Please report new issues athttps://github.com/DOCGroup
On invocation of a collocated AMI method, in-parameters become corrupted. A test-case is added to the test/AMI/client.cpp, that demonstrates the problem.
accept, regression is in svn
mine
Wed Apr 11 13:32:12 UTC 2007 Johnny Willemsen <jwillemsen@remedy.nl> * tests/AMI/ami_test_i.cpp: Fixed small bug in this test code that could lead to an incorrect return value when the parameters are checked.
Sun Apr 15 06:18:56 UTC 2007 Johnny Willemsen <jwillemsen@remedy.nl> * tests/AMI/client.cpp: * tests/AMI/AMI.mpc: * tests/AMI/run_test.pl: * tests/Bug_2805_Regression/*: Moved test for bug 2805 to its own regression, now the AMI test should succeed again, but 2805 is failing
Fri Apr 20 08:05:12 UTC 2007 Johnny Willemsen <jwillemsen@remedy.nl> * tao/PortableServer/Collocated_Arguments_Converter.cpp: * tao/PortableServer/Collocated_Arguments_Converter.h: * tao/PortableServer/DII_Arguments_Converter.cpp: * tao/PortableServer/DII_Arguments_Converter.h: Refactored the DII argument to a generic collocated argument converter. The issue that was solved for DII is also valid for AMI. Also made a method for converting the request and one for the reply, converting the reply is now not done yet but we do know that also fails. * tao/PortableServer/Upcall_Wrapper.cpp: Changed how the argument converter is retrieved. Instead of retrieving it using service configurator we put a pointer in the operation details which is set or not. * tao/DynamicInterface/DII_Arguments_Converter_Impl.cpp: * tao/DynamicInterface/DII_Arguments_Converter_Impl.h: Updated because the DII base class is renamed: * tao/DynamicInterface/DII_Invocation_Adapter.cpp: * tao/DynamicInterface/DII_Invocation_Adapter.h: Added new adapter for oneway DII calls so that this adapter can set the DII Argument Converter for DII oneway calls. * tao/DynamicInterface/Request.cpp: Use false instead of 0, layout changes, use new DII Oneway adapter. * tao/DynamicInterface/Request.h: * tao/Protocol_Factory.cpp: * tao/IIOP_Profile.cpp: * tao/PI/PolicyFactoryC.cpp: Layout changes * tao/DynamicInterface/Request.inl: Use false instead of 0 * tao/DynamicInterface/Server_Request.inl: Use static_cast instead of c-style cast * tao/IIOP_Connector.cpp: Fixed an inconsistent debug statement * tao/Invocation_Adapter.cpp: * tao/Invocation_Adapter.h: * tao/Invocation_Adapter.inl: Layout changes and no need to have a member whether we are a DII invocation or not * tao/Makefile.am: Updated for new files * tao/Messaging/AMI_Arguments_Converter_Impl.cpp: * tao/Messaging/AMI_Arguments_Converter_Impl.h: New AMI argument converter * tao/Messaging/Asynch_Invocation_Adapter.cpp: Use the new AMI argument converter which fixes bugzilla 2805. Also use the * tao/Messaging/Messaging_Loader.cpp: Added include of the AMI Argument converter * tao/operation_details.cpp: * tao/operation_details.h: * tao/operation_details.inl: Instead of storing whether we have a DII call or not, we store a pointer to a collocated argument converter that is set or not set * tao/ORB_Core.cpp: * tao/ORB_Core.h: * tao/params.cpp: * tao/params.h: * tao/params.inl: Added a new -ORBAMICollication which is default yes. If this is no, then when a collocated AMI invocation is made, the call is converted to a remote call so that another thread can handle the servant. Also the problems with argument corruption are then gone. This is for bugzilla 2351. * tao/Profile.h: Documentation fix * tao/Profile_Transport_Resolver.cpp: * tao/Profile_Transport_Resolver.h: * tao/PortableServer/POAManager.h: * tao/PortableServer/POAManagerFactory.cpp: Layout changes * tao/Stub.inl: * tao/TAO_Server_Request.cpp: * tao/TAO_Server_Request.h: * tao/TAO_Server_Request.inl: Bool changes * tao/TkResource/TkResource_Factory.h: * tao/QtResource/QtResource_Factory.h: Removed not used mutex * tao/Valuetype/AbstractBase.cpp: Bool change * docs/Options.html: Document new ORBAMICollocation option, this is default yes which means that when a collocated AMI call is made the servant is invoked using the client thread. When this option is no, when a collocated AMI call is made, this is converted to a remote call so that a different thread can be used to invoke the servant.
Johnny Willemsen 23-April-2007 12:46 AM US Central wrote: > The regression for 2805 is failing because 2806 is not fixed yet. I am > working on that but it will take probably until next week until I have made > all changes for fixing that bug. Reopened Bug 2805, which is now blocked by Bug 2806.
this one is fixed. 2806 itself is not fixed, but the problem with parameter corruption is, so closing this one