lava-dispatcher hang after "mkdir -p /lava/results"

[Spring Zhang]

lava-dispatcher will stop after executing "mkdir -p /lava/results" in cmd_run_abrek(), before export_display(), after a long time, it will send a cmd "reboot":

root@linaro:~# mkdir -p /lava/results
mkdir -p /lava/results
root@linaro:~# reboot
reboot

Log after lava-dispatcher finished:
root@master:/tmp/lava/results# cat /tmp//lava/results/stream.bundle | nc 10.192.244.171 50404Traceback (most recent call last):
  File "./lava-dispatch", line 22, in <module>
    job.run()
  File "/home/spring/linaro/lava/fix-unicode-encode-err/lava/dispatcher/__init__.py", line 43, in run
    action.run(**params)
  File "/home/spring/linaro/lava/fix-unicode-encode-err/lava/dispatcher/actions/abrek.py", line 12, in run
    response = TESTER_STR)
  File "/home/spring/linaro/lava/fix-unicode-encode-err/lava/dispatcher/client.py", line 88, in run_shell_command
    self.proc.expect(response, timeout=timeout)
  File "/usr/lib/python2.6/dist-packages/pexpect.py", line 1311, in expect
    return self.expect_list(compiled_pattern_list, timeout, searchwindowsize)
  File "/usr/lib/python2.6/dist-packages/pexpect.py", line 1325, in expect_list
    return self.expect_loop(searcher_re(pattern_list), timeout, searchwindowsize)
  File "/usr/lib/python2.6/dist-packages/pexpect.py", line 1409, in expect_loop
    raise TIMEOUT (str(e) + '\n' + str(self))
pexpect.TIMEOUT: Timeout exceeded in read_nonblocking().
<pexpect.spawn object at 0x8d2a0ec>
version: 2.3 ($Revision: 399 $)
command: /usr/bin/conmux-console
args: [u'/usr/bin/conmux-console', u'mx53loco01']
searcher: searcher_re:
    0: re.compile("android# ")
buffer (last 100 chars): ~# mkdir -p /lava/results
root@linaro:~#
before (last 100 chars): ~# mkdir -p /lava/results
root@linaro:~#
after: <class 'pexpect.TIMEOUT'>
match: None
match_index: None
exitstatus: None
flag_eof: False
pid: 8970
child_fd: 3
closed: False
timeout: 600
delimiter: <class 'pexpect.EOF'>
logfile: <closed file '<uninitialized file>', mode '<uninitialized file>' at 0x8cbc644>
logfile_read: None
logfile_send: None
maxread: 2000
ignorecase: False
searchwindowsize: None
delaybeforesend: 1
delayafterclose: 0.1
delayafterterminate: 0.1

[Spring Zhang (qzhang)]

cleaning up all .pyc doesn't help

[Paul Larson (pwlars)]

Yeah, I saw this as well... for some reason it's searching for the android prompt, not the linaro one:
    0: re.compile("android# ")

I saw this on my setup at home also, but not in the lab. I thought there was, perhaps, a stray import of TESTER_STR from android_config.py, but that doesn't seem to be the case. please make sure you remove ALL .pyc files in the tree. That worked for me, not sure what caused this, but since doing that I've gone through several rounds of alternately running android tests, and linaro tests, and I've yet to reproduce it.

[Spring Zhang (qzhang)]

seems the bug is fixed with a commit.

[Paul Larson (pwlars)]

Ah yes, forgot about this one, it's fixed now. There was really no reason to have a different prompt for android when in the test image, so I changed it to match so there's no possible confusion.