Add sleep after restarting usb ports and before checking for devices.

build/android/buildbot/bb_device_status_check.py

 #!/usr/bin/env python
 #
 # Copyright 2013 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 """A class to keep track of devices across builds and report state."""
 import logging
 import optparse
 import os
+import re
 import smtplib
 import subprocess
 import sys
-import re
+import time
 import urllib
 
 import bb_annotations
 import bb_utils
 
 sys.path.append(os.path.join(os.path.dirname(__file__),
                              os.pardir, os.pardir, 'util', 'lib',
                              'common'))
 import perf_tests_results_helper  # pylint: disable=F0401
 
@@ skipping 228 matching lines @@
   parser.add_option('--restart-usb', action='store_true',
                     help='Restart USB ports before running device check.')
   options, args = parser.parse_args()
   if args:
     parser.error('Unknown options %s' % args)
 
   if options.restart_usb:
     rc = RestartUsb()
     if rc:
       return 1
+    time.sleep(2)

[tonyg, 2013/10/24 01:10:59]

I think we want adb wait-for-device instead of time.sleep.

[Isaac (use chromium), 2013/10/24 01:20:37]

I think this actually does require sleep in worst case.  wait-for-device will
return after a single device comes back, I believe.  This is kind of a hack
(more efficient would be to poll adb devices until all devices were back) but
it's OK w/ me.

[frankf, 2013/10/24 01:30:12]

wait-for-device works for particular devices (using -s). Probably the cleanest
option is to add a method to android_commands.py and catch the timeout exception
here.


On 2013/10/24 01:20:37, Isaac (use chromium) wrote:

[navabi, 2013/10/24 02:42:35]

I talked to Frank about this, and I think sleep is the best solution here. As
Isaac pointed out there are multiple devices to wait for.

Some devices may come back and some may not, so we don't want to poll on all
'expected' devices. Say one device on the bot has turned off. Restarting USB
doesn't help in that case and we will poll forever waiting for it to come back.

We can sleep a little bit longer to make sure the device has a chance to come
back. Other than that, I'm not sure what a polling solution would look like.

[bulach, 2013/10/24 13:01:14]

FYI, one issue I saw was that there was a lingering adb process that would not
recover after the RestartUSB()...
I have this ready for your comments :)
https://codereview.chromium.org/39833003/

[bulach, 2013/10/24 13:12:17]

btw, dominik just tested and apparently 1s is not enough but 2 is.. :)
so perhaps we could poll in 1s interval up to say, 5s, and stop as soon as
_anything_ appears in GetAttachedDevices() ?

[navabi, 2013/10/24 18:56:24]

I don't think we want to stop as soon as anything appears. That would mean we
would not wait for devices to come back if the bot already had a device
attached.

I'm fine polling for the expected devices, and waiting up to 5s. I'm a little
worried that we will end up waiting 5s most of the time because there will be at
least 1 missing device, but as long as all of them come back we will save about
3s per build vs. just sleeping 5s.

 
   devices = android_commands.GetAttachedDevices()
   # TODO(navabi): Test to make sure this fails and then fix call
   offline_devices = android_commands.GetAttachedDevices(hardware=False,
                                                         emulator=False,
                                                         offline=True)
 
   types, builds, batteries, reports, errors = [], [], [], [], []
   fail_step_lst = []
   if devices:
@@ skipping 36 matching lines @@
     # devices with critically low battery or install speed. Remove those devices
     # from testing, allowing build to continue with good devices.
     return 1
 
   if not devices:
     return 1
 
 
 if __name__ == '__main__':
   sys.exit(main())