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

Issue 663889 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 663823
Owner: ----
Closed: Nov 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.4%-212.9% regression in memory.top_10_mobile_stress at 430534:430599

Project Member Reported by rsch...@chromium.org, Nov 9 2016

Issue description

See the link to graphs below.
 
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Nov 10 2016


===== BISECT JOB RESULTS =====
Status: completed


=== Bisection aborted ===
The bisect was aborted because The metric values for the initial "good" and "bad" revisions do not represent a clear regression.
Please contact the the team (see below) if you believe this is in error.

=== Warnings ===
The following warnings were raised by the bisect job:

 * Bisect failed to reproduce the regression with enough confidence.

===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N   Good?
chromium@430583  17517028  8762935  18  good
chromium@430590  17428584  8763893  18  bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 663889

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests memory.top_10_mobile_stress
Test Metric: memory:chrome:all_processes:reported_by_chrome:gpu:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 1.56%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3464
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8996436082828076784


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5872534661103616

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!
Project Member

Comment 8 by 42576172...@developer.gserviceaccount.com, Nov 10 2016


===== BISECT JOB RESULTS =====
Status: completed


=== Bisection aborted ===
The bisect was aborted because The metric values for the initial "good" and "bad" revisions do not represent a clear regression.
Please contact the the team (see below) if you believe this is in error.

=== Warnings ===
The following warnings were raised by the bisect job:

 * Bisect failed to reproduce the regression with enough confidence.

===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N   Good?
chromium@430583  64896869  2284464  18  good
chromium@430598  64419013  2544643  18  bad

Bisect job ran on: android_nexus5X_perf_bisect
Bug ID: 663889

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=https.www.google.co.uk.hl.en.q.science memory.top_10_mobile_stress
Test Metric: memory:chrome:all_processes:reported_by_chrome:java_heap:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 0.00%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5X_perf_bisect/builds/845
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8996425850305688976


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5279368539013120

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!
Project Member

Comment 9 by 42576172...@developer.gserviceaccount.com, Nov 10 2016

Mergedinto: 663823
Status: Duplicate (was: Untriaged)

===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Reland of Remove EventSender from HTMLStyleElement
Author  : hiroshige
Commit description:
  
This CL replaces EventSender in HTMLLinkElement with postTask():
Instead of calling HTMLStyleElement::dispatchPendingLoadEvents() in
Document::implicitClose() to enforce <style>'s onload events to be executed
before the document's onload event, this CL blocks document's onload until
<style>'s onload event using IncrementLoadEventDelayCount.

This CL changes the timing of body onload, particularly causes some of the
following events that were previously called AFTER document onload
to be called BEFORE document onload:
[A] The first layout.
[B] Latter half of FrameLoader::finishedParsing().
[C] JavaScript function |f| scheduled as |setTimeout(f, 0)| in the <script>
    at the end of <body>.

Notable test fixes (marked with the causes of failures):

fast/text/international/inline-plaintext-relayout-with-leading-neutrals.html:
  [A] Marked as flaky, because it depends on whether body onload is called
  before (Pass) or after (Failure) the first layout (crbug/651343).

fast/scrolling/scrollbar-tickmarks-styled.html:
  [A] Fixed by https://codereview.chromium.org/2450793002.

fast/scrolling/scrollbar-tickmarks-hittest.html:
  [A] Fixed by https://codereview.chromium.org/2484613003.

fast/forms/text/text-set-selection-crash.html:
  [A] There is race between "Blink Test Plugin: initializing" console log
  (which is asynchronously triggered by the first layout) and document onload
  (=test finish). Because the test should pass if not crashes, this CL make
  the test to ignore the log, by converting the test to use testharness.js
  and removing the -expected.txt.

virtual/rootlayerscrolls/fast/history/scroll-restoration/scroll-restoration-fragment-navigation-crossdoc.html:
  [B] Fixed by https://codereview.chromium.org/2467433002.

svg/animations/animate-end-attribute-numeric-precision.html:
  [C] After this CL, calling executeTest() in
  svg/animations/script-tests/animate-end-attribute-numeric-precision.js
  causes sampleAnimation() to be called before document onload and test to fail.
  We set |animationStartsImmediately| true instead, in order to ensure the test
  is started in document onload.

fast/harness/error-in-async-test.html:
  [C] Whether "Uncaught Exception" console log is shown depends on whether
  buggyAsyncCode() is executed before or after document onload. This CL
  deflakes the test by ensuring it is executed after document onload.

accessibility/inline-text-changes.html:
  [Race between the accessibility notification and document onload]
  Ignore the second call of the notification handler.
  Even before this CL, the notification is called multiple times, but the first
  notification terminated the test immediately by finishJSTest() because it is
  called after document onload.
  This CL makes the notification calls before document onload, and thus the test
  continues until document onload and we have to ignore the second call
  explicitly.

fast/block/float/marquee-shrink-to-avoid-floats.html:
  The layout tree changes slightly, but it looks like acceptable because the
  image expectation matches.

BUG= 624697 , 651343

Committed: https://crrev.com/52d00fd56184deadce99ff4fe599156ea1da3543
Review-Url: https://codereview.chromium.org/2269043002
Cr-Original-Commit-Position: refs/heads/master@{#430348}
Cr-Commit-Position: refs/heads/master@{#430572}
Commit  : b8cb16f79a95c8bc639a9c58785a8484c197aca9
Date    : Tue Nov 08 10:55:53 2016


===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N   Good?
chromium@430558  4101402   595366   12  good
chromium@430567  4429869   2526.31  5   good
chromium@430571  4430212   1992.04  8   good
chromium@430572  11249600  3155909  8   bad    <--
chromium@430573  11616611  3601400  8   bad
chromium@430575  12071863  243437   5   bad
chromium@430590  10728974  3487056  12  bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 663889

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=https.www.google.co.uk.hl.en.q.science memory.top_10_mobile_stress
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 166.90%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3465
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8996425825981890832


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5269347306569728

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!
Project Member

Comment 10 by 42576172...@developer.gserviceaccount.com, Nov 10 2016


===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Reland of Remove EventSender from HTMLStyleElement
Author  : hiroshige
Commit description:
  
This CL replaces EventSender in HTMLLinkElement with postTask():
Instead of calling HTMLStyleElement::dispatchPendingLoadEvents() in
Document::implicitClose() to enforce <style>'s onload events to be executed
before the document's onload event, this CL blocks document's onload until
<style>'s onload event using IncrementLoadEventDelayCount.

This CL changes the timing of body onload, particularly causes some of the
following events that were previously called AFTER document onload
to be called BEFORE document onload:
[A] The first layout.
[B] Latter half of FrameLoader::finishedParsing().
[C] JavaScript function |f| scheduled as |setTimeout(f, 0)| in the <script>
    at the end of <body>.

Notable test fixes (marked with the causes of failures):

fast/text/international/inline-plaintext-relayout-with-leading-neutrals.html:
  [A] Marked as flaky, because it depends on whether body onload is called
  before (Pass) or after (Failure) the first layout (crbug/651343).

fast/scrolling/scrollbar-tickmarks-styled.html:
  [A] Fixed by https://codereview.chromium.org/2450793002.

fast/scrolling/scrollbar-tickmarks-hittest.html:
  [A] Fixed by https://codereview.chromium.org/2484613003.

fast/forms/text/text-set-selection-crash.html:
  [A] There is race between "Blink Test Plugin: initializing" console log
  (which is asynchronously triggered by the first layout) and document onload
  (=test finish). Because the test should pass if not crashes, this CL make
  the test to ignore the log, by converting the test to use testharness.js
  and removing the -expected.txt.

virtual/rootlayerscrolls/fast/history/scroll-restoration/scroll-restoration-fragment-navigation-crossdoc.html:
  [B] Fixed by https://codereview.chromium.org/2467433002.

svg/animations/animate-end-attribute-numeric-precision.html:
  [C] After this CL, calling executeTest() in
  svg/animations/script-tests/animate-end-attribute-numeric-precision.js
  causes sampleAnimation() to be called before document onload and test to fail.
  We set |animationStartsImmediately| true instead, in order to ensure the test
  is started in document onload.

fast/harness/error-in-async-test.html:
  [C] Whether "Uncaught Exception" console log is shown depends on whether
  buggyAsyncCode() is executed before or after document onload. This CL
  deflakes the test by ensuring it is executed after document onload.

accessibility/inline-text-changes.html:
  [Race between the accessibility notification and document onload]
  Ignore the second call of the notification handler.
  Even before this CL, the notification is called multiple times, but the first
  notification terminated the test immediately by finishJSTest() because it is
  called after document onload.
  This CL makes the notification calls before document onload, and thus the test
  continues until document onload and we have to ignore the second call
  explicitly.

fast/block/float/marquee-shrink-to-avoid-floats.html:
  The layout tree changes slightly, but it looks like acceptable because the
  image expectation matches.

BUG= 624697 , 651343

Committed: https://crrev.com/52d00fd56184deadce99ff4fe599156ea1da3543
Review-Url: https://codereview.chromium.org/2269043002
Cr-Original-Commit-Position: refs/heads/master@{#430348}
Cr-Commit-Position: refs/heads/master@{#430572}
Commit  : b8cb16f79a95c8bc639a9c58785a8484c197aca9
Date    : Tue Nov 08 10:55:53 2016


===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N  Good?
chromium@430533  4230772   443265   5  good
chromium@430566  4430295   2960.88  5  good
chromium@430571  4265859   465381   8  good
chromium@430572  11807858  3847902  8  bad    <--
chromium@430573  11395472  3465650  8  bad
chromium@430575  12274388  172884   8  bad
chromium@430583  12866599  1713447  5  bad
chromium@430599  12872903  1699602  5  bad

Bisect job ran on: android_nexus5_perf_bisect
Bug ID: 663889

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=https.www.google.co.uk.hl.en.q.science memory.top_10_mobile_stress
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 204.27%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5_perf_bisect/builds/4329
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8996425898404981808


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5624150259924992

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!
Project Member

Comment 11 by 42576172...@developer.gserviceaccount.com, Nov 10 2016


===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Reland of Remove EventSender from HTMLStyleElement
Author  : hiroshige
Commit description:
  
This CL replaces EventSender in HTMLLinkElement with postTask():
Instead of calling HTMLStyleElement::dispatchPendingLoadEvents() in
Document::implicitClose() to enforce <style>'s onload events to be executed
before the document's onload event, this CL blocks document's onload until
<style>'s onload event using IncrementLoadEventDelayCount.

This CL changes the timing of body onload, particularly causes some of the
following events that were previously called AFTER document onload
to be called BEFORE document onload:
[A] The first layout.
[B] Latter half of FrameLoader::finishedParsing().
[C] JavaScript function |f| scheduled as |setTimeout(f, 0)| in the <script>
    at the end of <body>.

Notable test fixes (marked with the causes of failures):

fast/text/international/inline-plaintext-relayout-with-leading-neutrals.html:
  [A] Marked as flaky, because it depends on whether body onload is called
  before (Pass) or after (Failure) the first layout (crbug/651343).

fast/scrolling/scrollbar-tickmarks-styled.html:
  [A] Fixed by https://codereview.chromium.org/2450793002.

fast/scrolling/scrollbar-tickmarks-hittest.html:
  [A] Fixed by https://codereview.chromium.org/2484613003.

fast/forms/text/text-set-selection-crash.html:
  [A] There is race between "Blink Test Plugin: initializing" console log
  (which is asynchronously triggered by the first layout) and document onload
  (=test finish). Because the test should pass if not crashes, this CL make
  the test to ignore the log, by converting the test to use testharness.js
  and removing the -expected.txt.

virtual/rootlayerscrolls/fast/history/scroll-restoration/scroll-restoration-fragment-navigation-crossdoc.html:
  [B] Fixed by https://codereview.chromium.org/2467433002.

svg/animations/animate-end-attribute-numeric-precision.html:
  [C] After this CL, calling executeTest() in
  svg/animations/script-tests/animate-end-attribute-numeric-precision.js
  causes sampleAnimation() to be called before document onload and test to fail.
  We set |animationStartsImmediately| true instead, in order to ensure the test
  is started in document onload.

fast/harness/error-in-async-test.html:
  [C] Whether "Uncaught Exception" console log is shown depends on whether
  buggyAsyncCode() is executed before or after document onload. This CL
  deflakes the test by ensuring it is executed after document onload.

accessibility/inline-text-changes.html:
  [Race between the accessibility notification and document onload]
  Ignore the second call of the notification handler.
  Even before this CL, the notification is called multiple times, but the first
  notification terminated the test immediately by finishJSTest() because it is
  called after document onload.
  This CL makes the notification calls before document onload, and thus the test
  continues until document onload and we have to ignore the second call
  explicitly.

fast/block/float/marquee-shrink-to-avoid-floats.html:
  The layout tree changes slightly, but it looks like acceptable because the
  image expectation matches.

BUG= 624697 , 651343

Committed: https://crrev.com/52d00fd56184deadce99ff4fe599156ea1da3543
Review-Url: https://codereview.chromium.org/2269043002
Cr-Original-Commit-Position: refs/heads/master@{#430348}
Cr-Commit-Position: refs/heads/master@{#430572}
Commit  : b8cb16f79a95c8bc639a9c58785a8484c197aca9
Date    : Tue Nov 08 10:55:53 2016


===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N   Good?
chromium@430533  4430928   948.406  8   good
chromium@430566  4283578   425472   18  good
chromium@430571  4265322   465172   8   good
chromium@430572  12856298  1718529  5   bad    <--
chromium@430573  12976698  1671839  5   bad
chromium@430575  10364794  4002210  18  bad
chromium@430583  11949100  2863146  12  bad
chromium@430599  11022574  3203983  8   bad

Bisect job ran on: android_nexus5_perf_bisect
Bug ID: 663889

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=https.www.google.co.uk.hl.en.q.science memory.top_10_mobile_stress
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 132.43%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5_perf_bisect/builds/4330
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8996425840327250208


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6348869686788096

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!

Sign in to add a comment