[omniORB] [DIFF] SIGSEGV after CancelRequest

[Andrei Zaparii]

This 'cvs diff' is meant to address segmentation fault in 
omniCallHandle::upcall that occurs after CancelRequest message processed.
While this changes do not lead to memory leaking i'm not sure whether 
other resource management is intact. So please review this diff.
-- 
Regards,
Andrei Zaparii
-------------- next part --------------
Index: GIOP_S.cc
===================================================================
RCS file: /cvsroot/omniorb/omni/src/lib/omniORB/orbcore/Attic/GIOP_S.cc,v
retrieving revision 1.1.4.25
diff -u -r1.1.4.25 GIOP_S.cc
--- GIOP_S.cc	16 Jul 2003 14:22:38 -0000	1.1.4.25
+++ GIOP_S.cc	8 May 2004 11:34:34 -0000
@@ -553,7 +553,11 @@
   // an upcall by another thread. Therefore it is not possible to
   // cancel a request that has already been in progress. 
   omniORB::logs(5, "Received and ignored a CancelRequest message.");
-  pd_state = ReplyCompleted;
+  // assigning to state anything but WaitingForReply will brake assertion
+  // in the beginig of SendReply
+  pd_state = WaitingForReply;
+  // We should not send reply to client
+  response_expected(0);
   return 1;
 }
 
Index: giopStrand.cc
===================================================================
RCS file: /cvsroot/omniorb/omni/src/lib/omniORB/orbcore/Attic/giopStrand.cc,v
retrieving revision 1.1.4.20
diff -u -r1.1.4.20 giopStrand.cc
--- giopStrand.cc	30 Mar 2004 14:26:36 -0000	1.1.4.20
+++ giopStrand.cc	8 May 2004 11:35:43 -0000
@@ -537,7 +537,15 @@
     giop_s->giopStreamList::insert(servers);
   }
 
-  if (remove) delete giop_s;
+  // if we destroy GIOP_S instance while working in it -> servant
+  // will dump core (eg. when CancelRequest received)
+  // and we won't be able startIdleCounter if we don't destroy giop_s
+  if (remove && giop_s->state() != IOP_S::WaitingForReply) 
+  {
+    delete giop_s;
+  }
+  else
+    restart_idle = 0;
 
   if (restart_idle && !biDir) {
     CORBA::Boolean success = startIdleCounter();