ChangeLogTag:Wed May 23 18:22:41 2001 Steve Huston <shuston@riverace.com>

4 files changed, 49 insertions, 4 deletions

diff --git a/ChangeLog b/ChangeLog
index 7c4fd98e67f..aeb29982f2e 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,12 @@
+Wed May 23 18:22:41 2001  Steve Huston  <shuston@riverace.com>
+
+        * ace/Reactor.cpp (run_reactor_event_loop (ACE_Time_Value&, eh)):
+          It is possible for rounding/conversion fudges in calculating
+          the WFMO/select wait time to cause the wait to time out, but the
+          timer queue be not quite ready to expire the next timer. In this
+          case, the ACE_Time_Value won't have been reduced to 0, so the
+          handle_events call is repeated. Fixes Bugzilla # 153.
+
 Wed May 23 15:55:21 2001  Steve Huston  <shuston@riverace.com>
 
         * ace/Containers_T.h: Fixed comment directing reader to include
diff --git a/ChangeLogs/ChangeLog-02a b/ChangeLogs/ChangeLog-02a
index 7c4fd98e67f..aeb29982f2e 100644
--- a/ChangeLogs/ChangeLog-02a
+++ b/ChangeLogs/ChangeLog-02a
@@ -1,3 +1,12 @@
+Wed May 23 18:22:41 2001  Steve Huston  <shuston@riverace.com>
+
+        * ace/Reactor.cpp (run_reactor_event_loop (ACE_Time_Value&, eh)):
+          It is possible for rounding/conversion fudges in calculating
+          the WFMO/select wait time to cause the wait to time out, but the
+          timer queue be not quite ready to expire the next timer. In this
+          case, the ACE_Time_Value won't have been reduced to 0, so the
+          handle_events call is repeated. Fixes Bugzilla # 153.
+
 Wed May 23 15:55:21 2001  Steve Huston  <shuston@riverace.com>
 
         * ace/Containers_T.h: Fixed comment directing reader to include
diff --git a/ChangeLogs/ChangeLog-03a b/ChangeLogs/ChangeLog-03a
index 7c4fd98e67f..aeb29982f2e 100644
--- a/ChangeLogs/ChangeLog-03a
+++ b/ChangeLogs/ChangeLog-03a
@@ -1,3 +1,12 @@
+Wed May 23 18:22:41 2001  Steve Huston  <shuston@riverace.com>
+
+        * ace/Reactor.cpp (run_reactor_event_loop (ACE_Time_Value&, eh)):
+          It is possible for rounding/conversion fudges in calculating
+          the WFMO/select wait time to cause the wait to time out, but the
+          timer queue be not quite ready to expire the next timer. In this
+          case, the ACE_Time_Value won't have been reduced to 0, so the
+          handle_events call is repeated. Fixes Bugzilla # 153.
+
 Wed May 23 15:55:21 2001  Steve Huston  <shuston@riverace.com>
 
         * ace/Containers_T.h: Fixed comment directing reader to include
diff --git a/ace/Reactor.cpp b/ace/Reactor.cpp
index bf81a924f52..f8b65761cc9 100644
--- a/ace/Reactor.cpp
+++ b/ace/Reactor.cpp
@@ -191,10 +191,28 @@ ACE_Reactor::run_reactor_event_loop (ACE_Time_Value &tv,
 
       if (eh != 0 && (*eh)(0))
         continue;
-      else if (result == -1 && this->implementation_->deactivated ())
-        return 0;
-      else if (result <= 0)
-        return result;
+      else if (result == -1)
+        {
+          if (this->implementation_->deactivated ())
+            result = 0;
+          return result;
+        }
+      else if (result == 0)
+        {
+          // handle_events timed out without dispatching anything.
+          // Because of rounding and conversion errors and such, it could
+          // be that the wait loop (WFMO, select, etc.) timed out, but
+          // the timer queue said it wasn't quite ready to expire a
+          // timer. In this case, the ACE_Time_Value we passed into
+          // handle_events won't have quite been reduced to 0, and we
+          // need to go around again. If we are all the way to 0, just
+          // return, as the entire time the caller wanted to wait has been
+          // used up.
+          if (tv.usec () > 0)
+            continue;
+          return 0;
+        }
+      // Else there were some events dispatched; go around again
     }
 
   ACE_NOTREACHED (return 0;)