http://mrcote.info/phonedash/#/org.mozilla.fennec/throbberstop/local-twitter/2012-12-31/2013-01-07 shows a small (< 400 ms) increase in the time to throbber stop, for all phones, on Jan 3. For example, 78_d6_f0_89_af_9a_nexus_s has a time of 5241 at https://hg.mozilla.org/mozilla-central/rev/33064e13c3fd and a time of 5631 at https://hg.mozilla.org/mozilla-central/rev/4e18ac9b51e2. See also bug 825612 -- a similar regression on Dec 22.

Do we have some kind of script we can run to easily extract the times from local builds? It would make bisecting much easier. Also, why not make this a talos test now? I'd give up the current Ts to get this one used instead. The regression are too hard to bisect from daily builds.

I do not know of such a script. A talos test for "throbber start" / "throbber stop" would be cool, but of course would only run on tegras/pandas...I wonder if there is any chance of making phonedash test on every push? :mcote -- any thoughts?

You mean on mozilla-inbound right? Right now it's possible, and we also have the ability to manually run a range of builds if we want to file in some holes. However, this *won't* be possible once we deploy the unit tests, since they take a very long time to run on the phones. We're actually going to have to switch to running nightly, not even against m-c, at this point. We could conceivably keep a few phones strictly for the start-up tests, though, if we want results more often.

Chris, Brian is looking at bug 825612 which is similar, you may want to collaborate

filter on [mass-p5]

2013.