From dbed33a2ad65f5969fdd72bdb031a9504e8535bb Mon Sep 17 00:00:00 2001
From: TinaMengtingZhang <tina.mengting.zhang@huawei.com>
Date: Tue, 21 Sep 2021 17:53:01 -0400
Subject: [PATCH] Fix online debugger terminate with 1 issue: I3YAPB

---
 mindspore/ccsrc/debug/debugger/debugger.cc | 6 +++---
 mindspore/ccsrc/debug/debugger/debugger.h  | 2 +-
 mindspore/ccsrc/pipeline/jit/pipeline.cc   | 7 ++++++-
 mindspore/ccsrc/pipeline/jit/pipeline.h    | 6 +++++-
 4 files changed, 15 insertions(+), 6 deletions(-)

diff --git a/mindspore/ccsrc/debug/debugger/debugger.cc b/mindspore/ccsrc/debug/debugger/debugger.cc
index 031dc412ac5..7a341c55042 100644
--- a/mindspore/ccsrc/debug/debugger/debugger.cc
+++ b/mindspore/ccsrc/debug/debugger/debugger.cc
@@ -768,7 +768,7 @@ void Debugger::CommandLoop() {
         break;
       case DebuggerCommand::kExitCMD:
         MS_LOG(INFO) << "ExitCMD";
-        Exit();
+        Exit(true);
         // Used for debugger termination
         run = true;
         break;
@@ -1077,11 +1077,11 @@ std::list<TensorSummary> Debugger::LoadTensorsStat(const ProtoVector<TensorProto
   return tensor_summary_list;
 }
 
-void Debugger::Exit() {
+void Debugger::Exit(bool exit_success) {
   // debugger will notify main thread to exit because main thread can only exit at step boundary.
   MS_LOG(INFO) << "Exit Debugger";
   SetEnableHeartbeat(false);
-  pipeline::GraphExecutorPy::DebugTerminate(true);
+  pipeline::GraphExecutorPy::DebugTerminate(true, exit_success);
 }
 
 std::list<WatchpointHit> Debugger::CheckWatchpoints(const std::string &watchnode, const CNodePtr &kernel,
diff --git a/mindspore/ccsrc/debug/debugger/debugger.h b/mindspore/ccsrc/debug/debugger/debugger.h
index 1ac0c5e1a68..3eb5f940f02 100644
--- a/mindspore/ccsrc/debug/debugger/debugger.h
+++ b/mindspore/ccsrc/debug/debugger/debugger.h
@@ -223,7 +223,7 @@ class Debugger : public std::enable_shared_from_this<Debugger> {
   std::list<TensorSummary> LoadTensorsStat(const ProtoVector<TensorProto> &tensors) const;
 
   // terminate training process
-  void Exit();
+  void Exit(bool exit_success = false);
 
   // analyze tensors and check watchpoint conditions
   // return names of tensors and what condition they hit
diff --git a/mindspore/ccsrc/pipeline/jit/pipeline.cc b/mindspore/ccsrc/pipeline/jit/pipeline.cc
index 6fe4ea0d01a..6f891f3ab1e 100644
--- a/mindspore/ccsrc/pipeline/jit/pipeline.cc
+++ b/mindspore/ccsrc/pipeline/jit/pipeline.cc
@@ -107,6 +107,7 @@ GraphExecutorPyPtr GraphExecutorPy::executor_ = nullptr;
 std::mutex GraphExecutorPy::instance_lock_;
 #ifdef ENABLE_DEBUGGER
 bool GraphExecutorPy::debugger_terminate_ = false;
+bool GraphExecutorPy::exit_success_ = false;
 #endif
 
 std::unordered_map<abstract::AbstractBasePtrList, uint64_t, abstract::AbstractBasePtrListHasher,
@@ -1004,7 +1005,11 @@ void GraphExecutorPy::TerminateDebugger() {
   if (debugger_terminate_) {
     MS_LOG(INFO) << "Terminate debugger and clear resources!";
     ClearResAtexit();
-    exit(1);
+    if (exit_success_) {
+      exit(0);
+    } else {
+      exit(1);
+    }
   }
 }
 #endif
diff --git a/mindspore/ccsrc/pipeline/jit/pipeline.h b/mindspore/ccsrc/pipeline/jit/pipeline.h
index 4b4fb896e2a..1e655fde07d 100644
--- a/mindspore/ccsrc/pipeline/jit/pipeline.h
+++ b/mindspore/ccsrc/pipeline/jit/pipeline.h
@@ -109,7 +109,10 @@ class GraphExecutorPy : public std::enable_shared_from_this<GraphExecutorPy> {
   static void ClearRes();
 #ifdef ENABLE_DEBUGGER
   static bool GetDebugTerminate() { return debugger_terminate_; }
-  static void DebugTerminate(bool val) { debugger_terminate_ = val; }
+  static void DebugTerminate(bool val, bool exit_success) {
+    debugger_terminate_ = val;
+    exit_success_ = exit_success;
+  }
   void TerminateDebugger();
 #endif
 
@@ -131,6 +134,7 @@ class GraphExecutorPy : public std::enable_shared_from_this<GraphExecutorPy> {
   static std::mutex instance_lock_;
 #ifdef ENABLE_DEBUGGER
   static bool debugger_terminate_;
+  static bool exit_success_;
 #endif
   std::map<std::string, py::dict> stra_dict_;
   std::string phase_ = "";