ChangeLogTag: Sun Dec 30 10:31:09 2001 Balachandran Natarajan <bala@cs.wustl.edu>

3 files changed, 8 insertions, 28 deletions

diff --git a/TAO/tests/BiDirectional/README b/TAO/tests/BiDirectional/README
index 74deb74a8aa..ea17bd32f25 100644
--- a/TAO/tests/BiDirectional/README
+++ b/TAO/tests/BiDirectional/README
@@ -13,28 +13,7 @@ ready for callback. The server then callsback the client on the same
 connection <no_iterations> times. If the server creates a new
 connection the server would crash itself.
 
-Notes and Observations:
-----------------------
-It has been observed that when the server tries to call the client
-during the upcall on the same connection, there is a deadlock. To make
-things clearer, the scenario can be picturised as follows
-
-- the server receives a request from the client 
-- in the upcall the server makes a remote call to the client
-  (callback) 
-- the server received BiDirectional context info from the client (and
-  the server has the ability to call the client on the same
-  connection) 
-- then the call from the server to the client would be on the same
-  connection established by the client. 
-
-This situation would result in a deadlock. The deadlock would occur at
-the server end as the server does not read the reply from the
-client. This situation occurs because of the way the TP reactor (the
-default in TAO) works. The reactor suspends the handler during the
-upcall and so it is not in a position to react to the messages
-appearing on the handle. We are trying to look at ways to get around
-this situation.
+
 
 
 
diff --git a/TAO/tests/BiDirectional/server.cpp b/TAO/tests/BiDirectional/server.cpp
index 9ad9a5032b4..fd5c272ead5 100644
--- a/TAO/tests/BiDirectional/server.cpp
+++ b/TAO/tests/BiDirectional/server.cpp
@@ -139,7 +139,8 @@ main (int argc, char *argv[])
       int retval = 0;
       while (retval == 0)
         {
-          // Please see the README file on why this is being done
+          // Just process one upcall. We know that we would get the
+          // clients IOR in that call.
           CORBA::Boolean pending =
             orb->work_pending(TAO_ENV_SINGLE_ARG_PARAMETER);
           ACE_TRY_CHECK;
@@ -150,6 +151,8 @@ main (int argc, char *argv[])
               ACE_TRY_CHECK;
             }
 
+          // Now that hopefully we have the clients IOR, just start
+          // making remote calls to the client.
           retval = server_impl.call_client (TAO_ENV_SINGLE_ARG_PARAMETER);
           ACE_TRY_CHECK;
         }
diff --git a/TAO/tests/BiDirectional_NestedUpcall/README b/TAO/tests/BiDirectional_NestedUpcall/README
index 015f7693d5a..c40ad7424fd 100644
--- a/TAO/tests/BiDirectional_NestedUpcall/README
+++ b/TAO/tests/BiDirectional_NestedUpcall/README
@@ -1,9 +1,7 @@
 # $Id$
 
 This test is similar to the $TAO_ROOT/tests/BiDirectional test. The
-only difference is that, this test uses the select_mt reactor, which
-helps the server to make calls to client during the upcall.  Please
-see the README file in $TAO_ROOT/tests/BiDirectional to see why this
-would fail with a TP reactor. Also see bug id #575 in bugzilla to
-understand more about the TP reactors behavior.
+only difference being that the server as a part of the upcall sends a
+bunch of remote calls, to show some nested upcall behaviour. 
+