Support Forum discussions

[Resolved in Fx50] Facebook users, ideal problem to look at on Sumo Day

  • 8 Replies
  • Last reply by John99
  1. John99
    John99 3665 posts
    more options

    (I had initially posted this within the main Sumo day thread https://support.mozilla.org/en-US/forums/contributors/712157#post-70235 )


    Hi Everyone, Any of you with a Facebook account and able to try to reproduce a crash bug PLEASE ? Someone Bob has made scores of posts asking for our help and we have failed him so far it would help to get some more eyes on this.

    It is a simple thing to test it will take a few repeat key presses with some delay between them I think it would need 5-20 minutes of your time to test. No need to wait for Sumo day, try in advance if you wish.

    The Test I would Like you to try

    1. Open a facebook page and sign into facebook.com
    2. Open task manager (Ctrl+Shift+ Esc) and use the always on top option
    3. Open a new tab and use the newsfeed address https://www.facebook.com/?sk=nf
    4. Whilst tab https://www.facebook.com/?sk=nf has focus press the Keyboard Page down key repeatedly, until either Firefox Crashes or a count of 200 presses. Noting to pause keypressing if the page is still loading and the throbber spinning. 
    N.B. Nearly all of you will use the standard 32bit Firefox Release on Windows - there are no crashes on 64bit Firefox (No idea if Mac s & Linux 32bit Firefox will also crash) .

    For Info MY RESULT Firefox does not crash. I noted Firefox is using over 1GB memory per Task Manager and it is in total showing 3 of 3.5GB used 86% After something in the region of 40 keypresses I seem to have reached the end of the Newsfeed as it no longer scrolls down. Firefox still also seems responsive. I am able to post on this forum, and view video clips from the Facebook newsfeed.

    Boring Background details Explained slightly less briefly someone has multiple crash signatures and is able to very reliably crash Firefox. I and at least one other - Alex - have tried and not reproduced this. Bob has now reproduced it on three or four of his own PCs - differing Windows OS versions - and on another one he borrowed, also different Firefox versions and Channels. Possibly part of the problem is you need a lot of facebook friends so as to scroll through a large news section.

    Many you you will use Facebook a lot.

    Bob has been trying for a long while to prove and prevent this and has a very long thread as a result. He has repeatedly asked if we will test this out ourselves, but only the two of us have done so and reported back - we failed - but in my case because I had to borrow someone elses Facebook credentials, and it was only a low usage account I probably did not expect to be able to reproduce.

    We are presumably not testing this and then reporting back because

    • Everyone knows facebook does not crash
      • Well I am not sure that is so.
        The data in crash reports that could back that up is restricted access. It probably would show a lot of Facebook crashes, but then would be explained away by saying Facebook s used a lot, but is not necessarily the cause.
        The person produced Google search data to back up his position - however that was not as conclusive as it first looked.
      • People may have tested - but not reported back if it did not crash.
    • The crash signature in question is very low frequency.
      • OK so this may be a bit of an edge case. but Bob is getting multiple crash signatures, we are only concentrating on the one. He is getting OOM crash signatures, and OOM crashes are probably the biggest category of Firefox Crashes as far as numbers of crashes go.
    • We all know facebook would not have a memory leak, (I said similar to Bob in Your Machines or Firefox ? ) we even have a developer making that assumption#c8. After all that would never happen and go unnoticed it would be like Fx4 or top addon ABP having unrecognised memory leaks ;-)

    Personally I can not help thinking the one thing Bob has not tried is Windows Safe Mode. Not sure malware has been fully ruled out, but Bob did crash someone else's computer so unlikely. It could be a RAM issue, but vanishingly unlikely - multiple machines - about memory & Task manager are showing memory increase prior to crash.

    Of course if anyone else has ideas to test out or plausible solutions we are all ears.

    (I had initially posted this within the main Sumo day thread https://support.mozilla.org/en-US/forums/contributors/712157#post-70235 ) Hi Everyone, Any of you with a Facebook account and able to try to reproduce a crash bug PLEASE ? Someone '''''Bob''''' has made scores of posts asking for our help and we have failed him so far it would help to get some more eyes on this. It is a simple thing to test it will take a few repeat key presses with some delay between them I think it would need 5-20 minutes of your time to test. No need to wait for Sumo day, try in advance if you wish. '''The Test I would Like you to try''' #Open a facebook page and sign into facebook.com #Open task manager (Ctrl+Shift+ Esc) and use the always on top option #Open a new tab and use the newsfeed address https://www.facebook.com/?sk=nf #Whilst tab https://www.facebook.com/?sk=nf has focus press the Keyboard Page down key repeatedly, until either Firefox Crashes or a count of 200 presses. Noting to pause keypressing if the page is still loading and the throbber spinning. <sub>N.B. Nearly all of you will use the standard 32bit Firefox Release on Windows - there are no crashes on 64bit Firefox (No idea if Mac s & Linux 32bit Firefox will also crash) </sub>. <sub>[https://support.mozilla.org/en-US/questions/1123864?page=8#answer-908694 For Info] MY RESULT Firefox does not crash. I noted Firefox is using over 1GB memory per Task Manager and it is in total showing 3 of 3.5GB used 86% After something in the region of 40 keypresses I seem to have reached the end of the Newsfeed as it no longer scrolls down. Firefox still also seems responsive. I am able to post on this forum, and view video clips from the Facebook newsfeed. </sub> '''Boring Background details''' Explained slightly less briefly someone has multiple crash signatures and is able to very reliably crash Firefox. I and at least one other - Alex - have tried and not reproduced this. Bob has now reproduced it on three or four of his own PCs - differing Windows OS versions - and on another one he borrowed, also different Firefox versions and Channels. Possibly part of the problem is you need a lot of facebook friends so as to scroll through a large news section. Many you you will use Facebook a lot. Bob has been trying for a long while to prove and prevent this and has a very long thread as a result. He has repeatedly asked if we will test this out ourselves, but only the two of us have done so and reported back - we failed - but in my case because I had to borrow someone elses Facebook credentials, and it was only a low usage account I probably did not expect to be able to reproduce. We are presumably not testing this and then reporting back because * Everyone knows facebook does not crash ** Well I am not sure that is so. <br/>The data in crash reports that could back that up is restricted access. It probably would show a lot of Facebook crashes, but then would be explained away by saying Facebook s used a lot, but is not necessarily the cause. <br/>The person produced Google search data to back up his position - however that was not as conclusive as it first looked. ** People may have tested - but not reported back if it did not crash. * The crash signature in question is very low frequency. ** OK so this may be a bit of an edge case. but Bob is getting multiple crash signatures, we are only concentrating on the one. He is getting OOM crash signatures, and OOM crashes are probably the biggest category of Firefox Crashes as far as numbers of crashes go. * We all know facebook would not have a memory leak, (I said similar to Bob in ''[https://support.mozilla.org/en-US/questions/1123864?page=4#answer-902785 Your Machines or Firefox ?]'' ) we even have a developer making that assumption<sup>[https://bugzilla.mozilla.org/show_bug.cgi?id=1291294#c8 #c8]</sup>. After all that would never happen and go unnoticed it would be like Fx4 or top addon ABP having unrecognised memory leaks ;-) Personally I can not help thinking the one thing Bob has not tried is Windows Safe Mode. Not sure malware has been fully ruled out, but Bob did crash someone else's computer so unlikely. It could be a RAM issue, but vanishingly unlikely - multiple machines - about memory & Task manager are showing memory increase prior to crash. '''Of course if anyone else has ideas to test out or plausible solutions we are all ears.''' * Original thread, way too long to bother reading, unless you really do want to try to spend a long time assisting Bob: <br/> [/questions/1123864 Firefox crashes daily in plugin container on Facebook] * #c7 [https://bugzilla.mozilla.org/show_bug.cgi?id=1291294#c7 Bug 1291294 - Crash in OOM | large | mozalloc_abort | mozalloc_handle_oom | moz_xrealloc | NS_CycleCollectorSuspect3 ] with STR
  2. jscher2000 - Support Volunteer
    jscher2000 - Support Volunteer 833 posts
    more options

    I haven't performed that test myself. Would I need to disable NoScript and allow Flash?

    The Facebook news feed does consume a huge amount of memory in Firefox; that doesn't surprise me. If you spend 30 minutes loading more and more content into a single tab, that is to be expected. I don't know whether the speed of loading makes a difference. I generally use the keyboard shortcuts to skip up and down from story to story (j for up and k for down, if memory serves) rather than the arrow keys or PgUp/PgDown.

    I haven't performed that test myself. Would I need to disable NoScript and allow Flash? The Facebook news feed does consume a huge amount of memory in Firefox; that doesn't surprise me. If you spend 30 minutes loading more and more content into a single tab, that is to be expected. I don't know whether the speed of loading makes a difference. I generally use the keyboard shortcuts to skip up and down from story to story (j for up and k for down, if memory serves) rather than the arrow keys or PgUp/PgDown.
  3. Andrew
    Andrew 1232 posts
    more options

    I'll try to reproduce this on my Windows 10 laptop.

    Has Bob updated their Firefox version since?

    I also can't tell if add-ons are playing a factor into this, but are the computers that Bob is testing on similar in build? Similarity with graphics card, add-ons, version?

    Has he also completed windows updates as well (except for Windows 10)?

    One last thing: are videos set to automatically play for Bob on Facebook?

    I'll try to reproduce this on my Windows 10 laptop. Has Bob updated their Firefox version since? I also can't tell if add-ons are playing a factor into this, but are the computers that Bob is testing on similar in build? Similarity with graphics card, add-ons, version? Has he also completed windows updates as well (except for Windows 10)? One last thing: are videos set to automatically play for Bob on Facebook?
  4. Andrew
    Andrew 1232 posts
    more options

    After running it for about 5 minutes, my CPU went up to about 38%, memory at about 36%, disk ran as high as 98%.

    Unable to reproduce a crash.

    After running it for about 5 minutes, my CPU went up to about 38%, memory at about 36%, disk ran as high as 98%. Unable to reproduce a crash.
  5. John99
    John99 3665 posts
    more options

    First of all thanks Jefferson and Andrew.

    jscher2000 said

    I haven't performed that test myself. Would I need to disable NoScript and allow Flash? The Facebook news feed does consume a huge amount of memory in Firefox; that doesn't surprise me. If you spend 30 minutes loading more and more content into a single tab, that is to be expected. I don't know whether the speed of loading makes a difference. I generally use the keyboard shortcuts to skip up and down from story to story (j for up and k for down, if memory serves) rather than the arrow keys or PgUp/PgDown.

    Do you know Is this a known issue then that has a wont fix bug for it ? I would have thought Firefox should be able to handle Bobs STR without crashing. You seem to expect you will crash and expect to crash in this situation. Could you try that and if you are able to reproduce also get about:memory anonymised logs as you reproduce.

    If you can improve the STR and confirm it that would be great. The STR above are ones I wrote after trying to reproduce and slightly different from what bob had actually said.

    My philosophy has been to suggest testing with all plugins disabled and in Firefox's safe mode. A bug1291294 was filed for Bob's issue. That has had attention from Developers including Nicholas Nethercote who ran the Mozilla Memshrink project and wrote about:memory.

    It seems reasonable to me that Firefox should be resilient enough not to Crash in this situation. If it does crash I think it is worth trying to confirm that in the bug. We then at least can get expert opinion as to whether or not this is extreme use and expected to crash Firefox. I have not yet noticed anyone other than Bob able to reproduce a crash, but until Andrew tried Myself and Alex are the only ones that had reported back that we had tried the STR.

    Bob seems able to reproduce these crashes at will on I think it is 5 different computers & various Windows versions. He has certainly tried with all plugins disabled and in safemode, as verified (e.g bp-f709a9be-8459-433b-bfe3-d02302160810 ) by some of the Crash reports. What Bob is concerned about is

    • That it crashes in day to day use.
    • If he can make it crash so easily there must be a bug in Firefox

    It seems quite plausible using script or adblockers may reduce the likely hood of crashing. Trying to disentangle what addons could be involved would be impossible in Bob's (list) day to day setup

    This has been a fairly longstanding issue for Bob over several Firefox versions, and has also been tested on Developer Edition/Aurora. I have often thought the issue would turn out to be something other than directly caused by a Firefox bug or regression.

    Bob gets High memory use but does not get crashes in 64bit Firefox, and I guess that may be understandable. What is interesting in the last few posts from Bob is that 32bit Nightly is not crashing. So it may turn out to be a regression after all that has now been fixed.

    I do not personally use facebook, otherwise I would have tried to reproduce earlier myself. When I did try to reproduce using borrowed credentials I seem to have come to the end of the Newsfeed. I wondered if a particular minimum length of newsfeed is required to crash Firefox. I could only get to 40 key presses Bob apparently gets higher than that before crashing.

    Andrew Did you get to 200 key press scrolls or did you come to the end of the newsfeed before that ? Sorry I do not normally use Facebook so probably have not written very good STR. You ask 

    are videos set to automatically play

    The video setting could be important could you try it both ways please.

    Really what I am looking for is any way to get Facebook to reproducibly crash Firefox. If that is FlashPlayer enabled or video enabled if and it is reproducible then because facebook is so widely used I am sure it is worth documenting it if Firefox crashes. However as Flashplayer is so problematic anyhow successful STR with FlashPlayer disabled probably would be much more significant. The largest number of Firefox crashes are OOM crashes. Possibly some of those even a lot of those are caused by facebook but overlooked because facebook use is so common.

    First of all thanks Jefferson and Andrew. ''jscher2000 [[#post-70278|said]]'' <blockquote> I haven't performed that test myself. Would I need to disable NoScript and allow Flash? The Facebook news feed does consume a huge amount of memory in Firefox; that doesn't surprise me. If you spend 30 minutes loading more and more content into a single tab, that is to be expected. I don't know whether the speed of loading makes a difference. I generally use the keyboard shortcuts to skip up and down from story to story (j for up and k for down, if memory serves) rather than the arrow keys or PgUp/PgDown. </blockquote> Do you know Is this a known issue then that has a wont fix bug for it ? I would have thought Firefox should be able to handle Bobs STR without crashing. You seem to expect you will crash and expect to crash in this situation. Could you try that and if you are able to reproduce also get about:memory anonymised logs as you reproduce. If you can improve the STR and confirm it that would be great. The STR above are ones I wrote after trying to reproduce and slightly different from what bob had actually said. My philosophy has been to suggest testing with all plugins disabled and in Firefox's safe mode. A [https://bugzilla.mozilla.org/show_bug.cgi?id=1291294 bug1291294] was filed for Bob's issue. That has had attention from Developers including Nicholas Nethercote who ran the Mozilla Memshrink project and wrote about:memory. It seems reasonable to me that Firefox should be resilient enough not to Crash in this situation. If it does crash I think it is worth trying to confirm that in the bug. We then at least can get expert opinion as to whether or not this is extreme use and expected to crash Firefox. I have not yet noticed anyone other than Bob able to reproduce a crash, but until Andrew tried Myself and Alex are the only ones that had reported back that we had tried the STR. Bob seems able to reproduce these crashes at will on I think it is 5 different computers & various Windows versions. He has certainly tried with all plugins disabled and in safemode, as verified (e.g bp-f709a9be-8459-433b-bfe3-d02302160810 ) by some of the Crash reports. What Bob is concerned about is * That it crashes in day to day use. * If he can make it crash so easily there must be a bug in Firefox It seems quite plausible using script or adblockers may reduce the likely hood of crashing. Trying to disentangle what addons could be involved would be impossible in Bob's ([https://support.mozilla.org/en-US/questions/1123864?page=4#answer-902698 list]) day to day setup This has been a fairly longstanding issue for Bob over several Firefox versions, and has also been tested on Developer Edition/Aurora. I have often thought the issue would turn out to be something other than directly caused by a Firefox bug or regression. Bob gets High memory use but does not get crashes in 64bit Firefox, and I guess that may be understandable. What is interesting in the last few posts from Bob is that 32bit Nightly is not crashing. So it may turn out to be a regression after all that has now been fixed. I do not personally use facebook, otherwise I would have tried to reproduce earlier myself. When I did try to reproduce using borrowed credentials I seem to have come to the end of the Newsfeed. I wondered if a particular minimum length of newsfeed is required to crash Firefox. I could only get to 40 key presses Bob apparently gets higher than that before crashing. Andrew Did you get to 200 key press scrolls or did you come to the end of the newsfeed before that ? Sorry I do not normally use Facebook so probably have not written very good STR. You ask are videos set to automatically play The video setting could be important could you try it both ways please. Really what I am looking for is any way to get Facebook to reproducibly crash Firefox. If that is FlashPlayer enabled or video enabled if and it is reproducible then because facebook is so widely used I am sure it is worth documenting it if Firefox crashes. However as Flashplayer is so problematic anyhow successful STR with FlashPlayer disabled probably would be much more significant. The largest number of Firefox crashes are OOM crashes. Possibly some of those even a lot of those are caused by facebook but overlooked because facebook use is so common.

    Modified by John99 on

  6. jscher2000 - Support Volunteer
    jscher2000 - Support Volunteer 833 posts
    more options

    John99 said

    You seem to expect you will crash and expect to crash in this situation. Could you try that and if you are able to reproduce also get about:memory anonymised logs as you reproduce.

    No, I do not expect to reproduce with those STR. I never get OOM crashes myself, perhaps because it's an 8GB RAM system. Instead, when my 32-bit Firefox gets to 2.4GB-2.7GB "commit charge," I start to get NOT RESPONDING windows and black content areas. If I'm patient and use File > Exit to shut down Firefox, then I usually get a shutdown crash after 45-60 seconds. I report those but I doubt they are relevant.

    ''John99 [[#post-70283|said]]'' <blockquote> You seem to expect you will crash and expect to crash in this situation. Could you try that and if you are able to reproduce also get about:memory anonymised logs as you reproduce.</blockquote> No, I do not expect to reproduce with those STR. I never get OOM crashes myself, perhaps because it's an 8GB RAM system. Instead, when my 32-bit Firefox gets to 2.4GB-2.7GB "commit charge," I start to get NOT RESPONDING windows and black content areas. If I'm patient and use File > Exit to shut down Firefox, then I usually get a shutdown crash after 45-60 seconds. I report those but I doubt they are relevant.
  7. Andrew
    Andrew 1232 posts
    more options

    I had a long ways left to go, Facebook did need time to grab more posts into my feed.

    Having autoplay on for videos messes with Firefox's performance and sucks an incredible amount of data. Can you check with the user to see if they have it on before I test it myself?

    Thanks.

    I had a long ways left to go, Facebook did need time to grab more posts into my feed. Having autoplay on for videos messes with Firefox's performance and sucks an incredible amount of data. Can you check with the user to see if they have it on before I test it myself? Thanks.
  8. John99
    John99 3665 posts
    more options

    Bob was able to reproduce with FlashPlayer disabled. Would tht be the same or similar to having autoplay off ? I will cross post your question to make sure Bob has seen it, he has already been made aware of this thread.

    Bob was able to reproduce with FlashPlayer disabled. Would tht be the same or similar to having autoplay off ? I will cross post your question to make sure Bob has seen it, he has already been made aware of this thread.
  9. John99
    John99 3665 posts
    more options

    Thanks for your help. It seems Bob is quite satisfied that at present Fx50 Nightly does not crash.

    Bob did reply to say he was unsure where this setting about autoplay is in Facebook

    Thanks for your help. It seems Bob is quite satisfied that at present Fx50 Nightly does not crash. Bob did reply to say he was unsure where this setting about autoplay is in Facebook