moblab-generic-vm-pre-cq provision test failing due to wrong "found build" |
|||||
Issue descriptionhttps://luci-milo.appspot.com/buildbot/chromiumos.tryserver/pre_cq/91251 15:05:53 INFO | autoserv| [stderr] Suite job [ PASSED ] 15:05:53 INFO | autoserv| [stderr] provision [ FAILED ] 15:05:53 INFO | autoserv| [stderr] provision FAIL: Failed to update 192.168.231.100 to build R67-10509.0.0-b2405484; found build 10509.0.2018_03_22_1210 instead 15:05:53 INFO | autoserv| [stderr] 15:05:53 INFO | autoserv| [stderr] Suite timings: 15:05:53 INFO | autoserv| [stderr] Downloads started at 2018-03-22 12:54:58 15:05:53 INFO | autoserv| [stderr] Payload downloads ended at 2018-03-22 12:55:03 15:05:53 INFO | autoserv| [stderr] Suite started at 2018-03-22 12:55:11 15:05:53 INFO | autoserv| [stderr] Artifact downloads ended (at latest) at 2018-03-22 12:55:11 15:05:53 INFO | autoserv| [stderr] Testing started at 2018-03-22 12:55:17 15:05:53 INFO | autoserv| [stderr] Testing ended at 2018-03-22 13:03:55 15:05:53 INFO | autoserv| [stderr] 15:05:53 INFO | autoserv| [stderr] 15:05:53 INFO | autoserv| [stderr] Links to test logs: 15:05:53 INFO | autoserv| [stderr] Suite job http://localhost/tko/retrieve_logs.cgi?job=/results/1-moblab/ 15:05:53 INFO | autoserv| [stderr] provision http://localhost/tko/retrieve_logs.cgi?job=/results/2-moblab/ 15:05:53 INFO | autoserv| [stderr] provision FAIL: Failed to update 192.168.231.100 to build R67-10509.0.0-b2405484; found build 10509.0.2018_03_22_1210 instead So we expect to have a
,
Mar 22 2018
Nevermind, it's not really similar to issue 601625. I recall Yueherngl@ had an issue recently with a firmware builder he setup, let me take a closer look.
,
Mar 22 2018
This is caused by a bug between Autotest and some recent builder changes, more or less. I don't understand the mechanism for how this got into the system, though.
,
Mar 22 2018
Don, can you take a look?
,
Mar 22 2018
The logs from the test suite aren't adequate to show it, but basically,
this failure is a consequence of a code change that alters how builds
are named. dgarrett@ will know which change is involved.
The reason moblab-generic-vm-pre-cq is failing is because there's a
_second_ code path that makes a broken check comparing "build installed"
with "build name". The second broken check is similar to the first
broken check (the one we fixed), but it's only used by moblab.
What I can't explain is why we're only seeing this change _now_.
I suspect dgarrett@ can enlighten us.
For the record, the broken check is in (you guessed it) cros_host.py
(the root of all evil in the world not caused by a love of money).
Here's the relevant source:
def _post_update_processing(self, updater, expected_kernel=None):
# ...
# Check that we've got the build we meant to install.
if not updater.check_version_to_confirm_install():
raise autoupdater.ChromiumOSError(
'Failed to update %s to build %s; found build '
'%s instead' % (self.hostname,
updater.update_version,
self.get_release_version()))
,
Mar 22 2018
My current inclination is to outright disable the problem code rather than sort through the immediate fix; the check is redundant given the current verifier, which I expect _will_ work correctly.
,
Mar 23 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/third_party/autotest/+/50b69931fdcaa8ed3c42ad4f6741c20df419e8df commit 50b69931fdcaa8ed3c42ad4f6741c20df419e8df Author: Richard Barnette <jrbarnette@chromium.org> Date: Fri Mar 23 00:04:56 2018 [autotest] Delete a problematic check from machine_install(). In CrosHost.machine_install(), there's a sanity check that the build installed on the DUT is the same as the build that was requested. The check operates by comparing strings in /etc/lsb-release against portions of the URL that named the build; that check requires various heuristics and special cases. The check in question is only used by Moblab, not by Autotest in the full lab. A recent change to the way builds are named on the builders broke the check in machine_install(), and now the moblab pre-CQ builder is failing. The check is somewhat redundant, so stop the bleeding by deleting the check. BUG= chromium:824956 TEST=run this through the pre-cq Change-Id: Iccf19779b5f7a4c7a4e8e233e95c8c683cf6e990 Reviewed-on: https://chromium-review.googlesource.com/976644 Tested-by: Richard Barnette <jrbarnette@chromium.org> Trybot-Ready: Richard Barnette <jrbarnette@chromium.org> Reviewed-by: Shuqian Zhao <shuqianz@chromium.org> Reviewed-by: Prathmesh Prabhu <pprabhu@chromium.org> [modify] https://crrev.com/50b69931fdcaa8ed3c42ad4f6741c20df419e8df/server/hosts/cros_host.py
,
Apr 2 2018
I believe this is now fixed. |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by sha...@chromium.org
, Mar 22 2018