Improve buffer flush error handling (#416)

* Update include/rocprofiler-sdk/fwd.h

- add ROCPROFILER_STATUS_ERROR_FINALIZED error code

* Update lib/rocprofiler-sdk/rocprofiler.cpp

- status string for ROCPROFILER_STATUS_ERROR_FINALIZED

* Update lib/rocprofiler-sdk/buffer.cpp

- return error code if buffer flush invoked after finalized
- fatal error if task group destroyed
- error message if task runs after finalized
- improve join of task group

* Update lib/rocprofiler-sdk/counters/tests/evaluate_ast_tests.cpp

- Update lambdas to return reference due to strange -Warray-bounds and -Wstringop-overflow warnings with g++ (Ubuntu 13.1.0-8ubuntu1~20.04.2) 13.1.0

[ROCm/rocprofiler-sdk commit: 3547a45c0c]

projects/rocprofiler-sdk/source/include/rocprofiler-sdk/fwd.h

@@ -79,6 +79,7 @@ typedef enum  // NOLINT(performance-enum-size)
     ROCPROFILER_STATUS_ERROR_INVALID_ARGUMENT,  ///< Function invoked with one or more invalid
                                                 ///< arguments
     ROCPROFILER_STATUS_ERROR_METRIC_NOT_VALID_FOR_AGENT,  ///< Invalid metric supplied to agent.
+    ROCPROFILER_STATUS_ERROR_FINALIZED,  ///< invalid because rocprofiler has been finalized
     ROCPROFILER_STATUS_LAST,
 } rocprofiler_status_t;
 

projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/buffer.cpp

@@ -135,7 +135,14 @@ allocate_buffer()
 rocprofiler_status_t
 flush(rocprofiler_buffer_id_t buffer_id, bool wait)
 {
-    if(registration::get_fini_status() > 0) return ROCPROFILER_STATUS_SUCCESS;
+    if(registration::get_fini_status() > 0)
+    {
+        LOG(ERROR) << "ignoring rocprofiler buffer flush (handle=" << buffer_id.handle
+                   << ") request after finalization";
+        return ROCPROFILER_STATUS_ERROR_FINALIZED;
+    }
+
+    if(registration::get_fini_status() < 0 && !wait) wait = true;
 
     auto offset = get_buffer_offset();
 
@@ -146,7 +153,11 @@ flush(rocprofiler_buffer_id_t buffer_id, bool wait)
     auto* task_group =
         internal_threading::get_task_group(rocprofiler_callback_thread_t{buff->task_group_id});
 
-    if(wait && task_group) task_group->wait();
+    LOG_IF(FATAL, !task_group)
+        << "buffer (" << buffer_id.handle
+        << ") flush request received after the task group for handling request was destroyed";
+
+    if(wait) task_group->wait();
 
     // buffer is currently being flushed or destroyed
     if(buff->syncer.test_and_set())
@@ -162,6 +173,9 @@ flush(rocprofiler_buffer_id_t buffer_id, bool wait)
     auto idx = buff->buffer_idx++;
 
     auto _task = [buffer_id, idx, offset]() {
+        LOG_IF(ERROR, registration::get_fini_status() > 0)
+            << "executing buffer (" << buffer_id.handle << ") flush task finalization!";
+
         auto& buff_v          = CHECK_NOTNULL(get_buffers())->at(buffer_id.handle - offset);
         auto& buff_internal_v = buff_v->get_internal_buffer(idx);
 
@@ -197,14 +211,11 @@ flush(rocprofiler_buffer_id_t buffer_id, bool wait)
         buff_v->syncer.clear();
     };
 
-    if(task_group)
+    task_group->exec(_task);
+    if(wait)
     {
-        task_group->exec(_task);
-        if(wait) task_group->join();
-    }
-    else
-    {
-        _task();
+        while(task_group->size() > 0)
+            task_group->join();
     }
 
     return ROCPROFILER_STATUS_SUCCESS;

projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/counters/tests/evaluate_ast_test.cpp

@@ -591,7 +591,8 @@ TEST(evaluate_ast, counter_reduction_sum)
         return base_id;
     };
 
-    auto sum_vec = [](auto& a) {
+    auto sum_vec = [](auto& a) -> auto&
+    {
         for(size_t i = 1; i < a.size(); i++)
         {
             a[0].counter_value += a[i].counter_value;
@@ -671,7 +672,8 @@ TEST(evaluate_ast, counter_reduction_min)
         return base_id;
     };
 
-    auto min_vec = [](auto& a) {
+    auto min_vec = [](auto& a) -> auto&
+    {
         a[0].counter_value = std::min_element(a.begin(), a.end(), [](const auto& b, const auto& c) {
                                  return b.counter_value < c.counter_value;
                              })->counter_value;
@@ -750,7 +752,8 @@ TEST(evaluate_ast, counter_reduction_max)
         return base_id;
     };
 
-    auto max_vec = [](auto& a) {
+    auto max_vec = [](auto& a) -> auto&
+    {
         a[0].counter_value = std::max_element(a.begin(), a.end(), [](const auto& b, const auto& c) {
                                  return b.counter_value < c.counter_value;
                              })->counter_value;
@@ -829,7 +832,8 @@ TEST(evaluate_ast, counter_reduction_avg)
         return base_id;
     };
 
-    auto avg_vec = [](auto& a) {
+    auto avg_vec = [](auto& a) -> auto&
+    {
         for(size_t i = 1; i < a.size(); i++)
         {
             a[0].counter_value += a[i].counter_value;
@@ -910,7 +914,8 @@ TEST(evaluate_ast, evaluate_mixed_counters)
         return base_id;
     };
 
-    auto sum_vec = [](auto& a) {
+    auto sum_vec = [](auto& a) -> auto&
+    {
         for(size_t i = 1; i < a.size(); i++)
         {
             a[0].counter_value += a[i].counter_value;

projects/rocprofiler-sdk/source/lib/rocprofiler-sdk/rocprofiler.cpp

@@ -73,6 +73,8 @@ ROCPROFILER_STATUS_STRING(ROCPROFILER_STATUS_ERROR_INVALID_ARGUMENT,
                           "Function invoked with one or more invalid arguments")
 ROCPROFILER_STATUS_STRING(ROCPROFILER_STATUS_ERROR_METRIC_NOT_VALID_FOR_AGENT,
                           "Metric is not valid for the agent")
+ROCPROFILER_STATUS_STRING(ROCPROFILER_STATUS_ERROR_FINALIZED,
+                          "Invalid request because rocprofiler has finalized")
 
 template <size_t Idx, size_t... Tail>
 const char*