Olli asked me to file this after it came up in a Facebook conversation. Google Sheets leaks memory in Firefox (Dev Edition, currently 46) on Mac. It leaks it a LOT faster if your login cookie for Google expires while it's open. I remember looking into it at one point and it looks like the part when you get logged out is because of trying to connect to Google Talk for the chat engine. It repeatedly tries to re-connect, gets told to go away because you're not logged in, and keeps trying. And apparently doesn't throw away whatever objects it created during the attempt and just makes new ones on the next attempt. I can leave my laptop sit overnight and have it get logged out at some point during the night, and I come back in the morning to find Firefox using 13 GB of RAM, Firefox itself completely unresponsive, and the laptop dog slow until I kill Firefox. I'm not sure what's causing it when you're not logged in, but that's nowhere near as pronounced. The normal leakage takes 2 or 3 days to get to 3 or 4 GB of RAM usage, but it gets there eventually.

Does anyone happen to have an example spreadsheet. Is this perhaps the issue Nathan has seen.

unfortunately all of the ones I regularly reproduce this on contain private data :(

is this bug in the right product/component? its not composer like the old web page editor. thinking this should be core gecko, right? maybe general until we could identify which part of gecko might still have a leak? also probably needs some help tracking down a regression range inside 46 or previous.

Some example here would be great.

Also, justdave, could you perhaps create CC/GC logs once you see this happening, hopefully when the memory usage hasn't gone up much yet. Both verbose and concise in about:memory. The logs may contain something privacy sensitive, like URLs. Feel free to send link to the logs to me (and mccr8) privately.

Bug 1249739 is another bug about high memory usage on Google Sheets. Before cycle collector logs, I think a first step would be to get an about:memory report. You can make it anonymized if you would like.

I see a regressionwindow-wanted tag on here, not sure if this is a regression, seems like it's been this way for a long time. Anyway, just had it happen again. See attached memory dump (anonymized). Interestingly enough, the top few tabs in the list were docs.google.com, not sheets.google.com. But I do see something interesting in the system process list (see next attachment)

screenshot of process list showing Google Talk plugin using massive memory alongside Firefox.

And after generating that report and attempting to quit Firefox, it crashed while attempting to quit. bp-2be10b89-66d9-46a9-9c04-393e72160330

Could this have been a dupe of bug 1249739? Have you seen this recently, Dave?

(In reply to Andrew Overholt [:overholt] from comment #10) > Could this have been a dupe of bug 1249739? Have you seen this recently, > Dave? That's quite plausible. I have not seen it recently, but I've either not had spreadsheets left open or have been restarting Firefox frequently for other reasons. I'll try to leave something open intentionally and see if I can trigger it again.

I'm still seeing this in 51.0a2 (2016-10-31) (64-bit), so probably not a dupe.

Thanks for confirming, Dave.

There wasn't anything that looked out of the ordinary in the about:memory log. Does this reproduce in Chrome? The page may just hold onto a lot of data under some circumstances.

I have had a Google Spreadsheet open for a few days now and haven't seen any memory issues. Can you provide a copy of the spreadsheet you've seen this with, Dave? Maybe with anonymized data in it? I can try again to reproduce your exact setup.

I actually haven't seen this happen in a while myself.

(In reply to Dave Miller [:justdave] (justdave@bugzilla.org) from comment #16) > I actually haven't seen this happen in a while myself. I guess that's a good outcome :) Please continue reporting things as they're super-helpful. Sorry we didn't get anywhere here.