Issue metadata
Sign in to add a comment
|
without network gmail leaks gigabytes of memory on chromium
Reported by
david.bo...@gmail.com,
May 16 2016
|
||||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/49.0.2623.108 Chrome/49.0.2623.108 Safari/537.36 Steps to reproduce the problem: 1. start chromium with gmail in one of the tabs 2. suspend, resume in an area without wifi (e.g. train) 3. gmail will quickly grow from ~200MiB to >2GiB What is the expected behavior? The gmail tab should stay at roughly the same size regardless if network is available or not. What went wrong? I commute by train every day and then I work on my laptop on local files, because the train does not have wifi. But I keep my browsers running. I noticed that during the train ride my laptop would always crawl to a halt after 10-15 minutes. The task manager (shift-esc) clearly identified the gmail tab as a giant memory hog, it consumed more than half of the available RAM on my laptop. Killing the gmail Tab always frees the memory and restores normal laptop performance. But it should not be necessary to kill the gmail tab every time I'm working offline for more than 5 minutes! Did this work before? N/A Chrome version: 49.0.2623.108 Channel: stable OS Version: linux mint 17.3 Flash Version: Shockwave Flash 11.2 r999 I do not know whether this problem existed earlier I am only using chromium since two weeks ago. In the past few years I was using Chrome instead, on linuxmint 16 until February, and on the horrible SuSE Leap 42.1 until two weeks ago. Chrome did not have this gmail-memory problem on those platforms: it did not blow up when running without network (and I practically always have gmail tab open). When I installed LinuxMint 17.3 two weeks ago I decided to try chromium instead of Chrome and then I started to notice this memory problem.
,
May 17 2016
Hi "a"! Thanks for the quick response. I did what you requested: switching off the extensions make screenshots. I looks like also without extensions there is a leak. I have a suspicion that it matters whether actively switch off networking or that that there are simply now wifi networks available. I have only two extensions: MP3 player 2.0.1 and the EFF privacy badger version 2016.4.13. I disabled them in the chromium settings page and restarted chromium, while I had a working wifi connection so that the pages could actually load. Then went home and on the train back I checked the task manager. It started out with its usual size (about 200 MiB), growing only very slowly, maybe 1 MiB per minute. I did not actively use gmail, except that my mouse probably hovered over it, randomly. After a few minutes I stopped watching and did other things (not using any browser; stuff in a terminal). Then 20 minutes later I noticed my laptop getting slow again and sure enough, the task manager showed that gmail had grown to 2,5 GiB. See attached screenshot, 21:17 is the in the beginning of the train ride (gmail is 200 MiB in taskmanager), 21:36 near the end (2,5 GiB in taskmanager). I edited the screen shots to blur out the full names of some of my friends that were visible in the gmail window. While I prepared this message I actively switched off wireless networking (toggle the wifi switch in the NetworkManager applet). And during 20 minutes the gmail app has not significantly grown. When I am on the train I am usually *not* switching off the wifi, so the NetworkManager app is still scanning and will connect if it finds a known wifi network. I wonder if this makes a difference for chromium. I guess it asks the OS if network is available, and maybe the answer is not a simple yes/no, but it could be something like "yes" / "nope networking has been disabled" / "no network available right now but I'm scanning and I'll let you know as soon as I find something". I will try a bit more tomorrow to see if actively disabling networking helps to prevent the gmail memory inflation. I hope that this somehow helps. Thanks, David
,
May 19 2016
Hi again, I did some additional experiments to find out under which conditions the gmail@chromium memory leak occurs and when it doesn't. My general impression is that the leak occurs whenever gmail@chromium thinks that network should/might be available while actually it isn't. * Wifi available and enabled: NO LEAK. * Ethernet available and enabled: NO LEAK. * wifi+ethernet available, but actively disabled in the NM applet: NO LEAK. * neither ethernet nor wifi available, but NM (fruitlessly) scans for wifi networks: LEAK. * ethernet not availabe, wifi available but disabled: SOMETIMES LEAK. This last scenario I need to check this better. In my previous message I claimed that this scenario did not result in a leak, but today it did. I suspect that it might depend on connectivity history, because today I entered this scenario in a different way than on Tuesday. * Tuesday: no ethernet plugged in, wifi available but disabled. Last available network was over wifi. Result: NO LEAK. * Today: no ethernet plugged in, wifi is available but disabled. Last available network was over ethernet. Result: LEAK. I am not sure what to make of this and if any these diagnostics are in any way helpful to you. Maybe none is. :) One other suspicion/idea I have is that it might matter whether gmail is in the "front tab" or not. In the TaskManager I see that the gmail tab memory is jumping up and down by a few MiB all the time when that tab is the front tab, but it seems to be a lot more stable when it is not the front tab. That might become the topic of a later message in this thread. (Let me reiterate that for all these observations I am not actively using gmail, it is just sitting there. The only interaction is that occasionally my mouse will *hover* over different parts of gmail. I am not clicking.) Have a nice day, David
,
May 20 2016
Minor update: my OS updated the chromium-browser package, the "about:" now says: ############################################################################## Chromium 50.0.2661.102 (Developer Build) Built on Ubuntu 14.04, running on LinuxMint 17.3 (64-bit) Revision 496034c629901db4564ba5613e2d736635eeb082 OS Linux Blink 537.36 (@496034c629901db4564ba5613e2d736635eeb082) JavaScript V8 5.0.71.48 Flash 11.2.999.999 User Agent Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/50.0.2661.102 Chrome/50.0.2661.102 Safari/537.36 Command Line /usr/lib/chromium-browser/chromium-browser --disable-new-tab-first-run --enable-user-scripts --ppapi-flash-path=/usr/lib/adobe-flashplugin/libpepflashplayer.so --ppapi-flash-version --enable-pinch --window-depth=24 --flag-switches-begin --flag-switches-end Executable Path /usr/lib/chromium-browser/chromium-browser Profile Path /home/boersma/.config/chromium/Default Variations 16e0dd70-3f4a17df cf5ed6e1-3f4a17df 775ebbd7-3f4a17df dd4da2fc-3f4a17df 64cbdfc2-3f4a17df 4ea303a6-3f4a17df fa99bb73-3f4a17df dbffab5d-f51b51 30e679f-3f4a17df c8b9b12d-d93a0620 867c4c68-3f4a17df 12a73824-3f4a17df b0dc61a1-3f4a17df ############################################################################## This update does not seem to change anything to fix the memory leak problems on my system: during my morning commute the gmail tab still managed to inflate itself to about 2 GiB before I killed it in the TaskManager.
,
May 20 2016
Thank you for the report. We are able to reproduce this issue across all platforms/channels and expecting a fix soon. We are tracking this issue through https://bugs.chromium.org/p/chromium/issues/detail?id=613658 internally (which you might not be able to access), however you can ping here for any further updates. Thanks again for providing the detailed repro steps.
,
May 21 2016
Wonderful! Glad to help. :) |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by ajha@chromium.org
, May 17 2016