All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=629227

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgyuiD_QoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgis-ZoQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgiuDqswoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg8sr59ggM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg8qKZ_wsM


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

mac

Looks like the mac bot is not measuring "sizes" correctly. See initial description (at top).

tapted, as sizes owner on Mac, can you take a look?

Hm

> Sizes seems to increase whenever "Updating trunk VERSION from X to Y". That seems bogus.
$ ls /Applications/Google\ Chrome.app/Contents/Versions/
53.0.2783.5/ 53.0.2785.8/

The sizes bot should probably be doing clobber builds, or at least `rm -r out/[Release]/Google\ Chrome.app`. sullivan@ - do you know where the bot config for the bot we care about is?

But there's little outside of the versioned subfolder that we care about (for me it's 828kB worth), we could probably adjust sizes.py to only read the most recent version without losing too much.

+dtu - I think we do non-clobber builds because it's faster (and we want to do a build per revision), right? Is removing Google Chrome.app an option?

Came across this while sheriffing. Dupe of  Issue 631409 
I think mark just fixed this in https://codereview.chromium.org/2179013002