InformAction Forums

Windows 7 x64
Firefox 4 RC1
NoScript 2.0.9.9

Steps to reproduce:
1. Load: http://cmap.m-plex.com/hb/hwymap.php?mt ... showint=30
2. View each layer in this order: Mapnik (starting point) > Osmarend > CHM > Yahoo Map > Yahoo Aerial > Bing Map > Bing Arerial > Bing Hybrid > Map > Satellite > Hybrid > Terrain > (back to) Map
3. Pan around the USA twice (around the border) using the mouse in the big window.
4. "center" the map back at the starting point via the hand in the middle of the panning circle.
5. Zoom in all the way with the "Zoom In" (the "+") button.
6. Zoom all the way out using the "Zoom Out" (the "-") button.
7. Repeat Step 5.
8. Repeat Step 6.
9. Firefox should dump some memory. (wait about 1-2 minutes)
10. Hit the Recenter button so it's zoomed back in.
11. Do another two laps panning around the USA (like in step 3) and then recenter again.


Safe Mode Memory Usage:
0. ~ 45,000 K
1. ~ 63,600 K
2. ~309,200 K
3&4. ~149,100 K (Hit max of ~425,000 K before dumping memmory half way in 2nd loop)
5. ~236,500 K
6. ~312,900 K
7. ~343,900 K
8. ~343,900 K
9. ~164,000 K
10. *didn't get this one because I wasn't paying attention*
11. ~181,400 K


Fx with only NoScript activated Memory Usage (no plugins active as well):
0. ~ 52,600 K
1. ~ 87,100 K
2. ~387,200 K
3&4. ~265,500 K (Hit max of ~625,000 K before dumping memmory half way in 2nd loop)
5. ~374,500 K
6. ~444,600 K
7. ~418,100 K
8. ~435,500 K
9. ~412,500 K (no major memory dump when NoScript is activated unlike in "Safe Mode")
10. ~430,700 K
11. ~891,600 K !!!!

Now, once I finish on the NoScript pass, I can continue panning if I want to after the recenter and the memory useage in Fx continues to go up. I stopped at around 1,118,100 K! If I continued going, I bet I could have broken 2GB worth of memory if I had felt like continuing. So, it seems that NoScript triggers a MAJOR memory leak in Fx when using this website somehow. I'm no computer programer, but if you think this also needs to be reported to the Fx people, please go ahead and do so and post a link to the report here so I can give them any extra info if needed.

It might be https://bugzilla.mozilla.org/show_bug.cgi?id=639186
Could you check whether this happens also with Adblock Plus + the Easylist filterset?

I've honestly never have had Adblock Plus installed in my copy of Firefox.

But I'll try it later today and report the results.

AB+ ONLY Memory Usage:
0. ~ 66,600 K
1. ~100,600 K
2. ~350,200 K
3&4. ~225,300 K (Hit max of ~574,000 K before dumping memory half way in 2nd loop)
5. ~323,500 K
6. ~386,800 K
7. ~399,000 K
8. ~411,500 K
9. ~407,200 K
10. ~419,800 K
11. ~474,400 K

After some more random panning in the USA, got it up to ~568,100 K. However, no really crazy runaway memory usage like when NoScript was running just by itself. It took a lot of extra panning around just to get it up that high.

AB+ & NS Memory Usage:
0. ~ 78,100 K
1. ~ 108,000 K
2. ~ 394,300 K
3&4. ~ 440,700 K (Hit max of ~650,000 K before dumping memory half way in 2nd loop)
5. ~ 557,500 K
6. ~ 639,900 K
7. ~ 668,100 K
8. ~ 699,200 K
9. ~ 519,300 K
10. ~ 532,400 K
11. ~1,011,800 K

Then with some random panning/zooming, I was able to get it over 2,300,000 K. However, as soon as I stopped and started to type this (in a text file), a few seconds later it dropped down to ~1,795,500 K. About a minute later, it dropped to ~1,168,800 K. Don't know it if would drop any more since I needed to do something else. At one point during this memory leak, I was using over 70% of my Physical Memory (8GB of DDR2).

So, there is a REALLY BAD memory leak here either in NS and/or AB+ and Fx.

The difference in numbers may be due to the fact NoScript checks and blocks much more stuff than ABP, but the underlying Firefox bug seems to be the same.

Could you please file a report in Bugzilla, adding a "maybe related to bug 639186?" clause and CCing some key people from that bug?

Done.

https://bugzilla.mozilla.org/show_bug.cgi?id=642472