All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=876041

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=57ceb0018705acb5b2ef0e5b1e600e8e4a414f6f1fcbaa3a1d17af7d3f1e2d94


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

Win 7 Perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/13d7bb9e640000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/13d7bb9e640000

The swarming task expired. The bots are likely overloaded, dead, or misconfigured.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1697b7e5640000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/1697b7e5640000

[CI] Cleanup paint invalidation flags by wangxianzhu@chromium.org
https://chromium.googlesource.com/chromium/src/+/8a53cc3a814bd1a1027118909458d4bf3936137c
8.323 → 10.23 (+1.902)

Separate TextMetrics advances from CharacterRanges (ligatures fixed). by davidqu@chromium.org
https://chromium.googlesource.com/chromium/src/+/3c61aca7b5c67ebdee22f5e146f215d724249fae
10.88 → 12.43 (+1.549)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

There are actually two regressions in the range. My CL did cause one of the regressions, as well as other performance regressions reported in  bug 875321 . The CL was reverted, and then relanded with the performance issue fixed.

I've checked and it doesn't look like my CL is related to the regression. Feel free to reassign if I'm missing something.

The regression should be re-triaged instead of being closed.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/144ac5a9640000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/144ac5a9640000

[CI] Cleanup paint invalidation flags by wangxianzhu@chromium.org
https://chromium.googlesource.com/chromium/src/+/8a53cc3a814bd1a1027118909458d4bf3936137c
8.562 → 10.44 (+1.882)

Separate TextMetrics advances from CharacterRanges (ligatures fixed). by davidqu@chromium.org
https://chromium.googlesource.com/chromium/src/+/3c61aca7b5c67ebdee22f5e146f215d724249fae
11.34 → 12.62 (+1.275)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

davidqu@ I re-kicked the bisect and the result was the same. It seems the page in question involves text rendering on the canvas: https://web.archive.org/web/20110430162652/http://ie.microsoft.com/testdrive/Performance/SpeedReading/Default.html so it seems like the cl could be related?

On the other hand it's as far as I can tell it's the one metric on the one bot on the one page so unless something something obvious jumps out at you let's close and save our energy for other regressions.

hjd@ Thank you for re-verifying, I'm still fairly new at chromium development that I wasn't quite sure how to check again. 

From my cl, there is a significant change in how we calculate advances, but since they are sill experimental (behind a flag) that shouldn't affect an old test rendering text. I could try to change some of the computations to be lazy loading, but even then I'm not sure how the test is using canvas measureText(). So as you said, nothing obvious jumps out at me, I'm ok with closing it and I will keep in mind that test to compare with when I try to optimize the TextMetrics advances.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14c78b28e40000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/14c78b28e40000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12882f28e40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12882f28e40000

[WIP] Fix regression by davidqu@chromium.org
https://chromium-review.googlesource.com/c/chromium/src/+/1256378/4
10.7 → 12.45 (+1.752)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  https://bit.ly/rendering-benchmarks

I did some digging and I think this metric is rather noisy. The biggest change in my CL is around how we calculate the TextMetrics advances when calling canvas.measureText. I thought that maybe we were doing some expensive operations but I couldn't find any obvious problems so I tried to run again the test with a patch that comments-out the main function. The performance should naturally be better but somehow it end up being worse:
https://pinpoint-dot-chromeperf.appspot.com/job/12882f28e40000