Closed Bug 763868 Opened 12 years ago Closed 11 years ago

Periodic hangs - CC > 3000 ms (due to zombie compartments?)

Categories

(Firefox :: General, defect)

13 Branch
x86
Windows XP
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: karl156, Assigned: mccr8)

References

(Depends on 1 open bug)

Details

User Agent: Mozilla/5.0 (Windows NT 5.1; rv:13.0) Gecko/20100101 Firefox/13.0 Build ID: 20120601045813 Steps to reproduce: As advised in bug 490122#c343 here my bug report about periodic hangs in Firefox. The last 8 Log entries (from error console): CC(T+5.4) duration: 3344ms, suspected: 782, visited: 358456 RCed and 2296053 GCed, collected: 530 RCed and 490 GCed (1020 waiting for GC) ForgetSkippable 7 times before CC, min: 0 ms, max: 94 ms, avg: 15 ms, total: 109 ms, removed: 1350 ---------- GC(T+10.1) TotalTime: 626,4ms, Type: global, MMU(20ms): 0%, MMU(50ms): 0%, Reason: CC_WAITING, NonIncrementalReason: GC mode, +chunks: 0, -chunks: 0 mark: 525,2, mark-roots: 31,7, mark-other: 382,2, sweep: 98,1, sweep-obj: 8,4, sweep-string: 0,4, sweep-script: 2,8, sweep-shape: 12,6, discard-code: 13,5, discard-analysis: 17,8, xpconnect: 14,7, deallocate: 0,1 ---------- GC(T+12.8) TotalTime: 616,6ms, Type: global, MMU(20ms): 0%, MMU(50ms): 0%, Reason: CC_FORCED, NonIncrementalReason: requested, +chunks: 0, -chunks: 0 mark: 519,3, mark-roots: 31,0, mark-other: 376,2, sweep: 94,9, sweep-obj: 8,2, sweep-string: 0,4, sweep-script: 2,8, sweep-shape: 12,5, discard-code: 13,7, discard-analysis: 18,2, xpconnect: 14,6, deallocate: 0,2 ---------- CC(T+15.5) duration: 3328ms, suspected: 264, visited: 357214 RCed and 2295596 GCed, collected: 146 RCed and 63 GCed (209 waiting for GC) ForgetSkippable 2 times before CC, min: 0 ms, max: 78 ms, avg: 39 ms, total: 78 ms, removed: 328 ---------- CC(T+46.6) duration: 2641ms, suspected: 403, visited: 350535 RCed and 2262682 GCed, collected: 30 RCed and 0 GCed (239 waiting for GC) ForgetSkippable 21 times before CC, min: 0 ms, max: 16 ms, avg: 2 ms, total: 47 ms, removed: 3058 ---------- CC(T+94.6) duration: 2656ms, suspected: 359, visited: 350335 RCed and 2262661 GCed, collected: 0 RCed and 0 GCed (239 waiting for GC) ForgetSkippable 26 times before CC, min: 0 ms, max: 16 ms, avg: 0 ms, total: 16 ms, removed: 2713 ---------- CC(T+103.3) duration: 2656ms, suspected: 428, visited: 350361 RCed and 2262641 GCed, collected: 0 RCed and 0 GCed (239 waiting for GC) ForgetSkippable 5 times before CC, min: 0 ms, max: 0 ms, avg: 0 ms, total: 0 ms, removed: 459 ---------- CC(T+129.0) duration: 2719ms, suspected: 400, visited: 350215 RCed and 2262138 GCed, collected: 8 RCed and 0 GCed (247 waiting for GC) ForgetSkippable 40 times before CC, min: 0 ms, max: 16 ms, avg: 1 ms, total: 62 ms, removed: 7873 Other Measurements 13,462,324 B ── canvas-2d-pixel-bytes 1,062,083,442 B ── explicit 0 B ── gfx-d2d-surfacecache 0 B ── gfx-d2d-surfacevram 4,928 B ── gfx-surface-image 15,452,472 B ── gfx-surface-win32 758,538,436 B ── heap-allocated 787,902,464 B ── heap-committed 3.72% ── heap-committed-fragmentation 3,125,248 B ── heap-dirty 52,009,818 B ── heap-unallocated 2,191,228 B ── images-content-used-uncompressed 7 ── js-compartments-system 55 ── js-compartments-user 292,552,704 B ── js-gc-heap 7,574,256 B ── js-main-runtime-analysis-temporary 32,243,240 B ── js-main-runtime-gc-heap-arena-unused 0 B ── js-main-runtime-gc-heap-chunk-clean-unused 0 B ── js-main-runtime-gc-heap-chunk-dirty-unused 16,961,536 B ── js-main-runtime-gc-heap-decommitted 0.08% ── js-main-runtime-gc-heap-unused-fraction 7,321,728 B ── js-main-runtime-mjit 163,822,072 B ── js-main-runtime-objects 181,396,592 B ── js-main-runtime-scripts 118,469,864 B ── js-main-runtime-shapes 25,664,114 B ── js-main-runtime-strings 7,755,440 B ── js-main-runtime-type-inference 0 ── low-commit-space-events 0 ── low-memory-events-physical 0 ── low-memory-events-virtual 1,148,592,128 B ── private 1,158,930,432 B ── resident 19,188,064 B ── storage-sqlite 1,309,847,552 B ── vsize 69,603,242 B ── window-objects-dom 69,791,600 B ── window-objects-layout-arenas 19,982,832 B ── window-objects-layout-style-sets 114,688 B ── window-objects-layout-text-runs 29,208,736 B ── window-objects-style-sheets
Blocks: 490122
Component: Untriaged → General
Version: 10 Branch → 13 Branch
Do you have any addons? Can you reproduce the problem if you disabled all the addons? Have you tried Nightly. Does it behave better?
Thanks for your fast answer. I am sure the hangs will be much less a problem with all addons disabled. But the main reason for using Firefox is the availability of such addons. At the moment (and the last months) I have to restart Firefox approximately twice a day. The snapshot above is taken after almost 11 hours uptime. After starting Firefox it uses ~600 MB RAM and is much more responsive. It feels like its getting slower with every new tab/window opened (and closed), but this feeling is more a guess. I have not tried nightly with my main browsing workspace yet (addon compatibility etc). I am also not sure if this will help, because this problem (slow responsiveness, high memory consumption) exists at least since Firefox 4. But if you think that there is any noteworthy improvement since Firefox 13, I will give it a try.
See also bug 646941, where I'm waiting for another occurrence to provide further debug info. Perhaps mark as duplicate of that?
Status: UNCONFIRMED → NEW
Ever confirmed: true
(In reply to karl155 from comment #2) > Thanks for your fast answer. I am sure the hangs will be much less a problem > with all addons disabled. But the main reason for using Firefox is the > availability of such addons. It would be helpful to know which addon causes the problems. Hopefully the addon developer can fix the issue. > At the moment (and the last months) I have to restart Firefox approximately > twice a day. That is not good. I can keep Firefox open for weeks are median CC time is around 10ms. > The snapshot above is taken after almost 11 hours uptime. > After starting Firefox it uses ~600 MB RAM and is much more responsive. It > feels like its getting slower with every new tab/window opened (and closed), > but this feeling is more a guess. That is very possible, especially if you have some addon which does something to each tab. > > I have not tried nightly with my main browsing workspace yet (addon > compatibility etc). I am also not sure if this will help, because this > problem (slow responsiveness, high memory consumption) exists at least since > Firefox 4. But if you think that there is any noteworthy improvement since > Firefox 13, I will give it a try. There is a major change in Nightly, well, actually even in Aurora related to addons handling. See Bug 695480. https://bugzilla.mozilla.org/show_bug.cgi?id=695480#c50 lists few addons which leak without that patch.
(In reply to Olli Pettay [:smaug] from comment #4) > That is not good. I can keep Firefox open for weeks are median CC time is > around 10ms. s/are/and/
If you can reproduce the problem easily, could you perhaps create a cc log. See https://bugzilla.mozilla.org/show_bug.cgi?id=646941#c18 (Note, cc log may contain some visited URLs)
What does your about:compartments page say? Odds are that one of your addons is leaking pages.
(when you are experiencing the problem, of course)
(In reply to Olli Pettay [:smaug] from comment #4) > There is a major change in Nightly, well, actually even in Aurora related to > addons handling. I just switched to Nightly (BuildID 20120612030527). Lets see what will happen the next few hours. (In reply to Andrew McCreight [:mccr8] from comment #7) > What does your about:compartments page say? Odds are that one of your > addons is leaking pages. I went to about:compartments in Firefox 13 yesterday or so. Nothing unusual. Nightly also looks good for now. When I open a new tab, a new compartment is created. When I close the tab, the compartment disappears a few seconds later. I will give you an update tomorrow. (Now: Resident 609 MB, GC 390ms, CC 463ms)
Something weired happened. Without restarting Nightly (see last comment) CC times fell to 16ms. (Resident 1069 MB, GC 424ms, CC 16ms) The browser still feels "fresh" (like just restarted). Only the about:blank usage looks a little odd to me. Does it really mean, that a blank page consumes 300 MB RAM? about:compartments System Compartments [System Principal], about:blank [78] about:memory 1,007,514,657 B (100.0%) -- explicit ├────586,763,026 B (58.24%) -- js │ ├──303,591,016 B (30.13%) -- compartment([System Principal], about:blank) │ │ ├──165,494,784 B (16.43%) -- gc-heap │ │ │ ├───57,136,760 B (05.67%) -- arena │ │ │ │ ├──56,282,632 B (05.59%) ── unused [78] │ │ │ │ ├─────478,912 B (00.05%) ── headers [38] │ │ │ │ └─────375,216 B (00.04%) ── padding [37] │ │ │ ├───48,258,752 B (04.79%) -- objects │ │ │ │ ├──29,397,488 B (02.92%) ── non-function [75] │ │ │ │ └──18,861,264 B (01.87%) ── function [76] │ │ │ ├───31,692,872 B (03.15%) -- shapes │ │ │ │ ├──23,439,984 B (02.33%) ── tree [76] │ │ │ │ ├───4,410,072 B (00.44%) ── dict [75] │ │ │ │ └───3,842,816 B (00.38%) ── base [75] │ │ │ ├───25,710,624 B (02.55%) ── scripts [75] │ │ │ ├────1,984,656 B (00.20%) ── strings [38] │ │ │ ├──────668,784 B (00.07%) ── sundries [77] │ │ │ └───────42,336 B (00.00%) ── type-objects [2] │ │ ├───92,651,256 B (09.20%) ── script-data [75] │ │ ├───22,817,248 B (02.26%) -- shapes-extra │ │ │ ├──13,618,432 B (01.35%) ── tree-tables [75] │ │ │ ├───4,530,368 B (00.45%) ── tree-shape-kids [75] │ │ │ ├───3,055,616 B (00.30%) ── compartment-tables [75] │ │ │ └───1,612,832 B (00.16%) ── dict-tables [74] │ │ ├───17,651,488 B (01.75%) -- objects │ │ │ ├──15,542,976 B (01.54%) ── slots [76] │ │ │ └───2,108,512 B (00.21%) ── elements [37] │ │ ├────1,664,904 B (00.17%) ── string-chars [39] │ │ ├────1,441,792 B (00.14%) ── analysis-temporary [2] │ │ ├──────730,320 B (00.07%) ── type-inference/script-main [74] │ │ ├──────704,512 B (00.07%) ── cross-compartment-wrappers [50] │ │ ├──────334,520 B (00.03%) ── other-sundries [78] │ │ └──────100,192 B (00.01%) ── mjit-data [2] 16,138,324 B ── canvas-2d-pixel-bytes 1,007,522,949 B ── explicit 0 B ── gfx-d2d-surfacecache 0 B ── gfx-d2d-surfacevram 4,880 B ── gfx-surface-image 18,655,843 B ── gfx-surface-win32 0 ── ghost-windows 660,178,772 B ── heap-allocated 696,168,448 B ── heap-committed 35,974,030 B ── heap-committed-unused 5.44% ── heap-committed-unused-ratio 3,641,344 B ── heap-dirty 66,477,178 B ── heap-unused 2,200,479 B ── images-content-used-uncompressed 475 ── js-compartments-system 425 ── js-compartments-user 342,884,352 B ── js-gc-heap 17,638,544 B ── js-main-runtime-analysis-temporary 185,020,256 B ── js-main-runtime-gc-heap-allocated 106,107,040 B ── js-main-runtime-gc-heap-arena-unused 5,242,880 B ── js-main-runtime-gc-heap-chunk-clean-unused 22,134,784 B ── js-main-runtime-gc-heap-chunk-dirty-unused 318,504,960 B ── js-main-runtime-gc-heap-committed 133,484,704 B ── js-main-runtime-gc-heap-committed-unused 72.14% ── js-main-runtime-gc-heap-committed-unused-ratio 24,379,392 B ── js-main-runtime-gc-heap-decommitted 494,424 B ── js-main-runtime-mjit 116,278,592 B ── js-main-runtime-objects 158,921,320 B ── js-main-runtime-scripts 82,686,784 B ── js-main-runtime-shapes 16,250,418 B ── js-main-runtime-strings 6,803,264 B ── js-main-runtime-type-inference 0 ── low-commit-space-events 0 ── low-memory-events-physical 0 ── low-memory-events-virtual 1,103,884,288 B ── private 1,107,111,936 B ── resident 33,988,584 B ── storage-sqlite 1,252,732,928 B ── vsize 0 B ── window-objects-dom-cdata-nodes 143,512 B ── window-objects-dom-comment-nodes 52,869,788 B ── window-objects-dom-element-nodes 2,193,712 B ── window-objects-dom-other 17,693,974 B ── window-objects-dom-text-nodes 66,959,552 B ── window-objects-layout-arenas 3,712,080 B ── window-objects-layout-pres-contexts 19,597,776 B ── window-objects-layout-style-sets 118,880 B ── window-objects-layout-text-runs 29,451,256 B ── window-objects-style-sheets
Looks like something is leaking system compartments. The number of about:blank compartments even raised after my last comment. (Resident: 1122 MB, GC 456ms, CC 31ms) Following tabs are really opened (among others) about:addons (once) about:memory (once) about:memory?verbose (once) about:compartments?verbose (once) Output of about:compartments Main Process User Compartments about:blank (...) System Compartments [System Principal], about:addons [2] [System Principal], about:blank [86] [System Principal], about:compartments [4] [System Principal], about:compartments?verbose [System Principal], about:memory [System Principal], about:memory?verbose (...) Ghost Windows (None)
> Something weired happened. Without restarting Nightly (see last comment) CC times > fell to 16ms. (Resident 1069 MB, GC 424ms, CC 16ms) > The browser still feels "fresh" (like just restarted). Probably what happened there is that there was some page that was leaking, slowing down the CC, that eventually was freed up. In the future if you get slow CCs again, you could take a CC dump and post it here or email it to me. It will contain URLs of pages you have open, or possibly those you have opened earlier in the session, if those are leaking: https://wiki.mozilla.org/Performance:Leak_Tools#Cycle_collector_heap_dump The about:blank problem is probably something different. See bug 759581. If you could post what you are seeing there, that would be useful, along with addons.
(In reply to Andrew McCreight [:mccr8] from comment #12) > Probably what happened there is that there was some page that was leaking, > slowing down the CC, that eventually was freed up. I went through my history of the last days and I saw some CC spikes (>1000ms) on this page: http://v6launch.ripe.net/cgi-bin/index.cgi After closing it again, it remains as zombie compartment even after minutes. about:compartments User Compartments http://v6launch.ripe.net/cgi-bin/index.cgi http://v6launch.ripe.net/cgi-bin/index.cgi, about:blank [3] about:memory 1,262,084,630 B (100.0%) -- explicit ├────758,434,792 B (60.09%) -- js │ ├──410,241,596 B (32.51%) -- compartment([System Principal], about:blank) │ ├───28,869,944 B (02.29%) -- compartment(http://v6launch.ripe.net/cgi-bin/index.cgi, about:blank) │ │ ├──23,486,464 B (01.86%) -- gc-heap │ │ │ ├──11,115,256 B (00.88%) -- arena │ │ │ │ ├──11,000,696 B (00.87%) ── unused [3] │ │ │ │ ├──────91,744 B (00.01%) ── headers [3] │ │ │ │ └──────22,816 B (00.00%) ── padding │ │ │ ├──10,879,648 B (00.86%) -- objects │ │ │ │ ├──10,225,056 B (00.81%) ── non-function [3] │ │ │ │ └─────654,592 B (00.05%) ── function [3] │ │ │ ├─────824,040 B (00.07%) -- shapes │ │ │ │ ├──367,656 B (00.03%) ── tree [3] │ │ │ │ ├──344,736 B (00.03%) ── dict [3] │ │ │ │ └──111,648 B (00.01%) ── base [3] │ │ │ ├─────465,024 B (00.04%) ── scripts [3] │ │ │ ├─────159,808 B (00.01%) ── type-objects [3] │ │ │ ├──────22,848 B (00.00%) ── sundries [3] │ │ │ └──────19,840 B (00.00%) ── strings │ │ ├───2,827,840 B (00.22%) -- objects │ │ │ ├──1,480,256 B (00.12%) ── elements [2] │ │ │ └──1,347,584 B (00.11%) ── slots [3] │ │ ├───1,548,952 B (00.12%) ── script-data [3] │ │ ├─────534,528 B (00.04%) -- shapes-extra │ │ │ ├──214,368 B (00.02%) ── tree-tables [3] │ │ │ ├──167,552 B (00.01%) ── dict-tables [3] │ │ │ ├───79,872 B (00.01%) ── compartment-tables [3] │ │ │ └───72,736 B (00.01%) ── tree-shape-kids [3] │ │ ├─────299,680 B (00.02%) ── analysis-temporary [3] │ │ ├─────136,096 B (00.01%) -- type-inference │ │ │ ├───96,800 B (00.01%) ── object-main [3] │ │ │ └───39,296 B (00.00%) ── tables [3] │ │ ├──────25,840 B (00.00%) ── string-chars │ │ └──────10,544 B (00.00%) ── other-sundries [3]
Now CC times went to >1000 ms again. I made a CC dump as requested in comment 12. https://hotfile.com/dl/159783939/6e78ea5/cc-edges.zip.html (9,4 MB unpacked) (not uploaded as attachment because it may contain private informations) If you need to know whats behind a masked URL, just let me know the line number. Something interesting from the dump: Counts how often an URL was found in the log 5923 http://v6launch.ripe.net/cgi-bin/http.cgi?embedded=1 5249 chrome://browser/content/browser.xul 1135 chrome://console2/content/console2.xul 360 https://developer.mozilla.org/en/CSS/-moz-image-rect 315 chrome://firebug/content/firefox/firebugFrame.xul 256 https://developer.mozilla.org/En/CSS/Text-overflow 195 https://developer.mozilla.org/en/css/-moz-border-image (...) from error console: Warning: Error: Failed to preserve wrapper of wrapped native weak map key. Source file: chrome://browser/content/places/treeView.js Line: 1215 ---------- Error: Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [nsIDOMWindow.content] ---------- CC(T+154963.9) duration: 32ms, suspected: 1029, visited: 10363 RCed and 13455 GCed, collected: 13 RCed and 0 GCed (36 waiting for GC) ForgetSkippable 16 times before CC, min: 0 ms, max: 0 ms, avg: 0 ms, total: 0 ms, removed: 11543 ---------- GC(T+154967.0) Total Time: 632,6ms, Compartments Collected: 745, Total Compartments: 745, MMU (20ms): 0%, MMU (50ms): 0%, Reason: SET_NEW_DOCUMENT, Nonincremental Reason: requested, Allocated: 387MB, +Chunks: 0, -Chunks: 0 Totals: Purge: 0,3ms, Mark: 471,5ms, Mark Roots: 32,7ms, Mark Other: 23,1ms, Finalize Start Callback: 2,2ms, Sweep: 150,5ms, Sweep Compartments: 68,5ms, Sweep Object: 17,6ms, Sweep String: 5,7ms, Sweep Script: 4,8ms, Sweep Shape: 25,7ms, Discard Code: 16,2ms, Discard Analysis: 23,9ms, Discard TI: 1,9ms, Sweep Types: 10,6ms, Clear Script Analysis: 4,4ms, Finalize End Callback: 19,9ms, Deallocate: 0,2ms, End Callback: 9,8ms ---------- GC(T+154971.3) Total Time: 585,7ms, Compartments Collected: 746, Total Compartments: 746, MMU (20ms): 0%, MMU (50ms): 0%, Reason: CC_FORCED, Nonincremental Reason: requested, Allocated: 379MB, +Chunks: 0, -Chunks: 0 Totals: Mark: 472,5ms, Mark Roots: 31,5ms, Mark Other: 23,2ms, Finalize Start Callback: 2,2ms, Sweep: 112,5ms, Sweep Compartments: 55,7ms, Sweep Object: 3,8ms, Sweep String: 0,4ms, Sweep Script: 4,3ms, Sweep Shape: 22,5ms, Discard Code: 15,0ms, Discard Analysis: 18,9ms, Discard TI: 0,9ms, Sweep Types: 9,4ms, Clear Script Analysis: 4,3ms, Finalize End Callback: 17,5ms, Deallocate: 0,4ms, End Callback: 0,4ms ---------- CC(T+154971.4) duration: 719ms, suspected: 664, visited: 35146 RCed and 76113 GCed, collected: 129 RCed and 9 GCed (138 waiting for GC) ForgetSkippable 8 times before CC, min: 0 ms, max: 46 ms, avg: 9 ms, total: 77 ms, removed: 2344 ---------- GC(T+154977.8) Total Time: 1197,8ms, Compartments Collected: 686, Total Compartments: 686, MMU (20ms): 0%, MMU (50ms): 0%, Reason: CC_FORCED, Nonincremental Reason: requested, Allocated: 379MB, +Chunks: 0, -Chunks: 0 Totals: Begin Callback: 0,1ms, Wait Background Thread: 32,0ms, Purge: 0,1ms, Mark: 942,7ms, Mark Roots: 64,2ms, Mark Other: 46,4ms, Finalize Start Callback: 4,3ms, Sweep: 221,7ms, Sweep Compartments: 109,0ms, Sweep Object: 7,6ms, Sweep String: 0,9ms, Sweep Script: 8,7ms, Sweep Shape: 45,1ms, Discard Code: 29,3ms, Discard Analysis: 36,7ms, Discard TI: 1,8ms, Sweep Types: 18,6ms, Clear Script Analysis: 8,5ms, Finalize End Callback: 34,4ms, Deallocate: 0,5ms, End Callback: 0,6ms ---------- ....cc-edges-2.3904.log ---------- CC(T+154978.4) duration: 1812ms, suspected: 1399, visited: 35521 RCed and 76434 GCed, collected: 51 RCed and 316 GCed (367 waiting for GC) ForgetSkippable 6 times before CC, min: 0 ms, max: 31 ms, avg: 10 ms, total: 62 ms, removed: 451 ---------- Warning: Error: A long running script was terminated Source file: chrome://browser/content/places/treeView.js Line: 1279 ---------- GC(T+154983.0) Total Time: 591,0ms, Compartments Collected: 682, Total Compartments: 682, MMU (20ms): 0%, MMU (50ms): 0%, Reason: CC_WAITING, Nonincremental Reason: requested, Allocated: 379MB, +Chunks: 0, -Chunks: 0 Totals: Mark: 476,7ms, Mark Roots: 31,7ms, Mark Other: 23,2ms, Finalize Start Callback: 2,1ms, Sweep: 113,3ms, Sweep Compartments: 54,6ms, Sweep Object: 4,2ms, Sweep String: 0,8ms, Sweep Script: 4,3ms, Sweep Shape: 22,8ms, Discard Code: 15,0ms, Discard Analysis: 18,3ms, Discard TI: 0,9ms, Sweep Types: 9,3ms, Clear Script Analysis: 4,2ms, Finalize End Callback: 18,5ms, Deallocate: 0,2ms, End Callback: 0,5ms ---------- GC(T+154985.9) Total Time: 1206,3ms, Compartments Collected: 684, Total Compartments: 684, MMU (20ms): 0%, MMU (50ms): 0%, Reason: CC_FORCED, Nonincremental Reason: requested, Allocated: 379MB, +Chunks: 0, -Chunks: 0 Totals: Begin Callback: 0,1ms, Wait Background Thread: 32,2ms, Purge: 0,1ms, Mark: 950,4ms, Mark Roots: 64,2ms, Mark Other: 46,7ms, Finalize Start Callback: 4,3ms, Sweep: 222,6ms, Sweep Compartments: 109,6ms, Sweep Object: 7,3ms, Sweep String: 1,2ms, Sweep Script: 8,5ms, Sweep Shape: 44,8ms, Discard Code: 30,4ms, Discard Analysis: 36,3ms, Discard TI: 1,7ms, Sweep Types: 18,4ms, Clear Script Analysis: 8,5ms, Finalize End Callback: 35,1ms, Deallocate: 0,4ms, End Callback: 0,4ms ---------- CC(T+154986.1) duration: 1360ms, suspected: 317, visited: 34863 RCed and 76174 GCed, collected: 0 RCed and 0 GCed (0 waiting for GC) ForgetSkippable 1 times before CC, min: 31 ms, max: 31 ms, avg: 31 ms, total: 31 ms, removed: 424
Ah, you have firebug. I think it is still leaking.
As far as performance is concerned, don't Firebug's leaks have the exact same effect as simply having a lot more tabs open? In other words, does it matter that it's Firebug, or could it just as well have been a bunch of open tabs causing slow CC?
(In reply to Roman from comment #16) > As far as performance is concerned, don't Firebug's leaks have the exact > same effect as simply having a lot more tabs open? No. Normally one tabs adds very tiny bit of stuff to the cycle collection graph. We're pretty good at optimizing stuff out from the graph. > In other words, does it > matter that it's Firebug, or could it just as well have been a bunch of open > tabs causing slow CC? IIRC, Firebug (and some other addons) does still cause runtime leaks, at least in FF13. I.e. keeps some documents alive too long in a way which is hard for optimize out in cycle collector. That causes bad CC times.
(In reply to Olli Pettay [:smaug] from comment #17) > (In reply to Roman from comment #16) > IIRC, Firebug (and some other addons) does still cause runtime leaks, at > least in FF13. I.e. keeps > some documents alive too long in a way which is hard for optimize out in > cycle collector. > That causes bad CC times. The dump was posted from Nightly. (I am running Nightly since comment 9) User Agent: Mozilla/5.0 (Windows NT 5.1; rv:16.0) Gecko/16.0 Firefox/16.0a1 Build ID: 20120612030527 Maybe interesting: I did not have Firebug open in this browsing session. Before starting Nightly, I deleted the Firebug folder from my profile, so Firebug is disabled for all sites. Do you think Firebug is responsible for the lots of about:blank compartments? (see comment 11) Now: about:compartments [System Principal], about:blank [121] about:memory 1,318.86 MB (100.0%) -- explicit ├────802.89 MB (60.88%) -- js │ ├──475.25 MB (36.04%) -- compartment([System Principal], about:blank)
It looks like there are a couple of problems here. First off, that IPv6 page seems to have something we don't optimize well. I'm getting CC times like this, which just that page open: CC(T+232.7) duration: 131ms, suspected: 215, visited: 10102 RCed and 47470 GCed, collected: 461 RCed and 2010 GCed (2471 waiting for GC) ForgetSkippable 18 times before CC, min: 0 ms, max: 12 ms, avg: 2 ms, total: 37 ms, removed: 2449 Initially, my CC times were okay, but at a certain point they started getting long. The second problem you are seeing is that the page isn't being cleaned up properly. I'm not seeing that locally. Maybe it is an addon? If you are using Nightly, it shouldn't be possible for an addon to do that, but maybe not. The IPv6 page is really huge, so I'm not surprised that you are seeing long CC times when it ends up as a zombie. I don't think Firebug is causing the about:blank, as we've seen other people having that problem.
I'll file a new bug for the http://v6launch.ripe.net/cgi-bin/index.cgi problem, as that seems like a nicely contained one. We can leave this bug for figuring out the source of your zombie compartments.
Summary: Periodic hangs - CC > 3000 ms → Periodic hangs - CC > 3000 ms (due to zombie compartments?)
Depends on: 764899
Assignee: nobody → continuation
I don't think there's anything actionable here, aside from bug 764899.
Status: NEW → RESOLVED
Closed: 11 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.