restarting adb server before each testing start will cause conflict for multiple emulators

build/android/run_tests.py

 #!/usr/bin/env python
 #
 # Copyright (c) 2012 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.
 
 """Runs all the native unit tests.
 
 1. Copy over test binary to /data/local on device.
 2. Resources: chrome/unit_tests requires resources (chrome.pak and en-US.pak)
@@ skipping 319 matching lines @@
         # Creates a temporary AVD for the extra emulators.
         avd_name = 'run_tests_avd_%d' % n
       buildbot_emulator = emulator.Emulator(avd_name, options.fast_and_loose)
       buildbot_emulator.Launch(kill_all_emulators=n == 0)
       t.Stop()
       buildbot_emulators.append(buildbot_emulator)
       attached_devices.append(buildbot_emulator.device)
     # Wait for all emulators to boot completed.
     map(lambda buildbot_emulator: buildbot_emulator.ConfirmLaunch(True),
         buildbot_emulators)
+    # Restart adb server as emulator will reset connection when booting,
+    # which will break the connection between adb server and emulator.

[bulach, 2012/07/17 10:26:08]

the comment seems to indicate this is needed per emulator.. perhaps something
like this in 337:
def WaitForEmulatorAndRestartAdb(emulator):
  emulator.ConfirmLaunch(True)
  android_commands.RestartAdbServer()
map(WaitForEmulatorAndRestartAdb, buildbot_emulators)

[John Grabowski, 2012/07/17 18:09:22]

agree

[Wei James(wistoch), 2012/07/20 04:07:01]

after applying several patches of bulach, I run the tests for several times, it
is stable even without this one. 
As we already handle the connection lost issue in
AndroidCommand::WaitForSystemBootCompleted, I think removing it here is ok. 

+    android_commands.RestartAdbServer()
   elif options.test_device:
     attached_devices = [options.test_device]
   else:
     attached_devices = android_commands.GetAttachedDevices()
 
   if not attached_devices:
     logging.critical('A device must be attached and online.')
     buildbot_report.PrintError()
     return 1
 
@@ skipping 142 matching lines @@
   # the batch (this happens because the exit status is a sum of all failures
   # from all suites, but the buildbot associates the exit status only with the
   # most recent step).
   if options.exit_code:
     return failed_tests_count
   return 0
 
 
 if __name__ == '__main__':
   sys.exit(main(sys.argv))