Just a point of information^*.

AdBlockPlus is currently known as causing memory issues with Firefox. Firefox and AdBlockPlus developers are aware of the issue.

• Bug 1001426 - 150MB memory usage spike by enabling Adblock Plus
• Bug 988266 - Very high style-sets memory usage with AdBlock Plus
• https://adblockplus.org/blog/on-the-adblock-plus-memory-consumption

I don't know what you call "currently", but this problem is not a problem that has suddenly appeared a few weeks ago. It has been increasingly bad for the past 2 years.

This issue is not something I am following in any detail, but it would not surprise me to find it has been a problem with AdBlockPluss (ABP) & Firefox for two or more years.

It may be something that affects some users more than others, for instance certain pages with many I frames may cause particular problems with the ABP issue IIRC .

Memory & CPU usage issues are possibly often multi faceted, I am just mentioning one factor that you may need to be aware of.

I believe upthread you considered restarting Firefox to reduce memory usage, & note you have tried using about:memory. Have you tried using the buttons within about:memory that reduce memory usage ? That may help you stay below the 1.5GB mark.

There's a bit of confusion about who said what :)

I'm the OP and I also mentioned about:memory

To answer the question -- under no circumstances I saw buttons in about:memory actually have any impact on Firefox memory usage. It seems that whatever automated mechanism is being used, it is comparable in its efficiency with those buttons. Or, in other words, these buttons in no way help to stay below 1.5 GB memory 'limit'.

Anyway, it now seems to me that 1.5 GB 'limit' is nothing like that -- it is merely an indication of the fact that you've opened quite a bunch of tabs.

The real problem seems to be that Firefox simply cannot properly handle many (25+) open tabs simultaneously. See, for example, reproduction steps I've posted on the clean profile.

I just tested the same with NO addons (even without FlashBlock) using about 25 rus.delfi.lv tabs and maps.google.com streetview. With only one google maps tab open panning in street view is very smooth (hold down LMB to grab picture and drag it back and forth). With 25 tabs of rus.delfi.lv in background -- street view is nowhere near as smooth.

I guess at this point this can be simply registered as a bug against Firefox since it is very easy to reproduce and happens every time.

P.S. wartab -- thanks for providing info about your own experience, it's helpful to know that it is not isolated to my own machine.

Hi solf, Others are already involved with answering this thread. I had only really intended to draw attention to ADB as it may be relevant to some of your problems.

Yes I had noticed you asked this question and had used about memory. I did not notice you mention use of about:memory's buttons.

My mention of the 1.5GB mark was only because you seem to have found that significant with your issue. (And I guess especially with memory fragmentation it is quite feasible to see performance hits on 32 bit Firefox at that level ). Are you noticing that Firefox is resorting to using memory paging ?

I do not often use (32 bit) Firefox Windows but I have used way over 25 tabs at times, & I have even tested getting up to hundreds of tabs on Windows , and have seen others mention use of hundreds of tabs. Ordinarily I would not expect Firefox to have severe problems with 25 tabs open.

The results possibly depend slightly on the tabs used, as I imagine some websites and actions could be a lot more memory hungry than others. Also it could be affected by the addons that are in use, ABP being one example that has noticeable affects on memory use.

I think in posts upthread, you explained the Steps To Reproduce (STR) & Moses tried your STR and got high memory use but not any problems with hangs or scrolling.

If you consider it is a bug in Firefox it probably helps to write STR to demonstrate is using all of the following as a starting point

1. Windows Safe mode with Networking
2. Firefox Safe Mode
3. A new clean profile with all plugins disabled.
4. Using the current Release (or better still a pre Release Nightly)

• If you consider it is a regression and Firefox has changed badly and developed a fault repeat and compare with an earlier version such as Firefox 28 in a similar configuration.

If it would assist you I am sure; if necessary; one of us will be able to give further suggestions on how to successfully file a bug

• It may help to look at
  • https://quality.mozilla.org/docs/bugzilla/starter-kit/how-to-write-a-proper-bug/
  • https://developer.mozilla.org/en-US/docs/Mozilla/QA/Bug_writing_guidelines

As another point of information. Consider turning on the Firefox Telemetry feature, that collects and reports anonymised data. Such data is available publicly, and is used by developers. IIRC the system was developed after a memory regression was introduced into Firefox 4 and missed; going without proper recognition for quite some time.

• Share data with Mozilla to help improve Firefox
  • Interactive public results from aggregated data (Dashboard) http://telemetry.mozilla.org/
  • Your own data stored locally (if FHR is on and has been collecting data for awhile, even if not submitting results) use about:health
    See also Firefox Health Report - understand your browser performance

Hello,

thank you for your reply John99!

First of all to answer you question -- in my case I do not have swap file at all -- I have 16GB of memory. So disk paging cannot be an issue.

I think the most likely cause of Moses not seeing any problem is that different people may have different sensitivity to 'scroll smoothness'.

So with that in mind let me propose another way to reproduce & see the problem that doesn't require any plugins or changes to your installation.

I would appreciate if people actually try this and report the results -- it shouldn't take much time at all. I'm also quite curious if this also happens on non-Windows systems.

Download Firefox Portable install. I'm using FirefoxPortable_29.0.1_English.paf.exe -- but this problem has been 'with us' at least since version 17 ESR or so (haven't paid much attention before) -- so let's say anything 17+ should work for this experiment.

Install Firefox Portable somewhere.

In the installation dir create FirefoxPortable.ini file and put the following lines there: [FirefoxPortable] AllowMultipleInstances=true

[optional] Install FlashBlock plugin for this portable installation -- problem is going to happen regardless, but FlashBlock demonstrates that it is not Flash-related.

Now copy installation to some other place -- so you have two independent copies.

Open both copies -- you should have two separate Firefox windows. In each one open maps.google.com and open street view for some place (preferably the same place).

Try panning in street view -- i.e. click and hold LMB and pan the view around -- see how smooth it is.

Now in one instance (without closing the street view) open http://rus.delfi.lv (the only reason I'm using this site is that it is quite 'heavy' and thus you can see the problem quicker).

From rus.delfi.lv page use middle mouse button (or whatever else) to open about 25 different links (so that you have about 25 different pages from rus.delfi.lv). You might want to avoid links that have "ВИДЕО:" in front as these might auto-play videos.

Now go through all the tabs that you have (I use ctrl-tab) and hit End-Home-End-Home -- this essentially causes each rus.delfi.lv page to attempt to load the entire content (it's lazy-loading normally). This is probably the most time-consuming step, so I'm open to suggestions as to what other content & JS heavy site can be used for the demonstration purpose.

Now go back to the street view page. Compare how panning compares between two instances (one that has only street view tab and one that additionally has 25+ rus.delfi.lv tabs). The difference should be quite noticeable.

If you feel that you can't see significant difference -- try opening more rus.delfi.lv tabs.

And let me stress that this is not rus.delfi.lv specific. It also happens in 'normal usage'. rus.delfi.lv simply allows me to easily reproduce the problem.

As a final test step you can also do the following -- select the last open tab and start quickly pressing ctrl+w (close tab). You should observe that Firefox 'freezes' (for something that feels like up to a second) when quickly closing tabs. The effect should become less and less (and eventually disappear) as there are fewer tabs still open.

Firefox portable is for Windows only.

I however have no problem running multiple versions of Firefox simultaneously or running the same version side by side with different profiles, however with only an old low powered laptop with 4GB ram I possibly may not expect good performance with 2x Firefox using >1.5GB Ram each. Also if you consider the issue to be related to Firefox 32 bit, I only use 64 bit versions on Linux.

Portable version is simply an another way to run multiple profiles.

For test you'd only need one firefox instance consuming ~1.5GB or memory -- the other instance only needs one tab.

It'd be interesting to know if your 64 bit version on Linux is affected.

Out of curiosity I've tried 64 bit nightly build for Windows (portable version). It seems to work much worse (less smooth, more cpu usage) even with a single tab and the problem is even more pronounced with many tabs open.

I have the exact same problem. Although I'm not surprised about the high RAM usage, with the amount of tabs I've got opened I would expect that, but every time after 1.5GB is passed Firefox becomes slow/sluggish even though there's plenty of RAM available, I've got 16GB, usually ~70% available.

In addition to that: A while after passing the 1.5GB, another problem occurs, something which started with Firefox v29: "This image cannot be displayed because it contains errors", that error starts to appear regularly after passing about 1.8GB. Firefox restart and everything is back to normal, till RAM usage gets above 1.5GB again....

Edit: And no, I do not have AB+, just noscript, download status bar, status4evar, classic theme restorer, session manager, click to play per element, showip, nothing known for memory leaks.

There is a version of portable Firefox for Mac OS. No idea about Linux.

As far as Google Maps goes, I find the classic interface better than the new one. You can get into that from here:

• https://www.google.com/maps?output=classic

I really ought to go and fill proper bug report on this.

But it is hard to find motivation since I find it unlikely that it'll get fixed. And if it will, it will probably be fixed in new versions -- and I detest their newest UI changes (not to mention that it'll break all the most important addons I need to make Firefox usable since it's rather useless (for advanced use) out-of the box).

I agree with that. That is why I am now using SeaMonkey as my regular browser. It has a much better interface. I even got 7 extensions for Firefox converted to work in SeaMonkey (in addition to the other 18 that were intended to be used with it).

I must say on some versions and profiles of Firefox I currently often seem to get pretty bad performance.

I am however having difficulty reproducing consistently or demonstrating any memory leak or any regression between Firefox Versions. I am certainly not yet ruling out an OS problem or interaction with some other software.

What I can say is that with many open tabs and windows Firefox may become unusable with even typing being so slow as to be useless. The notable factors are relatively high CPU use ca 50% and high memory >50% the most notable factor is swap file use requiring restart of Firefox or the OS before it will clear.