657759 - Firefox 4.0.1 Memory Leak

Status: RESOLVED INVALID

(Firefox :: General, defect)

Description

[E B]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0.1) Gecko/20100101 Firefox/4.0.1
Build Identifier: Mozilla/5.0 (Windows NT 5.1; rv:2.0.1) Gecko/20100101 Firefox/4.0.1

I opened up 15 TABs, each page having high graphic content (pictures). I watch MemInfo display RAM Usage climb toward 3 GB, I immediately disable a few add-ons > Restart.
It obviously starts reloading all 15 TABs upon restart, same problem with RAM.

As I see it go past 2 GB, I quickly Restart in Safe mode > It restarts and starts reloading all 15 TABs and sure enough memory usage climbs and climbs in safe mode with all add-ons disabled until it crashes Firefos 4.01. See below for details about what else I did to replicate the problem and isolate it to Firefox 4.0.1 without add-ons.

Reproducible: Always

Steps to Reproduce:
1. First I uninstalled Firefox 4.0.1 and removed all profiles, everything.
2. Then I installed Firefox 3.6.17. I watched it climb toward 2 GBs of RAM and I see it releasing memory back, it is handling all 15 TABs well, climbs up then releases RAM back. So definitely no problem there. 
3. Then I removed Firefox 3.6 and I installed Firefox 4.01, sure enough the problem was replicated on FF 4.0.1. I even used firefox.exe -profilemanager and used default profile and replicated the problem, Firefox 4.0.1 did not release memory back and continued to climb as more TABs were being added, Passed 2.6 GB out my 3 GB of RAM.

Actual Results:  
Memory Leak present with Firefox 4.0.1 but not with Firefox 3.6.17 on the same system.

Expected Results:  
No Memory Leak.

Only when the last open window is closed - does Firefox 4.0.1 release the memory back.

My system:
_____________________
Intel i7 950 [200] BCLK x 20 = 4.0 GHz @ [1.4000] CPU Voltage & [1.35000] QPI/DRAM Uncore Voltage, Batch 3029A40
3 x 1GB G.SKIL DDR3 1333 (PC3 10666) [DDR3-1691MHz] 10-10-10-24 @ 1.64 DRAM Bus Voltage  
ASUS P6T Deluxe v.1 [LGA 1366 Intel X58] BIOS 2209
Thermalright Ultra-120 eXtreme 1366 RT with 120mm Scythe S-Flex F fan
ASUS EAH4850 Radeon HD 4850 1GB DDR3 @ 625 MHz GPU & 1986 MHz Memory 
OCZ Agility 60GB SSD
Asus Xonar DX sound card
Antec nine hundred case, two front 120mm fans, one back 120mm Fan, one top 200mm fan 
Corsair CMPSU-750TX 750W

Comment 1

[E B]

Perhaps I did not use the word add-ons appropriately above.


I was referring to uninstalling *everything* and deleting all profiles, everything related to Firefox.


Then installing Firefox 3.6.17 from scratch and using its default profile testing for this problem, not replicating it.

Then uninstalling FF 3.6 and installing FF 4.0.1 from scratch and consistently being able to replicate the memory leak on FF 4.0.1's default profile by opening 15 tabs with pictures.

Comment 2

[Jo Hermans]

There's not much that can be done with this report, as there are no good steps to reproduce (URL's etc .. ). And besides - loading lots of tabs would produce such an enormous amount of data that a developer would not be able to see where the problem is. Memory leaks are investigated 1 URL at a time ! And we're using the data in <about:memory> for that.

This is not to say that there are not memory leaks, just that nothing can be done with this report in itself. Please see bug 640452 and its dependents for the current list of memory bugs where work is being done.

Note that Firefox 4.0 can hold up images in memory for up 120 seconds (worst case: 240 seconds), which is much larger than the 10 seconds in Firefox 3.6, but which is an advantage if you have to reload the images again (when browsing backwards and forwards through different pages on the same websites for instance). Keep that in mind if you're comparing, it can be annoying if you're opening websites with hundreds of images.

Also note that closing a tab will not always have the immediate effect that you're looking after. Often, the data-structures will be destroyed when a *new* tab is opened (it 'pushes' the old ones out), or after a certain timeout.

Comment 3

[VanillaMozilla]

Poster has provided more information here:  http://forums.mozillazine.org/viewtopic.php?f=38&t=2201945&p=10820747#p10820747

He writes, "Testing should be done by opening 10-20 TABs each of which loads about 10 or a dozen 2 MB pictures with res about 4752x3168."  Assuming 15 tabs, 10 images each, 4 bytes/pixel, that's 9 GB of data.  This is evidence that Fx 3.6 handles this differently than Fx 4, but NOT a memory leak.

Marking this INVALID.  Feel free to try again if you can write a correct bug report.

Comment 4

[VanillaMozilla]

Poster further reports that memory increases without any deallocation until Fx crashes, and that other people have confirmed it.

I could not verify.  Working with local images, it managed memory perfectly, deallocating at appropriate times.  I did not try with the links he provided privately.  He does not want to post links publicly.

I suspect he is just consuming all available virtual memory (and thus no bug), but have not seen evidence to the contrary.