summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorcoryan <coryan@ae88bc3d-4319-0410-8dbf-d08b4c9d3795>1997-11-23 02:19:01 +0000
committercoryan <coryan@ae88bc3d-4319-0410-8dbf-d08b4c9d3795>1997-11-23 02:19:01 +0000
commit910891301fccb54bc5825e28f1bdf3c0693477d2 (patch)
tree84055d5b7cb97e30b20dc6d7eb2ae22be180294c
parent1b9bf25ac6d0a926a880b6d4dbf65eafeb80e880 (diff)
downloadATCD-910891301fccb54bc5825e28f1bdf3c0693477d2.tar.gz
ChangeLogTag:Sat Nov 22 20:00:38 1997 Carlos O'Ryan <coryan@macarena.cs.wustl.edu>
-rw-r--r--ChangeLog-97b75
1 files changed, 74 insertions, 1 deletions
diff --git a/ChangeLog-97b b/ChangeLog-97b
index 4eec487f198..18710b2d484 100644
--- a/ChangeLog-97b
+++ b/ChangeLog-97b
@@ -1,3 +1,47 @@
+Sat Nov 22 20:00:38 1997 Carlos O'Ryan <coryan@macarena.cs.wustl.edu>
+
+ * ace/Select_Reactor.cpp:
+ The comparison for increasing the handle_limit was wrong, it
+ will only decrease it, not increase it, thanks to Arturo Montes
+ <mitosys@colomsat.net.co> for detecting this.
+
+ * ace/ACE.cpp:
+ Added support for RLIM_INFINITY in ACE::set_handle_limit().
+ Thanks to Arturo Montes <mitosys@colomsat.net.co> for suggesting
+ this.
+
+Sat Nov 22 19:16:35 1997 <irfan@TWOSTEP>
+
+ * tests/Process_Strategy_Test.cpp:
+
+ main: Removed signal handling code. It is unnecessary since we
+ do not use kill() anymore.
+
+ Process_Strategy: This subclass of ACE_Process_Strategy
+ overwrites the process creation method to include connection
+ counting.
+
+ connection_completed: Made sure that the Reactor is woken up
+ after all the connections have completed.
+
+ * ace/Priority_Reactor.cpp (init_bucket): Removed extra code.
+
+ * ace/Acceptor.cpp (activate_svc_handler): open svc_handler only
+ if result is 0. Also removed extra tracing code.
+
+ * ace/Handle_Set.i (reset): Purify is reporting a UMR
+ (uninitialized memory copy) when copying a Handle_Set. This
+ comes from the mask fd_set not being memset to 0 by FD_ZERO.
+ This is therefore a harmless error.
+
+ * tests/Priority_Reactor_Test.cpp (main): The ordering of
+ auto_ptrs is very subtle.... "auto_ptr<ACE_Reactor> reactor"
+ should come before "auto_ptr<ACE_Select_Reactor> impl", i.e.,
+ the implementation should be destroyed *before* the forwarding
+ class. This is apparent when the implementation class is
+ removing registered event handlers that hold a pointer to the
+ forwarding class.
+
Sat Nov 22 16:04:07 1997 Carlos O'Ryan <coryan@macarena.cs.wustl.edu>
* ace/Message_Queue.cpp:
@@ -100,6 +144,35 @@ Fri Nov 21 12:18:28 1997 Douglas C. Schmidt <schmidt@tango.cs.wustl.edu>
Fri Nov 21 02:23:47 1997 <irfan@TWOSTEP>
+ * tests/Process_Strategy_Test.cpp:
+
+ - The event loop execution was based on a fragile timeout
+ mechanism. This was changed to a more robust mechanism that
+ counts the connections.
+
+ - The unlinking of the file was moved from the server to the
+ client. This is necessary since there maybe multiple server
+ processes that cannot really coordinate when the file should
+ be removed.
+
+ * ace/Strategies_T.cpp (ACE_Concurrency_Strategy):
+ ACE_Concurrency_Strategy and all three of the subclasses
+ (ACE_Reactive_Strategy, ACE_Process_Strategy, and
+ ACE_Thread_Strategy) were messed up. Only ACE_Reactive_Strategy
+ was (trying) to set the correct blocking mode of the socket.
+ Therefore, moved the socket mode setting code to the base class,
+ ACE_Concurrency_Strategy, so that all the classes can share it.
+
+ This did create one problem though. If the implementation of
+ the reactor uses event associations (i.e., WFMO_Reactor), we
+ cannot try to set the socket mode since the registration of the
+ svc_handler caused an event association to be created for the
+ socket and by default (thanks Microslush) the socket is set to
+ non-blocking mode. If the user is excepting to have a socket in
+ blocking mode, he is in for a surprise. Maybe we should return
+ a failure if the user is expecting a socket in blocking mode.
+ This is probably too harsh for now.
+
* tests/Future_Test.cpp (open): Added THR_DETACHED since we don't
wait for this thread.
@@ -114,7 +187,7 @@ Fri Nov 21 02:23:47 1997 <irfan@TWOSTEP>
wakeup_all_threads_handler_. Unfortunately, the order of the
two registrations is very important since the optimization for
owner thread takes advantage of this ordering. Therefore,
- revert the order of register to notify_handler_ followed by
+ revert the order of registertion to notify_handler_ followed by
wakeup_all_threads_handler_.
* tests/version_tests/run_all_tests.bat: Added new bat file to