All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=729776

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpqKakwoM


Bot(s) for this bug's original alert(s):

android-webview-nexus5X

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8977590159007361216

=== BISECT JOB RESULTS ===
Bisect was unable to run to completion

Please try rerunning the bisect.


If failures persist contact the team (see below) and report the error.


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : get-attribute-rare/get-attribute-rare
  Change       : 9.79% | 94.8953087367 -> 85.6037487659

Revision             Result                   N
chromium@463289      94.8953 +- 0.854422      6      good
chromium@463335      85.6037 +- 19.2263       6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8977590159007361216

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5237165224624128


| 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 Speed>Bisection.  Thank you!

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8977404477681800400

Blocking on crbug.com/728293 for the failure

=== Auto-CCing suspected CL author bnc@chromium.org ===

Hi bnc@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : bnc
  Commit : 396d33b1f92f90f9cdecd880748bf489954691e7
  Date   : Mon Apr 10 16:40:18 2017
  Subject: Add operator<<(ostream, AlternativeService) for test output formatting.

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : get-attribute-rare/get-attribute-rare
  Change       : 2.78% | 91.5226523675 -> 91.174039876

Revision             Result                    N
chromium@463289      91.5227 +- 19.5969        14      good
chromium@463290      28.8603 +- 0.131971       6       bad       <--
chromium@463291      28.8584 +- 0.0675753      6       bad
chromium@463292      28.8516 +- 0.0398264      6       bad
chromium@463295      28.8584 +- 0.0545755      6       bad
chromium@463302      28.7712 +- 0.0425146      6       bad
chromium@463312      88.422 +- 31.4529         14      bad
chromium@463335      91.174 +- 4.34392         9       bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8977404477681800400

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5237165224624128


| 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 Speed>Bisection.  Thank you!

1. Suspected culprit does not make sense to me: in that CL, I introduce an operator<<() that gets compiled in the production binary, but it is never actually executed.  (Maybe the linker can remove it on that basis anyway?)  I cannot imagine the perf benchmark calling this operator, it is meant for pretty-printed output from failing tests.

2. In Comment #6, there's a regression from chromium@463289 to chromium@463290, but shortly after, at chromium@463335, result gets back to the same level.  Why would it do that if my CL was indeed causing it?

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8976986436667975424

Agreed with #7. Kicking off a bisect on a wider commit range because the graph is pretty noisy. cc-ing jbroman, owner of blink_perf.bindings, to see how deeply this should be investigated. Only seems to occur on one device.

=== BISECT JOB RESULTS ===
Bisect failed for unknown reasons

Please contact the team (see below) and report the error.


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : get-attribute-rare/get-attribute-rare


To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8976986436667975424

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5793229775044608


| 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 Speed>Bisection.  Thank you!

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8976879739219102032

=== BISECT JOB RESULTS ===
Bisect failed for unknown reasons

Please contact the team (see below) and report the error.


Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : blink_perf.bindings
  Metric       : get-attribute-rare/get-attribute-rare


To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.bindings

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8976879739219102032

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5793229775044608


| 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 Speed>Bisection.  Thank you!

Doesn't look like we'll be able to make progress here.