Merge pull request #10832 from typhoonzero/fix_server_port_file_conflict

fix dist test

paddle/fluid/operators/listen_and_serv_op.cc

@@ -12,6 +12,7 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
+#include <stdio.h>  // for removing the port file
 #include <fstream>
 #include <ostream>
 #include <thread>  // NOLINT
@@ -77,12 +78,14 @@ ListenAndServOp::ListenAndServOp(const std::string &type,
 void ListenAndServOp::Stop() {
   rpc_service_->Push(LISTEN_TERMINATE_MESSAGE);
   server_thread_->join();
+  auto file_path = string::Sprintf("/tmp/paddle.%d.port", ::getpid());
+  remove(file_path.c_str());
 }
 
-void ListenAndServOp::SavePort(const std::string &file_path) const {
+void ListenAndServOp::SavePort() const {
   // NOTE: default write file to /tmp/paddle.selected_port
   selected_port_ = rpc_service_->GetSelectedPort();
-
+  auto file_path = string::Sprintf("/tmp/paddle.%d.port", ::getpid());
   std::ofstream port_file;
   port_file.open(file_path);
   port_file << selected_port_.load();
@@ -331,7 +334,7 @@ void ListenAndServOp::RunImpl(const framework::Scope &scope,
   // Write to a file of server selected port for python use.
   std::string file_path = string::Sprintf("/tmp/paddle.%d.selected_port",
                                           static_cast<int>(::getpid()));
-  SavePort(file_path);
+  SavePort();
   if (sync_mode) {
     RunSyncLoop(&executor, program, &recv_scope, prefetch_block);
   } else {

paddle/fluid/operators/listen_and_serv_op.h

@@ -48,8 +48,7 @@ class ListenAndServOp : public framework::OperatorBase {
   void RunAsyncLoop(framework::Executor* executor,
                     framework::ProgramDesc* program) const;
 
-  void SavePort(
-      const std::string& file_path = "/tmp/paddle.selected_port") const;
+  void SavePort() const;
 
   void WaitServerReady();
 

python/paddle/fluid/tests/unittests/test_dist_train.py

@@ -34,7 +34,7 @@ class TestSendOp(unittest.TestCase):
         p.start()
 
         time.sleep(10)
-        with open("/tmp/paddle.%d.selected_port" % p.pid, "r") as fn:
+        with open("/tmp/paddle.%d.port" % p.pid, "r") as fn:
             selected_port = int(fn.readlines()[0])
         self.init_client(place, selected_port)