Closed Bug 470969 Opened 16 years ago Closed 15 years ago

Will eat all available memory just sitting open over night. No Flash

Categories

(Firefox :: General, defect)

Product:
Firefox
Component:
General
Version:
3.0 Branch
Platform:
x86
OS/2
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED WORKSFORME

People

(Reporter: ke6kzj, Unassigned)

Details

User-Agent:       Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.9.0.5) Gecko/2008121717 Firefox/3.0.5
Build Identifier: Mozilla/5.0 (OS/2; U; Warp 4.5; en-US; rv:1.9.0.5) Gecko/2008121717 Firefox/3.0.5

FireFox 3.0.3 with 1 or more tabs open and nothing else happening on the computer in 8 hours or so can grab all available memory. It is not totally consistent, but happens so often that I often revert back to using FF 2.x to keep from the problem. As far as I can tell there is no Flash running. All the tabs are fully loaded. The modem is hung up so nothing is updating.

Reproducible: Always

Steps to Reproduce:
1.Open FF3 and 10:30 PM. Load the following tabs:
http://www.wunderground.com/cgi-bin/findweather/getForecast?query=Big+sandy%2C+TX
http://www.intellicast.com/IcastPage/LoadPage.aspx?loc=usa&seg=LocalWeather&prodgrp=SevereWeather&product=RadarSummary&prodnav=none&pid=none
http://www.duats.com/graphics/sfwxcur.html
http://www.duats.com/graphics/sfwxpan1024.html
Available memory: 186844 KB
Check at 6 am next morning
Available Memory: 129696 KB
Nothing was running between times of checking.

At 8 am same morning after using the computer I added one tab:
http://www.intellicast.com/IcastPage/LoadPage.aspx?loc=usa&seg=LocalWeather&prodgrp=SurfaceMaps&product=SurfaceAnalysis&prodnav=none&pid=none
Available Memory now: 123396 KB.

Check at 6 pm:
Available Memory: 80800 KB

11:30 PM after using FF3 for a number of things closed all tabs including the
test 5 tabs. Available memory still 88000 KB. Close FF3 and available memory:
275000 KB. So with nothing even displaying it still was holding almost
200000KB of memory.

No Flash was running as far as I can tell. All tabs were fully loaded. Modem was hung up so no updating was going on.
2.
3.
Actual Results:  
See above.

Expected Results:  
No additional memory should have been used. Performed the same test with FF 2.0.0.19 and no memory leak at all. The same plugins and extensions are loaded in both versions of the browser.
Version: unspecified → 3.0 Branch
Thanks Cliff, for filing this bug. Will start a Firefox session with this now, and watch memory usage in Theseus every now and then for the next few hours.
The Theseus session didn't show me much. Memory use of Firefox increased a little bit over about 4 hours (about 1 MB, fluctuating a few hundred kB), but not even close to the 4 MB per hour as originally reported.

Cliff, what add-ons are installed in your setup?
I now tried with NoScript installed and switched off all scripts in the URLs mentioned in comment 0 and switched off my router so that nothing could be loaded from the web. Then let Firefox sit there for two hours.
Before I went away, there were 257 MB RAM available with 8 MB SWAP, after I came back, there were 264 MB/8MB.
Somehow, the very first page had still tried to reload itself and arrived at an error page, after I let it fully reload, I saw 251 MB RAM available with 8 MB SWAP.

Cliff, if you have a hint how to really keep the first page from trying to reload, then I can try this experiment again. These numbers don't really tell us much.

By the way, why does your computer not fall asleep (using APM or ACPI) when you are away for so long?
Peter,

To answer your questions in comment #3, I have the screen saver turn off the display, but don't see anywere to suspend the computer. I'm using a desktop and not a Laptop. 

The first page never reloads here. NoScript blocks the redirection that would reload the page. Also you can set off the Rapid Fire function on the WUnderground page although I had it on and it still didn't reload.

Addons:
ADBlock Plus
DownThemAll
FlagFox
Flashblock
MediaPlayerConnectivity
NoScript
PDF Download
ReminderFox
Scrapbook
TabMixPlus
BugMeNot - Disabled, but not uninstalled.

Hope this helps. I added my name to the Notify List so hopefully I'll know when comments are made.
OK, I have downloaded and installed all those extensions. Of course this opens up a lot of additional possible configuration differences, but let me see what happens.

By the way, when you say "No Flash" does this mean that the flash plugin is installed and enabled (listed under Tools -> AddOns -> Plugins) but blocked by Flashblock? Please be aware that Flashblock cannot block all flash content. So you could try to disable the Flash plugin itself in that dialog and try again to let it running over night.

P.S.: APM can also be used on desktops and workstations. I have it installed and it puts my workstation to sleep whenever I am away for extended periods of time.
No Flash means that I don't see any flash content on those pages. Scripts are disabled in NoScript. Just for testing sake I've now disabled the Flash Plugin as you suggested and I'll setup a test this evening and see what transpires.

P.S.: Thanks for the info on APM. I didn't realize that APM could be beneficial on a Desktop. I'll play with it.
OK, I had a run yesterday and indeed I could reproduce that the application grows with time when these pages are loaded, with up to a few MB per hour. But it still reloads the first URL from comment 0, so that could be a problem of this analysis. (Where exactly should I tell NoScript to stop that?)

I also used iptrace to track the traffic over parts of the run. What I saw there was, apart from wunderground.com, related to safe browsing. Not sure if that would actually increase the memory consumption but that feature could switched off for a test under Tools -> Options -> Security (the two options "Tell me if the site I'm visiting is a suspected..."). Firefox 2 didn't have the same type of safe-browsing yet, therefore this might be the difference we are looking for.

Finally, I found that it's really difficult for me to run these tests. When my computer is on, it always has something else to do which would mess with the memory readings. If it's off then I am too far away to read off the memory and other people are likely to use it, too. So it may be a while before I get another chance to try this further.
Tell NoScript to 'Forbid META redirections inside <NOSCRIPT> elements" on the "Notifications" tab in the NoScript settings.

I tried turning off those two Security options and it made no difference.

I understand it is hard to trouble shoot and do anything else which is why I let it sit overnight here to see what it does. 

At this point I'm suspecting ADBlock Plus as the culprit, but it will take some more testing to prove it. Hopefully this evenings testing will tell me something. I setup a new copy of FireFox 3.0.5 with a totally new profile and no plugins. No memory loss. Added Adblock Plus and it began taking memory. Disabled it and the memory loss stopped. This eve. I'll disable AdBlock Plus on the copy of FF3 with all the other plugins & extensions installed and see what happens. If this is it, strange that the same extension in FF2.x doesn't cause a problem.
Cliff, thanks for the further testing. I haven't had time to try to reproduce again myself, but from what you write Adblock Plus really seems to trigger this problem.

Wladimir, maybe this is of interest to you. If you have any hints how to debug this further, please tell us.
This can easily happen if the page is constantly creating new elements (and hence new content policy requests) - as long as the page isn't reloaded Adblock Plus has to keep track of everything that it loads. However, right now I have these four pages open and I don't see any new content policy requests coming in. I will wait some longer of course but I don't expect anything that could be responsible for such a huge memory leak. What might be the difference between me and Scott - I have the ads (quite a few) blocked on all these sites. Scott, do you also use Adblock Plus with a filter subscription like EasyList or are you using it without any filters (meaning that it doesn't actually block anything)?
I searched for repetitive actions in Adblock Plus and apparently found the issue - overlay.js will call abpReloadPrefs() every two seconds to update UI state. If I change the interval to two millisecond I see a steady increase in used memory (Minefield build 20090129 on Windows XP). I'll check what is leaking there.
Actually, it doesn't seem to be a leak - on Windows memory goes up to 100 MB and then jumps back to 40 MB. So apparently that function call needs to allocate memory and that memory isn't given back to the operating system immediately.
I could attribute most of the memory allocation to nsIIOService.newURI() calls, four of those are performed every time abpReloadPrefs() is called. There is some more memory allocation going on but no real leaks from what I can tell. I guess that wasn't it after all...
Waldimir,

I am using ADBlock Plus with EasyList. To test my theory that ADBlock was the problem I disabled it in my "full blown" version of FF3 with all the other extensions and plugins installed and FF3 still ate memory at about the same speed so even though in some circumstances ADBlock Plus seems to use extra memory either it eventually gives it back as you saw or at least it is not the major problem I am seeing.
Firefox 3.0.10 seems a little better. It will sometimes still eat memory, but not nearly so badly or quickly as earlier releases. I am waiting for an OS/2 version of the 3.5 beta to be released so I can test it. I'll report what I find.
Cliff, could you try the current Adblock Plus development build at http://adblockplus.org/development-builds/almost-all-feature-work-for-adblock-plus-11-done? I got rid of the repetitive prefs reloading.
Cliff, the Firefox nightly from http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/contrib/2009/05/2009-05-14-11-mozilla-1.9.1/firefox-3.5b5pre.en-US.os2.zip is now already a month old, but I think this is as close to a beta as we will get within the next week or so. Perhaps you want to give that a try.
Ok. I'll give the beta and the dev. build of Adblock Plus a try and report back.
Tried the dev. build of AdBlock+ and it killed firefox 3.5b5pre. FF3.5 would not start after installing the dev. build of AdBlock+. I don't remember the safe mode startup for FFox so had to redo the profile in order to get it to restart.

Do I dare try it in 3.0.11?
3.5b5pre? I suggest you update. There was a JavaScript engine bug that was indeed crashing Firefox 3.5b4 with Adblock Plus installed, this one was fixed a while ago however.
I've been using Firefox 3.5.3 lately and it doesn't seem to have the problem, at least not to the extent as before. I'm not going to worry about it anymore unless it gets worse.
That's good news, thanks for letting us know! If it doesn't come back in the next weeks, I'll resolve the bug as WFM.
OK, almost a month has gone by, so I assume that FF 3.5 really fixed this.
Status: UNCONFIRMED → RESOLVED
Closed: 15 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.