Propagate exit status if given a child program to execute

README.md

@@ -97,6 +97,10 @@ prmon [--pid PPP] [--filename prmon.txt] [--json-summary prmon.json] \
   and remaining arguments are passed to it; `prmon` will then monitor this process
   instead of being given a PID via `--pid`
 
+`prmon` will exit with `1` if there is a problem with inconsistent or 
+incomplete arguments. If `prmon` starts a program itself (using `--`) then
+`prmon` will exit with the exit code of the child process.
+
 
 ## Outputs
 

package/src/prmon.cpp

@@ -96,6 +96,7 @@ int ProcessMonitor(const pid_t mpid, const std::string filename,
   // Monitoring loop until process exits
   bool wroteFile = false;
   std::vector<pid_t> cpids{};
+  int return_code = 0;
   // Scope of 'monitors' ensures safety of bare pointer here
   auto wallclock_monitor_p = static_cast<wallmon*>(monitors["wallmon"].get());
   while (kill(mpid, 0) == 0 && prmon::sigusr1 == false) {
@@ -179,7 +180,7 @@ int ProcessMonitor(const pid_t mpid, const std::string filename,
       }
     }
     std::this_thread::sleep_for(std::chrono::milliseconds(200));
-    prmon::reap_children();
+    return_code = prmon::reap_children();
   }
   file.close();
 
@@ -192,7 +193,7 @@ int ProcessMonitor(const pid_t mpid, const std::string filename,
   file << std::setw(2) << json_summary << std::endl;
   file.close();
 
-  return 0;
+  return return_code;
 }
 
 int main(int argc, char* argv[]) {
@@ -335,8 +336,8 @@ int main(int argc, char* argv[]) {
     if (child == 0) {
       execvp(argv[child_args], &argv[child_args]);
     } else if (child > 0) {
-      ProcessMonitor(child, filename, jsonSummary, interval, store_hw_info,
-                     store_unit_info, netdevs);
+      return ProcessMonitor(child, filename, jsonSummary, interval,
+                            store_hw_info, store_unit_info, netdevs);
     }
   }
 

package/src/prmonutils.cpp

@@ -92,26 +92,29 @@ std::vector<pid_t> offspring_pids(const pid_t mother_pid) {
 
 void SignalCallbackHandler(int /*signal*/) { sigusr1 = true; }
 
-void reap_children() {
+int reap_children() {
   int status;
+  int return_code = 0;
   pid_t pid{1};
   while (pid > 0) {
     pid = waitpid((pid_t)-1, &status, WNOHANG);
     if (status && pid > 0) {
-      if (WIFEXITED(status))
+      if (WIFEXITED(status)) {
+        return_code = WEXITSTATUS(status);
         std::clog << "Child process " << pid
-                  << " had non-zero return value: " << WEXITSTATUS(status)
-                  << std::endl;
-      else if (WIFSIGNALED(status))
+                  << " had non-zero return value: " << return_code << std::endl;
+      } else if (WIFSIGNALED(status)) {
         std::clog << "Child process " << pid << " exited from signal "
                   << WTERMSIG(status) << std::endl;
-      else if (WIFSTOPPED(status))
+      } else if (WIFSTOPPED(status)) {
         std::clog << "Child process " << pid << " was stopped by signal"
                   << WSTOPSIG(status) << std::endl;
-      else if (WIFCONTINUED(status))
+      } else if (WIFCONTINUED(status)) {
         std::clog << "Child process " << pid << " was continued" << std::endl;
+      }
     }
   }
+  return return_code;
 }
 
 }  // namespace prmon

package/src/prmonutils.h

@@ -25,7 +25,7 @@ extern bool sigusr1;
 void SignalCallbackHandler(int);
 
 // Child process reaper
-void reap_children();
+int reap_children();
 
 // Precision specifier for average output, to truncate to an integer
 // for anything >avg_precision and round the fraction to

package/tests/CMakeLists.txt

@@ -38,6 +38,7 @@ script_install(SCRIPT testMEM.py)
 script_install(SCRIPT netBurner.py)
 script_install(SCRIPT httpBlock.py DESTINATION cgi-bin)
 script_install(SCRIPT testCOUNT.py)
+script_install(SCRIPT testEXIT.py)
 
 # Setup the target version of Python we will use for testing
 set(PYTHON_TEST "python3" CACHE STRING "Python binary to use for tests")
@@ -79,3 +80,6 @@ add_test(NAME basicCOUNT COMMAND ${PYTHON_TEST} testCOUNT.py --procs 2 --threads
 
 # Units check test
 add_test(NAME testUnits COMMAND ${PYTHON_TEST} testCPU.py --units --time 3 --slack 0 --interval 1) 
+
+# Test passing the child exit code works
+add_test(NAME testExitCode COMMAND ${PYTHON_TEST} testEXIT.py --exit-code 43)

package/tests/testEXIT.py

@@ -0,0 +1,38 @@
+#! /usr/bin/env python3
+#
+# Copyright (C) 2018-2020 CERN
+# License Apache2 - see LICENCE file
+
+import argparse
+import subprocess
+import sys
+import unittest
+
+def setupConfigurableTest(exit_code = 0):
+    '''Wrap the class definition in a function to allow arguments to be passed'''
+    class configurableProcessMonitor(unittest.TestCase):
+        def test_runTestWithParams(self):
+            child_cmd = ['sh', '-c', 'sleep 3 && exit {0}'.format(exit_code)]
+
+            prmon_cmd = ['../prmon', '--interval', '1']
+            prmon_cmd.append('--')
+            prmon_cmd.extend(child_cmd)
+            prmon_p = subprocess.Popen(prmon_cmd, shell = False)
+            prmon_rc = prmon_p.wait()
+
+            self.assertEqual(prmon_rc, exit_code, "Wrong return code from prmon (expected {0}".format(exit_code))
+
+    return configurableProcessMonitor
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description="Configurable test runner")
+    parser.add_argument('--exit-code', type=int, default=0)
+
+    args = parser.parse_args()
+    # Stop unittest from being confused by the arguments
+    sys.argv=sys.argv[:1]
+
+    cpm = setupConfigurableTest(args.exit_code)
+
+    unittest.main()