New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 833215 link

Starred by 2 users

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Mac
Pri: 3
Type: Bug



Sign in to add a comment

external/wpt/accelerometer/Accelerometer.https.html is flaky (in a strange way)

Project Member Reported by robertma@chromium.org, Apr 16 2018

Issue description

During a recent triage, I noticed this test behaving strangely.

It's pass-fail flaky on Linux. An example of failures:
https://test-results.appspot.com/data/layout_results/Linux_Tests/69207/layout-test-results/results.html

It's pass-timeout flaky on Mac 10.12 & 10.13 (including our Retina, which is currently 10.12). Notably, often when it times out, the output is actually complete and correct, but there are extra output in the head and tail (https://test-results.appspot.com/data/layout_results/Mac10_12_Tests/12101/layout-test-results/results.html):

Content-Type: text/plain
...(normal output)...
#EOF
#EOF

This looks extremely suspicious to me. Something is perhaps wrong with content_shell test runner of r-w-t. And I can't think of any reason it only happens on Mac 10.12 & 10.13.
 
Project Member

Comment 1 by bugdroid1@chromium.org, Apr 16 2018

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/1d0417b9d838c2208369ea5e00e5c91b44f986f3

commit 1d0417b9d838c2208369ea5e00e5c91b44f986f3
Author: Robert Ma <robertma@chromium.org>
Date: Mon Apr 16 03:07:25 2018

Clean up TestExpectaions

1. Some expectations can be consolidated (some tryjobs did not finish
   during the original imports, and expectations were added in later
   imports).
2. Filed a separate issue for wpt/accelerometer/Accelerometer.https.html

TBR=cbiesinger

Bug: 833215, 833093
Change-Id: If23c7188c0c74e525e362f8fd39dccc1c5f11aab
Reviewed-on: https://chromium-review.googlesource.com/1013739
Reviewed-by: Robert Ma <robertma@chromium.org>
Commit-Queue: Robert Ma <robertma@chromium.org>
Cr-Commit-Position: refs/heads/master@{#550929}
[modify] https://crrev.com/1d0417b9d838c2208369ea5e00e5c91b44f986f3/third_party/WebKit/LayoutTests/TestExpectations

Cc: alexande...@intel.com mikhail....@intel.com
Project Member

Comment 3 by bugdroid1@chromium.org, Apr 17 2018

Labels: merge-merged-testbranch
The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/1d0417b9d838c2208369ea5e00e5c91b44f986f3

commit 1d0417b9d838c2208369ea5e00e5c91b44f986f3
Author: Robert Ma <robertma@chromium.org>
Date: Mon Apr 16 03:07:25 2018

Clean up TestExpectaions

1. Some expectations can be consolidated (some tryjobs did not finish
   during the original imports, and expectations were added in later
   imports).
2. Filed a separate issue for wpt/accelerometer/Accelerometer.https.html

TBR=cbiesinger

Bug: 833215, 833093
Change-Id: If23c7188c0c74e525e362f8fd39dccc1c5f11aab
Reviewed-on: https://chromium-review.googlesource.com/1013739
Reviewed-by: Robert Ma <robertma@chromium.org>
Commit-Queue: Robert Ma <robertma@chromium.org>
Cr-Commit-Position: refs/heads/master@{#550929}
[modify] https://crrev.com/1d0417b9d838c2208369ea5e00e5c91b44f986f3/third_party/WebKit/LayoutTests/TestExpectations

The Linux flake looks like the Mojo mocks are getting into a bad start throwing "Error: Call reset() before initialize()."
Owner: mikhail....@intel.com
Status: Assigned (was: Untriaged)
 Issue 833087  has been merged into this issue.
Project Member

Comment 7 by chromium...@appspot.gserviceaccount.com, May 10 2018

Labels: Sheriff-Chromium
Detected 4 new flakes for test/step "external/wpt/accelerometer/Accelerometer.https.html". To see the actual flakes, please visit https://chromium-try-flakes.appspot.com/all_flake_occurrences?key=ahVzfmNocm9taXVtLXRyeS1mbGFrZXNyPgsSBUZsYWtlIjNleHRlcm5hbC93cHQvYWNjZWxlcm9tZXRlci9BY2NlbGVyb21ldGVyLmh0dHBzLmh0bWwM. This message was posted automatically by the chromium-try-flakes app. Since flakiness is ongoing, the issue was moved back into Sheriff Bug Queue (unless already there).
Labels: -Sheriff-Chromium Test-Disabled
Project Member

Comment 9 by bugdroid1@chromium.org, May 11 2018

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c0e275eb8c0a351b3baca3a44dd5b3c04d649af5

commit c0e275eb8c0a351b3baca3a44dd5b3c04d649af5
Author: Max Morin <maxmorin@chromium.org>
Date: Fri May 11 09:51:13 2018

Add expectations for flaky Accelerometer.https.html

Tbr: cbiesinger
No-Try: true
Bug: 833215
Change-Id: I3807d48df11e6b4416357896b7af4564b0ef7c6e
Reviewed-on: https://chromium-review.googlesource.com/1055448
Reviewed-by: Max Morin <maxmorin@chromium.org>
Commit-Queue: Max Morin <maxmorin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#557828}
[modify] https://crrev.com/c0e275eb8c0a351b3baca3a44dd5b3c04d649af5/third_party/WebKit/LayoutTests/TestExpectations

Cc: dpranke@chromium.org
Adding an owner for the Web Tests infrastructure. I'm similarly confused why the runner is expecting the output to contain the Content-Type and #EOF lines on this, and only this, test file.
The runner isn't *expecting* Content-Type of EOF.

Here's the expectation (baseline), which seems correct to me: https://chromium.googlesource.com/chromium/src/+/master/third_party/WebKit/LayoutTests/external/wpt/accelerometer/Accelerometer.https-expected.txt

Rather, it's the actual test output from content_shell that sometimes contains Content-Type and EOF, which I don't understand.
Strange. The extra text is the header and footer that content_shell writes for the test. Given that the harness thinks that the test is timing out, it's as if it doesn't see or recognize the text, but is getting it anyway.

This suggests that maybe something is out of sync in port/driver.py, but just for that one test. The tests before and after run fine.

I'm not sure what would cause this.
Sorry, I realized my comment in 11 might be a bit confusing w.r.t. my use of "expecting". content_shell prints the Content-Type header and the EOF footer, and run_web_tests understands them and strips them away.
(https://cs.chromium.org/chromium/src/third_party/blink/tools/blinkpy/web_tests/port/driver.py?l=471&rcl=a81931feabf43701dd4d8732e08276e376fccb57)

So in that sense, run_web_tests runner does expect to see the header and footer, but they should be handled within run_web_tests and opaque to Blink engineers, i.e. they shouldn't appear in the *-actual output or *-expected baselines.
Project Member

Comment 14 by sheriffbot@chromium.org, Jun 1 2018

Labels: Hotlist-Recharge-BouncingOwner
Owner: ----
Status: Untriaged (was: Assigned)
The assigned owner "mikhail.pozdnyakov@intel.com" is not able to receive e-mails, please re-triage.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Cc: -mikhail....@intel.com
Owner: raphael....@intel.com
Status: Assigned (was: Untriaged)
Status: Started (was: Assigned)
https://chromium-review.googlesource.com/c/chromium/src/+/1126388 should take care of the test timing out problem; the extra output needs to be taken care of separately.
Project Member

Comment 17 by bugdroid1@chromium.org, Jul 6

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2dc478a481992a16af70f690611e2b7e08740721

commit 2dc478a481992a16af70f690611e2b7e08740721
Author: Raphael Kubo da Costa <raphael.kubo.da.costa@intel.com>
Date: Fri Jul 06 20:49:22 2018

Mark external/wpt/accelerometer/Accelerometer.https.html as a slow test

This test often times out on the Mac bots, but it doesn't seem to be really
failing but rather just being slow: it currently tests 3 different
accelerometer sensors and for each of them we make several asynchronous
calls and wait for many events, which can often cause it to take longer than
6 seconds to run (even on a release Linux build it never takes less than 5
seconds to run).

Bug: 833215
Change-Id: I105a04e797f68b91739b631387d8e7d8da36e6da
Reviewed-on: https://chromium-review.googlesource.com/1126388
Commit-Queue: Raphael Kubo da Costa (CET) <raphael.kubo.da.costa@intel.com>
Reviewed-by: Reilly Grant <reillyg@chromium.org>
Cr-Commit-Position: refs/heads/master@{#573064}
[modify] https://crrev.com/2dc478a481992a16af70f690611e2b7e08740721/third_party/WebKit/LayoutTests/SlowTests
[modify] https://crrev.com/2dc478a481992a16af70f690611e2b7e08740721/third_party/WebKit/LayoutTests/TestExpectations

Cc: raphael....@intel.com
Owner: robertma@chromium.org
Status: Assigned (was: Started)
Reassignig to Robert for the infra side; the test itself should no longer be flaky.

Sign in to add a comment