>>>> 1.3209 * tao/Messaging/Asynch_Invocation.cpp: |