sysdeps test: Wait for spawned process to exit

On Windows, the "out" parameter for the process handle isn't set
unless we specify G_SPAWN_FLAGS_DO_NOT_REAP_CHILD. That means we
can't terminate it, and the child process is leaked. When running the
test individually, the leaked process is harmless apart from its
resource cost, but when running under CTest, it holds a file descriptor
open (or something) which causes CTest to not exit.

If we *do* specify G_SPAWN_FLAGS_DO_NOT_REAP_CHILD, we become
responsible for "reaping" the child process by waiting for its exit
status, which is an OS-specific action.

Signed-off-by: Simon McVittie <smcv@collabora.com>
Resolves: https://gitlab.freedesktop.org/dbus/dbus/issues/238

diff --git a/test/internals/sysdeps.c b/test/internals/sysdeps.c
index 992c4619356ac9acb0ff4cf5ba298e80c92cb9aa..33d04f91ba09a6337aa904b33b9310fed5dab2e6 100644 (file)
--- a/test/internals/sysdeps.c
+++ b/test/internals/sysdeps.c
@@ -43,13 +43,52 @@ setup (Fixture *f G_GNUC_UNUSED,
  * handle (a HANDLE, vaguely analogous to a file descriptor) on Windows.
  * For _dbus_command_for_pid() we need an actual process ID. */
 #ifdef G_OS_UNIX
+# include <errno.h>
+# include <sys/wait.h>
+# include <sys/types.h>
+
 # define NO_PROCESS 0
-# define terminate(process) kill (process, SIGTERM)
-# define get_pid(process) process
+# ifndef G_PID_FORMAT
+#   define G_PID_FORMAT "i"
+# endif
+
+static unsigned long
+get_pid (pid_t process)
+{
+  return process;
+}
+
+static void
+terminate_and_wait (pid_t process)
+{
+  int ret;
+
+  kill (process, SIGTERM);
+
+  do
+    {
+      ret = waitpid (process, NULL, 0);
+    }
+  while (ret == -1 && errno == EINTR);
+}
 #else
 # define NO_PROCESS NULL
-# define terminate(process) TerminateProcess (process, 1)
-# define get_pid(process) GetProcessId (process)
+# ifndef G_PID_FORMAT
+#   define G_PID_FORMAT "p"
+# endif
+
+static unsigned long
+get_pid (HANDLE process)
+{
+  return GetProcessId (process);
+}
+
+static void
+terminate_and_wait (HANDLE process)
+{
+  TerminateProcess (process, 1);
+  WaitForSingleObject (process, INFINITE);
+}
 #endif
 
 static void
@@ -73,11 +112,17 @@ test_command_for_pid (Fixture *f,
 
   argv[1] = g_strdup ("bees");
 
-  if (!g_spawn_async (NULL, argv, NULL, G_SPAWN_DEFAULT, NULL, NULL,
+  if (!g_spawn_async (NULL, argv, NULL,
+                      (G_SPAWN_STDOUT_TO_DEV_NULL |
+                       G_SPAWN_STDERR_TO_DEV_NULL |
+                       G_SPAWN_DO_NOT_REAP_CHILD),
+                      NULL, NULL,
                       &process, &error))
     g_error ("Unable to run %s: %s", argv[0], error->message);
 
   pid = get_pid (process);
+  g_test_message ("Process ID of process handle %" G_PID_FORMAT ": %lu",
+                  process, pid);
 
   if (!_dbus_string_init (&string))
     g_error ("out of memory");
@@ -99,8 +144,8 @@ test_command_for_pid (Fixture *f,
     }
   else
     {
-      g_test_message ("Unable to get command for process ID: %s: %s",
-                      d_error.name, d_error.message);
+      g_test_message ("Unable to get command for process %lu: %s: %s",
+                      pid, d_error.name, d_error.message);
       g_assert_nonnull (d_error.name);
       g_assert_nonnull (d_error.message);
       dbus_error_free (&d_error);
@@ -127,11 +172,11 @@ test_command_for_pid (Fixture *f,
     }
   else
     {
-      g_test_message ("Unable to get command for process ID");
+      g_test_message ("Unable to get command for process %lu", pid);
     }
 
   if (process != NO_PROCESS)
-    terminate (process);
+    terminate_and_wait (process);
 
   _dbus_string_free (&string);
   g_spawn_close_pid (process);