All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=608488

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICglNfBpgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICglJOdrwoM


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

Google Chrome Linux x64

This is a pretty small regression, but let's at least try to confirm where it's coming from.

The revision range is:
https://chromium.googlesource.com/chromium/src/+log/0b4adfd25e307a8a82541556690c9a3e0f4d8b30..5227a7901b098f82a15038cf0d3418fe0711d750

eae@, any change the HarfBuzz roll could have caused this?

Yeah, that seems likely. We're slowly moving some functionality out of Chrome and into harbuzz. During the transition the binary size will be a little higher due to code duplication.