I'm experiencing excessive CPU and ram usage on OS X on a regular basis. At the time About:memory was captured, FF was using 5gb ram and 105% cpu. (overachiever!) Machine is a 2012 Macbook Air w/2GHz Intel Core i7 and 8gb ram running Mountain Lion. About:memory?verbose - https://gist.github.com/a39152fc0be086729ace About:support - https://gist.github.com/0a16e2e403e79d4bd10b
The two things that stand out in the about:memory dump are the huge heap-unclassified (5GB/7.4GB) and the huge number of zombie compartments. I tried to reproduce this with most of her add-ons installed, and it's not especially easy to reproduce the zombie compartment issue. It seems to only happen with certain sites. I'll try to figure out what add-on is causing that when I can figure out how to reliably trigger it, though it may not matter after 15 is released. The heap-unclassified issue apparently takes some time to manifest.
Lisa, if you can narrow it down to a single add-on, that would be very helpful.
So, the crazy number of like button compartments seem to be coming from Google Reader. I get them without any add-ons installed, and they go away when the tab is closed. That might be enough to explain everything, but 1Password is a Jetpack add-on, so it's possible that bug 780391 is also coming into play here and we're leaking subframes until the Reader tab is closed, in which case we need to retriage that bug.
Also, it would be interesting to know if you are getting any messages in the error console that say "TypeError: can't access dead object".
Ok, so these compartments seem to pile up on Google Reader whenever a pane is refreshed with or without 1Password. However, with 1Password, we seem to be getting 3 content script sandboxes for every one of those compartments... Moving to Core > General and opening up for retriage, since this doesn't seem to be mainly an add-on issue. It might be a good idea to open another bug for the vast number of content compartments created by 1Password.
Thanks for the analysis, Kris! That kind of sounds like a dupe of bug 750048.
Agreed, it seems like a dup. I'm not sure if the heap unclassified issue is related, though. We can reopen if that seems to remain when bug 750048 is fixed.