All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=754847

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=df581b80b0fca18b6fec373920c71be950b9fa3b549bd1b13e71ab5881b53618


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

chromium-rel-win10

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8971524347669158064

=== Auto-CCing suspected CL author shaseley@google.com ===

Hi shaseley@google.com, 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 : Scott Haseley
  Commit : c3f876b18c07c50b360b93fb91ec67549b58deb8
  Date   : Thu Aug 10 00:40:36 2017
  Subject: Use swap time for all first paint times in TBM, UMA, UKM & web perf API

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : loading.desktop
  Metric       : timeToFirstContentfulPaint_avg/pcv1-warm/FDA
  Change       : 10.81% | 177.081833333 -> 196.223833323

Revision             Result                  N
chromium@493096      177.082 +- 10.7533      6      good
chromium@493147      176.66 +- 4.5934        6      good
chromium@493173      182.216 +- 8.7138       6      good
chromium@493186      181.788 +- 7.79929      6      good
chromium@493189      179.516 +- 12.7151      6      good
chromium@493190      177.847 +- 9.32694      6      good
chromium@493191      197.563 +- 4.87957      6      bad       <--
chromium@493192      199.427 +- 16.2136      6      bad
chromium@493198      196.224 +- 8.89118      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=FDA loading.desktop

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8971524347669158064


For feedback, file a bug with component Speed>Bisection

Hi,

This "regression" is due to  crbug.com/738235 , and this behavior is not unexpected. The CL causing the apparent regression changed where the underlying first contentful paint metric timestamp is captured. We now record the timestamp later in the rendering pipeline, which is causing the discrepancy. The reason for the change was to make the timestamps for First Paint, First Contentful Paint, and First Meaningful Paint more accurate, per web dev feedback. But, as a side effect it is causing the appearance of regressions since previous results were measured when the FCP timestamp was captured at an earlier point in the rendering pipeline. So, in this case performance only appeared to regress but did not actually regress.

I added a trace event that illustrates the difference between the old and new versions of FCP, and I've uploaded the data and an an image. The rail around 3090ms is the old timestamp, and the rail around 3097ms is the new timestamp. This discrepancy from moving the timestamp is why the test is failing.

Deleted: fcp_render_vs_swap 42.7 KB

fcp_render_vs_swap 42.7 KB View Download

Deleted: trace_fcp_regression.json.gz 1.3 MB

trace_fcp_regression.json.gz 1.3 MB Download

 Issue 755289  has been merged into this issue.

 Issue 755580  has been merged into this issue.

 Issue 755662  has been merged into this issue.

 Issue 759668  has been merged into this issue.

 Issue 760135  has been merged into this issue.

 Issue 754812  has been merged into this issue.