Dromaeo numbers keep dropping (higher is better in this case): http://graphs.mozilla.org/graph.html#tests=[[72,131,35]]&sel=none&displayrange=30&datatype=running There is a regression from revision 277bacd9f2f8 on June 11th: https://tbpl.mozilla.org/?tree=Mozilla-Inbound&fromchange=e54cc67cf015&tochange=c11c022b8495&jobname=Ubuntu%20HW%2012.04%20x64%20mozilla-inbound%20talos%20dromaeojs I did some retriggers and you can see the value of dromaeo css goes from 5150-5300 down to 5050-5150, a small but still a measurable regression.

njn: could this be a regression from JS memory reporter bug 1023719? The only other changesets in comment 0's pushlog that look like they might be related are h4writer's Tracelogger changes (bug 1009603).

I would be very suprised if this is Tracelogger, since it isn't enabled by default, the compiler should optimize all tracelogger function calls away.

I am pretty sure it is from the js memory reporter as there are enough retriggers with a good pattern of keeping all values in the lower range. Even with that said, there could be other randomness invoked in the world causing confusion.

> njn: could this be a regression from JS memory reporter bug 1023719? It seems extremely unlikely. The code changed by that patch only runs when memory reporting occurs. Also, the patch from that bug has been backed out. Did we see an corresponding un-regression?

I couldn't find the unregression- maybe there is something else in play here. Thanks for the info.

this didn't show up on the aurora uplift, I recommend resolving as wontfix

Marking WONTFIX as per comment 6.