Formatting

qa/L0_shared_memory/shared_memory_test.py

@@ -123,7 +123,7 @@ def tearDown(self):
         util.kill_server(self._server_process)
         # Restore stdout / stderr so we can print to console and see server
         # output in CI even after logs expire. Print test result to client
-        # before doing so.
+        # before doing so for legibility.
         if not self._test_passed:
             print("*\n*\n*\nTest Failed\n*\n*\n*\n")
         util.stream_to_console(self._original_stdout, self._original_stderr)
@@ -294,9 +294,9 @@ def test_valid_create_set_register(self, protocol):
         )
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 1)
+            self.assertEqual(len(shm_status), 1)
         else:
-            self.assertTrue(len(shm_status.regions) == 1)
+            self.assertEqual(len(shm_status.regions), 1)
         shm.destroy_shared_memory_region(shm_op0_handle)
         self._test_passed = True
 
@@ -316,9 +316,9 @@ def test_unregister_before_register(self, protocol):
         self._triton_client.unregister_system_shared_memory("dummy_data")
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 0)
+            self.assertEqual(len(shm_status), 0)
         else:
-            self.assertTrue(len(shm_status.regions) == 0)
+            self.assertEqual(len(shm_status.regions), 0)
         shm.destroy_shared_memory_region(shm_op0_handle)
         self._test_passed = True
 
@@ -340,9 +340,9 @@ def test_unregister_after_register(self, protocol):
         self._triton_client.unregister_system_shared_memory("dummy_data")
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 0)
+            self.assertEqual(len(shm_status), 0)
         else:
-            self.assertTrue(len(shm_status.regions) == 0)
+            self.assertEqual(len(shm_status.regions), 0)
         shm.destroy_shared_memory_region(shm_op0_handle)
         self._test_passed = True
 
@@ -371,9 +371,9 @@ def test_reregister_after_register(self, protocol):
             )
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 1)
+            self.assertEqual(len(shm_status), 1)
         else:
-            self.assertTrue(len(shm_status.regions) == 1)
+            self.assertEqual(len(shm_status.regions), 1)
         shm.destroy_shared_memory_region(shm_op0_handle)
         self._test_passed = True
 
@@ -400,9 +400,9 @@ def test_unregister_after_inference(self, protocol):
         self._triton_client.unregister_system_shared_memory("output0_data")
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 3)
+            self.assertEqual(len(shm_status), 3)
         else:
-            self.assertTrue(len(shm_status.regions) == 3)
+            self.assertEqual(len(shm_status.regions), 3)
         self._cleanup_server(shm_handles)
         self._test_passed = True
 
@@ -433,9 +433,9 @@ def test_register_after_inference(self, protocol):
         )
         shm_status = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(shm_status) == 5)
+            self.assertEqual(len(shm_status), 5)
         else:
-            self.assertTrue(len(shm_status.regions) == 5)
+            self.assertEqual(len(shm_status.regions), 5)
         shm_handles.append(shm_ip2_handle)
         self._cleanup_server(shm_handles)
         self._test_passed = True
@@ -502,15 +502,15 @@ def test_unregisterall(self, protocol):
 
         status_before = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(status_before) == 4)
+            self.assertEqual(len(status_before), 4)
         else:
-            self.assertTrue(len(status_before.regions) == 4)
+            self.assertEqual(len(status_before.regions), 4)
         self._triton_client.unregister_system_shared_memory()
         status_after = self._triton_client.get_system_shared_memory_status()
         if self._protocol == "http":
-            self.assertTrue(len(status_after) == 0)
+            self.assertEqual(len(status_after), 0)
         else:
-            self.assertTrue(len(status_after.regions) == 0)
+            self.assertEqual(len(status_after.regions), 0)
         self._cleanup_server(shm_handles)
         self._test_passed = True
 

qa/L0_shared_memory/test.sh

@@ -29,18 +29,10 @@ CLIENT_LOG="./client.log"
 SHM_TEST=shared_memory_test.py
 TEST_RESULT_FILE='test_results.txt'
 
-if [[ ${TEST_WINDOWS} == 1 ]]; then
-    TRITON_DIR=${TRITON_DIR:=c:/tritonserver}
-    SERVER=${SERVER:=c:/tritonserver/bin/tritonserver.exe}
-    BACKEND_DIR=${BACKEND_DIR:=c:/tritonserver/backends}
-    MODELDIR=${MODELDIR:=c:/}
-else
-    # Configure to support test on jetson as well
-    TRITON_DIR=${TRITON_DIR:="/opt/tritonserver"}
-    SERVER=${TRITON_DIR}/bin/tritonserver
-    BACKEND_DIR=${TRITON_DIR}/backends
-    MODELDIR=${MODELDIR:=`pwd`}
-fi
+# Configure to support test on jetson as well
+TRITON_DIR=${TRITON_DIR:="/opt/tritonserver"}
+SERVER=${TRITON_DIR}/bin/tritonserver
+BACKEND_DIR=${TRITON_DIR}/backends
 SERVER_ARGS_EXTRA="--backend-directory=${BACKEND_DIR}"
 source ../common/util.sh
 
@@ -49,7 +41,6 @@ rm -fr *.log
 
 for i in \
         test_invalid_create_shm \
-        test_invalid_registration \
         test_valid_create_set_register \
         test_unregister_before_register \
         test_unregister_after_register \
@@ -61,7 +52,7 @@ for i in \
         test_unregisterall \
         test_infer_offset_out_of_bound; do
     for client_type in http grpc; do
-        SERVER_ARGS="--model-repository=${MODELDIR} --log-verbose=1 ${SERVER_ARGS_EXTRA}"
+        SERVER_ARGS="--model-repository=`pwd`/models --log-verbose=1 ${SERVER_ARGS_EXTRA}"
         SERVER_LOG="./$i.$client_type.server.log"
         run_server
         if [ "$SERVER_PID" == "0" ]; then
@@ -78,10 +69,18 @@ for i in \
         if [ $? -ne 0 ]; then
             echo -e "\n***\n*** Test Failed\n***"
             RET=1
+        else
+            check_test_results $TEST_RESULT_FILE 1
+            if [ $? -ne 0 ]; then
+                cat $CLIENT_LOG
+                echo -e "\n***\n*** Test Result Verification Failed\n***"
+                RET=1
+            fi
         fi
         set -e
 
-        kill_server
+        kill $SERVER_PID
+        wait $SERVER_PID
     done
 done
 

qa/common/util.py

@@ -1,23 +1,29 @@
-from pathlib import Path
+import shutil
+import signal
 import subprocess
-import time
 import sys
-import signal
-import shutil
+import time
+from pathlib import Path
 
 from tritonclient.utils import InferenceServerException
 
+
 def run_server(server_executable: str, launch_command: str, log_file):
     if not Path(server_executable).is_file():
         raise Exception(f"{server_executable} does not exist")
     print(f"=== Running {launch_command}")
     if sys.platform == "win32":
-        server = subprocess.Popen(launch_command, text=True, stdout=log_file, stderr=log_file)
+        server = subprocess.Popen(
+            launch_command, text=True, stdout=log_file, stderr=log_file
+        )
     else:
-        server = subprocess.Popen(launch_command.split(), text=True, stdout=log_file, stderr=log_file) 
+        server = subprocess.Popen(
+            launch_command.split(), text=True, stdout=log_file, stderr=log_file
+        )
     time.sleep(3)
     return server
 
+
 def wait_for_server_ready(server_process, triton_client, timeout):
     start = time.time()
     while time.time() - start < timeout:
@@ -34,36 +40,47 @@ def wait_for_server_ready(server_process, triton_client, timeout):
                 print("=== Server is ready", flush=True)
                 return True
         except InferenceServerException:
-            pass # Host not ready
+            pass  # Host not ready
     raise Exception(f"=== Timeout {timeout} secs. Server not ready. ===")
 
+
 def kill_server(server_process):
     # Only kill process if it's stil running
     if server_process and not server_process.poll():
         print("*\n*\n*\nTerminating server\n*\n*\n*\n")
         # Terminate gracefully for Linux
         if sys.platform == "win32":
-            server_process.kill()  
+            server_process.kill()
         else:
             server_process.send_signal(signal.SIGINT)
         server_process.wait()
 
+
 def stream_to_log(client_log):
     original_stdout = sys.stdout
     original_sterr = sys.stderr
     sys.stdout = sys.stderr = client_log
     return original_stdout, original_sterr
 
+
 def stream_to_console(original_stdout, original_sterr):
     sys.stdout = original_stdout
     sys.stderr = original_sterr
 
+
 def remove_model_dir(model_dir_path: Path):
     if not model_dir_path.is_dir():
         return
     shutil.rmtree(model_dir_path)
 
-def create_model_dir(model_dir_path: Path, model_name: str, model_version: int, model_source_path: Path, model_config_path: Path):
+
+def create_model_dir(
+    model_dir_path: Path,
+    model_name: str,
+    model_version: int,
+    model_source_path: Path,
+    model_config_path: Path,
+):
     remove_model_dir(model_dir_path)
     model_dir_path = model_dir_path / model_name / str(model_version)
     model_dir_path.mkdir(parents=True)
@@ -72,11 +89,15 @@ def create_model_dir(model_dir_path: Path, model_name: str, model_version: int,
     shutil.copy(model_source_path, model_dir_path)
     shutil.copy(model_config_path, model_dir_path.parent)
 
-def replace_config_attribute(model_config_path: Path, current_attribute: str, desired_attribute: str):
+
+def replace_config_attribute(
+    model_config_path: Path, current_attribute: str, desired_attribute: str
+):
     original_config = model_config_path.read_text()
     new_config = original_config.replace(current_attribute, desired_attribute)
     model_config_path.write_text(new_config)
 
+
 def add_config_attribute(model_config_path: Path, new_attribute: str):
     with model_config_path.open("a") as f:
         f.write(new_attribute)

src/shared_memory_manager.cc

@@ -111,10 +111,11 @@ MapSharedMemory(
   if (*mapped_addr == NULL) {
     CloseSharedMemoryRegion(shm_handle);
     return TRITONSERVER_ErrorNew(
-        TRITONSERVER_ERROR_INTERNAL, std::string(
-                                         "unable to process address space: " +
-                                         std::to_string(GetLastError()))
-                                         .c_str());
+        TRITONSERVER_ERROR_INTERNAL,
+        std::string(
+            "unable to process address space, error code: " +
+            std::to_string(GetLastError()))
+            .c_str());
   }
   return nullptr;
 }
@@ -191,7 +192,6 @@ SharedMemoryManager::RegisterSystemSharedMemory(
   // Map and then close the shared memory handle
   TRITONSERVER_Error* err_map =
       MapSharedMemory(shm_handle, offset, byte_size, &mapped_addr);
-  // TODO: Test if we can close windows handles without invalidating mapping
   TRITONSERVER_Error* err_close = CloseSharedMemoryRegion(shm_handle);
   if (err_map != nullptr) {
     return TRITONSERVER_ErrorNew(
@@ -207,7 +207,7 @@ SharedMemoryManager::RegisterSystemSharedMemory(
         TRITONSERVER_ERROR_INVALID_ARG,
         std::string(
             "failed to register shared memory region '" + name +
-            "': " + std::to_string(GetLastError()))
+            "' with error code: " + std::to_string(GetLastError()))
             .c_str());
   }