Fix condition for stopping after N failures

Hlt/HLTScheduler/src/HLTControlFlowMgr.cpp

@@ -208,8 +208,7 @@ private:
   // the tbb "threadpool"
   std::unique_ptr<tbb::task_arena> m_taskArena;
 
-  /// atomic count of the number of finished events
-  mutable std::atomic<uint32_t> m_finishedEvt{0};
+  mutable Gaudi::Accumulators::Counter<> m_finishedEvt{this, "Processed events"};
   /// condition variable to wake up main thread when we need to create a new event
   mutable std::condition_variable m_createEventCond;
   /// mutex assoiciated with m_createEventCond condition variable
@@ -866,7 +865,7 @@ StatusCode HLTControlFlowMgr::nextEvent( int maxevt ) {
 
   if ( !endTimeAndTicks ) {
     endTimeAndTicks.emplace( Clock::now(), getTotalWorkerTicks() );
-    m_stopTimeAtEvt = m_finishedEvt;
+    m_stopTimeAtEvt = m_finishedEvt.value();
     m_alloc_tracker->endTracking();
   }
 
@@ -876,7 +875,7 @@ StatusCode HLTControlFlowMgr::nextEvent( int maxevt ) {
   releaseEvtSelContext().ignore();
 
   if ( !startTimeAndTicks ) {
-    info() << "---> Loop over " << m_finishedEvt << " Events Finished - "
+    info() << "---> Loop over " << m_finishedEvt.value() << " Events Finished - "
            << " WSS " << System::mappedMemory( System::MemoryUnit::kByte ) * 1. / 1024. << ", timing failed.."
            << endmsg;
   } else {
@@ -885,7 +884,7 @@ StatusCode HLTControlFlowMgr::nextEvent( int maxevt ) {
             .count();
     const auto timeDiff   = m_stopTimeAtEvt.value() - m_startTimeAtEvt.value();
     const auto evtsPerSec = ( totalTime > 0 ? timeDiff * 1. / totalTime * 1e3 : 0 );
-    info() << "---> Loop over " << m_finishedEvt << " Events Finished - "
+    info() << "---> Loop over " << m_finishedEvt.value() << " Events Finished - "
            << " WSS " << System::mappedMemory( System::MemoryUnit::kByte ) * 1. / 1024. << ", timed " << timeDiff
            << " Events: " << totalTime << " ms"
            << ", Evts/s = " << evtsPerSec << endmsg;
@@ -926,7 +925,7 @@ StatusCode HLTControlFlowMgr::eventFailed( EventContext const& eventContext ) co
   std::ostringstream ost;
   m_algExecStateSvc->dump( ost, eventContext );
   info() << "Dumping Alg Exec State for slot " << eventContext.slot() << ":\n" << ost.str() << endmsg;
-  if ( ++m_failed_evts_detected > m_stopAfterNFailures ) { m_shutdown_now = true; }
+  if ( ++m_failed_evts_detected >= m_stopAfterNFailures ) { m_shutdown_now = true; }
   return StatusCode::FAILURE;
 }
 

Hlt/HLTScheduler/tests/qmtest/scheduler_testEarlyAbort.qmt

@@ -18,13 +18,11 @@
 <argument name="exit_code"><integer>3</integer></argument>
 <argument name="validator"><text>
 
-def test(stdout):
-  expected_4 = "LAZY_AND: moore  #=4       Sum=4           Eff=|( 100.0000 +- 0.00000 )%|"
-  expected_5 = "LAZY_AND: moore  #=5       Sum=5           Eff=|( 100.0000 +- 0.00000 )%|"
-  expected_6 = "LAZY_AND: moore  #=6       Sum=6           Eff=|( 100.0000 +- 0.00000 )%|"
-  if all(e not in stdout for e in [expected_4, expected_5, expected_6]):
-    causes.append("early abortion didn't work correctly")
-test(stdout)
+from GaudiConf.QMTest.LHCbTest import extract_counters
+counters = extract_counters(stdout)
+n_processed = int(counters["HLTControlFlowMgr"]["Processed events"][0])
+if n_processed > 6:
+    causes.append("early abort didn't work correctly")
 
 </text></argument>
 </extension>

Hlt/HLTScheduler/tests/refs/event_selector.ref

@@ -124,3 +124,6 @@ LAZY_AND: top                 #=77      Sum=77          Eff=|( 100.0000 +- 0.000
 HLTControlFlowMgr    INFO Histograms converted successfully according to request.
 ApplicationMgr       INFO Application Manager Finalized successfully
 ApplicationMgr       INFO Application Manager Terminated successfully
+HLTControlFlowMgr    INFO Number of counters : 1
+ |    Counter                                      |     #     |    sum     | mean/eff^* | rms/err^*  |     min     |     max     |
+ | "Processed events"                              |        77 |

Hlt/HLTScheduler/tests/refs/write.ref

@@ -88,3 +88,6 @@ TFile**		write.legacy_histos.root	GAUDI Histograms
   KEY: TH2D	2;1	2D fix binning
   KEY: TH3D	3;1	3D fix binning
   KEY: TDirectoryFile	Prof;1	Prof
+HLTControlFlowMgr    INFO Number of counters : 1
+ |    Counter                                      |     #     |    sum     | mean/eff^* | rms/err^*  |     min     |     max     |
+ | "Processed events"                              |       100 |