Test testGoBackAndGoForward became flaky and fails for Mac and Linux platforms |
|||
Issue descriptionTests testAlertHandlingOnPageUnload and testGoBackAndGoForward became flaky and fail on Mac and Linux platforms Repro steps: 1. Get the latest sources from chromium repository 2. Do 'git checkout ce66ffd5325f41ec2ed4fa22dc83aa61737e7831' 3. Build chrome and chromedriver 4. Go to /src/chrome/test/chromedriver/test directory 5. Run the following command: python run_py_tests.py --chrome=<chromium_fetched_dir>/src/out/Default/chrome --chromedriver=<chromium_fetched_dir>/src/out/Default/chromedriver --filter=*.testGoBackAndGoForward --log-path=<log_file> Because of test testGoBackAndGoForward is flaky you should run this command until test fails with the following error: raise _ExceptionForLegacyResponse(response) UnknownError: unknown error: session deleted because of page crash from unknown error: cannot determine loading status from tab crashed (Session info: chrome=65.0.3324.0) (Driver info: chromedriver=2.35.530041 (c39e772a093b608128cf5f1eda0db21d92eecf37),platform=Linux 4.4.0-109-generic x86_64) ChromeDriver log file for this failure is attached. Bisecting reports this test started to fail after the following commit: https://chromium-review.googlesource.com/868327
,
Jan 23 2018
testAlertHandlingOnPageUnload as well became flaky due to the same commit https://crrev.com/530042
,
Jan 23 2018
,
Jan 24 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/a5893ea4171827f372fe6f892e4fbc9b2036abb7 commit a5893ea4171827f372fe6f892e4fbc9b2036abb7 Author: Artur Khachatryan <khachatryan@chromium.org> Date: Wed Jan 24 17:39:43 2018 [ChromeDriver] Disabling a flaky test for Linux build bot Test testAlertHandlingOnPageUnload became flaky on Linux build bot. Disabling test execution until the bug is fixed. Bug: 803678 Change-Id: Ib51bbeb95f4d3b64e71a9b48a8bb2dbfecfcfa55 Reviewed-on: https://chromium-review.googlesource.com/882197 Reviewed-by: John Chen <johnchen@chromium.org> Commit-Queue: Artur Khachatryan <khachatryan@chromium.org> Cr-Commit-Position: refs/heads/master@{#531590} [modify] https://crrev.com/a5893ea4171827f372fe6f892e4fbc9b2036abb7/chrome/test/chromedriver/test/run_py_tests.py
,
Jan 25 2018
The crash problem was fixed as a part of issue 804214. However, I stumbled upon a race which still makes these tests flaky - going to fix that.
,
Jan 25 2018
Thank you.
,
Jan 29 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/8a854532e7b2605ba975074b6e0673ea98319298 commit 8a854532e7b2605ba975074b6e0673ea98319298 Author: Dmitry Gozman <dgozman@chromium.org> Date: Mon Jan 29 02:35:36 2018 [DevTools] Fix a race between suspended and waiting messages in DevToolsSession If a message comes right before navigation, and we send it out but never get a response from the old renderer (e.g. it was killed), and then suspend messages due to navigation, the original message is stuck in waiting_for_response_messages_ and will not be dispatched when resuming upon navigation finish. To mitigate this, we move all waiting messages to suspended if we are not resending them to new renderer due to being currently suspended. Bug: 803678 Change-Id: I125609e0ec8a5e2b98ff2d81a9e71732fc8f6f0f Reviewed-on: https://chromium-review.googlesource.com/887435 Reviewed-by: Pavel Feldman <pfeldman@chromium.org> Reviewed-by: John Chen <johnchen@chromium.org> Commit-Queue: Dmitry Gozman <dgozman@chromium.org> Cr-Commit-Position: refs/heads/master@{#532307} [modify] https://crrev.com/8a854532e7b2605ba975074b6e0673ea98319298/chrome/test/chromedriver/test/run_py_tests.py [modify] https://crrev.com/8a854532e7b2605ba975074b6e0673ea98319298/content/browser/devtools/devtools_session.cc
,
Jan 29 2018
Marking this as fixed. Please reopen if the flakiness continues.
,
Jan 29 2018
The tests now appear stable on v66, but v65 is still flaky. Would it be possible to merge the fix to v65? Thanks.
,
Jan 30 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/62c93ab55fdfb7945a8e4320dd07b8e8601b1f3e commit 62c93ab55fdfb7945a8e4320dd07b8e8601b1f3e Author: Artur Khachatryan <khachatryan@chromium.org> Date: Tue Jan 30 20:30:17 2018 [ChromeDriver] Disable testGoBackAndGoForward for Chrome v65 Test testGoBackAndGoForward fails on Waterfall for Chrome v65. Disabling test execution in Waterfall build. Bug: 803678 Change-Id: I0e084797f727dbd911cb16032fbb642040a1de3a Reviewed-on: https://chromium-review.googlesource.com/893420 Reviewed-by: John Chen <johnchen@chromium.org> Commit-Queue: Artur Khachatryan <khachatryan@chromium.org> Cr-Commit-Position: refs/heads/master@{#533007} [modify] https://crrev.com/62c93ab55fdfb7945a8e4320dd07b8e8601b1f3e/chrome/test/chromedriver/test/run_py_tests.py
,
Feb 2 2018
Ping: Would it be possible to merge the fix to v65? Thanks.
,
Feb 5 2018
Unfortunately, the fix does not apply to M65 due to large code changes in between. I am hesitant to make a manual fix, since it could potentially break DevTools. Is this vital for 65, or it's just a matter of flaky tests?
,
Feb 5 2018
Thanks for the update. I'm OK with not fixing it in v65 if there is too much risk. |
|||
►
Sign in to add a comment |
|||
Comment 1 by bugdroid1@chromium.org
, Jan 19 2018