Fix flakey test_server.

See:
https://bugs.chromium.org/p/chromium/issues/detail?id=946475#c48

I believe the root cause of the flake is trying to test if the server
has been killed by trying to rebind to the port, although my test shows
that the port can get claimed for something else by the system.

Add a process.wait() to ensure server process has terminated and remove
the error from checking the port status. Could do a check on PID status
but system could reassign the PID, so that could lead to the same
flake as before except based on PID instead of port number.

Bug: 946475
Change-Id: I48911fffcab6b592ddf627e5999fc0609e0f0b52
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2339553Reviewed-by: Andrew Grieve <agrieve@chromium.org>
Reviewed-by: David Benjamin <davidben@chromium.org>
Commit-Queue: benjamin joyce <bjoyce@chromium.org>
Cr-Commit-Position: refs/heads/master@{#799803}

build/util/lib/common/chrome_test_server_spawner.py

@@ -270,6 +270,10 @@ class TestServerThread(threading.Thread):
     self.stop_event.wait()
     if self.process.poll() is None:
       self.process.kill()
+      # Wait for process to actually terminate.
+      # (crbug.com/946475)
+      self.process.wait()
+
     self.port_forwarder.Unmap(self.forwarder_device_port)
     self.process = None
     self.is_ready = False
@@ -387,8 +391,11 @@ class SpawningServerRequestHandler(BaseHTTPServer.BaseHTTPRequestHandler):
       self._SendResponse(200, 'OK', {}, 'killed')
       _logger.info('Test server on port %d is killed', port)
     else:
-      self._SendResponse(500, 'Test Server Error.', {}, '')
-      _logger.info('Encounter problem during killing a test server.')
+      # We expect the port to be free, but nothing stops the system from
+      # binding something else to that port, so don't throw error.
+      # (crbug.com/946475)
+      self._SendResponse(200, 'OK', {}, '')
+      _logger.warn('Port %s is not free after killing test server.' % port)
 
   def log_message(self, format, *args):
     # Suppress the default HTTP logging behavior if the logging level is higher