1313452 - Poor performance and excess memory usage viewing washingtonpost.com comments with 32bit OS

Status: NEW

(Firefox :: General, defect)

Description

[q1]

FF 49.0.1 performs poorly when viewing comments on www.washingtonpost.com . This bug is very similar to bug 1294161 , and, in fact, that bug also still occurs, though possibly less frequently than before it was fixed (https://bugzilla.mozilla.org/show_bug.cgi?id=1294161#c29 ).

The main symptoms of this bug are slowness, jankiness, memory consumption > 1GB, and an eventual OOM crash while viewing washingtonpost.com comments.

To reproduce:

1. Visit http://www.washingtonpost.com;
2. Click on a popular story;
3. Click on the "xxxx comments" box at the story's bottom;
4. Wait for the comments to appear. The story is suitable if new comments appear every few seconds or faster;
5. Wait for ~1 hr;
6. While waiting, notice (using, e.g., Process Explorer) that FF's private bytes" and "working set" both increase basically monotonically to 1GB+;
7. Return to the page and attempt to scroll through the comments. Notice that scrolling is slow and janky;
8. Notice that all other pages and tabs are also slow and janky;
9. Leave the browser running for several more hours. Eventually it will crash with an OOM (though perhaps only if you're using the 32-bit build).

I am using ABP 2.7.3, but no other add-ons.

Comment 1

[Mike Conley (:mconley) (:⚙️)]

Out of curiosity, are you able to reproduce with ABP disabled? If not, then this is likely a memory leak in ABP and we should report it to them.

Comment 2

[q1]

(In reply to Mike Conley (:mconley) - (digging out of review / needinfo backlog) from comment #1)
> Out of curiosity, are you able to reproduce with ABP disabled? If not, then
> this is likely a memory leak in ABP and we should report it to them.

I haven't tried yet, since the site is almost unusable without ABP. I will try later today.

Comment 3

[q1]

(In reply to Mike Conley (:mconley) - (digging out of review / needinfo backlog) from comment #1)
> Out of curiosity, are you able to reproduce with ABP disabled? If not, then
> this is likely a memory leak in ABP and we should report it to them.

I see the problem with ABP disabled, too.

Comment 4

[Michelle Funches - QA]

Mike, here is some gecko profiler performance info from Win7. 
https://cleopatra.io/#report=e2c0d6b840d298eb5b4801209f3bd7ed6a6e4f55
Version 	49.0.1
Build ID 	20160922113459
User Agent 	Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:49.0) Gecko/20100101 Firefox/49.0

Comment 5

[Mike Conley (:mconley) (:⚙️)]

The STR sounds like a memory leak.

Reporter, are you able to get an about:memory report posted when you're in this state, but before you OOM?

Comment 6

[q1]

(In reply to Mike Conley (:mconley) - (buried in review / needinfo - send help) from comment #5)
> The STR sounds like a memory leak.
> 
> Reporter, are you able to get an about:memory report posted when you're in
> this state, but before you OOM?

I'll do this, probably tomorrow.

Comment 7

[q1]

Ack! I got it into the bad state, but about:memory does nothing. I have created a process dump, which I will answer questions about (but not upload). I'll try again the next time. What can cause about:memory to do nothing? Fallible allocators failing? If so, probably FF should preallocate enough space for about:memory always to work.

Comment 8

[q1]

Attachment: about-memory.txt

about:memory from very janky session.

Comment 9

[Mike Conley (:mconley) (:⚙️)]

Hey erahm,

Is there enough information in attachment 8808258 [details] to do any kind of diagnosis here? Is washingtonpost.com just leaking memory?

Comment 10

[Eric Rahm [:erahm]]

(In reply to Mike Conley (:mconley) from comment #9)
> Hey erahm,
> 
> Is there enough information in attachment 8808258 [details] to do any kind
> of diagnosis here? Is washingtonpost.com just leaking memory?


I'm not seeing any obvious issues here, I would guess the OOM is due to address space fragmentation:

>   103,481,344 B ── vsize-max-contiguous

100MiB is about the danger zone where things start going awry. If we had a crash report I could confirm that. heap-unclassified was reasonable and there aren't any ghost windows.

Keeping ni? open until I can take a look at the STR.

Comment 11

[Eric Rahm [:erahm]]

I'm trying to repro on Windows, but I think Washington Post has updated so that not all comments are loaded at once (you have to click 'More' after a few entries), and can no longer reproduce high memory usage. I was able to see some spikes yesterday on Linux.

q1, can you confirm that this no longer reproduces?

Comment 12

[q1]

(In reply to Eric Rahm [:erahm] from comment #11)
> I'm trying to repro on Windows, but I think Washington Post has updated so
> that not all comments are loaded at once (you have to click 'More' after a
> few entries), and can no longer reproduce high memory usage. I was able to
> see some spikes yesterday on Linux.
> 
> q1, can you confirm that this no longer reproduces?

WaPo comments still work the same way for me -- new ones are added automatically -- and FF still behaves as per comment 0.

Comment 13

[Eric Rahm [:erahm]]

(In reply to q1 from comment #12)
> (In reply to Eric Rahm [:erahm] from comment #11)
> > I'm trying to repro on Windows, but I think Washington Post has updated so
> > that not all comments are loaded at once (you have to click 'More' after a
> > few entries), and can no longer reproduce high memory usage. I was able to
> > see some spikes yesterday on Linux.
> > 
> > q1, can you confirm that this no longer reproduces?
> 
> WaPo comments still work the same way for me -- new ones are added
> automatically -- and FF still behaves as per comment 0.

Okay I'll leave it up longer, but things seemed alright in 51. It's possible we fixed the issue since 49.

Comment 14

[q1]

(In reply to Eric Rahm [:erahm] from comment #13)
> (In reply to q1 from comment #12)
> > (In reply to Eric Rahm [:erahm] from comment #11)
> > > I'm trying to repro on Windows, but I think Washington Post has updated so
> > > that not all comments are loaded at once (you have to click 'More' after a
> > > few entries), and can no longer reproduce high memory usage. I was able to
> > > see some spikes yesterday on Linux.
> > > 
> > > q1, can you confirm that this no longer reproduces?
> > 
> > WaPo comments still work the same way for me -- new ones are added
> > automatically -- and FF still behaves as per comment 0.
> 
> Okay I'll leave it up longer, but things seemed alright in 51. It's possible
> we fixed the issue since 49.

Unless comments are auto-loading, you're not testing what I'm testing. Try some other stories.

Comment 15

[Eric Rahm [:erahm]]

I'm still unable to repro an OOM or jankinewss in a clean install of Firefox 49.0.2 32-bit on Windows 7 64-bit. It's possible the article I've chosen isn't updating quick enough, but it seems to be pretty active.

q1, any chance you can try in safe-mode? Maybe there's some interaction with your profile going on. Also it would be helpful if we could get an OOM crash report, there might be one listed under about:crashes

I'll leave it running to see if I can catch an OOM.

Comment 16

[q1]

(In reply to Eric Rahm [:erahm] from comment #15)
> I'm still unable to repro an OOM or jankinewss in a clean install of Firefox
> 49.0.2 32-bit on Windows 7 64-bit. It's possible the article I've chosen
> isn't updating quick enough, but it seems to be pretty active.

It will take much longer to OOM on a 64-bit OS than on a 32-bit OS, because the available user-mode address space is almost twice as large.

> q1, any chance you can try in safe-mode? Maybe there's some interaction with
> your profile going on.

I will try this.

> Also it would be helpful if we could get an OOM crash
> report, there might be one listed under about:crashes

Nope. I don't let the thing report crashes for privacy reasons.

> I'll leave it running to see if I can catch an OOM.

Thanks.

Comment 17

[Brindusa Tot, Desktop QA]

(In reply to q1 from comment #16)

> > q1, any chance you can try in safe-mode? Maybe there's some interaction with
> > your profile going on.
> 
> I will try this.

q1, did you get the chance to try this in safe-mode?


Are you still seeing the OOM on the latest Firefox 50.1.0?

Comment 18

[q1]

(In reply to Brindusa Tot[:brindusat] from comment #17)
> (In reply to q1 from comment #16)
> 
> > > q1, any chance you can try in safe-mode? Maybe there's some interaction with
> > > your profile going on.
> > 
> > I will try this.
> 
> q1, did you get the chance to try this in safe-mode?

I have just tried it. I get slow-script dialogs and very janky behavior immediately upon clicking on the Flynn-shared-secrets-without-permission article on the front page, and FF uses 100% of one CPU core continually, though the memory usage is OK. FWIW, I disabled third-party cookies, OCSP, reporting, and updates prior to trying this. This is with FF 49.0.1 and a new profile in safe mode.
 
> Are you still seeing the OOM on the latest Firefox 50.1.0?

I'm not using that version yet.

Comment 19

[q1]

OK, I tried creating a new profile and using safe mode without customizing anything. Opening the Flynn story off the WaPo front page still cause major jank and high CPU consumption -- all of one core -- though not excessive memory consumption.

Comment 20

[Brindusa Tot, Desktop QA]

I tried reproducing this on Win7 with 64bit on Firefox 49.0.2 and Firefox 50.1.0. Opened the http://www.washingtonpost.com/ open one story and after ~2h scrolled down on comments. I am unable to reproduce an OOM or jankinewss. Next days, I will try to reproduce this on a Win 7 with 32 bit or on a Win XP.

Comment 21

[q1]

(In reply to Brindusa Tot[:brindusat] from comment #20)
> I tried reproducing this on Win7 with 64bit on Firefox 49.0.2 and Firefox
> 50.1.0. Opened the http://www.washingtonpost.com/ open one story and after
> ~2h scrolled down on comments. I am unable to reproduce an OOM or
> jankinewss. Next days, I will try to reproduce this on a Win 7 with 32 bit
> or on a Win XP.

You need to open the comments, then wait awhile, then try to scroll the comments. Apologies if that is what you did.

Comment 22

[q1]

(In reply to q1 from comment #21)
> (In reply to Brindusa Tot[:brindusat] from comment #20)
> > I tried reproducing this on Win7 with 64bit on Firefox 49.0.2 and Firefox
> > 50.1.0. Opened the http://www.washingtonpost.com/ open one story and after
> > ~2h scrolled down on comments. I am unable to reproduce an OOM or
> > jankinewss. Next days, I will try to reproduce this on a Win 7 with 32 bit
> > or on a Win XP.
> 
> You need to open the comments, then wait awhile, then try to scroll the
> comments. Apologies if that is what you did.

Also the problem occurs only for stories that are rapidly accumulating comments.

Comment 23

[Brindusa Tot, Desktop QA]

Finally, I get some time to investigate this. 
I tried reproducing this on Win7 with 32 bit on Firefox 49.0.1 and Firefox 50.1.0, but still unable to reproduce the OOM. 
Opened the http://www.washingtonpost.com/ and opened one story(one that is rapidly accumulating comments), and after ~1h scrolled down on comments. I can say that after 1 hour of being idle, the browser is a little slower, for few seconds but then it is back to normal usage. Did not get the OOM and also did not observe high CPU consumption.

Q1, please try this on Firefox 50.1.0 and let us know if you can reproduce this.

Comment 24

[q1]

(In reply to Brindusa Tot[:brindusat] from comment #23)
> Finally, I get some time to investigate this. 
> I tried reproducing this on Win7 with 32 bit on Firefox 49.0.1 and Firefox
> 50.1.0, but still unable to reproduce the OOM. 
> Opened the http://www.washingtonpost.com/ and opened one story(one that is
> rapidly accumulating comments), and after ~1h scrolled down on comments. I
> can say that after 1 hour of being idle, the browser is a little slower, for
> few seconds but then it is back to normal usage. Did not get the OOM and
> also did not observe high CPU consumption.
> 
> Q1, please try this on Firefox 50.1.0 and let us know if you can reproduce
> this.

I have done this. I still see the same problem. Try opening more than one story, open the comments on each, and let it sit for a few hours.

Comment 25

[ovidiu boca[:Ovidiu]]

Hi q1,

I saw the long discussions from this bug, I also tried to reproduce it but unfortunately with no results, I'm not able to reproduce this issue. By any chance do you have antivirus on your OS? If yes can you please disable it and try to retest this problem? I so some similar issues where the problem was caused by antivirus. 
Until further informations that can help us making a decision about the appropriate component, I will move this to general. If you have any idea for a component that will fit better to this issue, please update it.     
Thanks

Comment 26

[q1]

(In reply to ovidiu boca[:Ovidiu] from comment #25)
> Hi q1,
> 
> I saw the long discussions from this bug, I also tried to reproduce it but
> unfortunately with no results, I'm not able to reproduce this issue.

Have you tried it on XP SP3 x86? Have you tried opening several WP stories that are quickly accumulating comments, and leaving them open for several hours? The problem appears consistently for me; I have to restart FF every day because of it. BTW, I also observe another, apparently-related issue, which is that, prior to FF becoming unusably slow, the client area of windows often appears entirely black. I just experienced this when attempting to create an about:memory report. Also, another peculiarity of my browsing might be relevant: I use private browsing almost exclusively.

> By any chance do you have antivirus on your OS? If yes can you please disable it
> and try to retest this problem? I so some similar issues where the problem
> was caused by antivirus. 

I do not use a realtime AV.

Comment 27

[q1]

Here is an about:memory log from when FF was very, very slow:

Main Process
Explicit Allocations

1,020.07 MB (100.0%) -- explicit
├────524.88 MB (51.46%) -- window-objects
│    ├──369.38 MB (36.21%) -- top(https://www.washingtonpost.com/blogs/post-partisan/wp/2017/02/15/james-comeys-behavior-looks-worse-and-worse/?***, id=94)
│    │  ├──307.66 MB (30.16%) -- active
│    │  │  ├──307.44 MB (30.14%) -- window(https://www.washingtonpost.com/blogs/post-partisan/wp/2017/02/15/james-comeys-behavior-looks-worse-and-worse/?utm_term=.***#comments)
│    │  │  │  ├──217.06 MB (21.28%) -- js-compartment(https://www.washingtonpost.com/blogs/post-partisan/wp/2017/02/15/james-comeys-behavior-looks-worse-and-worse/)
│    │  │  │  │  ├──216.76 MB (21.25%) -- classes
│    │  │  │  │  │  ├───88.09 MB (08.64%) -- class(Object)/objects
│    │  │  │  │  │  │   ├──54.42 MB (05.33%) -- malloc-heap
│    │  │  │  │  │  │   │  ├──47.71 MB (04.68%) ── slots
│    │  │  │  │  │  │   │  └───6.71 MB (00.66%) ── elements/normal
│    │  │  │  │  │  │   └──33.68 MB (03.30%) ── gc-heap
│    │  │  │  │  │  ├───59.19 MB (05.80%) -- class(Function)/objects
│    │  │  │  │  │  │   ├──45.64 MB (04.47%) ── gc-heap
│    │  │  │  │  │  │   └──13.55 MB (01.33%) ── malloc-heap/slots
│    │  │  │  │  │  ├───37.50 MB (03.68%) -- class(Call)/objects
│    │  │  │  │  │  │   ├──37.49 MB (03.68%) ── gc-heap
│    │  │  │  │  │  │   └───0.01 MB (00.00%) ── malloc-heap/slots
│    │  │  │  │  │  ├───27.26 MB (02.67%) -- class(Array)/objects
│    │  │  │  │  │  │   ├──26.45 MB (02.59%) ── gc-heap
│    │  │  │  │  │  │   └───0.81 MB (00.08%) ++ malloc-heap
│    │  │  │  │  │  └────4.71 MB (00.46%) ++ (9 tiny)
│    │  │  │  │  └────0.30 MB (00.03%) ++ (4 tiny)
│    │  │  │  ├───60.20 MB (05.90%) -- dom
│    │  │  │  │   ├──48.33 MB (04.74%) ── orphan-nodes
│    │  │  │  │   ├──10.89 MB (01.07%) ── element-nodes
│    │  │  │  │   └───0.99 MB (00.10%) ++ (4 tiny)
│    │  │  │  ├───27.42 MB (02.69%) ++ layout
│    │  │  │  └────2.76 MB (00.27%) ++ (2 tiny)
│    │  │  └────0.21 MB (00.02%) ++ (2 tiny)
│    │  └───61.72 MB (06.05%) -- js-zone(0x11018000)
│    │      ├──36.73 MB (03.60%) -- strings
│    │      │  ├──24.88 MB (02.44%) ++ (114 tiny)
│    │      │  └──11.84 MB (01.16%) ++ string(<non-notable strings>)
│    │      ├──21.27 MB (02.09%) -- shapes
│    │      │  ├──14.98 MB (01.47%) -- gc-heap
│    │      │  │  ├──11.85 MB (01.16%) ── dict
│    │      │  │  └───3.12 MB (00.31%) ++ (2 tiny)
│    │      │  └───6.29 MB (00.62%) ++ malloc-heap
│    │      └───3.72 MB (00.36%) ++ (8 tiny)
│    ├───95.98 MB (09.41%) ++ (30 tiny)
│    ├───36.03 MB (03.53%) -- top(https://www.washingtonpost.com/news/powerpost/paloma/daily-202/2017/02/15/daily-202-it-s-bigger-than-flynn-new-russia-revelations-widen-trump-s-credibility-gap/***/?utm_term=.***, id=25)
│    │   ├──23.99 MB (02.35%) ++ active
│    │   └──12.04 MB (01.18%) ++ js-zone(0x11e1e000)
│    ├───12.03 MB (01.18%) ++ top(https://www.nytimes.com/, id=134)
│    └───11.45 MB (01.12%) ++ top(http://emma.msrb.org/***, id=175)
├────294.29 MB (28.85%) ── heap-unclassified
├─────69.90 MB (06.85%) -- js-non-window
│     ├──39.13 MB (03.84%) -- zones
│     │  ├──27.03 MB (02.65%) ++ zone(0x831000)
│     │  ├──10.91 MB (01.07%) -- zone(0x82b800)
│     │  │  ├──10.72 MB (01.05%) ++ strings/string(<non-notable strings>)
│     │  │  └───0.18 MB (00.02%) ++ (5 tiny)
│     │  └───1.19 MB (00.12%) ++ (19 tiny)
│     ├──25.69 MB (02.52%) ++ runtime
│     └───5.08 MB (00.50%) ++ gc-heap
├─────42.67 MB (04.18%) -- heap-overhead
│     ├──33.20 MB (03.26%) ── bin-unused
│     └───9.47 MB (00.93%) ++ (2 tiny)
├─────32.69 MB (03.21%) -- images
│     ├──29.94 MB (02.94%) -- content
│     │  ├──27.13 MB (02.66%) ++ raster/used
│     │  └───2.81 MB (00.28%) ++ vector/used
│     └───2.76 MB (00.27%) ++ (2 tiny)
├─────24.74 MB (02.43%) -- workers
│     ├──18.48 MB (01.81%) -- workers(pdf.js)/worker(../build/pdf.worker.js, 0x21026000)
│     │  ├──13.84 MB (01.36%) -- zone(0x25ca2800)
│     │  │  ├──12.47 MB (01.22%) -- compartment(web-worker)
│     │  │  │  ├──12.25 MB (01.20%) ++ classes
│     │  │  │  └───0.22 MB (00.02%) ++ (3 tiny)
│     │  │  └───1.37 MB (00.13%) ++ (11 tiny)
│     │  └───4.64 MB (00.46%) ++ (3 tiny)
│     └───6.26 MB (00.61%) ++ workers(chrome)
├─────16.32 MB (01.60%) ++ (19 tiny)
└─────14.57 MB (01.43%) -- gfx
      ├──10.39 MB (01.02%) ── heap-textures
      └───4.18 MB (00.41%) ++ (6 tiny)

Other Measurements

2,047.94 MB (100.0%) -- address-space
├──1,307.83 MB (63.86%) -- commit
│  ├──1,157.75 MB (56.53%) -- private
│  │  ├──1,155.77 MB (56.44%) ── readwrite(segments=13010)
│  │  └──────1.98 MB (00.10%) ++ (5 tiny)
│  ├────110.40 MB (05.39%) -- image
│  │    ├───41.57 MB (02.03%) ── execute(segments=22)
│  │    ├───41.51 MB (02.03%) ── readonly(segments=343)
│  │    ├───24.00 MB (01.17%) ── execute-read(segments=137)
│  │    └────3.32 MB (00.16%) ++ (2 tiny)
│  └─────39.68 MB (01.94%) -- mapped
│        ├──26.68 MB (01.30%) ── readonly(segments=27)
│        └──13.00 MB (00.63%) ++ (2 tiny)
├────550.80 MB (26.90%) -- reserved
│    ├──538.41 MB (26.29%) ── private(segments=12124)
│    └───12.39 MB (00.61%) ── mapped(segments=7)
└────189.31 MB (09.24%) ── free(segments=407)

55.40 MB (100.0%) -- decommitted
├──38.81 MB (70.06%) -- workers
│  ├──36.90 MB (66.61%) ── workers(pdf.js)/worker(../build/pdf.worker.js, 0x21026000)/gc-heap/decommitted-arenas
│  └───1.91 MB (03.45%) -- workers(chrome)
│      ├──0.65 MB (01.17%) ── worker(resource://gre/modules/PageThumbsWorker.js, 0x25913c00)/gc-heap/decommitted-arenas
│      ├──0.64 MB (01.15%) ── worker(resource:///modules/sessionstore/SessionWorker.js, 0xe0d3400)/gc-heap/decommitted-arenas
│      └──0.63 MB (01.13%) ── worker(resource://gre/modules/osfile/osfile_async_worker.js, 0xbf34000)/gc-heap/decommitted-arenas
└──16.59 MB (29.94%) ── js-non-window/gc-heap/decommitted-arenas

19,901 (100.0%) -- event-counts
└──19,901 (100.0%) -- window-objects
   ├───4,794 (24.09%) -- top(https://www.washingtonpost.com/blogs/post-partisan/wp/2017/02/15/james-comeys-behavior-looks-worse-and-worse/?utm_term=.***#comments, id=94)/active
   │   ├──4,787 (24.05%) -- window(https://www.washingtonpost.com/blogs/post-partisan/wp/2017/02/15/james-comeys-behavior-looks-worse-and-worse/?utm_term=.***#comments)/dom
   │   │  ├──4,779 (24.01%) ── event-listeners
   │   │  └──────8 (00.04%) ── event-targets
   │   └──────7 (00.04%) ++ (2 tiny)
   ├───1,338 (06.72%) -- top(https://www.washingtonpost.com/news/powerpost/paloma/daily-202/2017/02/15/daily-202-it-s-bigger-than-flynn-new-russia-revelations-widen-trump-s-credibility-gap/***/?utm_term=.***, id=25)/active
   │   ├────822 (04.13%) ++ (15 tiny)
   │   └────516 (02.59%) -- window(https://www.washingtonpost.com/news/powerpost/paloma/daily-202/2017/02/15/daily-202-it-s-bigger-than-flynn-new-russia-revelations-widen-trump-s-credibility-gap/***/?utm_term=.***)/dom
   │        ├──507 (02.55%) ── event-listeners [2]
   │        └────9 (00.05%) ── event-targets [2]
   ├───1,282 (06.44%) -- top(chrome://browser/content/browser.xul, id=12)/active
   │   ├──1,281 (06.44%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,242 (06.24%) ── event-listeners
   │   │  └─────39 (00.20%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,264 (06.35%) -- top(chrome://browser/content/browser.xul, id=215)/active
   │   ├──1,263 (06.35%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,242 (06.24%) ── event-listeners
   │   │  └─────21 (00.11%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,227 (06.17%) -- top(chrome://browser/content/browser.xul, id=194)/active
   │   ├──1,226 (06.16%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,212 (06.09%) ── event-listeners
   │   │  └─────14 (00.07%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,216 (06.11%) -- top(chrome://browser/content/browser.xul, id=151)/active
   │   ├──1,215 (06.11%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,202 (06.04%) ── event-listeners
   │   │  └─────13 (00.07%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,206 (06.06%) -- top(chrome://browser/content/browser.xul, id=131)/active
   │   ├──1,205 (06.05%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,171 (05.88%) ── event-listeners
   │   │  └─────34 (00.17%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,205 (06.05%) -- top(chrome://browser/content/browser.xul, id=165)/active
   │   ├──1,204 (06.05%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,181 (05.93%) ── event-listeners
   │   │  └─────23 (00.12%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,187 (05.96%) -- top(chrome://browser/content/browser.xul, id=3)/active
   │   ├──1,186 (05.96%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,180 (05.93%) ── event-listeners
   │   │  └──────6 (00.03%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,182 (05.94%) -- top(chrome://browser/content/browser.xul, id=119)/active
   │   ├──1,181 (05.93%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,170 (05.88%) ── event-listeners
   │   │  └─────11 (00.06%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├───1,179 (05.92%) -- top(chrome://browser/content/browser.xul, id=140)/active
   │   ├──1,178 (05.92%) -- window(chrome://browser/content/browser.xul)/dom
   │   │  ├──1,171 (05.88%) ── event-listeners
   │   │  └──────7 (00.04%) ── event-targets
   │   └──────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├─────791 (03.97%) ++ (18 tiny)
   ├─────558 (02.80%) -- top(http://emma.msrb.org/***, id=175)/active/window(http://emma.msrb.org/***)/dom
   │     ├──555 (02.79%) ── event-listeners
   │     └────3 (00.02%) ── event-targets
   ├─────508 (02.55%) -- top(https://www.washingtonpost.com/local/education/influential-conservative-group-trump-devos-should-dismantle-education-department-and-bring-god-into-classrooms/2017/02/15/196bf872-f2df-11e6-8d72-263470bf0401_story.html?utm_term=.***, id=105)/active
   │     ├──507 (02.55%) -- window(https://www.washingtonpost.com/local/education/influential-conservative-group-trump-devos-should-dismantle-education-department-and-bring-god-into-classrooms/2017/02/15/196bf872-f2df-11e6-8d72-263470bf0401_story.html?utm_term=.***)/dom
   │     │  ├──498 (02.50%) ── event-listeners [2]
   │     │  └────9 (00.05%) ── event-targets [2]
   │     └────1 (00.01%) ── window(about:blank)/dom/event-targets
   ├─────501 (02.52%) -- top(https://www.washingtonpost.com/news/powerpost/wp/2017/02/15/trump-reveled-in-leaks-that-hurt-hillary-clinton-he-now-calls-administration-disclosures-un-american/?utm_term=.***, id=87)/active
   │     ├──500 (02.51%) -- window(https://www.washingtonpost.com/news/powerpost/wp/2017/02/15/trump-reveled-in-leaks-that-hurt-hillary-clinton-he-now-calls-administration-disclosures-un-american/?utm_term=.***)/dom
   │     │  ├──493 (02.48%) ── event-listeners [2]
   │     │  └────7 (00.04%) ── event-targets [2]
   │     └────1 (00.01%) ── window(about:blank)/dom/event-targets
   └─────463 (02.33%) -- top(https://www.washingtonpost.com/, id=15)/active/window(https://www.washingtonpost.com/)/dom
         ├──441 (02.22%) ── event-listeners [2]
         └───22 (00.11%) ── event-targets [2]

760.94 MB (100.0%) -- heap-committed
├──718.27 MB (94.39%) ── allocated
└───42.67 MB (05.61%) ── overhead

32.69 MB (100.0%) -- images
├──29.94 MB (91.57%) -- content
│  ├──27.13 MB (82.97%) -- raster/used
│  │  ├──15.13 MB (46.27%) ── decoded-heap
│  │  └──12.00 MB (36.70%) ── source
│  └───2.81 MB (08.61%) -- vector/used
│      ├──2.81 MB (08.60%) ── source
│      └──0.00 MB (00.00%) ── decoded-heap
├───1.70 MB (05.19%) -- chrome
│   ├──1.12 MB (03.44%) -- vector/used
│   │  ├──1.12 MB (03.43%) ── source
│   │  └──0.00 MB (00.00%) ── decoded-heap
│   └──0.57 MB (01.76%) -- raster/used
│      ├──0.46 MB (01.40%) ── decoded-heap
│      └──0.12 MB (00.35%) ── source
└───1.06 MB (03.24%) -- uncached
    ├──0.91 MB (02.80%) -- raster
    │  ├──0.91 MB (02.79%) -- used
    │  │  ├──0.51 MB (01.55%) ── decoded-heap
    │  │  └──0.41 MB (01.25%) ── source
    │  └──0.00 MB (00.00%) ── unused/source
    └──0.14 MB (00.44%) ── vector/used/source

424.93 MB (100.0%) -- js-main-runtime
├──269.73 MB (63.48%) -- compartments
│  ├──253.34 MB (59.62%) -- classes/objects
│  │  ├──175.70 MB (41.35%) ── gc-heap
│  │  └───77.65 MB (18.27%) -- malloc-heap
│  │      ├──68.06 MB (16.02%) ── slots
│  │      ├───9.41 MB (02.21%) ── elements/normal
│  │      └───0.18 MB (00.04%) ── misc
│  ├────8.86 MB (02.08%) -- scripts
│  │    ├──5.68 MB (01.34%) ── gc-heap
│  │    └──3.17 MB (00.75%) ── malloc-heap/data
│  └────7.53 MB (01.77%) ++ (12 tiny)
├──124.44 MB (29.28%) -- zones
│  ├───53.98 MB (12.70%) -- strings
│  │   ├──38.44 MB (09.05%) -- malloc-heap
│  │   │  ├──20.95 MB (04.93%) ── two-byte
│  │   │  └──17.50 MB (04.12%) ── latin1
│  │   └──15.54 MB (03.66%) -- gc-heap
│  │      ├──11.00 MB (02.59%) ── latin1
│  │      └───4.54 MB (01.07%) ── two-byte
│  ├───43.21 MB (10.17%) -- shapes
│  │   ├──27.96 MB (06.58%) -- gc-heap
│  │   │  ├──13.89 MB (03.27%) ── dict
│  │   │  ├──12.91 MB (03.04%) ── tree
│  │   │  └───1.15 MB (00.27%) ── base
│  │   └──15.25 MB (03.59%) -- malloc-heap
│  │      ├───8.67 MB (02.04%) ── tree-tables
│  │      ├───4.47 MB (01.05%) ── dict-tables
│  │      └───2.12 MB (00.50%) ── tree-kids
│  ├───12.51 MB (02.94%) ++ (8 tiny)
│  ├────5.20 MB (01.22%) ++ lazy-scripts
│  ├────4.77 MB (01.12%) ── type-pool
│  └────4.76 MB (01.12%) ++ scopes
├───25.69 MB (06.05%) ── runtime
└────5.08 MB (01.20%) ++ gc-heap

473 (100.0%) -- js-main-runtime-compartments
├──388 (82.03%) -- system
│  ├──347 (73.36%) ++ (343 tiny)
│  ├───23 (04.86%) ── [System Principal], inProcessTabChildGlobal?ownedBy=chrome://browser/content/browser.xul [23]
│  └───18 (03.81%) ── [System Principal], about:blank [18]
└───85 (17.97%) ++ user

245.41 MB (100.0%) -- js-main-runtime-gc-heap-committed
├──241.78 MB (98.52%) -- used
│  ├──234.51 MB (95.56%) -- gc-things
│  │  ├──175.70 MB (71.59%) ── objects
│  │  ├───26.80 MB (10.92%) ── shapes
│  │  ├───15.54 MB (06.33%) ── strings
│  │  ├────5.68 MB (02.32%) ── scripts
│  │  ├────4.24 MB (01.73%) ── lazy-scripts
│  │  ├────3.78 MB (01.54%) ── object-groups
│  │  └────2.77 MB (01.13%) ++ (4 tiny)
│  ├────4.08 MB (01.66%) ── chunk-admin
│  └────3.19 MB (01.30%) ── arena-admin
└────3.64 MB (01.48%) -- unused
     ├──2.64 MB (01.07%) ++ gc-things
     └──1.00 MB (00.41%) ++ (2 tiny)

1,154 (100.0%) -- low-memory-events
├──1,154 (100.0%) ── virtual
└──────0 (00.00%) ++ (2 tiny)

780 (100.0%) -- message-manager
└──780 (100.0%) -- referent
   ├──706 (90.51%) -- global-manager
   │  ├──706 (90.51%) ── strong
   │  └────0 (00.00%) ++ weak
   ├───59 (07.56%) -- parent-process-manager
   │   ├──59 (07.56%) ── strong
   │   └───0 (00.00%) ++ weak
   └───15 (01.92%) -- child-process-manager
       ├──14 (01.79%) ── strong
       └───1 (00.13%) ++ weak

2,978 (100.0%) -- observer-service
└──2,978 (100.0%) -- referent
   ├──2,505 (84.12%) ── strong
   └────473 (15.88%) -- weak
        ├──473 (15.88%) ── alive
        └────0 (00.00%) ── dead

1,696 (100.0%) -- observer-service-suspect
├────453 (26.71%) ── referent(topic=xpcom-shutdown)
├────447 (26.36%) ── referent(topic=memory-pressure)
├────118 (06.96%) ── referent(topic=chrome-flush-skin-caches)
├────113 (06.66%) ── referent(topic=agent-sheet-added)
├────113 (06.66%) ── referent(topic=agent-sheet-removed)
├────113 (06.66%) ── referent(topic=author-sheet-added)
├────113 (06.66%) ── referent(topic=author-sheet-removed)
├────113 (06.66%) ── referent(topic=user-sheet-added)
└────113 (06.66%) ── referent(topic=user-sheet-removed)

1,430 (100.0%) -- preference-service
└──1,430 (100.0%) -- referent
   ├──1,240 (86.71%) ── strong
   └────190 (13.29%) -- weak
        ├──190 (13.29%) ── alive
        └────0 (00.00%) ── dead

171.30 MB (100.0%) -- window-objects
├───77.90 MB (45.47%) -- dom
│   ├──48.76 MB (28.47%) ── orphan-nodes
│   ├──22.40 MB (13.08%) ── element-nodes
│   ├───4.60 MB (02.68%) ── text-nodes
│   ├───2.04 MB (01.19%) ── other
│   └───0.09 MB (00.05%) ++ (3 tiny)
├───66.17 MB (38.63%) -- layout
│   ├──17.87 MB (10.43%) ── style-structs
│   ├──13.65 MB (07.97%) ── style-sets
│   ├───8.74 MB (05.10%) ── frames
│   ├───8.01 MB (04.67%) ── pres-shell
│   ├───6.03 MB (03.52%) ── style-contexts
│   ├───4.67 MB (02.73%) ── rule-nodes
│   ├───2.85 MB (01.66%) ── text-runs
│   ├───2.21 MB (01.29%) ── pres-contexts
│   └───2.14 MB (01.25%) ── line-boxes
├───27.03 MB (15.78%) ── style-sheets
└────0.20 MB (00.12%) ── property-tables

    0.00 MB ── gfx-d2d-vram-draw-target
    0.00 MB ── gfx-d2d-vram-source-surface
    0.17 MB ── gfx-surface-win32
    0.00 MB ── gfx-textures
    0.00 MB ── gfx-textures-peak
    0.00 MB ── gfx-tiles-waste
          0 ── ghost-windows
  718.27 MB ── heap-allocated
    1.00 MB ── heap-chunksize
1,098.00 MB ── heap-mapped
   11.78 MB ── imagelib-surface-cache-estimated-locked
   11.78 MB ── imagelib-surface-cache-estimated-total
          0 ── imagelib-surface-cache-overflow-count
    2.14 MB ── js-main-runtime-temporary-peak
1,209.14 MB ── private
1,210.21 MB ── resident
1,206.11 MB ── resident-unique
   74.89 MB ── system-heap-allocated
1,858.08 MB ── vsize
    7.55 MB ── vsize-max-contiguous

End of Main Process

Comment 28

[q1]

My subjective impression is that the problem is worse on FF 51.0.1 than on the recent previous versions.

Comment 29

[ovidiu boca[:Ovidiu]]

(In reply to q1 from comment #26)

> Have you tried it on XP SP3 x86? Have you tried opening several WP stories
> that are quickly accumulating comments, and leaving them open for several
> hours? The problem appears consistently for me; 

I tried on a Windows XP x64 SP 2 with FF release 51 and FF Nightly 54.0a1 and I can't reproduce this. I opened http://www.washingtonpost.com/ and choose 2 stories that where accumulating a lot of comments. After I waited for an hour and then I started to scroll the comments, I did not observe a high CPU consumption.

Comment 30

[q1]

(In reply to ovidiu boca[:Ovidiu] from comment #29)
> (In reply to q1 from comment #26)
> 
> > Have you tried it on XP SP3 x86? Have you tried opening several WP stories
> > that are quickly accumulating comments, and leaving them open for several
> > hours? The problem appears consistently for me; 
> 
> I tried on a Windows XP x64 SP 2...

I am using XP SP3 *x86*. x64 supports far more available virtual (and much more available physical) memory, so memory-related problems are less likely to be reproducible on that platform. Please try XP SP3 x86.

Comment 31

[Wayne Mery (:wsmwk)]

Better with the 3GB switch?
https://stackoverflow.com/questions/639540/how-much-memory-can-a-32-bit-process-access-on-a-64-bit-operating-system

Comment 32

[q1]

(In reply to Wayne Mery (:wsmwk) from comment #31)
> Better with the 3GB switch?
> https://stackoverflow.com/questions/639540/how-much-memory-can-a-32-bit-
> process-access-on-a-64-bit-operating-system

No. I always used that switch. BTW, I no longer use XP, so I cannot test further workarounds. FWIW, the slowdown problem is less-pronounced on my new platform (Win10 x64) and newer FF (55.x, 57.x), but memory usage is still excessive (often > 2GB).

Comment 33

[Wayne Mery (:wsmwk)]

There are quite a few recent bug reports involving washington post
https://mzl.la/2Cw3eUU

Comment 34

[danny92]

"Viewing comments on washingtonpost.com with a 32-bit OS might lead to poor performance and high memory usage. Upgrading to a 64-bit system could improve your browsing experience." https://washingmachinerepairkuwait.com/ We are Kuwait's leading experts in fully automatic washer and dryer repairs, with a skilled team of Indian and Pakistani technicians renowned for their service efficiency and professionalism