InformAction Forums

Hello,

I wish to report a bug :

Linux Ubuntu 14 Mate x64
Memory : 8 Go
CPU : intel i5
Noscript : 2.9.5.2
Firefox : 50.0.2
Adobe Flash : Shockwave Flash 11.2 r202

With noscript enable, set as allow global script, there is a memory leak with any website using Adobe Flash, in the plugin container of Flash. Easy to reproduce with fresh firefox profil with noscript installed.

So I go here, http://www.adobe.com/software/flash/about/, & what am I supposed to be seeing with plugin-container ?


Do you have e10s enabled?

With noscript 2.9.5.2 module enabled, set as allow global, when I go to http://www.adobe.com/software/flash/about/ then I have a quick memory leak :
Capture before : http://hpics.li/4403f38
Capture during leak : http://hpics.li/865fe14
When I disabled noscript module, then no memory leak with http://www.adobe.com/software/flash/about/
I can reproduce it with a fresh new profil in firefox.
In firefox, e10s is enable.

When this issue occurs, do you see anything related in the Browser Console? (Ctrl-Shift-J)
(if you don't know what's related, turn off CSS warnings and post everything else you see)

Win7 x64
SeaMonkey 2.46 x86 (clip is open, running, though I did not otherwise test in SeaMonkey)
FF 50.0.2 x86
NoScript 2.9.5.2rc5
Flash 23.0.0.207

URL: http://www.flashyourweb.com/gallery3/FY ... _720p_h264


1. FF + e10 + Safe Mode
2. FF + e10 + NoScript
3. FF + Safe Mode
4. FF + NoScript

(Oh, & that first instance of firefox.exe, 36 MB, that is a stand-alone version that I use as a backend in order to run ChatZilla on XULRunner.)

So... just what am I seeing?

...

Appears that NoScript greatly increases memory usage - in Firefox?
Or at least memory usage with NoScript installed in far more then with Firefox in Safe Mode.

But I am not seeing any meaningful effect to plugin-container.exe / Flash.

And of course you ask about NoScript 2.9.0.14rc1.


1. Safe Mode
2. NoScript

I don't understand your approach. You come to share that you do not have my bug, great so much better for you and the others. On my side, I have this bug which I want to inform the members of noscript. If you wish to help me to solve my issue, maybe there is another method of exchanges between you and me to achieve it?

And what do we have with that... ?

Appears the memory increase by using NoScript is unchanged (materially) between 2.9.0.14 & 2.9.5.2.

But ... what about FF Safe Mode?
About a 12 MB difference?
Is that a meaningful difference?
Could the fact that FF (in the first instance) "knew" that it could do e10s have made a difference - even in Safe Mode, compared to FF (in the second instance) "knowing" that it could not do e10s (with 2.9.0.14 not being e10 compatible)?
So knowing it could do e10s, it threw a switch, set other Prefs, that "by default" uses more RAM?

Don't know?


And then more importantly, kind of, is how NoScript might affect memory usage when you ramp up the number of windows...

So we see an initial bump on a new load with only "big bug bunny" running, but how does that scale up? If we have 10 windows open & 10 tabs in each? One time bump or might each window/tab be similarly affected, or at least to some degree?

And then once we're past that little hurdle, what happens when we have hundreds of windows/tabs open? With that, you run into a bit of a catch 22, because you're not to likely to even be able to load that number of windows/tabs without NoScript installed... If they were 100's of tabs opened in only a single window... By default, FF would load only the focused tab (fully) from startup. (SeaMonkey, you'd have to change the Pref before your attempt.)

That will have to wait, but feel free .

Well, maybe its something Linux specific, which I can't test, so you'll have to wait for others to do so.

And I think I totally missed something in your screenshots (& with what you wrote)...

Capture before : http://hpics.li/4403f38
Capture during leak : http://hpics.li/865fe14

OK, so you're saying the capture before - is before Flash initiated.
And the capture during - is while Flash is running?

Ah, if that is repeatable (with Flash using 5.7 G), that is certainly an issue.

In firefox, e10s is enable.

And if you disable e10s (browser.tabs.remote.autostart, maybe browser.tabs.remote.autostart.# too)?

And if you disable all other Plugins (Java?)?

NoScript 2.9.5.2
Firefox 50.0.2
Linux x86_64
Flash 11.2.202.632

I confirm the bug. With e10s enabled, Flash chews up so much memory that my test VM completely freezes and grinds away at the virtual HDD. I had to force power it off.

With e10s disabled, memory usage is just 21 MB for plugin-container.

Given how bad this is on my end, I'm afraid I can't investigate the bug, sorry.

DavidTerp, merci pour le rapport. Je suggère que vous juste désactivez e10s quand vous voulez utiliser Flash.

As you suggest, I disabled e10s and no more memory leak with flash and noscript installed. Done !
Thanks for your help, and I hope that my little report can help.