All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=889061

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=3f56d6001856db3c4efdd771420a8fd66efb93eb6a23832b92fa258cfe16da00


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

Android Nexus5 Perf
Android Nexus6 WebView Perf
android-nexus5x-perf

blink_perf.parser - Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a90757640000

📍 Found significant differences after each of 4 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14a90757640000

Reland "Roll clang 340925:342523." by thakis@chromium.org
https://chromium.googlesource.com/chromium/src/+/bbb140b822e8ae819be4fae81c2c7e713a7a5e8e
6.31e+04 → 6.572e+04 (+2625)

Remove support for ETC1 tiles by piman@chromium.org
https://chromium.googlesource.com/chromium/src/+/09c2f5373c61735db0dfedffb96608212f0d9e44
6.513e+04 → 6.58e+04 (+671.4)

Remove the runtime flag for test reports. by paulmeyer@chromium.org
https://chromium.googlesource.com/chromium/src/+/96a4668d39b28f933b44801b594aa91cdd32f87b
6.538e+04 → 6.586e+04 (+479.4)

Prevent renderer initiated back navigation to cancel a browser one. by arthursonzogni@chromium.org
https://chromium.googlesource.com/chromium/src/+/a7d715ae5b654d1f98669fd979a00282a7229044
6.586e+04 → 6.715e+04 (+1291)

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

Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

Wow, the pinpoint UI is so great now, congrats!

The code my CL modifies should not be reached. These are blink parser tests, they aren't doing any history navigation.
Maybe we can launch pinpoint only on my CL and see if it reproduce again, can we?

I don't see why pinpoint finds a significant differences with the other CLs except with the one about using a different clang version. It looks quite likely to have changed how the code is optimized.

This is likely the clang roll. We had to temporarily disable ThinLTO, so this is expected. It should recover when we turn it back on.