Improve failure detection and reporting in the Tests job. (#14729)

* Fix exception formatting when a dependency of our chip-tool process
  has died.

* Detect startup crashes in the server app without having to wait for
  the entire 10-second timeout.

* Call apps_register.uninit() before sys.exit(), because otherwise on
  failure we end up sitting there waiting for an unjoined thread until
  the test job times out.

scripts/tests/chiptest/runner.py

@@ -101,8 +101,8 @@ def RunSubprocess(self, cmd, name, wait=True, dependencies=[]):
             for dependency in dependencies:
                 if dependency.poll() is not None:
                     s.kill()
-                    raise Exception("Unexpected return %d for %r",
-                                    dependency.poll(), dependency)
+                    raise Exception("Unexpected return %d for %r" %
+                                    (dependency.poll(), dependency))
 
         code = s.wait()
         if code != 0:

scripts/tests/chiptest/test_definition.py

@@ -41,7 +41,7 @@ def start(self, discriminator):
             self.process = None
             process, outpipe, errpipe = self.__startServer(
                 self.runner, self.command, discriminator)
-            self.__waitForServerReady(outpipe)
+            self.__waitForServerReady(process, outpipe)
             self.__updateSetUpCode(outpipe)
             self.process = process
             self.stopped = False
@@ -95,11 +95,15 @@ def __startServer(self, runner, command, discriminator):
         app_cmd = command + ['--discriminator', str(discriminator)]
         return runner.RunSubprocess(app_cmd, name='APP ', wait=False)
 
-    def __waitForServerReady(self, outpipe):
+    def __waitForServerReady(self, server_process, outpipe):
         logging.debug('Waiting for server to listen.')
         start_time = time.time()
         server_is_listening = outpipe.CapturedLogContains("Server Listening")
         while not server_is_listening:
+            if server_process.poll() is not None:
+                died_str = 'Server died during startup, returncode %d' % server_process.returncode
+                logging.error(died_str)
+                raise Exception(died_str)
             if time.time() - start_time > 10:
                 raise Exception('Timeout for server listening')
             time.sleep(0.1)

scripts/tests/run_test_suite.py

@@ -212,6 +212,7 @@ def cmd_run(context, iterations, all_clusters_app, tv_app):
                 test_end = time.time()
                 logging.exception('%s - FAILED in %0.2f seconds' %
                                   (test.name, (test_end - test_start)))
+                apps_register.uninit()
                 sys.exit(2)
 
     apps_register.uninit()