All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=793460

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=118c2dafa87c02832a42f3f9d99594d54a36ccc2357aaf2071912119b617bc4a


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

chromium-rel-mac11

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a5e22c040000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/15f219a8840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/13a0c044840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12906ad8840000

Kicking a few more bisects since this does seem to repro, but not sure what can be done 1.5 months after landing for a relatively small regression.

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/13a0c044840000

Avoid unnecessary background image tiling
By fmalita@chromium.org · Wed Dec 06 18:00:13 2017
chromium @ d46daceb7047d0ab92a164a062fa26e31baae377

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

I'm guessing that the tiny regression here is due to skipping the rounded-rect shader optimization (drawing border-radius backgrounds using shaders instead of clips).

Since the regression is small and the behavior change is intentional (we do want to avoid shaders when possible due to texture size restrictions on mobile devices), I propose we let this one slide.

📍 Found significant differences after each of 7 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/15f219a8840000

Added |is_fatal_cert_error| flag to ssl_info
By carlosil@chromium.org · Wed Dec 06 17:18:45 2017
chromium @ 811333893d25ec4c391e6dd968e4960c08251d4c

[Chrome Home Survey] Return early if infobar tab is null
By danielpark@chromium.org · Wed Dec 06 17:35:45 2017
chromium @ d4571659f6984d4f888b8ee8fffde41f4568e1cd

Replace base::MakeUnique with std::make_unique in content/
By gyuyoung.kim@lge.com · Wed Dec 06 17:47:08 2017
chromium @ f4e2e65212a72810646656c41e68b359c426fa0d

Avoid unnecessary background image tiling
By fmalita@chromium.org · Wed Dec 06 18:00:13 2017
chromium @ d46daceb7047d0ab92a164a062fa26e31baae377

Reland: Media Controls: Double tap on either side to jump
By beccahughes@chromium.org · Wed Dec 06 18:16:11 2017
chromium @ f2c186263631aa5acd49fda880c21630474e7dbf

Add base::debug::ScopedCrashKeyString.
By rsesek@chromium.org · Wed Dec 06 18:46:16 2017
chromium @ 5799f48f4759cc4e47125f35310d266253f659f6

OOP HP: Add a flag to keep small allocations.
By erikchen@chromium.org · Wed Dec 06 19:06:25 2017
chromium @ 8bb76b5830f9187f17a1c1f8ead4b9e8c179a11f

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

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12906ad8840000

Reland "Snap at GestureScrollEnd on main thread."
By sunyunjia@chromium.org · Wed Dec 06 19:18:03 2017
chromium @ db77c44916a84b82327d64608bfb8ef2dc4ff090

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

The CL found does not have any relation to the regression. Marking as Wontfix again per comment #8.

In fact it is unfortunate the bisects has been finding irrelevant commits as in #9, and #11. I wonder if this is because the metric is noisy? /cc vmpstr@chromium.org, wkorman@chromium.org

Removing everyone else which got cc'd incorrectly to this issue.

vmpstr@chromium.org, wkorman@chromium.org: See above for my concern about this metric.

+dtu for ideas from bisect side for reducing metric noise

The graph in comment 9 -- the line doesn't go up and down randomly; there are clear steps in the graph. Is it possible that all of those CLs really do have a minuscule effect on this metric?

The graph in comment 11 does appear to move up and down randomly. But I suspect if Pinpoint was more sensitive to small changes in the metric, it would look more like the graph in comment 9.