veyron_rialto-release: No module named telemetry.core |
|||||
Issue descriptionFailing for ~5days First failure: https://luci-milo.appspot.com/buildbot/chromeos/veyron_rialto-release/1947 " [Test-Logs]: graphics_Idle: ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: login_Cryptohome: ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: login_CryptohomeIncognito: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: login_LoginSuccess: ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: login_LogoutProcessCleanup: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: security_NetworkListeners: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: security_ProfilePermissions.guest: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: security_ProfilePermissions.login: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core [Test-Logs]: security_SandboxStatus: retry_count: 1, ERROR: Unhandled ImportError: No module named telemetry.core " I wonder if it's related to https://chromium-review.googlesource.com/c/chromiumos/overlays/chromiumos-overlay/+/875208 which had similar issues. Timing does not exactly fit, but several canary builds before #1947 failed because of infra issues, so it's still possible that the CL broke veyron_rialto but the build only started failing because of it 1-2 days later.
,
Feb 22 2018
probably caused by this: https://crrev.com/c/875208/4/chromeos-base/autotest-tests-cryptohome/autotest-tests-cryptohome-9999.ebuild#46 The cryptohome tests package now includes tests that rely on telemetry to emulate Chrome logins. should it belong to a different package?
,
Feb 22 2018
Maybe login_Cryptohome and login_CryptohomeIncognito should move to autotest-tests-cryptohome instead of autotest-chrome, but that won't change the other tests that are failing (e.g. security_*), right? My impression is that https://crrev.com/c/875208/4/chromeos-base/autotest-tests-cryptohome/autotest-tests-cryptohome-9999.ebuild#46 overrides (or maybe rather is executed after?) a similar method in autotest-chrome: https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/master/chromeos-base/autotest-chrome/autotest-chrome-9999.ebuild#247
,
Feb 22 2018
I added both that custom src_prepare() and DEPEND on autotest-chrome in a single update. Maybe just DEPEND was enough. W/o those it was (understandably) failing - see all those errors for patchset 3 in CL:857208. After that it passed Pre-CQ and CQ. But that means that for the majority of the boards it works fine even when both autotest-chrome and autotest-tests-cryptohome do this installation of telemetry into a tempdir. Likely, the issue is a little different: for veyron_rialto we only had autotest-tests-chryptohome. We didn't have autotest-chrome (or a test that needs telemetry in autotest-tests-chryptohome). It now stumbles on autotest-chrome, but would stumble on the updated autotest-tests-chryptohome if it was coming first in the dependencies tree. I guess, the telemetry package is built differently for veyron_rialto? And it simply doesn't install telemetry.core? And if we need to support telemetry-less boards like this, I need to move platform_InitLoginPerf out of autotest-tests-chryptohome to some other package, and revert all changes to autotest-tests-chryptohome ebuild. The question is what should that other package that is only included for boards with Chrome be?
,
Feb 22 2018
platform_InitLoginPerf may need to go to autotest-chrome. Looks like we did exactly that for tests_platform_Pkcs11InitOnLogin, which is included in autotest-chrome ebuild for similar reasons? So, reverting CL:875208 and adding platform_InitLoginPerf to autotest-chrome sounds like the right approach. Not sure who owns autotest-chrome to confirm. +cywang for opinion.
,
Feb 22 2018
Yes, I believe we put the autotests import telemetry to test chrome in autotest-chrome.
,
Feb 22 2018
,
Feb 22 2018
Note: I couldn't reproduce the issue on the local veyron_rialto build. Though I noticed that it built Chrome from scratch to complete build_packages. Are those failing builders incremental - that could affect it. Still, it apparently adds chrome to the set of built packages when dependencies are resolved correctly. So, moving platform_InitLoginPerf to autotest-chrome (which is not included in case of USE=chromeless_tests, as on veyron_rialto) is the right move. Submitted https://crrev.com/c/930323 for review.
,
Feb 23 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/0ddaa4bcccd08cfb9f751b57ddc4f811dab65543 commit 0ddaa4bcccd08cfb9f751b57ddc4f811dab65543 Author: Andrey Pronin <apronin@chromium.org> Date: Fri Feb 23 22:16:02 2018 Move platform_InitLoginPerf to autotest-chrome autotest-tests-cryptohome can be used on a headless target like veyron_rialto and shouldn't utilize telemetry. Move platform_InitLoginPerf to autotest-chrome and remove all dependencies on telemetry from autotest-tests-cryptohome. BUG= chromium:814514 TEST=build_packages --board=veyron_rialto emerge-eve autotest-chrome autotest-tests-cryptohome Change-Id: I2b09e111678d804e9cd8b126ef26db326cf5c0c3 Reviewed-on: https://chromium-review.googlesource.com/930323 Commit-Ready: Andrey Pronin <apronin@chromium.org> Tested-by: Andrey Pronin <apronin@chromium.org> Reviewed-by: Achuith Bhandarkar <achuith@chromium.org> Reviewed-by: Chung-yih Wang <cywang@chromium.org> [modify] https://crrev.com/0ddaa4bcccd08cfb9f751b57ddc4f811dab65543/chromeos-base/autotest-tests-cryptohome/autotest-tests-cryptohome-9999.ebuild [modify] https://crrev.com/0ddaa4bcccd08cfb9f751b57ddc4f811dab65543/chromeos-base/autotest-chrome/autotest-chrome-9999.ebuild
,
Feb 25 2018
The 5 last builds have been green |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by norvez@chromium.org
, Feb 21 2018