In the work to move android Panda talos tests to run with mozharness and mozpool in bug 829211, aki suggested that instead of using talos.zip we should use mobile_talos.json (not talos.json). jmaher generously offered to make a first attempt at this file in #ateam

Did we end up agreeing that we would try to always bump the desktop's revision at the same time as the mobile's revision? Or we would try to fix any discrepancies as the code paths diverge and try to catch up? I think we could keep the revision in one file and the test definitions in two different files. Or even all together in one single file. I will not fuzz/complain if we stay with current approach.

Armen: My understanding is that the plan is to try to keep the same revision. Here is the discussion from #ateam starting at 18:38 http://logbot.glob.com.au/?c=mozilla%23ateam&s=10+Sep+2013&e=10+Sep+2013

I have successfully cleaned up the talos.json in m-c for desktop in bug 916045, now it is time to tackle this bug.

and this looks good on try server as well: https://tbpl.mozilla.org/?tree=Try&rev=c782de38014c

Example of what talos.json will look like with the mobile tests added to it. While I was not able to test it this should be a good starting point for when we have code that actually uses it. I will ask for review once we have the talos changes landed.

Comment on attachment 804985 [details] [diff] [review] update talos test defaults for mobile Review of attachment 804985 [details] [diff] [review]: ----------------------------------------------------------------- lgtm

landed on talos: http://hg.mozilla.org/build/talos/rev/473c3513c97e

this will require that we land 917315 first, then we can land this on inbound and let it roll to the other branches.

Comment on attachment 806034 [details] [diff] [review] update talos.json to include mobile talos definitions (1.0) jmaher: Isn't remote-tsvg missing from the mobile-suites?

remote-tsvgx has replaced remote-tsvg. I realized this week that tsvg is still running on buildbot, that was my mistake.

I just got a Talos perf regression (probably the same one you just got) about a 100% regression on Mac OSX 10.8 CanvasMark. The regression range included my patches (which shouldn't have affected perf either way) and this patch, which does touch talos. Any chance you think it was this patch?

Oh, hah, looking at http://mzl.la/14jEJVd, the reading went to 0.

yes, I should have mentioned that in here. CanvasMark went from broken (i.e. maxint which is a great number for it to post), to working (i.e. 7000) which ends up as a regression, but it really isn't.

Great, phew :)