Enhance execution error info (#18482)

* enhance execution error info
test=develop

paddle/fluid/platform/init.cc

@@ -13,6 +13,8 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 #include <string.h>  // for strdup
 #include <algorithm>
+#include <fstream>
+#include <iostream>
 #include <memory>
 #include <set>
 #include <stdexcept>
@@ -30,7 +32,6 @@ limitations under the License. */
 #include "paddle/fluid/platform/init.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/string/piece.h"
-
 #if defined(PADDLE_WITH_DGC)
 #include "dgc/dgc.h"
 #endif
@@ -202,10 +203,27 @@ void InitDevices(bool init_p2p, const std::vector<int> devices) {
 #endif
 }
 
+#ifndef _WIN32
+static void SignalHandle(const char *data, int size) {
+  auto file_path = string::Sprintf("/tmp/paddle.%d.dump_info", ::getpid());
+  try {
+    std::ofstream dump_info;
+    dump_info.open(file_path, std::ios::app);
+    dump_info << std::string(data, size);
+    dump_info.close();
+  } catch (...) {
+  }
+}
+#endif
+
 void InitGLOG(const std::string &prog_name) {
   // glog will not hold the ARGV[0] inside.
   // Use strdup to alloc a new string.
   google::InitGoogleLogging(strdup(prog_name.c_str()));
+#ifndef _WIN32
+  google::InstallFailureSignalHandler();
+  google::InstallFailureWriter(&SignalHandle);
+#endif
 }
 
 #if defined(PADDLE_WITH_DGC)

python/paddle/fluid/executor.py

@@ -17,6 +17,7 @@ from __future__ import print_function
 import logging
 import os
 import multiprocessing
+import sys
 import numpy as np
 from .wrapped_decorator import signature_safe_contextmanager
 import six
@@ -627,6 +628,23 @@ class Executor(object):
 
             list(numpy.array): fetch result according to fetch_list.
         """
+        try:
+            return self._run_impl(
+                program=program,
+                feed=feed,
+                fetch_list=fetch_list,
+                feed_var_name=feed_var_name,
+                fetch_var_name=fetch_var_name,
+                scope=scope,
+                return_numpy=return_numpy,
+                use_program_cache=use_program_cache)
+        except Exception as e:
+            if not isinstance(e, core.EOFException):
+                print("An exception was thrown!\n {}".format(str(e)))
+            raise e
+
+    def _run_impl(self, program, feed, fetch_list, feed_var_name,
+                  fetch_var_name, scope, return_numpy, use_program_cache):
 
         if self._closed:
             raise RuntimeError("Attempted to use a closed Executor")
@@ -639,7 +657,7 @@ class Executor(object):
         compiled = isinstance(program, compiler.CompiledProgram)
         # For backward compatibility, run directly.
         if not compiled:
-            return self._run(
+            return self._run_program(
                 program,
                 self._default_executor,
                 feed=feed,
@@ -672,7 +690,7 @@ class Executor(object):
             # TODO(panyx0718): executor should be able to run graph.
             assert program._program, "CompiledProgram is compiled from graph, can only run with_data_parallel."
             # use_program_cache is not valid with CompiledProgram
-            return self._run(
+            return self._run_program(
                 program._program,
                 self._default_executor,
                 feed=feed,
@@ -683,8 +701,8 @@ class Executor(object):
                 return_numpy=return_numpy,
                 use_program_cache=False)
 
-    def _run(self, program, exe, feed, fetch_list, feed_var_name,
-             fetch_var_name, scope, return_numpy, use_program_cache):
+    def _run_program(self, program, exe, feed, fetch_list, feed_var_name,
+                     fetch_var_name, scope, return_numpy, use_program_cache):
 
         if feed is None:
             feed = {}