Excessive CPU usage with 23 extensions present

RESOLVED WORKSFORME

Status

()

RESOLVED WORKSFORME
3 years ago
10 months ago

People

(Reporter: hrdubwd, Unassigned)

Tracking

(Depends on: 1 bug, {crash})

43 Branch
x86_64
Windows 7
crash
Points:
---
Dependency tree / graph

Firefox Tracking Flags

(Not tracked)

Details

(Whiteboard: [memshrink:P3])

Attachments

(6 attachments, 7 obsolete attachments)

(Reporter)

Description

3 years ago
User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101 Firefox/43.0
Build ID: 20160105164030

Steps to reproduce:

I have 18 tabs pinned for very frequently used pages, and then may open anything from a few to many tabs in the course of my work (academic stuff).  FF works well most of the time, but occasionally, but with monotonous regularity, it will go dead slow, consuming up to ~36% CPU, and so making FF use difficult to say the least, but also slowing the machine appreciably.  I have let it run in this state for hours, and it never resolves.  For comparison, the normal background tickover is around 2 or 3%, with occasional extra activity to say 10%.  No problem with this at all.

What seems to be happening is that memory is being shuffled around, or the contents thereof.  The fact is that there is some 20 GB of free RAM, of which about 2 GB is being used (the numbers tick up and down continuously during this kind of event).  I have put the cache into RAM on a virtual drive, along with all Windows temp files, so that should not be a problem - there is plenty of memory there as well, 20 GB!, no indexing, no compression.  (After crashing and restarting to write this up, memory usage is around 1.3 GB, CPU usage around 2 - 5% - no problem to use.)

This is under W7 Pro, 64-bit.  It has been occurring for all of at least a few of the previous versions of FF.

The only way to deal with it is to crash FF and reload, when after the loading settling down, all is well again, until some random point in the future.  This has now happened many times.  I have held off reporting thinking that I might identify the problem with a specific tab.  I cannot.)

I cannot see any similar reports of this kind of behaviour, except for some very specific single-tab problems, for example.


Actual results:

Excessive CPU usage slows FF to a crawl, and interferes with the machine generally, by causing slow responses.  


Expected results:

Memory clean-up from time to time would not be a problem, but locked in a seemingly unending cycle is not functional.
(Reporter)

Updated

3 years ago
OS: Unspecified → Windows 7
Hardware: Unspecified → x86_64

Comment 1

3 years ago
Please enter the address "about:memory?verbose" (with a lower-case "v") in the address bar and attach (using the "Add an attachment" link above) the output here.
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 2

3 years ago
Created attachment 8710630 [details]
memory-report.json.gz
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 3

3 years ago
Is that what you need?  It was not generated directly.
Thanks.
(Reporter)

Comment 4

3 years ago
It did it again - unusable ... had to restart.
Reporter, can you please attach the crash reports from about:crashes?
Flags: needinfo?(hrdubwd)
Keywords: crash
Whiteboard: [memshrink]
Chris, could you please take a look at the memory report attached in Comment #2?
Flags: needinfo?(cpeterson)
(Reporter)

Comment 7

3 years ago
(In reply to Brindusa Tot from comment #5)
> Reporter, can you please attach the crash reports from about:crashes?

This event has never (noticeably) caused a crash.  All crash reports have already been submitted anyway, and are not on this machine now.

Further info:
I have been exploring the effects of the pinned tabs. 
I found one that took a lot of CPU (http://www.speedtest.net/), but not in the sustained blocking fashion reported above.  I have got it down to 2 possibles, but it seems that the condition only arises after a long period, and so far I have no definitive suggestions as to what is the cause.
It might help if there was a way to see which tab was responsible for how much CPU use, but I cannot find anything reported - am I missing something (Chrome can do this, apparently).

In the meantime I have put the cache into RAM, but with no obvious effect on the present problem (it was on a RAM disk before, as mentioned).  

I shall continue my testing, but it is necessarily slow!
Flags: needinfo?(hrdubwd)
I think Nicholas would be a better person to analyze about:memory reports. :)
Flags: needinfo?(cpeterson) → needinfo?(n.nethercote)
The memory usage doesn't seem out of the ordinary for that amount of tabs.

At least one page seems to be using a ton of events, it's possible this is related to the high CPU usage.

> 35,849 (100.0%) -- event-counts
> ├──35,782 (99.81%) -- window-objects
> │  ├──16,764 (46.76%) -- top(<anonymized-509>, id=509)/active
> │  │  ├──16,755 (46.74%) -- window(<anonymized-1225>)/dom
> │  │  │  ├──16,754 (46.73%) ── event-listeners
> │  │  │  └───────1 (00.00%) ── event-targets
> │  │  └───────9 (00.03%) ++ window(<anonymized-511>)/dom

That page also correlates to the top memory consumer:
> │    ├──135.03 MB (09.08%) -- top(<anonymized-509>, id=509)

If it's not sensitive, can you let us know which site this is?

Also we think switching to the 64-bit version of Firefox might help with memory pressure issues which could lead to high CPU usage as well. You can download that here: https://www.mozilla.org/firefox/all/
Flags: needinfo?(n.nethercote) → needinfo?(hrdubwd)
> Also we think switching to the 64-bit version of Firefox might help with
> memory pressure issues which could lead to high CPU usage as well. You can
> download that here: https://www.mozilla.org/firefox/all/

Yes, even if you have 20 GB of physical RAM, if you're running a 32-bit version of Firefox -- and you probably are, because that's the standard one -- then Firefox can use at most 2 or 4 GB of RAM, depending on the machine configuration.

Also, judging by the memory reports you have quite a few extensions installed -- 14 or more?  In our experience, when users have 10 or more extensions installed and they have high memory and/or CPU usage, it's one or more of the extensions that it the cause. Unfortunately the only way to confirm this is to selectively disable extensions in order to determine which one is at fault; then you can either report the problem to the extension author or decide if you can live without that extension.

If you can post the full list of extensions (visit about:support, click on "Copy text to clipboard", paste in a comment here) we might be able to give you suggestions as to which ones are more likely to be causing problems.
Summary: Excessive CPU usage → Excessive CPU usage with 14+ extensions present
(Reporter)

Comment 11

3 years ago
1.

(In reply to Nicholas Nethercote [:njn] from comment #10)

> Also, judging by the memory reports you have quite a few extensions ...

Advice welcome, thanks.


Application Basics
------------------

Name: Firefox
Version: 43.0.4
Build ID: 20160105164030
Update Channel: release
User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101 Firefox/43.0
Multiprocess Windows: 0/2 (default: false)
Safe Mode: false

Crash Reports for the Last 3 Days
---------------------------------

Report ID: bp-97b6f96e-6fdb-45b4-a5fa-199582160127
Submitted: 9 hours ago

All Crash Reports

Extensions
----------

Name: Adblock Plus
Version: 2.7.1
Enabled: true
ID: {d10d0bf8-f5b5-c8b4-a8b2-2b9879e08c5d}

Name: Adobe Acrobat - Create PDF
Version: 2.0
Enabled: true
ID: web2pdfextension@web2pdf.adobedotcom

Name: Blur
Version: 5.3.1983
Enabled: true
ID: donottrackplus@abine.com

Name: Capture & Print
Version: 0.1.9.3.1-signed
Enabled: true
ID: {146f1820-2b0d-49ef-acbf-d85a6986e10c}

Name: Copy Plain Text 2
Version: 1.4
Enabled: true
ID: copyplaintext@teo.pl

Name: Duplicate in Tab Context Menu
Version: 1.0.9.1-signed
Enabled: true
ID: DuplicateInTabContext@schuzak.jp

Name: FireFTP
Version: 2.0.26
Enabled: true
ID: {a7c6cf7f-112c-4500-a7ea-39801a327e5f}

Name: Flashblock
Version: 1.5.20
Enabled: true
ID: {3d7eb24f-2740-49df-8937-200b1cc08f8a}

Name: HTTPS-Everywhere
Version: 5.1.2
Enabled: true
ID: https-everywhere-eff@eff.org

Name: IE Tab 2 (FF 3.6+)
Version: 5.12.12.1.1-signed
Enabled: true
ID: {1BC9BA34-1EED-42ca-A505-6D2F1A935BBB}

Name: JavaScript Deobfuscator  [this I will remove, recently added, does not do what I thought] 
Version: 2.0.2
Enabled: true
ID: jsdeobfuscator@adblockplus.org

Name: Lightbeam
Version: 1.3.0.1-signed
Enabled: true
ID: jid1-F9UJ2thwoAm5gQ@jetpack

Name: More About
Version: 1.2.13
Enabled: true
ID: MoreAbout@schuzak.jp

Name: NoScript
Version: 2.9.0.2
Enabled: true
ID: {73a6fe31-595d-460b-a920-fcc0f8843232}

Name: Padlock
Version: 0.5.0.1-signed
Enabled: true
ID: {d09e32df-8610-4b33-b929-1e631b764130}

Name: Privacy Badger
Version: 1.0.5
Enabled: true
ID: jid1-MnnxcxisBPnSXQ-eff@jetpack

Name: PubPeer                   [only Added today]
Version: 0.1.5.1-signed
Enabled: true
ID: jid1-rU2mNakSg7IiSQ@jetpack

Name: Restart
Version: 1.2.8
Enabled: true
ID: Restart@schuzak.jp

Name: SSL Version Control
Version: 0.4.1-signed
Enabled: true
ID: jid1-ZM3BerwS6FsQAg@jetpack

Name: The Camelizer - Price Tracker
Version: 2.8.1
Enabled: true
ID: izer@camelcamelcamel.com

Name: Trustwave SecureBrowsing
Version: 3.722.1-signed
Enabled: true
ID: securebrowsing@m86security.com

Name: Vacuum Places Improved
Version: 1.2.1-signed
Enabled: true
ID: VacuumPlacesImproved@lultimouomo-gmail.com

Name: Xmarks
Version: 4.3.7.1-signed
Enabled: true
ID: foxmarks@kei.com

Name: Garmin Communicator
Version: 4.1.0.1-signed
Enabled: false
ID: {195A3098-0BD5-4e90-AE22-BA1C540AFD1E}

Name: Microsoft .NET Framework Assistant
Version: 1.3.1.1-signed
Enabled: false
ID: {20a82645-c095-46ed-80e3-08825760534b}

Name: selectivecookiedelete
Version: 4.1.1-signed
Enabled: false
ID: selectivecookiedelete@siju.mathew

Name: Tab Auto Reload
Version: 2.0.17
Enabled: false
ID: TabAutoReload@schuzak.jp

Name: WOT
Version: 20151208
Enabled: false
ID: {a0d7ccb3-214d-498b-b4aa-0e8fda9a7bf7}

Graphics
--------

Adapter Description: NVIDIA Quadro 4000
Adapter Description (GPU #2): NVIDIA Quadro NVS 290
Adapter Drivers: nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter Drivers (GPU #2): nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um
Adapter RAM: 2048
Adapter RAM (GPU #2): 256
Asynchronous Pan/Zoom: none
Device ID: 0x06dd
Device ID (GPU #2): 0x042f
Direct2D Enabled: true
DirectWrite Enabled: true (6.2.9200.17568)
Driver Date: 2-3-2015
Driver Date (GPU #2): 2-3-2015
Driver Version: 9.18.13.4144
Driver Version (GPU #2): 9.18.13.4144
GPU #2 Active: false
GPU Accelerated Windows: 2/2 Direct3D 11 (OMTC)
Subsys ID: 078010de
Subsys ID (GPU #2): 0000000c
Supports Hardware H264 Decoding: Yes
Vendor ID: 0x10de
Vendor ID (GPU #2): 0x10de
WebGL Renderer: Google Inc. -- ANGLE (NVIDIA Quadro 4000 Direct3D11 vs_5_0 ps_5_0)
windowLayerManagerRemote: true
AzureCanvasBackend: direct2d 1.1
AzureContentBackend: direct2d 1.1
AzureFallbackCanvasBackend: cairo
AzureSkiaAccelerated: 0

Important Modified Preferences
------------------------------

accessibility.typeaheadfind: true
accessibility.typeaheadfind.flashBar: 0
browser.cache.disk.capacity: 4194302
browser.cache.disk.enable: false
browser.cache.disk.filesystem_reported: 1
browser.cache.disk.hashstats_reported: 1
browser.cache.disk.metadata_memory_limit: 2500
browser.cache.disk.parent_directory: R:\FFcache
browser.cache.disk.smart_size_cached_value: 358400
browser.cache.disk.smart_size.enabled: false
browser.cache.disk.smart_size.first_run: false
browser.cache.disk.smart_size.use_old_max: false
browser.cache.frecency_experiment: 4
browser.cache.memory.capacity: 4194302
browser.cache.memory.max_entry_size: 51200
browser.download.folderList: 2
browser.download.importedFromSqlite: true
browser.download.manager.alertOnEXEOpen: true
browser.history_expire_days.mirror: 180
browser.places.smartBookmarksVersion: 7
browser.search.useDBForOrder: true
browser.sessionstore.restore_on_demand: false
browser.sessionstore.upgradeBackup.latestBuildID: 20160105164030
browser.startup.homepage: about:home
browser.startup.homepage_override.buildID: 20160105164030
browser.startup.homepage_override.mstone: 43.0.4
browser.tabs.warnOnClose: false
browser.tabs.warnOnOpen: false
browser.urlbar.userMadeSearchSuggestionsChoice: true
dom.apps.reset-permissions: true
dom.disable_window_open_feature.status: false
dom.disable_window_status_change: false
dom.ipc.plugins.enabled.npietab2.dll: true
dom.max_chrome_script_run_time: 40
dom.max_script_run_time: 0
dom.mozApps.used: true
dom.w3c_touch_events.expose: false
extensions.lastAppVersion: 43.0.4
font.internaluseonly.changed: false
general.useragent.extra.microsoftdotnet: ( .NET CLR 3.5.30729; .NET4.0E)
gfx.blacklist.suggested-driver-version: 257.21
gfx.crash-guard.d3d11layers.appVersion: 43.0.4
gfx.crash-guard.d3d11layers.deviceID: 0x06dd
gfx.crash-guard.d3d11layers.driverVersion: 9.18.13.4144
gfx.crash-guard.d3d11layers.feature-d2d: true
gfx.crash-guard.d3d11layers.feature-d3d11: true
gfx.crash-guard.glcontext.gfx.driver-init.direct3d11-angle: true
gfx.crash-guard.glcontext.gfx.driver-init.webgl-angle: true
gfx.crash-guard.glcontext.gfx.driver-init.webgl-angle-force-d3d11: false
gfx.crash-guard.glcontext.gfx.driver-init.webgl-angle-force-warp: false
gfx.crash-guard.glcontext.gfx.driver-init.webgl-angle-try-d3d11: true
gfx.crash-guard.status.d3d11layers: 2
gfx.crash-guard.status.d3d9video: 2
gfx.crash-guard.status.glcontext: 2
gfx.direct3d.last_used_feature_level_idx: 0
gfx.driver-init.appVersion: 42.0
gfx.driver-init.deviceID: 0x06dd
gfx.driver-init.driverVersion: 9.18.13.4144
gfx.driver-init.feature-d2d: true
gfx.driver-init.feature-d3d11: true
gfx.driver-init.status: 2
keyword.URL: https://uk.search.yahoo.com/search?fr=greentree_ff1&ei=utf-8&ilc=12&type=242154&p=
media.directshow.enabled: false
media.gmp-eme-adobe.abi: x86-msvc
media.gmp-eme-adobe.lastUpdate: 1446718157
media.gmp-eme-adobe.version: 15
media.gmp-gmpopenh264.abi: x86-msvc
media.gmp-gmpopenh264.lastUpdate: 1451577599
media.gmp-gmpopenh264.version: 1.5.3
media.gmp-manager.buildID: 20160105164030
media.gmp-manager.lastCheck: 1453905307
media.hardware-video-decoding.failed: false
media.youtube-ua.override.to: 43
network.auth.allow-subresource-auth: 2
network.cookie.prefsMigrated: true
network.http.keep-alive.timeout: 300
network.predictor.cleaned-up: true
permissions.default.image: 0
places.database.lastMaintenance: 1453925907
places.history.expiration.transient_current_max_pages: 104858
places.history.expiration.transient_optimal_database_size: 139484692
places.last_vacuum: 1300785292
plugin.disable_full_page_plugin_for_types: application/vnd.adobe.xfdf,application/vnd.fdf,application/vnd.adobe.xdp+xml,chemical/x-chemdraw,audio/aiff,audio/x-aiff
plugin.importedState: true
plugin.state.flash: 1
plugin.state.npdeployjava: 0
plugin.state.nppdf: 2
print.print_printer: Lexmark E250d
print.printer_Adobe_PDF.print_bgcolor: false
print.printer_Adobe_PDF.print_bgimages: false
print.printer_Adobe_PDF.print_colorspace:
print.printer_Adobe_PDF.print_command:
print.printer_Adobe_PDF.print_downloadfonts: false
print.printer_Adobe_PDF.print_duplex: 896
print.printer_Adobe_PDF.print_edge_bottom: 0
print.printer_Adobe_PDF.print_edge_left: 0
print.printer_Adobe_PDF.print_edge_right: 0
print.printer_Adobe_PDF.print_edge_top: 0
print.printer_Adobe_PDF.print_evenpages: true
print.printer_Adobe_PDF.print_footercenter: &D
print.printer_Adobe_PDF.print_footerleft: &U
print.printer_Adobe_PDF.print_footerright: &P
print.printer_Adobe_PDF.print_headercenter:
print.printer_Adobe_PDF.print_headerleft:
print.printer_Adobe_PDF.print_headerright:
print.printer_Adobe_PDF.print_in_color: true
print.printer_Adobe_PDF.print_margin_bottom: 0.196527779102325
print.printer_Adobe_PDF.print_margin_left: 0.196527779102325
print.printer_Adobe_PDF.print_margin_right: 0.196527779102325
print.printer_Adobe_PDF.print_margin_top: 0.196527779102325
print.printer_Adobe_PDF.print_oddpages: true
print.printer_Adobe_PDF.print_orientation: 0
print.printer_Adobe_PDF.print_page_delay: 50
print.printer_Adobe_PDF.print_pagedelay: 500
print.printer_Adobe_PDF.print_paper_data: 9
print.printer_Adobe_PDF.print_paper_height: 11.00
print.printer_Adobe_PDF.print_paper_name:
print.printer_Adobe_PDF.print_paper_size_type: 0
print.printer_Adobe_PDF.print_paper_size_unit: 1
print.printer_Adobe_PDF.print_paper_width: 8.50
print.printer_Adobe_PDF.print_plex_name:
print.printer_Adobe_PDF.print_resolution: 7168
print.printer_Adobe_PDF.print_resolution_name:
print.printer_Adobe_PDF.print_reversed: false
print.printer_Adobe_PDF.print_scaling: 0.90
print.printer_Adobe_PDF.print_shrink_to_fit: false
print.printer_Adobe_PDF.print_to_file: false
print.printer_Adobe_PDF.print_unwriteable_margin_bottom: 0
print.printer_Adobe_PDF.print_unwriteable_margin_left: 0
print.printer_Adobe_PDF.print_unwriteable_margin_right: 0
print.printer_Adobe_PDF.print_unwriteable_margin_top: 0
print.printer_Brother_FAX-2820_USB_Printer.print_bgcolor: false
print.printer_Brother_FAX-2820_USB_Printer.print_bgimages: false
print.printer_Brother_FAX-2820_USB_Printer.print_command:
print.printer_Brother_FAX-2820_USB_Printer.print_downloadfonts: false
print.printer_Brother_FAX-2820_USB_Printer.print_edge_bottom: 0
print.printer_Brother_FAX-2820_USB_Printer.print_edge_left: 0
print.printer_Brother_FAX-2820_USB_Printer.print_edge_right: 0
print.printer_Brother_FAX-2820_USB_Printer.print_edge_top: 0
print.printer_Brother_FAX-2820_USB_Printer.print_evenpages: true
print.printer_Brother_FAX-2820_USB_Printer.print_footercenter:
print.printer_Brother_FAX-2820_USB_Printer.print_footerleft:
print.printer_Brother_FAX-2820_USB_Printer.print_footerright:
print.printer_Brother_FAX-2820_USB_Printer.print_headercenter:
print.printer_Brother_FAX-2820_USB_Printer.print_headerleft:
print.printer_Brother_FAX-2820_USB_Printer.print_headerright:
print.printer_Brother_FAX-2820_USB_Printer.print_in_color: true
print.printer_Brother_FAX-2820_USB_Printer.print_margin_bottom: 0.236111119389534
print.printer_Brother_FAX-2820_USB_Printer.print_margin_left: 0.472222238779068
print.printer_Brother_FAX-2820_USB_Printer.print_margin_right: 0.472222238779068
print.printer_Brother_FAX-2820_USB_Printer.print_margin_top: 0.236111119389534
print.printer_Brother_FAX-2820_USB_Printer.print_oddpages: true
print.printer_Brother_FAX-2820_USB_Printer.print_orientation: 0
print.printer_Brother_FAX-2820_USB_Printer.print_page_delay: 50
print.printer_Brother_FAX-2820_USB_Printer.print_pagedelay: 500
print.printer_Brother_FAX-2820_USB_Printer.print_paper_data: 9
print.printer_Brother_FAX-2820_USB_Printer.print_paper_height: 11.00
print.printer_Brother_FAX-2820_USB_Printer.print_paper_size_type: 0
print.printer_Brother_FAX-2820_USB_Printer.print_paper_size_unit: 1
print.printer_Brother_FAX-2820_USB_Printer.print_paper_width: 8.50
print.printer_Brother_FAX-2820_USB_Printer.print_reversed: false
print.printer_Brother_FAX-2820_USB_Printer.print_scaling: 0.80
print.printer_Brother_FAX-2820_USB_Printer.print_shrink_to_fit: false
print.printer_Brother_FAX-2820_USB_Printer.print_to_file: false
print.printer_Brother_FAX-2820_USB_Printer.print_unwriteable_margin_bottom: 0
print.printer_Brother_FAX-2820_USB_Printer.print_unwriteable_margin_left: 0
print.printer_Brother_FAX-2820_USB_Printer.print_unwriteable_margin_right: 0
print.printer_Brother_FAX-2820_USB_Printer.print_unwriteable_margin_top: 0
print.printer_HP_Color_LaserJet_CP1215.print_bgcolor: false
print.printer_HP_Color_LaserJet_CP1215.print_bgimages: false
print.printer_HP_Color_LaserJet_CP1215.print_command:
print.printer_HP_Color_LaserJet_CP1215.print_downloadfonts: false
print.printer_HP_Color_LaserJet_CP1215.print_edge_bottom: 0
print.printer_HP_Color_LaserJet_CP1215.print_edge_left: 0
print.printer_HP_Color_LaserJet_CP1215.print_edge_right: 0
print.printer_HP_Color_LaserJet_CP1215.print_edge_top: 0
print.printer_HP_Color_LaserJet_CP1215.print_evenpages: true
print.printer_HP_Color_LaserJet_CP1215.print_footercenter:
print.printer_HP_Color_LaserJet_CP1215.print_footerleft:
print.printer_HP_Color_LaserJet_CP1215.print_footerright:
print.printer_HP_Color_LaserJet_CP1215.print_headercenter:
print.printer_HP_Color_LaserJet_CP1215.print_headerleft:
print.printer_HP_Color_LaserJet_CP1215.print_headerright:
print.printer_HP_Color_LaserJet_CP1215.print_in_color: true
print.printer_HP_Color_LaserJet_CP1215.print_margin_bottom: 0.236111119389534
print.printer_HP_Color_LaserJet_CP1215.print_margin_left: 0.472222238779068
print.printer_HP_Color_LaserJet_CP1215.print_margin_right: 0.472222238779068
print.printer_HP_Color_LaserJet_CP1215.print_margin_top: 0.236111119389534
print.printer_HP_Color_LaserJet_CP1215.print_oddpages: true
print.printer_HP_Color_LaserJet_CP1215.print_orientation: 0
print.printer_HP_Color_LaserJet_CP1215.print_page_delay: 50
print.printer_HP_Color_LaserJet_CP1215.print_pagedelay: 500
print.printer_HP_Color_LaserJet_CP1215.print_paper_data: 9
print.printer_HP_Color_LaserJet_CP1215.print_paper_height: 11.00
print.printer_HP_Color_LaserJet_CP1215.print_paper_size_type: 0
print.printer_HP_Color_LaserJet_CP1215.print_paper_size_unit: 1
print.printer_HP_Color_LaserJet_CP1215.print_paper_width: 8.50
print.printer_HP_Color_LaserJet_CP1215.print_reversed: false
print.printer_HP_Color_LaserJet_CP1215.print_scaling: 0.80
print.printer_HP_Color_LaserJet_CP1215.print_shrink_to_fit: false
print.printer_HP_Color_LaserJet_CP1215.print_to_file: false
print.printer_HP_Color_LaserJet_CP1215.print_unwriteable_margin_bottom: 0
print.printer_HP_Color_LaserJet_CP1215.print_unwriteable_margin_left: 0
print.printer_HP_Color_LaserJet_CP1215.print_unwriteable_margin_right: 0
print.printer_HP_Color_LaserJet_CP1215.print_unwriteable_margin_top: 0
print.printer_HP_LaserJet_1018.print_bgcolor: false
print.printer_HP_LaserJet_1018.print_bgimages: false
print.printer_HP_LaserJet_1018.print_colorspace:
print.printer_HP_LaserJet_1018.print_command:
print.printer_HP_LaserJet_1018.print_downloadfonts: false
print.printer_HP_LaserJet_1018.print_duplex: 896
print.printer_HP_LaserJet_1018.print_edge_bottom: 0
print.printer_HP_LaserJet_1018.print_edge_left: 0
print.printer_HP_LaserJet_1018.print_edge_right: 0
print.printer_HP_LaserJet_1018.print_edge_top: 0
print.printer_HP_LaserJet_1018.print_evenpages: true
print.printer_HP_LaserJet_1018.print_footercenter: &D
print.printer_HP_LaserJet_1018.print_footerleft: &U
print.printer_HP_LaserJet_1018.print_footerright: &P
print.printer_HP_LaserJet_1018.print_headercenter:
print.printer_HP_LaserJet_1018.print_headerleft:
print.printer_HP_LaserJet_1018.print_headerright:
print.printer_HP_LaserJet_1018.print_in_color: true
print.printer_HP_LaserJet_1018.print_margin_bottom: 0.196527779102325
print.printer_HP_LaserJet_1018.print_margin_left: 0.196527779102325
print.printer_HP_LaserJet_1018.print_margin_right: 0.196527779102325
print.printer_HP_LaserJet_1018.print_margin_top: 0.196527779102325
print.printer_HP_LaserJet_1018.print_oddpages: true
print.printer_HP_LaserJet_1018.print_orientation: 0
print.printer_HP_LaserJet_1018.print_page_delay: 50
print.printer_HP_LaserJet_1018.print_paper_data: 9
print.printer_HP_LaserJet_1018.print_paper_height: 11.00
print.printer_HP_LaserJet_1018.print_paper_name:
print.printer_HP_LaserJet_1018.print_paper_size_type: 0
print.printer_HP_LaserJet_1018.print_paper_size_unit: 1
print.printer_HP_LaserJet_1018.print_paper_width: 8.50
print.printer_HP_LaserJet_1018.print_plex_name:
print.printer_HP_LaserJet_1018.print_resolution: 7168
print.printer_HP_LaserJet_1018.print_resolution_name:
print.printer_HP_LaserJet_1018.print_reversed: false
print.printer_HP_LaserJet_1018.print_scaling: 0.60
print.printer_HP_LaserJet_1018.print_shrink_to_fit: false
print.printer_HP_LaserJet_1018.print_to_file: false
print.printer_HP_LaserJet_1018.print_unwriteable_margin_bottom: 0
print.printer_HP_LaserJet_1018.print_unwriteable_margin_left: 0
print.printer_HP_LaserJet_1018.print_unwriteable_margin_right: 0
print.printer_HP_LaserJet_1018.print_unwriteable_margin_top: 0
print.printer_HP_LaserJet_CP_1025nw.print_bgcolor: false
print.printer_HP_LaserJet_CP_1025nw.print_bgimages: false
print.printer_HP_LaserJet_CP_1025nw.print_colorspace:
print.printer_HP_LaserJet_CP_1025nw.print_command:
print.printer_HP_LaserJet_CP_1025nw.print_downloadfonts: false
print.printer_HP_LaserJet_CP_1025nw.print_duplex: 896
print.printer_HP_LaserJet_CP_1025nw.print_edge_bottom: 0
print.printer_HP_LaserJet_CP_1025nw.print_edge_left: 0
print.printer_HP_LaserJet_CP_1025nw.print_edge_right: 0
print.printer_HP_LaserJet_CP_1025nw.print_edge_top: 0
print.printer_HP_LaserJet_CP_1025nw.print_evenpages: true
print.printer_HP_LaserJet_CP_1025nw.print_footercenter: &D
print.printer_HP_LaserJet_CP_1025nw.print_footerleft: &U
print.printer_HP_LaserJet_CP_1025nw.print_footerright: &P
print.printer_HP_LaserJet_CP_1025nw.print_headercenter:
print.printer_HP_LaserJet_CP_1025nw.print_headerleft:
print.printer_HP_LaserJet_CP_1025nw.print_headerright:
print.printer_HP_LaserJet_CP_1025nw.print_in_color: true
print.printer_HP_LaserJet_CP_1025nw.print_margin_bottom: 0.196527779102325
print.printer_HP_LaserJet_CP_1025nw.print_margin_left: 0.196527779102325
print.printer_HP_LaserJet_CP_1025nw.print_margin_right: 0.196527779102325
print.printer_HP_LaserJet_CP_1025nw.print_margin_top: 0.196527779102325
print.printer_HP_LaserJet_CP_1025nw.print_oddpages: true
print.printer_HP_LaserJet_CP_1025nw.print_orientation: 0
print.printer_HP_LaserJet_CP_1025nw.print_page_delay: 50
print.printer_HP_LaserJet_CP_1025nw.print_paper_data: 9
print.printer_HP_LaserJet_CP_1025nw.print_paper_height: 11.00
print.printer_HP_LaserJet_CP_1025nw.print_paper_name:
print.printer_HP_LaserJet_CP_1025nw.print_paper_size_type: 0
print.printer_HP_LaserJet_CP_1025nw.print_paper_size_unit: 1
print.printer_HP_LaserJet_CP_1025nw.print_paper_width: 8.50
print.printer_HP_LaserJet_CP_1025nw.print_plex_name:
print.printer_HP_LaserJet_CP_1025nw.print_resolution: 7168
print.printer_HP_LaserJet_CP_1025nw.print_resolution_name:
print.printer_HP_LaserJet_CP_1025nw.print_reversed: false
print.printer_HP_LaserJet_CP_1025nw.print_scaling: 0.58
print.printer_HP_LaserJet_CP_1025nw.print_shrink_to_fit: true
print.printer_HP_LaserJet_CP_1025nw.print_to_file: false
print.printer_HP_LaserJet_CP_1025nw.print_unwriteable_margin_bottom: 0
print.printer_HP_LaserJet_CP_1025nw.print_unwriteable_margin_left: 0
print.printer_HP_LaserJet_CP_1025nw.print_unwriteable_margin_right: 0
print.printer_HP_LaserJet_CP_1025nw.print_unwriteable_margin_top: 0
print.printer_ImagePrinter.print_bgcolor: false
print.printer_ImagePrinter.print_bgimages: false
print.printer_ImagePrinter.print_colorspace:
print.printer_ImagePrinter.print_command:
print.printer_ImagePrinter.print_downloadfonts: false
print.printer_ImagePrinter.print_duplex: 896
print.printer_ImagePrinter.print_edge_bottom: 0
print.printer_ImagePrinter.print_edge_left: 0
print.printer_ImagePrinter.print_edge_right: 0
print.printer_ImagePrinter.print_edge_top: 0
print.printer_ImagePrinter.print_evenpages: true
print.printer_ImagePrinter.print_footercenter: &D
print.printer_ImagePrinter.print_footerleft: &U
print.printer_ImagePrinter.print_footerright: &P
print.printer_ImagePrinter.print_headercenter:
print.printer_ImagePrinter.print_headerleft:
print.printer_ImagePrinter.print_headerright:
print.printer_ImagePrinter.print_in_color: true
print.printer_ImagePrinter.print_margin_bottom: 0.196527779102325
print.printer_ImagePrinter.print_margin_left: 0.196527779102325
print.printer_ImagePrinter.print_margin_right: 0.196527779102325
print.printer_ImagePrinter.print_margin_top: 0.196527779102325
print.printer_ImagePrinter.print_oddpages: true
print.printer_ImagePrinter.print_orientation: 0
print.printer_ImagePrinter.print_page_delay: 50
print.printer_ImagePrinter.print_paper_data: 9
print.printer_ImagePrinter.print_paper_height: 11.00
print.printer_ImagePrinter.print_paper_name:
print.printer_ImagePrinter.print_paper_size_type: 0
print.printer_ImagePrinter.print_paper_size_unit: 1
print.printer_ImagePrinter.print_paper_width: 8.50
print.printer_ImagePrinter.print_plex_name:
print.printer_ImagePrinter.print_resolution: 7168
print.printer_ImagePrinter.print_resolution_name:
print.printer_ImagePrinter.print_reversed: false
print.printer_ImagePrinter.print_scaling: 0.60
print.printer_ImagePrinter.print_shrink_to_fit: true
print.printer_ImagePrinter.print_to_file: false
print.printer_ImagePrinter.print_unwriteable_margin_bottom: 0
print.printer_ImagePrinter.print_unwriteable_margin_left: 0
print.printer_ImagePrinter.print_unwriteable_margin_right: 0
print.printer_ImagePrinter.print_unwriteable_margin_top: 0
print.printer_Lexmark_E250d.print_bgcolor: false
print.printer_Lexmark_E250d.print_bgimages: false
print.printer_Lexmark_E250d.print_command:
print.printer_Lexmark_E250d.print_downloadfonts: false
print.printer_Lexmark_E250d.print_edge_bottom: 0
print.printer_Lexmark_E250d.print_edge_left: 0
print.printer_Lexmark_E250d.print_edge_right: 0
print.printer_Lexmark_E250d.print_edge_top: 0
print.printer_Lexmark_E250d.print_evenpages: true
print.printer_Lexmark_E250d.print_footercenter:
print.printer_Lexmark_E250d.print_footerleft:
print.printer_Lexmark_E250d.print_footerright:
print.printer_Lexmark_E250d.print_headercenter:
print.printer_Lexmark_E250d.print_headerleft:
print.printer_Lexmark_E250d.print_headerright:
print.printer_Lexmark_E250d.print_in_color: true
print.printer_Lexmark_E250d.print_margin_bottom: 0.236111119389534
print.printer_Lexmark_E250d.print_margin_left: 0.472222238779068
print.printer_Lexmark_E250d.print_margin_right: 0.472222238779068
print.printer_Lexmark_E250d.print_margin_top: 0.236111119389534
print.printer_Lexmark_E250d.print_oddpages: true
print.printer_Lexmark_E250d.print_orientation: 0
print.printer_Lexmark_E250d.print_page_delay: 50
print.printer_Lexmark_E250d.print_pagedelay: 500
print.printer_Lexmark_E250d.print_paper_data: 9
print.printer_Lexmark_E250d.print_paper_height: 11.00
print.printer_Lexmark_E250d.print_paper_size_type: 0
print.printer_Lexmark_E250d.print_paper_size_unit: 1
print.printer_Lexmark_E250d.print_paper_width: 8.50
print.printer_Lexmark_E250d.print_reversed: false
print.printer_Lexmark_E250d.print_scaling: 0.80
print.printer_Lexmark_E250d.print_shrink_to_fit: false
print.printer_Lexmark_E250d.print_to_file: false
print.printer_Lexmark_E250d.print_unwriteable_margin_bottom: 0
print.printer_Lexmark_E250d.print_unwriteable_margin_left: 0
print.printer_Lexmark_E250d.print_unwriteable_margin_right: 0
print.printer_Lexmark_E250d.print_unwriteable_margin_top: 0
privacy.clearOnShutdown.cache: false
privacy.clearOnShutdown.cookies: false
privacy.clearOnShutdown.downloads: false
privacy.clearOnShutdown.formdata: false
privacy.clearOnShutdown.history: false
privacy.clearOnShutdown.sessions: false
privacy.cpd.cache: false
privacy.cpd.cookies: false
privacy.cpd.extensions-betterprivacy: false
privacy.cpd.formdata: false
privacy.cpd.sessions: false
privacy.donottrackheader.enabled: true
privacy.donottrackheader.value: 1
privacy.sanitize.migrateFx3Prefs: true
privacy.sanitize.timeSpan: 0
security.disable_button.openCertManager: false
security.disable_button.openDeviceManager: false
security.OCSP.disable_button.managecrl: false
security.ssl.errorReporting.automatic: true
security.warn_viewing_mixed: false
storage.vacuum.last.index: 1
storage.vacuum.last.places.sqlite: 1453294100

Important Locked Preferences
----------------------------

JavaScript
----------

Incremental GC: true

Accessibility
-------------

Activated: false
Prevent Accessibility: 0

Library Versions
----------------

NSPR
Expected minimum version: 4.10.10
Version in use: 4.10.10

NSS
Expected minimum version: 3.20.2 Basic ECC
Version in use: 3.20.2 Basic ECC

NSSSMIME
Expected minimum version: 3.20.2 Basic ECC
Version in use: 3.20.2 Basic ECC

NSSSSL
Expected minimum version: 3.20.2 Basic ECC
Version in use: 3.20.2 Basic ECC

NSSUTIL
Expected minimum version: 3.20.2
Version in use: 3.20.2

Experimental Features
---------------------
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 12

3 years ago
Interesting to see that!
   print.printer_Lexmark_E250d.print
   print.printer_Brother_FAX-2820
   print.printer_HP_Color_LaserJet_CP1215
- should not be present, uninstalled long ago!
How do I remove the traces?
(Reporter)

Comment 13

3 years ago
2.  

(In reply to Nicholas Nethercote [:njn] from comment #10)
> > Also we think switching to the 64-bit version of Firefox might help ...

OK, can try that later.  Thanks.
This is most just a hunch, but I'd suggest disabling Trustwave SecureBrowsing, IE Tab 2, and maybe The Camelizer to begin with, to see if that helps.

(In reply to Dr B W Darvell from comment #12)
> Interesting to see that!
>    print.printer_Lexmark_E250d.print
>    print.printer_Brother_FAX-2820
>    print.printer_HP_Color_LaserJet_CP1215
> - should not be present, uninstalled long ago!
> How do I remove the traces?

You can edit and delete preferences in about:config. It probably doesn't matter to have them there, though.
Summary: Excessive CPU usage with 14+ extensions present → Excessive CPU usage with 23 extensions present
(Reporter)

Comment 15

3 years ago
3.

(In reply to Eric Rahm [:erahm] from comment #9)
> If it's not sensitive, can you let us know which site this is?

Sure - how do I identify it?  (Forgive the ignorance.)
(Reporter)

Comment 16

3 years ago
(In reply to Nicholas Nethercote [:njn] from comment #14)
> This is most just a hunch, but I'd suggest disabling Trustwave
> SecureBrowsing, IE Tab 2, and maybe The Camelizer to begin with, to see if
> that helps.

Done.  I'll see what happens tomorrow.

> (In reply to Dr B W Darvell from comment #12)

> > How do I remove the traces?
> 
> You can edit and delete preferences in about:config. It probably doesn't
> matter to have them there, though.

Noted, thanks.  Might as well, though, they are useless.
> > If it's not sensitive, can you let us know which site this is?
> 
> Sure - how do I identify it?  (Forgive the ignorance.)

You'll have to re-measure, but with the "anonymize" box unchecked. The measurements won't be the same, but hopefully will be broadly similar so you can tell what site "<anonymized-509>" actually was.
(Reporter)

Comment 18

3 years ago
(In reply to Nicholas Nethercote [:njn] from comment #17)
> You'll have to re-measure, but with the "anonymize" box unchecked. The
> measurements won't be the same, but hopefully will be broadly similar so you
> can tell what site "<anonymized-509>" actually was.

OK, will play tomorrow.  Thanks.
Comment hidden (off-topic)
Comment hidden (off-topic)
Comment hidden (off-topic)
Comment hidden (off-topic)
Comment hidden (off-topic)
(Reporter)

Comment 24

3 years ago
1.  Running FF-64, and have returned all extensions except Camelizer, with all previous pinned tabs in place.  (I also added PrintEdit, to assist info. capture.) Normal business otherwise - quite a lot of tabs, diverse sites.
   BTW: The Plug-in container now seems to be much more efficient (watching with Process Explorer).

2.  Now set cache to use 4 GB RAM - FF presently using total of 2.8 GB, plus 250 MB for plug-in container.  (It was formerly set at 1 GB RAMdisk, although actually running at 2.0 - 2.3 GB)

About:Cache
memory
Number of entries: 	6801
Maximum storage size: 	4194302 KiB
Storage in use: 	73605 KiB
Storage disk location: 	none, only stored in memory 

3.  I have checked about:memory for event counts  a few times but not found anything untoward; I cannot identify what caused the problem identified earlier, and no sign of a memory hog either.
This is after a full day yesterday, and 4 h today (after "Sleep" - which is not a problem, then, either, I guess).  Total CPU FF = 2 h 30 min, PIC = 30 min.
4.  I have now enabled the Camelizer again (which to me was most suspicious), and will see if that has any effect.  
I'll report again next week.

Q: Could I raise again the question of whether a per-tab CPU usage can be shown?  That would help in such circs, I would imagine.

Thanks.
> Q: Could I raise again the question of whether a per-tab CPU usage can be
> shown?  That would help in such circs, I would imagine.

There's an experimental performance tool that finds expensive add-ons and tabs. You see it by typing "about:performance" into the address bar. However, it's currently only enabled on Nightly builds. I can't work out how to turn it on in release builds like 43.

Yoric, can you tell us how to do it? Thank you.
Flags: needinfo?(dteller)
Sorry, it can't be turned on in a release build, it's #ifdef-away.
Flags: needinfo?(dteller)
Ok. Dr B W Darvell, if you're feeling motivated, you could try out a Nightly build (from https://nightly.mozilla.org/) and try using about:performance there. I'm not sure if Nightly uses a different profile to release Firefox. If it does, you'll need to re-install your add-ons in the Nightly profile, which I understand is annoying.
(Reporter)

Comment 28

3 years ago
(In reply to Nicholas Nethercote [:njn] from comment #27)
> Ok. Dr B W Darvell, if you're feeling motivated, 
To be honest, I have taken on so much that to add another task would be to the detriment of work.  This thread was motivated because the problem was interfering.
>#ifdef-away
I take it that means disabled ...
I'll just have to wait for it to appear, then.

Anyway, I'll keep monitoring and let you know if the problem recurs.

One thing that I notice now, though, on restarting - only pinned tabs were reloaded, all others had gone (I did not close them before shutting down).  I did not change any setting (knowingly).  It's not because of RAM cache, is it?  I'll test later.
(In reply to Dr B W Darvell from comment #28)
> One thing that I notice now, though, on restarting - only pinned tabs were
> reloaded, all others had gone (I did not close them before shutting down). 
> I did not change any setting (knowingly).

I have that problem too, sometimes. When you start the browser, it will show the about:home page, which should have a button underneath the search box that says something like "click here to restore tabs from the previous session". You can also go to the history menu option and try something like "restore previous session" and maybe it will bring those tabs back.
(Reporter)

Comment 30

3 years ago
Thanks for that - I did not notice the button, and the restore option is now greyed out.
I'll keep an eye on it.
(Reporter)

Comment 31

3 years ago
A curious observa tion:
Blank tab, Google search:
https://www.google.co.uk/search?q=Sky+tv+view&ie=utf-8&oe=utf-8&gws_rd=cr&ei=XZ2vVoPeFcHSesTRsdAG#cns=0&gws_rd=cr
- CPU activity then runs at around 12% continuously, with frequent excursions to 30%.  Switch to a blank tab - almost immediately CPU becomes negligible 2 ~3%, with occasional other activity peaks, but nothing big.
So:
https://www.google.co.uk/search?q=dummy+search&ie=utf-8&oe=utf-8&gws_rd=cr&ei=0aSvVrqvFIa6Uf2VtZAC
- which settles down after a few seconds some activity, but nothing problematic.  Switch back to the Sky tab: steady activity again at 8 ~10%, but then stops when I switch away.  Reproducible.
I have Adblock Plus operating, so no ads show.  I do not see many 'events' for this tab.
Could it be that some search tabs could continue to be active when not in the foreground to have caused the initiating problem?
Another example: Search on "Monet exhibition"- background soon very quiet ...  switch the Sky search, busy again.  Definitely something there.
(Reporter)

Comment 32

3 years ago
I think I may have sussed it.
I had set the cache in RAM at 4 GB, as I said above.  All was fine - lots of tabs coming and going (about 30 open now).  All the while I was monitoring the FF process window in Process Explorer.  When Private Bytes in the Performance Graph tab reached 4 GB the CPU activity went high and stayed there for a while.  It is now showing 4.2 GB ...
Closing tabs makes no difference.
It seems to me that the dumping of old cache items is not very efficient, and while there may be some improvement now using RAM over RAM disk, it still slowed me to a crawl for too long, and is running at a much higher than quiescent level still 20 minutes later.

About:memory showed this:

100,598 (100.0%) -- event-counts
├──100,529 (99.93%) -- window-objects
│  ├───63,427 (63.05%) -- top(none)/detached
│  │   ├──44,821 (44.55%) ── window(about:memory)/dom/event-listeners [2]
│  │   ├──16,598 (16.50%) ── window(chrome://browser/content/browser.xul)/dom/event-listeners [15]
│  │   └───2,008 (02.00%) ++ (21 tiny)
│  ├───14,824 (14.74%) ++ (79 tiny)

(There is only one About: tab open.)
I have not seen any tab make excessive calls other than that shown here (having checked several times).

I shall now increase the cache allowed and see what happens.
Whiteboard: [memshrink] → [memshrink:P3]
(Reporter)

Comment 33

3 years ago
Am I right in thinking that FF is not capable of running threads on other CPUs, or is there a thread limit, or overall throttling to limit the max. CPU% that it can take?

I ask because no matter how busy it seems to peak now at around 30%, when it appears that there is plenty of CPU capacity remaining.  If it were processor number-aware, and subbed off cache-maintenance to a separate thread, perhaps it would become more responsive?  Or is all that totally naive because the system handles it?

Thanks.

Meanwhile, I have reached about 5 GB in the RAM cache, with occasional but annoying very slow response, but so far it has not gone into that extended unresponsive state.
(Reporter)

Comment 34

3 years ago
Have we lost interest, chaps?

Anyway, I have just found this:
http://www.ghacks.net/2013/09/23/mozilla-launches-firefoxs-new-caching-back-end-nightly-version/
http://www.janbambas.cz/mozilla-firefox-new-http-cache-is-live/
and others.
browser.cache.use_new_backend was "0" - I have now set it to "1".
browser.cache.use_new_backend_temp is (still) True

I get the impression that it should have been turned on by default by now.  Is this a relevant setting, do you think?

Are we missing anything else?
If you let Firefox manage the cache on its own (rather than setting your own custom values), does the CPU usage stay more stable over time?
(Reporter)

Comment 36

3 years ago
No, that is certainly not my impression. It is even better now, I think, that I have allocated 8 GB.  At one point I saw that memory usage had reached 7.2 GB, but I have not experienced the extreme slowness that I was complaining about at the outset.  There are still episodes of higher activity, but they are relatively short and do not seem to interfere too much.

Noticing that the memory content goes up and down (in small amounts) all the time, as the CPU activity varies as well, even though I am not doing anything, I get the impression that it was the major cache clean-up triggered when full that caused the huge and sustained CPU usage.

Having just checked, I have about 84 tabs open (I know - but this is a consequence of the work I do), with the memory at about 2.2 GB.  CPU variable, but no lock-ups.
Hi Darvell, thanks for hanging and providing all the details related to this issue. Unfortunately, problems like these are difficult to diagnose and debug. 
If you have some time, would you be willing to try collecting a profile to see if anything stands out there? The link below has directions for how to collect and submit a profile.
https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Profiling_with_the_Built-in_Profiler

Thanks!
(Reporter)

Comment 38

3 years ago
Hi,
TFTM
I did Tools | Web Developer | Performance, ran it until the buffer was full, and saved the profile - which gave a .json file (20 MB), no .sym (which is what I was expecting, based on my limited understanding of that page you linked.
What am I missing?
Thanks,

BWD

FF was very busy running that, and it still is ...
(Reporter)

Comment 39

3 years ago
BTW: running that process and trying to save results in FF locking up for long periods.
Oddly, since the last update (47.0.1), CPU usage seems to have dropped ... it could be a coincidence.

BWD
(In reply to Dr B W Darvell from comment #39)
> Oddly, since the last update (47.0.1), CPU usage seems to have dropped ...
> it could be a coincidence.
> 
> BWD

Darvell, from your last comment my understanding is that this issue is not reproducible on the latest release version 47.0.1. Is this right?
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 41

3 years ago
No, I cannot say that.  It appears to be a coincidence, that is all.  I am still getting very slow occasions, but possibly less severe than when I started this.
I am still looking forward to a multi-thread, multi-cpu version.
( https://bugzilla.mozilla.org/show_bug.cgi?id=392073 , https://wiki.mozilla.org/Electrolysis )
Hi Darvell, thanks again for you patience and for providing all the details related to this issue. 

If you have some time, would you be willing to try collecting a Cleopatra profile that hopefully would provide useful information for diagnose and debug.

The link below has explicit directions for how to collect and submit a profile.
https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Reporting_a_Performance_Problem

Thanks!
(Reporter)

Comment 43

3 years ago
Hi,
Sorry for the delay - I have been travelling.
Sure, I'll give that a go, but in a day or so - a few jobs to do first.

BWD
(Reporter)

Comment 44

3 years ago
Apologies, it has not been possible to devote time to this yet.  I'll have to return to the job in October.

BWD
Hi Darvell, 

When time permits, could you please try to collect a Cleopatra profile?

Thanks,
Brindusa
I am marking this as Incomplete due to lack of response from reporter.

Darvell, when have the time, please feel free to reopen it and provide the Cleopatra report if the issue is still reproducible on the latest Firefox version (current 49.0.2)
Status: UNCONFIRMED → RESOLVED
Last Resolved: 2 years ago
Resolution: --- → INCOMPLETE
(Reporter)

Comment 47

2 years ago
Hi,
My apologies, I am just not able to do this for a while - snowed under and multitasking.
I will endeavour to do this in about 10 - 12 days.  Things should be clearer then.

BWD
(Reporter)

Comment 48

2 years ago
Hi,
I cleared my work queue and started to try to do as requested.
Immediately (even after restarting) I had dreadful delays - up to 40% CPU reported - that took ages to die down.
I assume I was supposed load the Gecko Profiler extension - correct?
So I tried a run, and received a warning in Cleopatra (after clicking the globe icon and clicking 'share') about my Profile being publicly visible.  I cannot see what file is to be uploaded (i.e., I cannot check to see what content I am revealing).  If it is supposed to be a json file, then there is no new one anywhere that I can see to correspond to allow me to check.

Please advise:
1.  Am I doing the right thing?
2.  Where do I find the file that is to be uploaded.
3.  Does this contain personal information of any kind that compromises my security in any way?
4.  The profiler is said to collect a few seconds-worth of data - how many is a few?  If I miss the behaviour we are trying to understand it will be pointless, of course.  How long have I got?
5.  I take it that the profiler is running continuously and should be disabled when not required, but then catching the action when FF is unresponsive is clearly not very practical.

Thanks,

BWD
(Reporter)

Comment 49

2 years ago
Status: Resolved?  Not having had a reply to my last I though you had just given up ...
the problem has appeared to be less, but it has not gone away.

Anyway, using Process Explorer, I have found an offending thread:  firefox.exe|GetHandleVerifier
This has been taking a full core's cycles for many minutes at a time lately - locking FF completely.
I cannot see that I am doing anything in particular to trigger this, it appears to be a spontaneous matter.  In fact, it is running continuously anyway, unevenly up to 20%, typically around 5%
This is at one address (0x442c) while another instance (0x24d8) appears to be completely quiescent.  (There is only one window.)
Does this focus attention helpfully?  It certainly seems to be the major process running.

What I miss is a function that used to be present - suspending a tab.  If this were feasible, it would be easier to identify an offending site or function.  Unloading everything sequentially is just too time-consuming and interferes with work too much.  Is that option still present?
(There appears to be no add-on for this simple function.  I have no need to save memory.)
The only way I can see to achieve it is to click for reload and then immediately stop the process.  Not very handy.  Anyway, the few active tabs that I have tried this on have had no effect whatsoever on the offending thread.

Thanks,

BWD
I think a profile from the Gecko Profiler add-on is the best way to proceed here to determine what's happening.

(In reply to Dr B W Darvell from comment #48)
> Please advise:
> 1.  Am I doing the right thing?

Presumably, yes. The profiler needs to be running during the bad performance behaviour, and then you need to dump the profile. The profile will be visible in a tab called "Cleopatra" (or perf.html for more modern versions of the add-on). Then you can share the profile with us.

> 2.  Where do I find the file that is to be uploaded.

The file is in memory, but can be saved to disk from the Cleopatra / perf.html interface.

The profile is uploaded to storage backed by Google App Engine. The source code that does this sharing for perf.html is here:

https://github.com/devtools-html/perf.html/blob/c5b71f0869f95475f655d4bc57342f14f5aecd03/src/content/profile-store.js

> 3.  Does this contain personal information of any kind that compromises my
> security in any way?

The type of data that the profile will contain:

1) JS stacks which might include stack frames from inside your add-ons, so your add-on collection can be inferred
2) JS stacks for sites that are running JS - so the tabs you currently have open can possibly be inferred.
3) Version information on your OS, which version of Firefox you're running

> 4.  The profiler is said to collect a few seconds-worth of data - how many
> is a few?  If I miss the behaviour we are trying to understand it will be
> pointless, of course.  How long have I got?

It has less to do with time, and more to do with space. There is a circular buffer that is recording the profile. I'd say in general though that a good rule of thumb is to assume that about 5 - 10 seconds of data will be collected.

> 5.  I take it that the profiler is running continuously and should be
> disabled when not required, but then catching the action when FF is
> unresponsive is clearly not very practical.

Leaving the profiler running should be low cost. I suggest having it running constantly until you experience the behaviour, and then dump the profile as soon as possible.
(Reporter)

Comment 51

2 years ago
Hi,
  Thanks for the reply.
  
OK, so it can be left running, which is fine.  But if FF has locked up for many minutes at a time (as it has now started to do in the last few days - and all due to that single process, it appears) how do I get in to dump the profile?  I have to catch it within 5 or 10 s of it becoming responsive?  That might be tricky.

Ah - "Save to Local File" - correct? In a test just now, 13 Mb, no extension.  Sound right?

One last query: upload where?  Is that at "Add an attachment (proposed patch, testcase, etc.)" on the top of this page?

Thanks,

BWD
(In reply to Dr B W Darvell from comment #51)
> Hi,
>   Thanks for the reply.
>   
> OK, so it can be left running, which is fine.  But if FF has locked up for
> many minutes at a time (as it has now started to do in the last few days -
> and all due to that single process, it appears) how do I get in to dump the
> profile?  I have to catch it within 5 or 10 s of it becoming responsive? 

Unfortunately, yes.

> That might be tricky.

Unfortunately, yes. :(

It might be simpler to attempt to bisect your add-ons to determine which (if any) are causing this problem. Disable half, and see if the problem goes away. If not, disable half of the remaining, etc, until you determine which add-on(s) are contributing to the issue, and then we can examine it in isolation.

> 
> Ah - "Save to Local File" - correct? In a test just now, 13 Mb, no
> extension.  Sound right?
> 

Yes, that sounds right.

> One last query: upload where?  Is that at "Add an attachment (proposed
> patch, testcase, etc.)" on the top of this page?

You can upload it to this bug if you'd like. Cleopatra / perf.html has a built-in to compress and upload the profile to that Google App Engine server, at which time it'll give you a URL you can post in this bug as an alternative.
(Reporter)

Comment 53

2 years ago
Upload: OK - the "Share" button.  
(Not really feasible to check for 'personal' content in such a file!)

Add-ons: I tried disabling but that had no clear outcome, certainly no finger of blame.  The problem is that the effect is intermittent, and unpredictable. I could wait ages, and absence of evidence is not evidence of absence.

I will try to catch an unreponsive period - I have a process window open from Process Explorer (Performance Graph) alongside while I work so that I can see what is happening a little more easily.

If I can capture this well enough, I might then try disabling some add-ons, although this could perhaps only narrow the field somewhat, as perhaps you suggest.

Anyway, would you expect the GetHandleVerifier to be running such as to block all other activity?  The priority must be very high.
(Reporter)

Comment 54

2 years ago
Well, it took a few days, but I fianlly caught an event - seized up for over 5 minutes.
I think I have captured the necessary material:
https://cleopatra.io/#report=8d3dd9ed2a64c1c2d889652a27fd99b7f474df94
I hope this helps, but I can confirm that it was GetHandleVerifier that was taking all available cycles.
Please advise whether the file is usable.

Thanks,

BWD
(Reporter)

Comment 55

2 years ago
Done it again - 20 min later, for about 6 minutes.
https://cleopatra.io/#report=e385e632e8186bfd053d5cf72271400e670ed23b

BWD
Thanks BWD,

Looking at both of these profiles, I see indications of large amounts of cycle collection and garbage collection, which suggests to me that one or more of your add-ons are leaking.

A memory report from about:memory would be the next step, in order to try to figure out where the memory is going.
(Reporter)

Comment 57

2 years ago
Created attachment 8842085 [details]
memory-report.json.gz
(Reporter)

Comment 58

2 years ago
OK, thanks.  New file uploaded.
However, would it be most relevant when a freeze event occurs (which would have to be just after)?
If there are options I should use for this, please advise.
(Reporter)

Comment 59

2 years ago
BTW: what does GetHandleVerifier do and why is it running so much in the one instance - and the second apparently always quiet. (see comment 49)

Incidentally, GC and CC in about:memory have no visible effect.
Flags: needinfo?(mconley)
(Reporter)

Comment 60

2 years ago
Sorry - I do not understand: you need info from me?
No, sorry. I set the needinfo flag on myself to come back around to this and examine your about:memory report.
(Reporter)

Comment 62

2 years ago
Oh, OK.
I really think the best thing we can try at this point is selectively disabling add-ons. After looking at your memory report the most likely culprits are Adblock Plus (which you might try upgrading to the latest version) and Blur, possibly NoScript. Anything related to tabs may be an issue (IE Tab, Auto reload tab, etc) as well.

Generally we suggest a divide and conquer methodology:
  - Disable half your add-ons, if that fixes things re-enable half, repeat
  - If that doesn't fix things, disable half of the remaining add-ons, repeat

Also it looks like you have about 115 tabs open, does that sound right? 115 tabs with 23 add-ons is very severe load, this could definitely lead to performance issues.

Can you let us know which version of Firefox you're on at this point? Firefox 52 has quite a few memory leak fixes and other enhancements you might want to try out (it's currently Beta, will be promoted to release/ESR in a week). It also has e10s enabled which might help with overall performance.

Interesting snippets:

> ├──1,418.58 MB (37.92%) -- window-objects
> │  ├────918.09 MB (24.54%) ++ (111 tiny)
> │  ├────192.00 MB (05.13%) ++ top(<anonymized-62>, id=62)
> │  ├────137.38 MB (03.67%) ++ top(<anonymized-60>, id=60)
> │  ├─────66.48 MB (01.78%) ++ top(<anonymized-4948>, id=4948)
> │  ├─────65.60 MB (01.75%) -- top(none)
> │  │     ├──63.64 MB (01.70%) ++ ghost
> │  │     └───1.97 MB (00.05%) ++ detached
> │  └─────39.03 MB (01.04%) ++ top(<anonymized-14>, id=14)

A lot of active windows, a few ghosts (often add-on related). This is the majority of the memory.

> ├────666.40 MB (17.81%) -- js-non-window
> │    ├──543.00 MB (14.52%) ++ zones
> │    ├──106.06 MB (02.84%) ++ runtime
> │    └───17.34 MB (00.46%) ++ gc-heap

A lot dedicated to non-window zones. Much of this is add-on related. Recent versions have reduced overhead of Firefox's system zones, so a new version might help.

> ├────618.51 MB (16.53%) -- layout
> │    ├──617.88 MB (16.52%) ── rule-processor-cache
> │    └────0.63 MB (00.02%) ++ (2 tiny)

This looks like a bug, but could be poor interaction with ABP. For example my browser is using 0.62 MB with ~70 tabs. I'll file a follow up for this.

> 5 (100.0%) -- ghost-windows
> ├──1 (20.00%) ── <anonymized-1628>
> ├──1 (20.00%) ── <anonymized-1639>
> ├──1 (20.00%) ── <anonymized-2794>
> ├──1 (20.00%) ── <anonymized-2938>
> └──1 (20.00%) ── <anonymized-4642>

5 ghost windows, this indicates add-on issues.

> 1,465 (100.0%) -- js-main-runtime-compartments
> ├────772 (52.70%) ++ user
> └────693 (47.30%) ++ system

A *ton* of zones, this can cause large garbage collection and cycle collection pauses. I believe we've made improvements in recent versio
ns.
(In reply to Eric Rahm [:erahm] from comment #63)
> > 1,465 (100.0%) -- js-main-runtime-compartments
> > ├────772 (52.70%) ++ user
> > └────693 (47.30%) ++ system
> 
> A *ton* of zones, this can cause large garbage collection and cycle
> collection pauses. I believe we've made improvements in recent versio
> ns.

Bug 1324176 is one example of a fix in this area which landed in 53.
(Reporter)

Comment 65

2 years ago
Thanks for all that.  
The version is 51.0.1 (64-bit) at the moment - and the next does sound better in many respects.  I'll be patient.

I do end up with lots of tabs open because of the nature of the work I do - much scientific literature searching, linked with lots of Google searches.  I have many threads to pursue, interlinked in many ways, but also several projects on at once - which does not help.  It is simply not possible to avoid this situation arising because I have to cross-check so much.

However, I will try the selective disabling, as you suggest, and see whether that makes any difference.  Again, though, I might have to wait for some time to see what is going on.

AdBlock plus is 2.8.2 and no update is found on searching.  

I'll start with Blur at least, I am not convinced about that, and 10 others that I do not rely on much.  

The best thing to hope for is the problem recurs with what's left - that at least would be positive.

Thanks again.
(In reply to Eric Rahm [:erahm] from comment #63)
> > ├────618.51 MB (16.53%) -- layout
> > │    ├──617.88 MB (16.52%) ── rule-processor-cache
> > │    └────0.63 MB (00.02%) ++ (2 tiny)
> 
> This looks like a bug, but could be poor interaction with ABP. For example
> my browser is using 0.62 MB with ~70 tabs. I'll file a follow up for this.

I filed bug 1343387 for this.
(Reporter)

Comment 67

2 years ago
Created attachment 8842990 [details]
memory-report.json.gz

After disabling 11 extensions.
(Reporter)

Comment 68

2 years ago
I wonder whether you can see any significant changes in the new memory report.  The memory usage (according to Process Explorer) has been pretty stable around 1.8 - 2.2 GB, and I have seen no freeze since (maybe not enough time ...).  Meanwhile, the CPU usage appears to be much lower on average, althoug GetHandleVBerifier is still the busiest thread.

Thanks.
(In reply to Dr B W Darvell from comment #68)
> I wonder whether you can see any significant changes in the new memory
> report.  The memory usage (according to Process Explorer) has been pretty
> stable around 1.8 - 2.2 GB, and I have seen no freeze since (maybe not
> enough time ...).  Meanwhile, the CPU usage appears to be much lower on
> average, althoug GetHandleVBerifier is still the busiest thread.
> 
> Thanks.

The memory usage definitely looks better (note: you have fewer tabs open, so that helps too) and if you're not seeing freezes that's great!

I still see a high 'rule-processor-cache' entry as well as ghost windows so it looks like you still have a misbehaving add-on enabled. Can you provide the add-on list from about:support and try disabling a few more?
(Reporter)

Comment 70

2 years ago
I'll try diabling later, buit this is the current list:

Adblock Plus	2.8.2	true	{d10d0bf8-f5b5-c8b4-a8b2-2b9879e08c5d}
Application Update Service Helper	1.0	true	aushelper@mozilla.org
Block site	1.1.8.1-signed.1-signed	true	{dd3d7613-0246-469d-bc65-2a3cc1668adc}
Capture & Print	0.1.9.3.1-signed.1-signed	true	{146f1820-2b0d-49ef-acbf-d85a6986e10c}
Diagnostics	1.0	true	diagnostics@mozilla.org
Disconnect	3.15.3.1-signed.1-signed	true	2.0@disconnect.me
Download Virus Checker	0.1.1	true	{4785b360-ef55-4868-a385-c4ca829ba576}
Flashblock	1.5.20	true	{3d7eb24f-2740-49df-8937-200b1cc08f8a}
geckoprofiler	1.16.25	true	jid0-edalmuivkozlouyij0lpdx548bc@jetpack
HTTP/2 and SPDY indicator	2.3	true	spdyindicator@chengsun.github.com
IE Tab 2 (FF 3.6+)	6.2.18.1	true	{1BC9BA34-1EED-42ca-A505-6D2F1A935BBB}
Multi-process staged rollout	1.7	true	e10srollout@mozilla.org
NoScript	2.9.5.3	true	{73a6fe31-595d-460b-a920-fcc0f8843232}
Padlock	0.5.0.1-signed.1-signed	true	{d09e32df-8610-4b33-b929-1e631b764130}
Pocket	1.0.5	true	firefox@getpocket.com
Print Edit	17.9	true	printedit@DW-dev
Send HSTS Priming Requests	1.0	true	hsts-priming@mozilla.org
SHA-1 deprecation staged rollout	1.3	true	disableSHA1rollout@mozilla.org
Tree Style Tab	0.18.2016111701	true	treestyletab@piro.sakura.ne.jp
Tree Style Tabs Toplevel	1.0	true	jid1-h4lBQP9pp2k89i@jetpack
Trustwave SecureBrowsing	3.722.1-signed.1-signed	true	securebrowsing@m86security.com
Vacuum Places Improved	1.2.1-signed.1-signed	true	VacuumPlacesImproved@lultimouomo-gmail.com
Web Compat	1.0	true	webcompat@mozilla.org
Xmarks	4.4.1	true	foxmarks@kei.com
(Reporter)

Comment 71

2 years ago
Small update:  I have disabled 5 more add-ons, and on restart there are zero ghosts:

       45.13 MB ── d3d11-shared-textures
        0.00 MB ── gfx-d2d-vram-draw-target
        0.00 MB ── gfx-d2d-vram-source-surface
        0.02 MB ── gfx-surface-win32
        0.00 MB ── gfx-textures
        0.00 MB ── gfx-textures-peak
        0.00 MB ── gfx-tiles-waste
              0 ── ghost-windows
      250.05 MB ── gpu-committed
      320.13 MB ── gpu-dedicated
       48.65 MB ── gpu-shared
    1,065.12 MB ── heap-allocated
        1.00 MB ── heap-chunksize
    1,170.00 MB ── heap-mapped
              1 ── host-object-urls
        1.45 MB ── imagelib-surface-cache-estimated-locked
        1.75 MB ── imagelib-surface-cache-estimated-total
              0 ── imagelib-surface-cache-overflow-count
        9.21 MB ── js-main-runtime-temporary-peak
    1,601.10 MB ── private
    1,621.50 MB ── resident
    1,598.04 MB ── resident-unique
       19.77 MB ── system-heap-allocated
    2,254.73 MB ── vsize
4,206,544.69 MB ── vsize-max-contiguous
             18 ── webgl-buffer-count
       15.75 MB ── webgl-buffer-memory
              1 ── webgl-context-count
              1 ── webgl-renderbuffer-count
        8.62 MB ── webgl-renderbuffer-memory
             20 ── webgl-shader-count
             16 ── webgl-texture-count
       26.08 MB ── webgl-texture-memory

I will see whether any return when I get back to proper work.

This item is still large (although smaller):

├────288.69 MB (19.84%) -- layout
│    ├──288.16 MB (19.81%) ── rule-processor-cache
│    └────0.53 MB (00.04%) -- (2 tiny)
│         ├──0.49 MB (00.03%) ── style-sheet-cache
│         └──0.04 MB (00.00%) ── style-sheet-service


Are there other such specific pointers I can look at?

Thanks.
(Reporter)

Comment 72

2 years ago
I have just found this:

1 (100.0%) -- ghost-windows
└──1 (100.0%) ── http://www.sigmaaldrich.com/catalog/product/sigma/32922?lang=en&region=GB

This is unaffected by GC, CC or Memory minimization.

So I went to the website and followed a number of random links, then closed the tab tree: no ghosts (i.e that previous one has disappeared).
The exact same URL, opened and closed, also leaves no ghost.

Curious.

Also I have now:

├────459.00 MB (21.89%) -- layout
│    ├──458.42 MB (21.86%) ── rule-processor-cache
│    └────0.58 MB (00.03%) -- (2 tiny)
│         ├──0.53 MB (00.03%) ── style-sheet-cache
│         └──0.04 MB (00.00%) ── style-sheet-service

939 (100.0%) -- js-main-runtime-compartments
├──550 (58.57%) -- system
│  ├──484 (51.54%) ++ (470 tiny)
│  └───66 (07.03%) ── [System Principal], inProcessTabChildGlobal?ownedBy=chrome://browser/content/browser.xul [66]
└──389 (41.43%) ++ user

About 60 tabs.
(Reporter)

Comment 73

2 years ago
Now isn't this peculiar -- these have emerged after the check above showed nothing:

20 (100.0%) -- ghost-windows
├───2 (10.00%) ── http://www.sigmaaldrich.com/catalog/product/sigma/32922?lang=en&region=GB [2]
├───2 (10.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=12873256 [2]
├───2 (10.00%) ── http://www.sigmaaldrich.com/labware/learning-center.html [2]
├───2 (10.00%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├───1 (05.00%) ── http://www.sigmaaldrich.com/analytical-chromatography/labware-and-equipment.html
├───1 (05.00%) ── http://www.sigmaaldrich.com/analytical-chromatography/titration/karl-fischer-titration.html
├───1 (05.00%) ── http://www.sigmaaldrich.com/catalog/product/aldrich/z555525?lang=en&region=GB
├───1 (05.00%) ── http://www.sigmaaldrich.com/customer-service/quality-systems.html
├───1 (05.00%) ── http://www.sigmaaldrich.com/customer-service/quality-systems/iso-certification.html
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=16683011
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=17182794
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=17187090
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=9576922
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=9582543
├───1 (05.00%) ── http://www.sigmaaldrich.com/labware/learning-center/electrode-selection.html
└───1 (05.00%) ── http://www.sigmaaldrich.com/site-level/email-tech-service-us.html?ProductNo=z555525&Brand=ALDRICH

I shall do nothing deliberate that might affect these to see if they persist.
Comment hidden (offtopic)

Comment 75

2 years ago
Created attachment 8843591 [details]
memory-report.json-2017-03-04-10_11AM.gz

about:memory report for comment 74.
Steve S: Please file a new bug. It is very difficult to track what is going on with multiple people in a single bug. You clone the bug if you want, and put this bug in the see also field.
> Now isn't this peculiar -- these have emerged after the check above showed nothing:

When you close a window (by closing a tab with a webpage open , for instance), it becomes top(none)/detached. Usually, it will go away soon after a few GC/CCs happen. If it stays in that state for a minute, then it becomes a ghost window. Generally speaking, if a window remains top(none)/detached after you do minimize memory usage, it is going to become a ghost window, so you don't really need to wait the full minute.

It sounds like you have a way to reproduce a ghost window now. It would be good to figure out which addon is causing that (if any). I'd try disabling AdblockPlus and NoScript first. I fixed bug 1336811 in Firefox 52, which could be the same issue you are seeing. So you could try installing Firefox Beta, which will contain this fix, or wait until next week, when Firefox will be updated to version 52.
Status: RESOLVED → REOPENED
Ever confirmed: true
Resolution: INCOMPLETE → ---
(Reporter)

Comment 78

2 years ago
OK, noted.
I'll leave it for an hour or so, to see if any more emerge (I have just found that http://www.telegraph.co.uk/ does the same.)  I'll then do the disabling systematically.

Comment 79

2 years ago
(In reply to Andrew McCreight [:mccr8] from comment #76)
> Steve S: Please file a new bug. It is very difficult to track what is going
> on with multiple people in a single bug. You clone the bug if you want, and
> put this bug in the see also field.

OK. It just seemed like the same issue, so I thought having my info might reveal some common elements that might help track down the cause. I'll probably just wait and see if this bug gets resolved.
(In reply to Steve S from comment #79)
> 
> OK. It just seemed like the same issue, so I thought having my info might
> reveal some common elements that might help track down the cause. I'll
> probably just wait and see if this bug gets resolved.

It is possible, and we appreciate the fact that you noticed that the symptoms are similar.

Having said that, it's easier to effectively merge two bug reports (by marking one as a duplicate of another) than it is to split one bug report, so we prefer to err on the side of filing bugs separately.
(Reporter)

Comment 81

2 years ago
Overnight (no changes made), the several ghosts for http://www.sigmaaldrich.com remained, as did those for http://www.telegraph.co.uk/.
Opening a new tab and going to http://www.sigmaaldrich.com causes all the related ghosts to disappear.
Likewise, opening a new tab for http://www.telegraph.co.uk/ causes those ghosts to go, immediately - zero count reported
So I close both tabs.

After about 2 minutes, ALL the previous ghosts reappear - with multiples ...

39 (100.0%) -- ghost-windows
├───6 (15.38%) ── http://www.sigmaaldrich.com/united-kingdom.html [6]
├───3 (07.69%) ── http://www.telegraph.co.uk/ [3]
├───2 (05.13%) ── http://www.sigmaaldrich.com/catalog/product/sigma/32922?lang=en&region=GB [2]
├───2 (05.13%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=12873256 [2]
├───2 (05.13%) ── http://www.sigmaaldrich.com/labware/learning-center.html [2]
├───2 (05.13%) ── http://www.telegraph.co.uk/opinion/letters/ [2]
├───1 (02.56%) ── http://www.sigmaaldrich.com/analytical-chromatography/labware-and-equipment.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/analytical-chromatography/titration/karl-fischer-titration.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/catalog/product/aldrich/z555525?lang=en&region=GB
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/quality-systems.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/quality-systems/iso-certification.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/services/basic-research.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/services/facility-operations.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/services/mfg-production.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/customer-service/services/regulatory-compliance.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=16683011
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=17182794
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=17187090
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=9576922
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/labware-products.html?TablePage=9582543
├───1 (02.56%) ── http://www.sigmaaldrich.com/labware/learning-center/electrode-selection.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/safc/bioprocess.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/safc/testing-and-support-services.html
├───1 (02.56%) ── http://www.sigmaaldrich.com/site-level/email-tech-service-us.html?ProductNo=z555525&Brand=ALDRICH
├───1 (02.56%) ── http://www.telegraph.co.uk/football/2017/03/04/manchester-united-vs-bournemouth-premier-league-watch-live-score/
├───1 (02.56%) ── http://www.telegraph.co.uk/news/uk/
├───1 (02.56%) ── http://www.telegraph.co.uk/opinion/2017/03/01/telegraph-cartoons-march-2017/
└───1 (02.56%) ── http://www.telegraph.co.uk/opinion/cartoons/

Clearly, despite the memory report, the allocations persist, and are identifiable.   So the memory report is failing somewhere to see what has been labelled.  Is that another bug?

I will now repeat the exercise with AdBlock disabled.
(Reporter)

Comment 82

2 years ago
Restarted, and a random selection of links from the two root URLs, close all, two minutes later:

42 (100.0%) -- ghost-windows
├───8 (19.05%) ── http://www.sigmaaldrich.com/united-kingdom.html# [8]
├───5 (11.90%) ── http://gateway.answerscloud.com/sigmaaldrich/production/trigger/frameWorker.html?v=w0571rt [5]
├───3 (07.14%) ── http://www.telegraph.co.uk/news/#source=refresh [3]
├───2 (04.76%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/#source=refresh [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/business/ [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/vendor/lightslider/css/lightslider.min.css [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/vendor/lightslider/js/lightslider.min.js [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/culture/#source=refresh [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/lifestyle/ [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/money/consumer-affairs/new-1-coin-should-hoard-existing-1-coins/ [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/opinion/#source=refresh [2]
├───2 (04.76%) ── http://www.telegraph.co.uk/sport/ [2]
├───1 (02.38%) ── http://shares.telegraph.co.uk/iframe/indiceRisersAndFallers2.php
├───1 (02.38%) ── http://shares.telegraph.co.uk/iframe/indices2.php
├───1 (02.38%) ── http://www.telegraph.co.uk/culture/
├───1 (02.38%) ── http://www.telegraph.co.uk/opinion/
├───1 (02.38%) ── http://www.telegraph.co.uk/travel/
└───1 (02.38%) ── https://cf-particle-html.eip.telegraph.co.uk/852113fc-8752-4c39-9efb-4c23191fadfe.html?ref=http://www.telegraph.co.uk/money/consumer-affairs/new-1-coin-should-hoard-existing-1-coins/&title=New%20%C2%A31%20coin:%20but%20should%20I%20hoard%20my%20existing%20%C2%A31%20coins?

(BTW: I did not open 42 windows - only about 12 or 15.)

I might as well try with all extensions disabled ...
(Reporter)

Comment 83

2 years ago
... no ghosts.  After 10 minutes, just to make sure.  Repeat the tab openings and closings, still no ghosts.
New memory report will be uploaded after this.
(Reporter)

Comment 84

2 years ago
Created attachment 8843688 [details]
memory-report.json.gz
(Reporter)

Comment 85

2 years ago
Normal start.

Open tab for top level domain only, and about 1.5 min after closing get this:

6 (100.0%) -- ghost-windows
├──4 (66.67%) ── http://www.sigmaaldrich.com/united-kingdom.html [4]
├──1 (16.67%) ── http://gateway.answerscloud.com/sigmaaldrich/production/trigger/frameWorker.html?v=w0571rt
└──1 (16.67%) ── http://www.sigmaaldrich.com/

A minute later is this:

4 (100.0%) -- ghost-windows
└──4 (100.0%) ── http://www.sigmaaldrich.com/united-kingdom.html [4]


Repeat with the telegraph site:

6 (100.0%) -- ghost-windows
├──4 (66.67%) ── http://www.sigmaaldrich.com/united-kingdom.html [4]
└──2 (33.33%) ── http://www.telegraph.co.uk/ [2]

no other changes with time.
(Reporter)

Comment 86

2 years ago
With NoScript disabled, for the two test URLs, after about 1 min of closing, there is this:

5 (100.0%) -- ghost-windows
├──2 (40.00%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├──2 (40.00%) ── http://www.telegraph.co.uk/ [2]
└──1 (20.00%) ── http://gateway.answerscloud.com/sigmaaldrich/production/trigger/frameWorker.html?v=w0571rt

A minute later and there are none.  This seems to be reproducible.
At one point I had both an apparently active window and a ghost for the same item, presumably just a timing accident.

Thoughts?
(Reporter)

Comment 87

2 years ago
I re-enabled most extensions, except for Blur and Noscript.
After opening and closing the two test sites, after a minute I get only one:

1 (100.0%) -- ghost-windows
└──1 (100.0%) ── http://www.telegraph.co.uk/

So, try them again (the ghost disappears immediately on opening the URL).  This time I get:

5 (100.0%) -- ghost-windows
├──3 (60.00%) ── http://www.telegraph.co.uk/ [3]
└──2 (40.00%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]

I would surmise that NoScript is not to blame.

Now for AdBlock again ...
(Reporter)

Comment 88

2 years ago
Well, well.  Adblock disabled, NoScript running:

7 (100.0%) -- ghost-windows
├──2 (28.57%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├──1 (14.29%) ── http://www.telegraph.co.uk/
├──1 (14.29%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/custom/non-iab/v2/main.2.5.js
├──1 (14.29%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/custom/non-iab/v2/main.2.6.css
├──1 (14.29%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/vendor/lightslider/css/lightslider.min.css
└──1 (14.29%) ── http://www.telegraph.co.uk/content/dam/spark/component-assets/vendor/lightslider/js/lightslider.min.js

but only on the second attempt.  It is looking like it is an intermittent fault.  Decidely erratic.
(Reporter)

Comment 89

2 years ago
I found a couple more sites to produce the effect 

3 (100.0%) -- ghost-windows
├──1 (33.33%) ── http://www.fco.gov.uk/en/about-the-fco/embassies-and-posts/find-an-embassy-overseas/middle-east-and-north-africa/kuwait
├──1 (33.33%) ── http://www.fco.gov.uk/en/travelling-and-living-overseas/travel-advice-by-country/middle-east-north-africa/kuwait
└──1 (33.33%) ── https://www.wunderground.com/global/stations/40582.html

but after 2 minutes, these had gone.
Retrying all four sites, after a minute or so:

10 (100.0%) -- ghost-windows
├───2 (20.00%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├───2 (20.00%) ── http://www.telegraph.co.uk/ [2]
├───2 (20.00%) ── https://www.gov.uk/government/organisations/foreign-commonwealth-office [2]
├───2 (20.00%) ── https://www.wunderground.com/ [2]
├───1 (10.00%) ── http://gateway.answerscloud.com/sigmaaldrich/production/trigger/frameWorker.html?v=w0571rt
└───1 (10.00%) ── https://www.wunderground.com/MAR/

A couple of minutes later:

3 (100.0%) -- ghost-windows
├──1 (33.33%) ── https://www.gov.uk/government/organisations/foreign-commonwealth-office
├──1 (33.33%) ── https://www.wunderground.com/
└──1 (33.33%) ── https://www.wunderground.com/MAR/

later again:

2 (100.0%) -- ghost-windows
├──1 (50.00%) ── https://www.gov.uk/government/organisations/foreign-commonwealth-office
└──1 (50.00%) ── https://www.wunderground.com/

and again, a minute or so:

1 (100.0%) -- ghost-windows
└──1 (100.0%) ── https://www.wunderground.com/

... and a minute or so later, all gone.
It seems to have reached the point where I cannot sensibly diagnose what is going on because it would take many minutes every time I check an extension to be sure that it is not a passing state.  Is there any way to track this automatically?
If I am not reporting accurately, it clearly can be of no help to you.  I now do not know how persistent all the previous ghosts have been as I have not waited ages each time.

Advice?
(Reporter)

Comment 90

2 years ago
At 80 minutes after closing the tab:

1 (100.0%) -- ghost-windows
└──1 (100.0%) ── http://www.telegraph.co.uk/

I think that counts as persistent.
(Reporter)

Comment 91

2 years ago
12 h later, still there.
Quite a bit of data to unpack here. I'm going to try to summarize:

BWD: am I boiling this down correctly when I say that you're seeing permanent ghost windows on these sites (telegraph.co.uk for example) with NoScript enabled?

If so, yeah, sounds like we've got a leak here. I wonder if this is related at all to bug 1326095, which should be fixed in tomorrow's release.
Flags: needinfo?(mconley)
I apologize - bug 1326095 is _not_ fixed in tomorrow's release. Bug 1336811, which was similar, is.
(Reporter)

Comment 94

2 years ago
re: 92: Yes, apologies - I was trying v hard to nail it down, and failed.

permanent:  With and without NoScript (and, I thought, with and without AdBlock, but now doubts as I may not have checked for long enough).  Will revisit these possibilities later.

I have found another permanent (well, several hours, anyway) site in Dilbert:

3 (100.0%) -- ghost-windows
├──1 (33.33%) ── http://dilbert.com/
├──1 (33.33%) ── http://www.sigmaaldrich.com/
└──1 (33.33%) ── http://www.telegraph.co.uk/

There must be something odd with these sites to cause this.  I am trying to find more by checking lots of old bookmarks in batches.  I have been through a hundred or so, so far.

re: next release: I'll not have much chance to do anything for a few days, but will return to this when I can.

Thanks.
(Reporter)

Comment 95

2 years ago
Another site (out of a couple of hundred tested) that produces a persistent ghost:

5 (100.0%) -- ghost-windows
├──1 (20.00%) ── http://dilbert.com/
├──1 (20.00%) ── http://www.sigmaaldrich.com/
├──1 (20.00%) ── http://www.telegraph.co.uk/
├──1 (20.00%) ── https://www.omni-inc.com/
└──1 (20.00%) ── https://www.omni-inc.com/dounce-tissue-grinder-1ml-11x48mm.html#

What could it be about them that is special that causes this?

Anyway, shutting down shortly.
I'm having trouble investigating dilbert.com because I hit an assertion. I'll check the other sites, though. Also be sure to check for updates to Firefox, as that should get you version 52 which has one or two fixes which might help.
Depends on: 1342877
Depends on: 1345260
(Reporter)

Comment 97

2 years ago
I have just updated to FF52, loaded my test set of tabs, waited for completion, and then closed them all before running the memory check, when I get:

+++
WARNING: the following values are negative or unreasonably large.

    explicit/js-non-window/gc-heap/unused-arenas
    js-main-runtime/gc-heap/(2 tiny)/unused-arenas
    js-main-runtime-gc-heap-committed/unused/arenas 

This indicates a defect in one or more memory reporters. The invalid values are highlighted. 
+++
The negative values were all -0.98 MB


FF was then very unresponsive with a long (2 ~ 3 minutes or so) write to disk of over 200 MB, although CPU was only about 3%.

Rerunning the memory check was then uneventful - no warnig, no disk activity of note.  It was also much faster than on the previous version.

I will do a rerun and see what happens.
(Reporter)

Comment 98

2 years ago
Created attachment 8846332 [details]
memory-report.json.gz
(Reporter)

Comment 99

2 years ago
Created attachment 8846333 [details]
memory-report.json.gz
Attachment #8846332 - Attachment is obsolete: true
(Reporter)

Comment 100

2 years ago
Rerun ... 

2 (100.0%) -- ghost-windows
├──1 (50.00%) ── http://dilbert.com/
└──1 (50.00%) ── https://blog.scopus.com/

but these then disappeared later.
(I forgot to say, there were no ghosts in the previous run.)

And again:

5 (100.0%) -- ghost-windows
├──2 (40.00%) ── http://www.sigmaaldrich.com/united-kingdom.html [2]
├──1 (20.00%) ── http://dilbert.com/
├──1 (20.00%) ── http://www.hsc.edu.kw/jQueryNewsTicker/default.aspx
└──1 (20.00%) ── http://www.telegraph.co.uk/#source=refresh

after a couple of minutes, but these too disappeared.
It may be that the problem has been fixed.
I'll keep an eye on it, and especially the CPU usage.  GetHandleVerifier is still dominant by a very long way.

I have upload the current memory profile in case there is more to be seen.
(Reporter)

Comment 101

2 years ago
I have tested a couple of hundred more sites, and none has left a persistent ghost.  I think that particular aspect has now been sorted out.

It still strikes me that there is a lot of CPU activity when nothing is happening.  But I notice that the memory usage is not about 2.9 GB after minimizing memory, which only removed about 0.1 GB, when the start condition was about 2.1 - 2.2 GB.  Still leaks?

Comment 102

2 years ago
A few clues --- 
I've had persistent performance issues in FFox as well, and found this thread by searching for "GetHandleVerifier", which I found was completely freezing FFox for many minutes at a time, and the only real solution is to restart FFox.

I've had the identical problem on multiple PCs, in Win 7, 8, 8.1 10, 10-Ann.  My PCs have been single processors with low RAM, and today, I'm running 16 GB with the fastest true Intel quad with SSD you can get. The problem still persists, regardless of hardware or OS.

TABS and SESSION managers -- I had very few plugins in common with the OP.  But, I have 60-100 tabs open as a rule, use a TREE type tab manager, AND a session manager, as well as ~25 other plugins.  My CPU peaks at ~13% when freezing, which is probably typical for a single threaded process on a quad processor, and I'm only using about 3.5 GB of RAM.

I use "Tab Tree" by Sergey Zelentsov instead though, ( https://addons.mozilla.org/en-US/firefox/addon/tab-tree/ ), instead of "Tree STYLE Tabs".  However, I believe there's another PC I was using extensively, and it may have started showing the same problems, about the time I started using "trees", and I think that one was "Tree STYLE Tabs" plugin.  So, seemingly common problem. (I haven't checked which process is hung on that PC).

In "Tab Tree" plugin config page it states: "This extension was conceived as an alternative to Tree Style Tab to provide better performance, stability and usability."  This is the first clue.  Next, he also says: "Incompatible with TMP's session manager."  Hmmm.  Heard that before.  But, it wasn't isolated to just TMP's session manager, it was any session manager.     

Searching on TMP session manager, I found this quote:  "I do use TMP and used to use it on Windows before FF had its own session manager, but found the TMP session restore to cause my machine to perform poorly, a view I've seen echoed in other comments recently--but granted I haven't tried it myself with the latest FF and on this system. Do you have any experience with TMP's session manager--do you know whether it's stable and whether it slows the machine down much? – Philip Jan 3 '10 at 22:46"  https://superuser.com/questions/90851/how-can-i-make-firefox-remember-my-session-while-still-clearing-the-browsing-his

TMP's home page shows in the first few reviews, that it has performance problems:  https://superuser.com/questions/90851/how-can-i-make-firefox-remember-my-session-while-still-clearing-the-browsing-his

Here's another one that was SOLVED by disabling ONE of these:  session manager - tab manager.  https://support.mozilla.org/t5/Firefox/opens-with-tabs-from-previous-session/m-p/631957

I use "Session Manager" by Michael Kraft, ( http://sessionmanager.mozdev.org/documentation.html ), which I'm going to guess uses resources that conflict with ANY tab manager.  Although It has it's own problems, it is essential to save and retrieve sessions that contain a ton of research about a specific subject.  (Session Manager saves a separate file for the complete session, so I sometimes have GB's of files, which makes it nearly impossible to load that list!)

Trees are essential for another reason--I need a LIST that I can see all my tabs, which is impossible / impractical by adding a bunch of rows, without seriously reducing my vertical available browser window.  (Plus, good luck scanning a list of horizontal tabs, even if visible). -- There's a reason outlines are NEVER presented as a comma-delimited list--think about how dumb that would be.  Yet, we think a ton of tabs should be horizontal, with 16 x 9 screens?? c'mon Devs--have you thought this through??)

I really hope someone gets to the bottom of this.  IMO, this is a common resource conflict issue, that can't be solved by the add-on DEVS.  OTOH, we NEED these add-ons, or we wouldn't ever put up with the real pain of trying to coax FFox to perform even 50% of the time without crashing, or having us force a crash.
(Reporter)

Comment 103

2 years ago
Interesting.
For the record, I have never used an add-on session manager, and to be honest I have no idea what it is in FF either (sorry, guys!).  A quick check: it has never been activated, deliberately at least.

I use the horizontal tabs for very much the same reasons as awelsh, it seems: across the top is quite unusable, and tree structure essential - how I struggled before I found Tree Style Tabs!  (Judging from the remarks on its page, I am not inclined now to test that alternative.)

As for GetHandleVerifier, it remains the only process that churns away even when FF is quiescent - no loading, no refreshing, no animations.  Always there are two instances, one with zero activity whenever I have looked.  And nobody has responded to my enquiries about this, which is a little disappointing.  Is it so banal and obvious?  If this is just churning uselessly it would be good to tame it.

Incidentally, I am not seeign any ghost windows (except for the brief intermediate period after closing one), and memory usage does not steadily creep up as it used to, so the changes for this version have been beneficial.

Further to my comment 49: a way of suspending all activity (whatever it might be) in a tab would be a very useful way of diagnosing nwhere problems are occurring.  Am I missing something or can this be reinstated?

Thanks.

+++++++
The usual stack for the GHV thread, almost always (while I am doing nothing):

ntoskrnl.exe!memset+0x64a
ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
ntoskrnl.exe!KeWaitForSingleObject+0x19f
ntoskrnl.exe!PoStartNextPowerIrp+0xbd4
ntoskrnl.exe!PoStartNextPowerIrp+0x186d
ntoskrnl.exe!KeWaitForMultipleObjects+0xf5d
ntoskrnl.exe!KeWaitForMultipleObjects+0x26a
ntoskrnl.exe!NtWaitForSingleObject+0x40f
ntoskrnl.exe!NtWaitForSingleObject+0x77e
ntoskrnl.exe!KeSynchronizeExecution+0x3a23
ntdll.dll!ZwWaitForMultipleObjects+0xa
KERNELBASE.dll!GetCurrentProcess+0x40
kernel32.dll!WaitForMultipleObjectsEx+0xb3
USER32.dll!GetScrollBarInfo+0x1dd
USER32.dll!MsgWaitForMultipleObjectsEx+0x2e
xul.dll!mozilla::net::LoadInfo::GetPrincipalToInherit+0xd1af
xul.dll!mozilla::scache::PathifyURI+0x68f0
xul.dll!mozilla::net::LoadInfo::GetAboutBlankInherits+0x3ff3d
xul.dll!mozilla::scache::PathifyURI+0x6766
xul.dll!mozilla::scache::PathifyURI+0x6b1f
xul.dll!mozilla::scache::PathifyURI+0x60a2
xul.dll!mozilla::scache::PathifyURI+0x6423
xul.dll!mozilla::net::LoadInfo::GetLoadingDocument+0x20277
xul.dll!mozilla::net::LoadInfo::GetLoadingDocument+0x2023a
xul.dll!XRE_GetProcessType+0x10b80
xul.dll!XRE_GetProcessType+0x10840
xul.dll!XRE_GetProcessType+0x107eb
xul.dll!mozilla::net::LoadInfo::PrincipalToInherit+0x17787
xul.dll!XRE_InitCommandLine+0x23a
xul.dll!XRE_main+0x55
firefox.exe+0x198a
firefox.exe!GetHandleVerifier+0x1069
firefox.exe!GetHandleVerifier+0x4105
kernel32.dll!BaseThreadInitThunk+0xd
ntdll.dll!RtlUserThreadStart+0x21
(Reporter)

Comment 104

2 years ago
Well, well:

1 (100.0%) -- ghost-windows
└──1 (100.0%) ── https://www.youtube.com/embed/usYTH_D0wxc?rel=0?ecver=2

- and I don't remember even looking at this!
But of course, checking in FF makes the ghost go away, only to return - and persist overnight - when the window is closed again.
(Reporter)

Comment 105

2 years ago
It is doing it again ...
https://cleopatra.io/#report=e65016b645b47176f5566b17d9a53894e0e9b6e0
25 - 30% CPU, essentially locked out for several minutes at a time.
No ghost windows, though, but 7.1 GB "private bytes"!
Flags: needinfo?(hrdubwd)
(In reply to Dr B W Darvell from comment #105)
> It is doing it again ...
> https://cleopatra.io/#report=e65016b645b47176f5566b17d9a53894e0e9b6e0
> 25 - 30% CPU, essentially locked out for several minutes at a time.
> No ghost windows, though, but 7.1 GB "private bytes"!

For this one, I see a lot of time being spent inside of DevTools... can I assume you had the Developer Tools open in some tab?
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 107

2 years ago
Not knowingly.  I did have a 'logged to console' message at one point that it transpired I need to use DTs to view, but I thought I had closed all that as unhelpful.  (I had apparently exceeded a Google maps API call limit of 25,000!)  

Does DT put that much load on?  Bear in mind that this was not while I was trying to use any of it at that point (I think).
Flags: needinfo?(hrdubwd)
Hi Darvell, I know this is a hard one, and you helped developers with different investigations around it. Did you see any changes on the latest Release 53.0?
(Reporter)

Comment 109

2 years ago
Hi,
I have not noticed any major problems since the last report - no CPU-hogging that has prevented my doing anything at all.  I keep an eye on it from time to time, although I did not pay attention as to when 53.0 was installed.  Even with the profiler running on the offchance of catching a problem it has been seemingly unproblematic.  I have also checked occasionally for ghost windows and found none that lasted more than a short while, as it seems is to be expected.  And all this with many tabs loaded and much the same list of extensions.

Overall, I have been unaffected and therefore happy enough.  I will continue to make checks, just in case.

Looking now at Process Explorer's graph, I see more I/O and variation in private bytes than previously, even though there is nothing obviously running (but who knows what is happening in non-viewed tabs?).

Thanks for following up.
Darvell, thank you for all the support provided!

For the time beeing and based on the previous comment, I will mark this bug Resolved-WFM, but please feel free to reopen it if this shows up again.
Status: REOPENED → RESOLVED
Last Resolved: 2 years ago2 years ago
Resolution: --- → WORKSFORME
(Reporter)

Comment 111

2 years ago
OK, thanks.
I certainly have not seen any excessive CPU and lock-out for a while now, and no ghost windows.
Mind you, this might be helped by the twin quad processors now!  (Since a week ago, that is.)

If I notice anything, I'll certainly come back to this thread.

Thanks,

BWD

Comment 112

2 years ago
(In reply to Brindusa Tot[:brindusat] from comment #110)

I'm getting this issue - I found this bug searching for GetHandleVerifier.
ProcessExplorer shows me it's stuck on 23% CPU for 2.5 trillion cycles (just over 20 minutes)
It gives me this stack:
ntoskrnl.exe!KeSynchronizeExecution+0x3f26
ntoskrnl.exe!KeWaitForMultipleObjects+0x10d7
ntoskrnl.exe!KeWaitForMultipleObjects+0xb3f
ntoskrnl.exe!KeWaitForMutexObject+0x377
ntoskrnl.exe!PsGetCurrentThreadProcessId+0xa10
ntoskrnl.exe!KeSynchronizeExecution+0x44ef
ntoskrnl.exe!KeSynchronizeExecution+0x4142
ntoskrnl.exe!KeSynchronizeExecution+0x3c8d
xul.dll+0x6acc5a
xul.dll+0x6accfa
xul.dll+0xf4985
xul.dll+0xf4c39
xul.dll+0x6d104f


Can't say I know any C++, but there's a possibly relevant comment in the source for ActiveVerifier::InstallVerifier
https://dxr.mozilla.org/mozilla-central/source/security/sandbox/chromium/base/win/scoped_handle.cc#144 
  // If you are reading this, wondering why your process seems deadlocked, take
  // a look at your DllMain code and remove things that should not be done
  // there, like doing whatever gave you that nice windows handle you are trying
  // to store in a ScopedHandle.

Comment 113

2 years ago
(In reply to rob.swdev from comment #112)
> (In reply to Brindusa Tot[:brindusat] from comment #110)
> 
> I'm getting this issue - I found this bug searching for GetHandleVerifier.
> ProcessExplorer shows me it's stuck on 23% CPU for 2.5 trillion cycles (just
> over 20 minutes)
> It gives me this stack:
> ntoskrnl.exe!KeSynchronizeExecution+0x3f26
> ntoskrnl.exe!KeWaitForMultipleObjects+0x10d7
> ntoskrnl.exe!KeWaitForMultipleObjects+0xb3f
> ntoskrnl.exe!KeWaitForMutexObject+0x377
> ntoskrnl.exe!PsGetCurrentThreadProcessId+0xa10
> ntoskrnl.exe!KeSynchronizeExecution+0x44ef
> ntoskrnl.exe!KeSynchronizeExecution+0x4142
> ntoskrnl.exe!KeSynchronizeExecution+0x3c8d
> xul.dll+0x6acc5a
> xul.dll+0x6accfa
> xul.dll+0xf4985
> xul.dll+0xf4c39
> xul.dll+0x6d104f
> 
> 
> Can't say I know any C++, but there's a possibly relevant comment in the
> source for ActiveVerifier::InstallVerifier
> https://dxr.mozilla.org/mozilla-central/source/security/sandbox/chromium/
> base/win/scoped_handle.cc#144 
>   // If you are reading this, wondering why your process seems deadlocked,
> take
>   // a look at your DllMain code and remove things that should not be done
>   // there, like doing whatever gave you that nice windows handle you are
> trying
>   // to store in a ScopedHandle.

Forgot to mention I have Firefox 53.0.3 64 bit.
(Reporter)

Comment 114

2 years ago
On the same vbersion of FF, I have this at the moment:

ntoskrnl.exe!memset+0x64a
ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
ntoskrnl.exe!KeWaitForMutexObject+0x19f
ntoskrnl.exe!_misaligned_access+0xbd4
ntoskrnl.exe!_misaligned_access+0x186d
ntoskrnl.exe!KeWaitForMultipleObjects+0xf5d
ntoskrnl.exe!KeWaitForMultipleObjects+0x26a
ntoskrnl.exe!NtWaitForSingleObject+0x40f
ntoskrnl.exe!NtWaitForSingleObject+0x77e
ntoskrnl.exe!KeSynchronizeExecution+0x3a23
ntdll.dll!ZwWaitForMultipleObjects+0xa
KERNELBASE.dll!GetCurrentProcess+0x40
kernel32.dll!WaitForMultipleObjectsEx+0xb3
USER32.dll!GetScrollBarInfo+0x1dd
USER32.dll!MsgWaitForMultipleObjectsEx+0x2e
xul.dll+0x41b1f4
xul.dll+0xf6d52
xul.dll+0x74830d
xul.dll+0xf6b0e
xul.dll+0xf6edb
xul.dll!workerlz4_compress+0x1381a
xul.dll+0x22985a
xul.dll+0x29dad8

some 2 ~ 7 % of CPU, continuous.
The second instance of GetHandleVerifier mentioned a few times above is present at the moment.  Has something changed?
But it still seems to me to be odd that it runs all the time like this when there is minimal other activity.  Nobody has said what it is for, despite requests.

Other than the occasional autorefresh, is it not reasonable to expect a program to be quiescent when not in use?

It may be an oversight as implied at 112 above.
(Reporter)

Comment 115

2 years ago
Strangely, just now, on a laptop using Win7 32-bit, the same version of FF gave >>60% CPU for a long period - all down to the same source: GetHandleVerifier.  Nothing else could be done while this was running. The was only the one instance present.  

(No ghost windows, only a few tabs open, nothing active particularly.  Memory not a problem.)

There is something weird in there.  Could this be investigated, please?

Thanks.
(Reporter)

Comment 116

2 years ago
Doing it again, on the 64-bit machine - typing this is a struggle!  GHV has 612,717,000,000,000 cycles logged!  Kernel time 1 hour 56 min!
Others I have spoken to complain about FF being slow - it is not just me.
(In reply to Dr B W Darvell from comment #116)
> Doing it again, on the 64-bit machine - typing this is a struggle!  GHV has
> 612,717,000,000,000 cycles logged!  Kernel time 1 hour 56 min!
> Others I have spoken to complain about FF being slow - it is not just me.

Can you please try to gather another performance profile while in this state using the newest version of the Gecko Profiler add-on, available at https://perf-html.io/ ?
Note that I believe you'll need to do this using the Nightly version of the browser.
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 119

2 years ago
I am not sure I can dedicate the time to do this, if not a standard set up.  The events are erratic, and I do not know what triggers it.  It has taken a long time for this to reoccur, and I am out a lot in the coming couple of weeks.

That said, I cannot see that the Nightly is required - it has installed without hitch.
Flags: needinfo?(hrdubwd)
(In reply to Dr B W Darvell from comment #119)
> That said, I cannot see that the Nightly is required - it has installed
> without hitch.

Yes, it will install - however, the information we get through the Profiler from Nightly is richer and more usable than what we get through the other channels because of how it's been configured at build time.
(Reporter)

Comment 121

2 years ago
In that case, I am probably not going to be able to help.  I cannot risk getting in a tangle with trial versions.

But, tell me this, please: why does nobody say anything in response to the comments about GetHandleVerifier?  Is this a stupid question?  Am I missing something because I am not a programmer?  Something that is so obvious that it is beneath contempt?

Humour me, please: what does it do, and why does it need to run so much?
5 - 7% CPU (5 billion cycles a second!) when nearly everything else is at <0.01% CPU does seem very strange.

Thanks.
(In reply to Dr B W Darvell from comment #121)
> Humour me, please: what does it do, and why does it need to run so much?
> 5 - 7% CPU (5 billion cycles a second!) when nearly everything else is at
> <0.01% CPU does seem very strange.
> 

I don't know what it does, exactly. Seems related to sandboxing, but I'm really not sure.

Looks like bobowen touched the sandbox code around [1], so maybe he can give more insight.

[1]: http://searchfox.org/mozilla-central/rev/66d9eb3103e429127c85a7921e16c5a02458a127/security/sandbox/chromium/base/win/scoped_handle.cc#139-147
Flags: needinfo?(bobowencode)
(In reply to Dr B W Darvell from comment #121)
> But, tell me this, please: why does nobody say anything in response to the
> comments about GetHandleVerifier?  Is this a stupid question?  Am I missing
> something because I am not a programmer?  Something that is so obvious that
> it is beneath contempt?
> 

Also, I apologize I hadn't answered this question directly earlier. I was so intent on getting a profile, I missed it despite you asking so many times. It's certainly not a stupid question, it's a perfectly valid one.
(Reporter)

Comment 124

2 years ago
re #122:  Ah ... OK,
   If the mentioned deadlock (or similar) is underlying this, i.e. a fault in webpage code [???], is there no way to detect this and bypass the problem?  We can hope for a response from bobowen (but that has been there a long time, no? 2012).  Error trapping used to be the thing to ensure when I was working in BASIC (yes, I know ...).  It does look like a lot of looping is going on.

re #123:  OK, thank you.  We can but hope for a definitive answer from someone else.
(In reply to Dr B W Darvell from comment #121)
> In that case, I am probably not going to be able to help.  I cannot risk
> getting in a tangle with trial versions.
> 
> But, tell me this, please: why does nobody say anything in response to the
> comments about GetHandleVerifier?  Is this a stupid question?  Am I missing
> something because I am not a programmer?  Something that is so obvious that
> it is beneath contempt?

Not a stupid question (they rarely are) just things being typically confusing. :-)

I haven't read too far down the bug (it's rather long), but I think you are seeing a thread in Process Explorer with a start address of something like firefox.exe!GetHandleVerifier+0x27ff.
I'm pretty sure this is just down to not having the correct symbols.
By the way, even though I think this is wrong, this is the start address of the thread not what it is doing now, you would need to look at the stack to see that.

If I set up Process Explorer to use my symbol cache (which I use for VisualStudio and windbg) the thread's start address is firefox.exe!wmainCRTStartup. This is the main thread.
So all this means is that whatever is taking a fairly steady percentage of CPU is happening on the main thread, which is not too surprising and would explain why Firefox isn't very responsive.

You could try adding http://symbols.mozilla.org/firefox into the "Symbols Path:" for Process Explorer under Options->Configure Symbols... and then looking at the stack, but that never works very well for me.
I don't generally set up Process Explorer to download from the online symbol store as it quite often locks up.
Flags: needinfo?(bobowencode)
(Reporter)

Comment 126

2 years ago
Hi,
Thanks for responding.
1.  OK, thanks.
2.  Start address is similar, yes.
3.  Symbol cache: OK, done that.  The stack (at a random point) shows:

0	ntoskrnl.exe!memset+0x64a
1	ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
2	ntoskrnl.exe!KeWaitForMutexObject+0x19f
3	ntoskrnl.exe!_misaligned_access+0xbd4
4	ntoskrnl.exe!_misaligned_access+0x186d
5	ntoskrnl.exe!KeWaitForMultipleObjects+0xf5d
6	ntoskrnl.exe!KeWaitForMultipleObjects+0x26a
7	ntoskrnl.exe!NtWaitForSingleObject+0x40f
8	ntoskrnl.exe!NtWaitForSingleObject+0x77e
9	ntoskrnl.exe!KeSynchronizeExecution+0x3a23
10	ntdll.dll!ZwWaitForMultipleObjects+0xa
11	KERNELBASE.dll!GetCurrentProcess+0x40
12	kernel32.dll!WaitForMultipleObjectsEx+0xb3
13	USER32.dll!GetScrollBarInfo+0x1dd
14	USER32.dll!MsgWaitForMultipleObjectsEx+0x2e
15	xul.dll+0x41b1f4
16	xul.dll+0xf6d52
17	xul.dll+0x74830d
18	xul.dll+0xf6b0e
19	xul.dll+0xf6edb
20	xul.dll+0x41516e
21	xul.dll+0xf671f
22	xul.dll+0x4e5173
23	xul.dll+0x4e5136
24	xul.dll+0x4e4e1c
25	xul.dll+0x4e4adc
26	xul.dll+0x4e4a87
27	xul.dll+0x5ebe3b
28	xul.dll+0x644fbc
29	xul.dll+0x7ced39
30	firefox.exe+0x2a36
31	firefox.exe+0x1180
32	firefox.exe!GetHandleVerifier+0x2a3d
33	kernel32.dll!BaseThreadInitThunk+0xd
34	ntdll.dll!RtlUserThreadStart+0x21


lines 3 and 4 look ominous.  Very similar a few minutes later:

0	ntoskrnl.exe!memset+0x64a
1	ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
2	ntoskrnl.exe!KeWaitForMutexObject+0x19f
3	ntoskrnl.exe!_misaligned_access+0xbd4
4	ntoskrnl.exe!_misaligned_access+0x186d
5	ntoskrnl.exe!IoFreeErrorLogEntry+0x287
<snip>

... and again:

0	ntoskrnl.exe!memset+0x64a
1	ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
2	ntoskrnl.exe!KeWaitForMutexObject+0x19f
3	ntoskrnl.exe!_misaligned_access+0xbd4
4	ntoskrnl.exe!_misaligned_access+0x186d
5	ntoskrnl.exe!IoFreeErrorLogEntry+0x287
6	xul.dll+0x27b270
7	xul.dll+0x27b082
8	xul.dll+0x753a85
9	xul.dll+0x27ac10
10	xul.dll+0x350ca6
11	xul.dll+0x27a202
12	xul.dll+0x3dd197
13	xul.dll+0x3dcf79
14	xul.dll+0x36230a
15	xul.dll+0x362b36
16	xul.dll+0xabb3b
17	xul.dll+0xab73f
18	xul.dll+0xf701b
19	xul.dll+0xf66ba
20	xul.dll+0x4e5173
21	xul.dll+0x4e5136
22	xul.dll+0x4e4e1c
23	xul.dll+0x4e4adc
24	xul.dll+0x4e4a87
25	xul.dll+0x5ebe3b
26	xul.dll+0x644fbc
27	xul.dll+0x7ced39
28	firefox.exe+0x2a36
29	firefox.exe+0x1180
30	firefox.exe!GetHandleVerifier+0x2a3d
31	kernel32.dll!BaseThreadInitThunk+0xd
32	ntdll.dll!RtlUserThreadStart+0x21

Does this help?

What am I looking for?
(In reply to Dr B W Darvell from comment #126)
...
> 25	xul.dll+0x4e4adc

This is still not using the correct symbols, that 0x4e4adc would be resolved to a function something like xul.dll!MessageLoop::RunHandler+0x20

I was forgetting that I think you also need a full version of dbghelp.dll configured in the same dialog, you would need to install some development tools to get that. For example:

https://developer.microsoft.com/en-us/windows/downloads/windows-10-sdk

I think you only need to install the debugging tools and then look for dbghelp.dll.

> Does this help?

To be honest, trying to catch the problem by looking at stacks in Process Explorer is probably not the best approach, although it can sometimes help.

The best way is probably by trying to get the profile data mconley was talking about, this will give a broader set of data, which means we're more likely to be able to spot the problem (or problems).
(Reporter)

Comment 128

2 years ago
C:\Windows\system32\dbghelp.dll is rferenced in the PE Configure Symbols - wrong version?  I'll check later.

I will wait for another lockup and see if I can capture what is needed.  As I said, I do not have the time to go through a nightly installation, with the risks involved in all that.
(Reporter)

Comment 129

2 years ago
C:\Windows\system32\dbghelp.dll is rferenced in the PE Configure Symbols - wrong version?  I'll check later.

I will wait for another lockup and see if I can capture what is needed.  As I said, I do not have the time to go through a nightly installation, with the risks involved in all that.
(Reporter)

Comment 130

2 years ago
- perhaps a reboot is required for symbols to be used?  This machine is on for long periods.
(In reply to Dr B W Darvell from comment #128)
> C:\Windows\system32\dbghelp.dll is rferenced in the PE Configure Symbols -
> wrong version?  I'll check later.

This is not a full version unfortunately, I think you need one from development tools.
You shouldn't need to reboot.
(Reporter)

Comment 132

2 years ago
OK, installed ...
GHV has disappeared from the Threads list, I have to suppose that it is renamed as well: firefox.exe|wmainCRTStartup - since this is taking most CPU.  Is that true?  The stack for this looks like this:

ntoskrnl.exe!memset+0x64a
ntoskrnl.exe!KeWaitForMultipleObjects+0xd52
ntoskrnl.exe!KeWaitForMutexObject+0x19f
ntoskrnl.exe!_misaligned_access+0xbd4
ntoskrnl.exe!_misaligned_access+0x186d
ntoskrnl.exe!IoFreeErrorLogEntry+0x287
xul.dll!nsIDocument::GetInnerWindow+0x7
xul.dll!MarkContentViewer+0x58
xul.dll!MarkDocShell+0x67
xul.dll!MarkDocShell+0x17e
xul.dll!MarkWindowList+0x84
xul.dll!nsCCUncollectableMarker::Observe+0xd8
xul.dll!nsObserverList::NotifyObservers+0x62
xul.dll!nsObserverService::NotifyObservers+0xf5
xul.dll!XPCJSContext::PrepareForForgetSkippable+0x31
xul.dll!nsCycleCollector::ForgetSkippable+0x61
xul.dll!FireForgetSkippable+0x6a
xul.dll!CCTimerFired+0x83
xul.dll!nsTimerImpl::Fire+0x307
xul.dll!nsTimerEvent::Run+0xf3
xul.dll!nsThread::ProcessNextEvent+0x26b
xul.dll!mozilla::ipc::MessagePump::Run+0x9e
xul.dll!MessageLoop::RunHandler+0x1b
xul.dll!MessageLoop::Run+0x3e
xul.dll!nsBaseAppShell::Run+0x3c
xul.dll!nsAppShell::Run+0x2c
xul.dll!nsAppStartup::Run+0x27
xul.dll!XREMain::XRE_mainRun+0x62b
xul.dll!XREMain::XRE_main+0x2e4
xul.dll!XRE_main+0x55
firefox.exe!NS_internal_main+0x676
firefox.exe!wmain+0x140
firefox.exe!__scrt_common_main_seh+0x11d
kernel32.dll!BaseThreadInitThunk+0xd
ntdll.dll!RtlUserThreadStart+0x21


- does that make sense?  Is it usable info?
(In reply to Dr B W Darvell from comment #132)
> OK, installed ...
> GHV has disappeared from the Threads list, I have to suppose that it is
> renamed as well: firefox.exe|wmainCRTStartup - since this is taking most
> CPU.  Is that true?

Yes that is the correct start address now and it appears the symbols are being picked up correctly.
Looks like it is doing some sort of garbage collection, but that is just a point in time, so it's possible that isn't what is taking all the CPU time.

mconley might have more idea about this as I know he's been looking at performance issues.
Flags: needinfo?(mconley)
(Reporter)

Comment 134

2 years ago
OK, good.  Thanks.

Are line such as 
ntoskrnl.exe!_misaligned_access+0xbd4
a problem?  To my eye it suggests an error.


[Off now for a few hours.]

Comment 135

2 years ago
I have a stack trace from windbg of the thread that is continuously spinning, and presumably that causes my hang. 
Seems to be doing something with regular expressions. I assume that this is the UI thread


xul!js::irregexp::InterpretCode<unsigned char>+0x139 [c:\builds\moz2_slave\m-rel-w64-00000000000000000000\build\src\js\src\irregexp\regexpinterpreter.cpp @ 271] 
xul!js::RegExpShared::execute+0x2f1 [c:\builds\moz2_slave\m-rel-w64-00000000000000000000\build\src\js\src\vm\regexpobject.cpp @ 1176] 
xul!ExecuteRegExpImpl+0x27 [c:\builds\moz2_slave\m-rel-w64-00000000000000000000\build\src\js\src\builtin\regexp.cpp @ 127] 
xul!ExecuteRegExp+0x21d [c:\builds\moz2_slave\m-rel-w64-00000000000000000000\build\src\js\src\builtin\regexp.cpp @ 971] 
xul!js::RegExpTesterRaw+0x23 [c:\builds\moz2_slave\m-rel-w64-00000000000000000000\build\src\js\src\builtin\regexp.cpp @ 1199] 
0x0000021a`9d7776bc
0xf962f5b0
0x0000004f`00000000
0x0000021a`00000000
0x2844
0x0000004f`767f86e4
0x0000004f`767f86e4

The only modules I don't have symbols for are the following:
<Unloaded>	00007ff8`bc200000	00007ff8`bc205000	Sat Jul 16 03:28:55 2016 (57899be7)	0000659a	None	KBDUK.DLL
<Unloaded>	00007ff8`bd380000	00007ff8`bd39b000	Sat Jul 16 03:26:29 2016 (57899b55)	0001dff8	None	resourcepolicyclient.dll
<Unloaded>	00007ff8`bc210000	00007ff8`bc238000	Wed Nov 04 21:16:19 2015 (563a75a3)	00035e64	None	C:\WINDOWS\SYSTEM32\atiuxp64.dll
We seem to be going in circles a bit - because if you're finding that you're GC'ing a lot, then the next thing I'm going to ask for is an about:memory report, but I see you've been posting those for a while now.

Perhaps we've got a new leak here? When you in this state again, are you able to produce yet another about:memory report and post it in this bug?
Flags: needinfo?(mconley) → needinfo?(hrdubwd)
(Reporter)

Comment 137

2 years ago
Created attachment 8873564 [details]
memory-report.json.gz

For reference - no problem at this point.
(Reporter)

Comment 138

2 years ago
I can do that, of course.  An assumed no-problem report just uploaded, in case comparison helps.  On the other had, if something shows now ...
(Reporter)

Comment 139

2 years ago
Created attachment 8874215 [details]
Firefox 2017-06-04 15.52 profile.sps.json

Capture during a slow response period, albeit fairly short, just in case there is something of interest.
Attachment #8710630 - Attachment is obsolete: true
Attachment #8842085 - Attachment is obsolete: true
Attachment #8842990 - Attachment is obsolete: true
Attachment #8843688 - Attachment is obsolete: true
Attachment #8846333 - Attachment is obsolete: true
Attachment #8873564 - Attachment is obsolete: true
Flags: needinfo?(hrdubwd)
(Reporter)

Comment 140

2 years ago
Created attachment 8874222 [details]
Firefox 2017-06-04 15.54 profile.sps.json

And one a minute or so later.
(Reporter)

Comment 141

2 years ago
Created attachment 8874226 [details]
Firefox 2017-06-04 16.15 profile.sps.json

... and again.
(Reporter)

Comment 142

2 years ago
Created attachment 8874353 [details]
Firefox 2017-06-05 09.05 profile.sps.json

Another period of high CPU.
Unfortunately, these profiles are of limited value because they're not being captured on Nightly. From them, however, I can confirm that we're still spending a lot of time sweeping / GC'ing. An about:memory report rather than a profile when you're in this state might be better.
(Reporter)

Comment 144

2 years ago
OK, I'll try to do that.

re: Nightly - what are the risks, where do I get it, and how would I return to the main release version - seamlessly?  Does it mean I have to do a fresh install every day?

*  Can you verify that GetHandleVerifier is now renamed  firefox.exe|wmainCRTStartup  because of the symbol loading?
*  Can you say what that routine is supposed to be doing and why it takes so much CPU continuously?
*  Could you comment on what this means:  ntoskrnl.exe!_misaligned_access ?  (This is always present, twice.)
*  I have just found this in the stack for GHV:  
xul.dll!js::detail::HashTable<js::HashMapEntry<js::CrossCompartmentKey,js::detail::UnsafeBareReadBarriered<JS::Value> >,js::HashMap<js::CrossCompartmentKey,js::detail::UnsafeBareReadBarriered<JS::Value>,js::CrossCompartmentKey::Hasher,js::SystemAllocPolic
- which looks like another error.  True?

Thanks.
Flags: needinfo?(mconley)
(In reply to Dr B W Darvell from comment #144)
> OK, I'll try to do that.
> 
> re: Nightly - what are the risks, where do I get it, and how would I return
> to the main release version - seamlessly?  Does it mean I have to do a fresh
> install every day?

Yes, there are some risks. There are times when changes land in Nightly where it becomes hard to migrate _back_ to an earlier version. Most recently, bug 977177 introduced such a change.

In times like this, it is sometimes more convenient to _clone_ your user profile, and use one in Nightly, and the other in Release.

Here is a document regarding backing up / restoring your user profile that might help:

https://support.mozilla.org/en-US/kb/back-and-restore-information-firefox-profiles

Nightly can be downloaded and installed from https://www.mozilla.org/en-US/firefox/channel/desktop/#nightly

As a browser, it's really quite stable. Tens of thousands (including myself) use it as our everyday browser.

> *  Can you verify that GetHandleVerifier is now renamed 
> firefox.exe|wmainCRTStartup  because of the symbol loading?
> *  Can you say what that routine is supposed to be doing and why it takes so
> much CPU continuously?
> *  Could you comment on what this means:  ntoskrnl.exe!_misaligned_access ? 
> (This is always present, twice.)
> *  I have just found this in the stack for GHV:  
> xul.dll!js::detail::HashTable<js::HashMapEntry<js::CrossCompartmentKey,js::
> detail::UnsafeBareReadBarriered<JS::Value>
> >,js::HashMap<js::CrossCompartmentKey,js::detail::UnsafeBareReadBarriered<JS:
> :Value>,js::CrossCompartmentKey::Hasher,js::SystemAllocPolic
> - which looks like another error.  True?

I'm afraid I don't know enough about this level of Firefox or Gecko to say with any certainty - redirecting needinfo to bobowen.
Flags: needinfo?(mconley) → needinfo?(bobowencode)
(In reply to Dr B W Darvell from comment #144)
...
> *  Can you verify that GetHandleVerifier is now renamed 
> firefox.exe|wmainCRTStartup  because of the symbol loading?

Yes, that is the same thread ... the main thread.
The GetHandleVerifier start address was due to incorrect symbol resolution by Process Explorer the wmainCRTStartup address is correct.

> *  Can you say what that routine is supposed to be doing and why it takes so
> much CPU continuously?

It isn't, it is the starting function for that thread, not what it is doing at the point you are experiencing an issue.
Lots of things process on the main thread, tracking down which thing (or things) is causing this issue is what mconley is trying to do with the profiles he is requesting.

> *  Could you comment on what this means:  ntoskrnl.exe!_misaligned_access ? 
> (This is always present, twice.)

I believe it is something to do with the kernel re-aligning variables for other APIs that require it, although I'm not at all sure.
I really don't know if that could be the cause of your issues, but either way we would need to find the parts of our code that are causing this (if it is the underlying cause).
That brings us back to getting profiles.

> *  I have just found this in the stack for GHV:  
> xul.dll!js::detail::HashTable<js::HashMapEntry<js::CrossCompartmentKey,js::
> detail::UnsafeBareReadBarriered<JS::Value>
> >,js::HashMap<js::CrossCompartmentKey,js::detail::UnsafeBareReadBarriered<JS:
> :Value>,js::CrossCompartmentKey::Hasher,js::SystemAllocPolic
> - which looks like another error.  True?

This is just code within the JS engine.
I don't know too much about that, but if you are referring to the "Unsafe" part of the function name, I don't think that indicates an error.
Flags: needinfo?(bobowencode)
(Reporter)

Comment 147

2 years ago
(in reply to #145)
OK, thanks - will try to find time to get that organized.

(in reply to #146)
And also thanks - all noted.  The GHV part now makes much more sense!  I still find it odd that it is running so much when everything is (supposedly) quiescent, but no matter. One day I might try a stripped-out run ...

Comment 148

2 years ago
I also had this problem and investigated a little bit.. I killed FF then restarted, restored crashed session, provided master password, same behaviour... saved a first report. Then I closed firefox, started it back again, it had a couple hiccups and it worked nicely... saved a second report. then I thought I should click the tabs in the background, at least to have the same active tabs and boom, it`s back again... it seems visiting the page http://www.munchkin.com/latchtm-transition-cup.html (no advertising here) caused this behavior. I have also visited FB, this page and some google searched (closed already) + care.com page -- see it in report. I have some company-server support pages but they were simply "ghost" tabs (don`t know how to call them... they were restored, I did not clicked them to have ff reload the page).

so, here I create a 3rd report (this time I am authenticated in bugzilla).

I`ll also add some screenshots, maybe they are of help for anyone... otherwise someone please erase them.

Comment 149

2 years ago
Well surprise.. closing the munchkin webpage solved the issue... can`t believe it.. I`ll create a 4th report.

Comment 150

2 years ago
Created attachment 8877058 [details]
4 firefox reports, 1 before restarting, 3 in a single session, plus 18 screen captures with firefox details in process explorer
(Reporter)

Comment 151

2 years ago
Having updated to 54.0.1 I was hoping to see a dramatic chnage in behaviour, but while activity is spread over 8 cores, and the background CPU use seems lower than it used to be, I do not have the multiprocess window - apparently due to add-ons interfering.  How do I tell which ones are the problem?  Do I have to go through the the whole lot systematically again?
Thanks.
(Reporter)

Comment 152

2 years ago
Restarting with add-ons disabled I still get:

Multiprocess Windows 	0/1 (Disabled by add-ons)

The only ones enabled are the hidden ones:

Application Update Service Helper	2.0	true	aushelper@mozilla.org
Firefox Screenshots	                6.6.0	true	screenshots@mozilla.org
Multi-process staged rollout	        1.50	true	e10srollout@mozilla.org
Pocket	                                1.0.5	true	firefox@getpocket.com
Web Compat	                        1.1	true	webcompat@mozilla.org

So why can I not see multiprocess enabled?

I had set  dom.ipc.processCount  to 2, but even that seems to have been ignored.

What am I missing?
(Reporter)

Comment 153

2 years ago
Looking further (about:config|e10s), I find this:

elOs.rollout.cohort 					user set 	string	 	disqualified-test
elOs.rollout.cohortSample 				user set 	string 		0.458686
extensions.bootstrappedAddons 				user set 	string 		{"spdyindicator@cheng•...
extensions.elOs.rollout.blocklist 			user set 	string
extensions.elOs.rollout.hasAddon 			user set 	boolean 	false
extensions.elOs.rollout.policy 				user set 	string 		50allmpc
extensions.elOsBlockedByAddons 				user set 	boolean 	true
extensions.elOsBlocksEnabling 				default 	boolean 	true
extensions.elOsMuftiBlockedByAddons 			user set 	boolean 	true
extensions.elOsMultiBlocksEnabling 			default 	boolean 	true
extensions .. elOsrollout@mozilla.org.install-event-fired 	user set 	boolean 	true

- which suggests to my untrained eye that I was not supposed to have multi-.  True?
When will this happen, then?
(Reporter)

Comment 154

2 years ago
Ah:
https://hacks.mozilla.org/2017/06/firefox-54-e10s-webextension-apis-css-clip-path/#comment-21286

"Currently, e10s-multi is only turned on by default for users without any add-ons installed."

I shall have to wait, I suppose.

Comment 155

10 months ago
Like yourself, I had the same persistent issues in Firefox 32-bit on Windows 7 Pro x64, for about the last 3 months.  In the end, the fix was to update to 64-bit Firefox, and replace FireGestures with Mouse Gesture Events.  No other changes to my profile were necessary.

I'm currently using ESR 52.7.3 with over 20 add-ons, and E10S-multi is not enabled.

There are no more daily crashes after 6-8 hours and ~2.1 to ~2.2 GB of RAM usage, which was typical under the 32-bit edition.  Now RAM usage by Firefox goes as high as 2.6 GB for me, typically, in a 9 hour session.  I restart my PC daily at the end of my work shift.


Application Basics
------------------

Name: Firefox
Version: 52.7.3
Build ID: 20180322140748
Update Channel: esr
User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:52.0) Gecko/20100101 Firefox/52.0
OS: Windows_NT 6.1
Multiprocess Windows: 0/12 (Disabled by add-ons)
Safe Mode: false

Crash Reports for the Last 3 Days
---------------------------------

All Crash Reports

Extensions
----------

Name: Add to Bookmark Ninja
Version: 1.0.0.2
Enabled: true
ID: {f5e8db67-96ed-47fb-a39f-534e1d56aadf}

Name: Application Update Service Helper
Version: 2.0
Enabled: true
ID: aushelper@mozilla.org

Name: Bookmark Dupes
Version: 5.7
Enabled: true
ID: bookmarkdupes@martin-vaeth.org

Name: Bookmark Ninja
Version: 1.0.0.1
Enabled: true
ID: {47e7fc3c-b553-4aa7-bf22-9bdaa4339301}

Name: CanvasBlocker
Version: 0.4.4b
Enabled: true
ID: CanvasBlocker@kkapsner.de

Name: Classic Reload-Stop-Go Button
Version: 1.1
Enabled: true
ID: crsg@ArisT2_Noia4dev

Name: Cookie Controller
Version: 6.1
Enabled: true
ID: {ac2cfa60-bc96-11e0-962b-0800200c9a66}

Name: Copy Urls Expert
Version: 2.6.1
Enabled: true
ID: copy-urls-expert@kashiif-gmail.com

Name: DownThemAll!
Version: 3.0.8
Enabled: true
ID: {DDC359D1-844A-42a7-9AA1-88A850A938A8}

Name: Font Finder
Version: 0.1.8
Enabled: true
ID: {a658a273-612e-489e-b4f1-5344e672f4f5}

Name: Google search link fix
Version: 1.6.7
Enabled: true
ID: jid0-XWJxt5VvCXkKzQK99PhZqAn7Xbg@jetpack

Name: Informational Tab
Version: 0.5.2017061501
Enabled: true
ID: informationaltab@piro.sakura.ne.jp

Name: Mouse Gesture Events
Version: 2.3
Enabled: true
ID: @mousegesture

Name: Multi-process staged rollout
Version: 1.10
Enabled: true
ID: e10srollout@mozilla.org

Name: Personas Plus
Version: 2.0.1
Enabled: true
ID: personas@christopher.beard

Name: Pocket
Version: 1.0.5
Enabled: true
ID: firefox@getpocket.com

Name: PriceBlink Coupons and Price Comparison
Version: 6.6
Enabled: true
ID: info@priceblink.com

Name: Reload Plus
Version: 5.2.3
Enabled: true
ID: reloadplus@blackwind

Name: Restart
Version: 3.0.2
Enabled: true
ID: Restart@schuzak.jp

Name: Secure Login (Embedded WebExtension)
Version: 0.2.1
Enabled: true
ID: @slogin

Name: Tampermonkey
Version: 4.6.5757
Enabled: true
ID: firefox@tampermonkey.net

Name: Tree Style Tab
Version: 0.19.2017090601
Enabled: true
ID: treestyletab@piro.sakura.ne.jp

Name: uBlock Origin
Version: 1.16.2
Enabled: true
ID: uBlock0@raymondhill.net

Name: UnMHT
Version: 8.3.2
Enabled: true
ID: {f759ca51-3a91-4dd1-ae78-9db5eee9ebf0}

Name: Web Compat
Version: 1.0
Enabled: true
ID: webcompat@mozilla.org

Name: DuckDuckGo Plus
Version: 2017.11.15
Enabled: false
ID: jid1-ZAdIEUB7XOzOJw@jetpack

Name: User Agent Switcher
Version: 0.7.3.1-signed.1-signed
Enabled: false
ID: {e968fc70-8f95-4ab9-9e79-304de2a71ee1}

Graphics
--------

Features
Compositing: Direct3D 11
Asynchronous Pan/Zoom: none
WebGL Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics Direct3D11 vs_5_0 ps_5_0)
WebGL2 Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics Direct3D11 vs_5_0 ps_5_0)
Hardware H264 Decoding: Yes; Using D3D9 API
Audio Backend: wasapi
Direct2D: true
DirectWrite: true (6.2.9200.22164)
GPU #1
Active: Yes
Description: Intel(R) HD Graphics
Vendor ID: 0x8086
Device ID: 0x0152
Driver Version: 9.17.10.3040
Driver Date: 2-22-2013
Drivers: igdumd64 igd10umd64 igd10umd64 igdumd32 igd10umd32 igd10umd32
Subsys ID: 05771028
RAM: Unknown
GPU #2
Active: No
Description: AMD Radeon HD 7470
Vendor ID: 0x1002
Device ID: 0x6778
Driver Version: 8.922.0.0
Driver Date: 12-6-2011
Drivers: atiu9p64 atiuxp64 atiuxp64 atiu9pag atiuxpag atiuxpag atiumdva atiumd6a atitmm64
Subsys ID: 21201028
RAM: 1024
Diagnostics
ClearType Parameters: DISPLAY1 [ Gamma: 2.2 Pixel Structure: BGR ClearType Level: 50 Enhanced Contrast: 50 ] DISPLAY2 [ Gamma: 2.2 Pixel Structure: RGB ClearType Level: 0 Enhanced Contrast: 50 ] DISPLAY4 [ Gamma: 2.2 Pixel Structure: BGR ClearType Level: 0 Enhanced Contrast: 100 ]
AzureCanvasAccelerated: 0
AzureCanvasBackend: direct2d 1.1
AzureContentBackend: direct2d 1.1
AzureFallbackCanvasBackend: cairo
ClearType Parameters: DISPLAY1 [ Gamma: 2.2 Pixel Structure: BGR ClearType Level: 50 Enhanced Contrast: 50 ] DISPLAY2 [ Gamma: 2.2 Pixel Structure: RGB ClearType Level: 0 Enhanced Contrast: 50 ] DISPLAY4 [ Gamma: 2.2 Pixel Structure: BGR ClearType Level: 0 Enhanced Contrast: 100 ]
Decision Log
D3D9_COMPOSITING:
disabled by default: Disabled by default




Important Modified Preferences
------------------------------

accessibility.typeaheadfind.flashBar: 0
browser.cache.disk.capacity: 358400
browser.cache.disk.filesystem_reported: 1
browser.cache.disk.hashstats_reported: 1
browser.cache.disk.smart_size.enabled: false
browser.cache.disk.smart_size.first_run: false
browser.cache.disk.smart_size.use_old_max: false
browser.cache.frecency_experiment: 4
browser.download.importedFromSqlite: true
browser.download.useDownloadDir: false
browser.places.smartBookmarksVersion: 8
browser.sessionstore.restore_on_demand: false
browser.sessionstore.upgradeBackup.latestBuildID: 20180322140748
browser.startup.homepage: https://start.duckduckgo.com
browser.startup.homepage_override.buildID: 20180322140748
browser.startup.homepage_override.mstone: 52.7.3
browser.tabs.remote.autostart.2: true
browser.urlbar.daysBeforeHidingSuggestionsPrompt: 0
browser.urlbar.lastSuggestionsPromptDate: 20171122
browser.urlbar.maxRichResults: 20
browser.urlbar.suggest.history: false
dom.ipc.processCount: 4
dom.push.userAgentID: <SCRUBBED>
extensions.lastAppVersion: 52.7.3
font.name.serif.x-western: Calibri
font.size.fixed.x-western: 12
general.autoScroll: false
gfx.color_management.mode: 1
gfx.crash-guard.d3d11layers.appVersion: 52.7.3
gfx.crash-guard.d3d11layers.deviceID: 0x0152
gfx.crash-guard.d3d11layers.driverVersion: 9.17.10.3040
gfx.crash-guard.d3d11layers.feature-d2d: true
gfx.crash-guard.d3d11layers.feature-d3d11: true
gfx.crash-guard.status.d3d11layers: 2
gfx.crash-guard.status.d3d9video: 2
gfx.font_rendering.cleartype.always_use_for_content: true
media.benchmark.vp9.fps: 193
media.benchmark.vp9.versioncheck: 1
media.gmp-gmpopenh264.abi: x86_64-msvc-x64
media.gmp-gmpopenh264.lastUpdate: 1523364906
media.gmp-gmpopenh264.version: 1.6
media.gmp-manager.buildID: 20180322140748
media.gmp-manager.lastCheck: 1524747216
media.gmp-widevinecdm.abi: x86_64-msvc-x64
media.gmp-widevinecdm.lastUpdate: 1523364907
media.gmp-widevinecdm.version: 1.4.8.903
media.gmp.storage.version.observed: 1
media.hardware-video-decoding.failed: false
media.webrtc.debug.log_file: <SCRUBBED>\AppData\Local\Temp\WebRTC.log
network.cookie.cookieBehavior: 3
network.cookie.lifetimePolicy: 2
network.cookie.prefsMigrated: true
network.predictor.cleaned-up: true
places.database.lastMaintenance: 1524151136
places.history.expiration.transient_current_max_pages: 71962
plugin.disable_full_page_plugin_for_types: application/pdf
plugin.state.flash: 1
plugin.state.npadobeaamdetect: 2
plugin.state.npadobeexmandetectx: 2
plugin.state.npdeployjava: 0
plugin.state.npgoogletalk: 2
plugin.state.npo1d: 2
plugin.state.nppdf: 0
print.printer_CutePDF_Writer.print_bgcolor: true
print.printer_CutePDF_Writer.print_bgimages: true
print.printer_CutePDF_Writer.print_duplex: -437918235
print.printer_CutePDF_Writer.print_edge_bottom: 0
print.printer_CutePDF_Writer.print_edge_left: 0
print.printer_CutePDF_Writer.print_edge_right: 0
print.printer_CutePDF_Writer.print_edge_top: 0
print.printer_CutePDF_Writer.print_evenpages: true
print.printer_CutePDF_Writer.print_footercenter:
print.printer_CutePDF_Writer.print_footerleft: &PT
print.printer_CutePDF_Writer.print_footerright: &D
print.printer_CutePDF_Writer.print_headercenter:
print.printer_CutePDF_Writer.print_headerleft: &T
print.printer_CutePDF_Writer.print_headerright: &U
print.printer_CutePDF_Writer.print_in_color: true
print.printer_CutePDF_Writer.print_margin_bottom: 0
print.printer_CutePDF_Writer.print_margin_left: 0
print.printer_CutePDF_Writer.print_margin_right: 0
print.printer_CutePDF_Writer.print_margin_top: 0
print.printer_CutePDF_Writer.print_oddpages: true
print.printer_CutePDF_Writer.print_orientation: 0
print.printer_CutePDF_Writer.print_page_delay: 50
print.printer_CutePDF_Writer.print_paper_data: 1
print.printer_CutePDF_Writer.print_paper_height: -1.00
print.printer_CutePDF_Writer.print_paper_name:
print.printer_CutePDF_Writer.print_paper_size_unit: 0
print.printer_CutePDF_Writer.print_paper_width: -1.00
print.printer_CutePDF_Writer.print_resolution: 600
print.printer_CutePDF_Writer.print_reversed: false
print.printer_CutePDF_Writer.print_scaling: 0.70
print.printer_CutePDF_Writer.print_shrink_to_fit: true
print.printer_CutePDF_Writer.print_to_file: false
print.printer_CutePDF_Writer.print_unwriteable_margin_bottom: 0
print.printer_CutePDF_Writer.print_unwriteable_margin_left: 0
print.printer_CutePDF_Writer.print_unwriteable_margin_right: 0
print.printer_CutePDF_Writer.print_unwriteable_margin_top: 0
privacy.clearOnShutdown.cache: false
privacy.clearOnShutdown.cookies: false
privacy.clearOnShutdown.downloads: false
privacy.clearOnShutdown.history: false
privacy.cpd.cookies: false
privacy.cpd.downloads: false
privacy.cpd.formdata: false
privacy.cpd.history: false
privacy.cpd.offlineApps: true
privacy.donottrackheader.enabled: true
privacy.sanitize.sanitizeOnShutdown: true
privacy.sanitize.timeSpan: 0
security.sandbox.content.tempDirSuffix: {2116e583-0c18-4d52-b9f4-ac5823e01f15}
services.sync.declinedEngines: forms,history,tabs
services.sync.engine.history: false
services.sync.engine.prefs.modified: false
services.sync.engine.tabs: false
services.sync.lastPing: 1524747757
services.sync.lastSync: Thu Apr 26 2018 12:02:14 GMT-0500 (Central Standard Time)
services.sync.numClients: 3
storage.vacuum.last.index: 1
storage.vacuum.last.places.sqlite: 1524245127
storage.vacuum.last.queue.sqlite: 1522164427

user.js Preferences
-------------------

Your profile folder contains a user.js file, which includes preferences that were not created by Firefox.

Important Locked Preferences
----------------------------

Places Database
---------------

JavaScript
----------

Incremental GC: true

Accessibility
-------------

Activated: false
Prevent Accessibility: 0

Library Versions
----------------

NSPR
Expected minimum version: 4.13.1
Version in use: 4.13.1

NSS
Expected minimum version: 3.28.6
Version in use: 3.28.6

NSSSMIME
Expected minimum version: 3.28.6
Version in use: 3.28.6

NSSSSL
Expected minimum version: 3.28.6
Version in use: 3.28.6

NSSUTIL
Expected minimum version: 3.28.6
Version in use: 3.28.6

Experimental Features
---------------------

Sandbox
-------

Content Process Sandbox Level: 1


(In reply to Dr B W Darvell from comment #0)
> User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101
> Firefox/43.0
> Build ID: 20160105164030
> 
> Steps to reproduce:
> 
> I have 18 tabs pinned for very frequently used pages, and then may open
> anything from a few to many tabs in the course of my work (academic stuff). 
> FF works well most of the time, but occasionally, but with monotonous
> regularity, it will go dead slow, consuming up to ~36% CPU, and so making FF
> use difficult to say the least, but also slowing the machine appreciably.  I
> have let it run in this state for hours, and it never resolves.  For
> comparison, the normal background tickover is around 2 or 3%, with
> occasional extra activity to say 10%.  No problem with this at all.
> 
> What seems to be happening is that memory is being shuffled around, or the
> contents thereof.  The fact is that there is some 20 GB of free RAM, of
> which about 2 GB is being used (the numbers tick up and down continuously
> during this kind of event).  I have put the cache into RAM on a virtual
> drive, along with all Windows temp files, so that should not be a problem -
> there is plenty of memory there as well, 20 GB!, no indexing, no
> compression.  (After crashing and restarting to write this up, memory usage
> is around 1.3 GB, CPU usage around 2 - 5% - no problem to use.)
> 
> This is under W7 Pro, 64-bit.  It has been occurring for all of at least a
> few of the previous versions of FF.
> 
> The only way to deal with it is to crash FF and reload, when after the
> loading settling down, all is well again, until some random point in the
> future.  This has now happened many times.  I have held off reporting
> thinking that I might identify the problem with a specific tab.  I cannot.)
> 
> I cannot see any similar reports of this kind of behaviour, except for some
> very specific single-tab problems, for example.
> 
> 
> Actual results:
> 
> Excessive CPU usage slows FF to a crawl, and interferes with the machine
> generally, by causing slow responses.  
> 
> 
> Expected results:
> 
> Memory clean-up from time to time would not be a problem, but locked in a
> seemingly unending cycle is not functional.
(Reporter)

Comment 156

10 months ago
Just to update, I have not encountered any problems for quite some time now.  e10s seems to work well (8 cores all busy), I have not suffered any FF-related slowdowns, that I have noticed, at any rate.  My cache is now mostly in memory (and as RAMdisk as well).  The only limitation now seems to be the (lack of) responsiveness of remote servers.  I have many tabs open.

I had forgotten about this as the problems faded.  I suppose this bug can be closed as resolved, although the resolution was just part of the (extra!)ordinary development of FF - for which I am grateful (thanks all ... ).  Perhaps the disabling of incompatible add-ons played a part.

BWD
You need to log in before you can comment on or make changes to this bug.