Firefox video dropouts

Rafe Needleman asks today “Why can’t they fix the Flash/Firefox bug?”, pointing to a lengthy set of Bugzilla comments about intermittent halts of audio/video streaming in some Firefox 2 and 3 installations. The problem is not yet reproducible on demand by others, and so has been difficult to address.

I don’t have the answer, but I do have some context, observations. (Warning: This is long, and the only useful info in it is how to think about problem description. If you’ve got real work to do, then don’t waste time reading my blogpost here…. πŸ˜‰

First and most important, the way to confirm that you have addressed an issue is by being able to make the problem occur on demand, and to tell others how they can also make it happen on demand. That way they can test whether they can stop it from occurring.

A “Steps to Reproduce” is not what you must do to see the problem. A “Steps to Reproduce” is what an engineer needs to do to see the problem. You may not be able to instruct others completely, and knowing what you did is certainly a first step, but a steps-to-repro description should be written from the reader’s point of view, not the writer’s.

And intermittent problems are certainly the most difficult and timeconsuming to address. We need to be able to make the problem happen with assurance, in order to assure that it has been truly removed.

Check out the comments in the thread — “I read in a forum that someone else had the problem too” — that’s not a useful comment. If a capability is not working on your system, then we all believe you, and want to improve things. There’s no need to prove that other people have it. What we need to do is to be able to see it ourselves, so that we can test whether we can make it go away successfully. You don’t need to validate yourself. We all need to explore the argument. It’s not you, it’s it. Relax, we believe you.

There’s one comment from the original poster (identified solely as “M Z”) that “No, I have not managed to reproduce it in safe mode.” This is potentially a killer bit of info. I’m assuming he means “Windows Safe Mode”, an F8-key start which disables many system customizations. If the problem actually *never* occurs with system customizations turned off, then we know to look more closely at the system customizations. But unfortunately, the description is amibiguous… might mean that he rarely tests in OS “safe mode” just as well. Tantalizing, but still less-than-fully-useful.

Someone asks “What Firefox extensions do you have installed?” and then various lists are produced. It’s more useful to know whether the problem still occurs with a stock Firefox installation. If you ever see the symptom on that system when not running any browser customizations, then we’ll have more info than knowing which brands and versions of extensions you’ve customized the browser with. Key refactoring: “Have you ever seen the problem when all Firefox extensions have been turned off?” Even one such incident would exculpate all extensions.

Someone identifying themself only as “el3000@gmail.com” offers another potentially useful bit of detail: “I encounter it only after FF has been running for a while (>60 min).” The original-poster needs to be asked whether he has ever seen the problem immediately at system/browser startup, or whether it also needs a significant period of browser use before the problem has ever appeared. If so, I would also ask them to look at their Windows Task Manager, to see how much memory their copy of Firefox is currently using. In the past, media dropout has often been associated with low-memory situations. It wouldn’t be too hard to quantify the current reports, to see whether the well-known Firefox memory consumption issues are in play when they lose audio/video.

Subsequent comments of “I have the same problem” do not help at all. No one doubts the original reporter. But these low-info confirmations just muddy the discussion, making a resolution more difficult to reach.

There are additional contributions such as “the problem is your adobe flash player version”. That’s it, no citation, no reason offered. This is a great example of why public bugbases should be scrubbed for buggy comments. The conversation may be open to this person’s participation, but he is increasing readership costs for everyone else. A group needs a smart mix of inclusion and exclusion in order to function well.

Mike Beltzner, a Mozilla staffer, makes a little progress with this comment: “Can we at least call this Windows only? I haven’t seen any reports of it happening in other OSes.” He’s trying to craft a recipe for reproduction of the problem. If we can be sure it’s Windows-only, then we’d know an engineer shouldn’t bother trying to reproduce it on a Mac. It’s a start.

Let’s switch back to Rafe’s article. It’s got the headline “Why can’t they fix the Flash/Firefox bug?” This sets off warning lights for me, because of his use of the word “the”… implies that there is only a single problem, and that it’s famous. In reality, it’s not even yet well-defined. There are also semantic issues with the rhetorical “why”, as well as the cognitive issue of knowing what the problem actually is before starting to think about whether it is possible to fix it. When I read a loaded phrase like that, I start wondering how well the writer has started thinking about what they will be writing about. Not a big flag, but a small warning flag of possible confusion ahead.

“Both Mozilla and Adobe have been aware of the issue since late May, but as yet no solution has been found.” It would be fairer to see that no way to reproduce it on demand has yet been found. The current stopping-block is in the original descriptions, not in any lack of effort by people writing code.

“One workaround solution is to install the Flash 10 player, which is still in beta.” I have no assurance that this changes the problem… in the Bugzilla talk we haven’t seen that anyone who has had the symptom has been able to make it go away by using Astro, and make it return by going back to Player 9.

Matter of fact, after reading Rafe’s article we don’t know whether he has been able to see the problem himself. I sorta suspect he might have, which would explain his interest in this one not-yet-fully-formed Bugzilla entry, but there’s zip info on his personal work at reproduction of the issue.

(Later: Yes! Down at the very bottom he says that he sees the problem too, but that he doesn’t see it in Internet Explorer. Not much more detail, but it’d be useful if he contributed to the solution.)

There’s a quote from Mike Beltzner that implies “ah, if only all Player code were published, then the problem would be easy to solve.” Baloney. [expletives deleted] Mike should know, from Mozilla’s experience with the Tamarin Project, the massive study curve that even brilliant new engineers need to do to get up to speed, to understand what is going on. And Tamarin is just one small part of the tiny engineering marvel which is Adobe Flash Player.

“He also took a minute to trumpet Mozilla’s open-source philosophy. Since Firefox’s code is open, Adobe can look at it to try to determine what is going on. But Mozilla’s team can’t look into Flash. Beltzner didn’t blame Adobe for the bug itself, but he did say that Adobe’s traditional closed software architecture is slowing down their investigation. ‘We hit a wall when it’s a closed-source solution,’ he said.”

The truth is that you simply need to distill the public complaint into an actionable item. The problem actually lies in Bugzilla’s conversational style. Right now it’s just “Oh I saw someone on a forum describe a similar thing.” You need to show engineers how they can see it. Playing the “proprietary” card instead is just weak. I’m watching my language here, but….

Comments at Webware are interesting. Too bad they close it off by registration (yeah, like I’m going to open new accounts and track new passwords for each special little site), and too bad some commenters hide their identity when commenting on others. (Tip to indy Silverlight evangelists: Including a verifiable identity will reduce the taint of possible astroturfing.) The comments section is not very useful overall, but there’s some realistic thought in there, which I appreciate, thanks.

I’m with Rafe completely on his penultimate paragraph: “Finger pointing is common in software troubleshooting, and I give both Mozilla and Adobe credit for only generally waving, not pointing, their fingers at each other. Unfortunately, neither team seems to have developers who can reproduce this issue, which just keeps the ping-pong game going.” Making the problem occur on demand is the first necessary step in making sure the problem has really gone away.

But his final paragraph seems like rankest fantasy and fairytale to me: “What I find most interesting is the way the differing philosophies of Mozilla and Adobe are slowing down resolution of this issue. If both companies were open then any developer–at Mozilla, Adobe, or elsewhere–could get into things and start experimenting to find a fix. If both companies had closed philosophies then their engineers could swear each other to the secrecy, swap source code, and together fix the issue.”

To solve the problem quickly, focus on what it is.

Summary: From the little I can see in the descriptions, I’d really want to check reporters’ system memory consumption when the problem occurs… not a sure thing, but a quick and easy diagnostic that may zero-in on the cause of the problem. (To put it gently, Firefox is rather famous for its memory issues.)

Bugzilla needs (imho) to tighten down, get rid of the conversational bloat. Doing tech support is an acquired skill, and not everyone can think directly about a problem, but a good bugbase would instruct new contributors on how to help isolate the true problem, how to describe things so that others can usefully attempt to reproduce it. Readers should not have to read through stream-of-consciousness from strangers. Refactor it, make it functional.

And finally, that line “but it’s proprietary” needs to go away. It’s a replacement for branding issues. Even Apple, the most proprietary, closed, secretive company of them all, reflexively reaches for it when they don’t know what else to say. You and I have near-zero chance to influence the W3C or Mozilla to do something — they are not more “open” in process than Adobe, or even Microsoft. “Opensource” code tweakability means more for things which run on your own machines (Linux, Apache) rather than on everybody else’s machines, when “predictability” becomes more valuable. I’m tired of conversations getting derailed when someone resorts to this weak “proprietary” tactic. Think. We need you to think. Just be honest and think. Quit the blaming and think.

And check your system memory if video stops. Not a guarantee, but it’s a start.

Update Oct 13: I’ve closed comments on this. I was amazed that some of the early comments just added anecdotal noise to the basic “you have to break it before you can fix it” principle, but more amazed these kept coming in for months later — seems pretty obvious this post has become a top Google link for “firefox video problem”, and anonymous people just want to talk without listening. They’d do better by reading, and communicating.

21 Responses to Firefox video dropouts

  1. Great response to my article! Linking it now.

  2. Emmy Huang says:

    Hi JD,
    Here is a link to an open bug with the video symptoms in the Adobe Flash Player bugbase. If people have reproducible steps and system info, QE is ready to look at it.
    https://bugs.adobe.com/jira/browse/FP-522
    There was another bug with similar symptoms that Mozilla fixed: https://bugzilla.mozilla.org/show_bug.cgi?id=429903
    best,
    Emmy Huang
    Adobe Flash Player

  3. Matt says:

    The user M Z at bugzilla was referring to safe-mode in Firefox not Windows.
    “firefox.exe -safe-mode”

  4. duke dougal says:

    Ummmm.. well I know this only adds to the vague information about this problem, but I definitely had it on Windows for a long time. It happened in Firefox but not Safari (both on Windows). I’m active in changing and reconfiguring my system and somewhere along the way it stopped being a problem. I guess this provides no valuable diagnostic info but it does verify the existence of the problem.

  5. Danny Miller says:

    Oh that bug? A friend of mine on a forum said h-
    JUST KIDDING. But I do think that bugzilla needs a revamp. People tend to post “Happens to me too” because they don’t know they can “vote” for a bug to be fixed. If people were more aware of that and a snippet said “Please don’t comment ‘This happens to me too'”

  6. Martijn says:

    Bugzilla is really an indispensable tool for Mozilla.
    Not all comments are useful, but in general people try to be useful.
    Of course people have influence! Especially people that have proven to be trustworthy contributors.

  7. Ole Lunddahl says:

    John,
    What a refreshing reminder to us all.
    Thanks for taking the time to give many examples of what you mean by that – a lengthy post but done well.
    (stick with the issue, stop the loose talk and instead document how do help the developers reproduce in an environment they can get to / set up.)
    Ole

  8. Jason says:

    First, companies outsourced tech support. Now they’re outsourcing beta testing.
    Have Adobe made any attempt to narrow down this problem in house? Are you really unable to reproduce this error in your QA department? [jd sez: Like the above couple pages of text say, we’ve been trying.] It seems relatively commonplace on Windows XP machines with Flash and Firefox installed.
    Your response was more appropriate for a bug that affects 5 people, not 5,000. I’ve read so many reports of people being affected by this bug that I think it would do Adobe well to take a more active approach into researching and reproducing the issue.
    I get the sense from your post that Adobe hasn’t tried very hard and is waiting for other people to provide them testing info rather than use its internal QA department to do some testing themselves.

  9. Rick says:

    I’ve seen the “few seconds” Flash stoppage problem reported by a handful of *Opera* users for several months now, so I think it’s more likely a Flash plugin problem than a browser problem, unless by coincidence these are separate issues. Both browsers do use the same Flash plugin.

  10. Eduardo Ramos says:

    Hi! I am quite sure I’ve had this problem under Windows Vista and Internet Explorer. Exactely the described problem. It has gone away maybe some six months ago.
    I don’t know what caused it, but it was somewhat intermitent. But I am sure it used to happen with Internet Explorer, but doesn’t hapen anymore.
    Never had it under Mac, both with Safari and Firefox.

  11. not quite dumb enough? says:

    can you “hack together” a background recording” tool (which will of course introduce it’s own factors) to maybe record what’s going wrong in these erratically occurring incidents? (personally i avoid flash, so haven’t tried it yet in ff3)

  12. Matthias Versen says:

    Bugzilla itself is great because all people can report bug reports and other people can help to get the needed informations for the developers.
    The downside is that many people think that bugzilla is a support forum and they add useless comments in bugs. Most User don’t know what a kind of information a developer needs or that a bug with the status “new” doesn’t need a comment “I see this still in Version XYZ”.
    There is no way to hide comments and some bugs are full off Spam comments.
    For this problem:
    I help in bugzilla triage bug reports but i don’t have this problem and I don’t know what they did to see this bug. There is no debug output for the flash plugin availible to see what’s wrong.
    Finding a solution for the problem would help to find out what’s causing this that’s the reason why comments about how to fix this helps. The “install flash10beta” solution doesn’t help much to find the problem of course…

  13. As a developer I often find myself in the spot of hard to reproduce faults. I really do understand the frustration when you cannot find it or reproduce it no matter how you try.
    However, when those problems occure en masse. When so many people have the same problem over and over again you have to take other actions.
    Does it really have to be on the development machine with debug mode code for you to pinpoint the problem?
    Can’t you provide some of the reporters with an action list of what to check when it frezes the next time, for example whip out Process explorer for some in depth information. Memory stacks etc, that would provide you with a lot of information.
    When it is so frequent at There is also the possibility to find someone nearby, politly ask if you can make a personal visit and checkout the problem in place. That would provide you with all the information you need to pinpoint this problem. And I for one would gladly let you play around for a few hours to see if you can find a fix.
    By the way, some information for you regarding my setup:
    – Vista 64 bit OS.
    – Running first version 10 of both the plugin and activeX flash player.
    – Problem in FireFox 3, IE 7, Opera 9
    – If the problem occurs in FF it will as well in IE
    – Restart of system will clear it up for a moment.
    – After reinstalling older 9.0.47 plugin it still happens but not as frequent restarting FF often solves it for a while.
    – If you would like more information feel free to contact me, I would be happy to have an action list to check the next time it occurs.
    Haqwin

  14. Mark says:

    Is this at all related to the glitch/setting the BBC recently found with Flash Media Server which causes certain hardware routers to reset?
    http://www.theregister.co.uk/2008/08/28/bt_router_bbc_iplayer_bug/

  15. FlakyGenius86 says:

    [jd sez: Anonymous person, abusive, still focused on proving his personal validity rather than answering the previous clear diagnostic questions. Weird. How can we help such people along?]
    –load of crap, Adobe… as always…
    -k, u want details, how’s this (not the best, but supports my/MANYs’ arguments; -will doc exact results soon) : 8 fully-updated puters here (at home, + puters I fix), incl. 3 laptops, ranging from brand new (several) to 6 yrs old; -os’s: xp home+pro, vista home premium 64 (incl. some multi-boot); -did fresh, “clean” os reinstalls (*not* Factory/oem restores), installed FF3 clean (no prior ff2) and b4 any protection (a/v, etc); -in *EVERY* case, this WELL KNOWN bug showed itself roughly 33% of the time, *especially on certain sites*… [and I’ve never seen this happen w/FF2] -IMHO, your “reproducible” excuse is (yet again) obvious Adobe GARBAGE, completely typical, and the poster who questioned Adobe’s “in-house QA” hit the (yet again, OBVIOUS) nail on the head… -just look at Adobe’s INFAMOUS track record for a clue as to why this hasn’t been fixed yet… (Micro$hit, Adobe, AOL : will go down as history’s worst programming; -and yes, I’ve been programming hardcore since age 13, now 40); ENUF EXCUSES, JUST GET IT DONE.
    (ps: if u actually post this msg, unchopped, I’ll be amazed; if u also “own up” to it, and get things FIXED, I just might be shocked into cardiac arrest, so there’s sum motivation for ya, lol)
    >;]
    (pps: -btw, I’d *LUV* for u to *try* to explain Foxit, wink wink, u know what I mean)

  16. Joeflash says:

    [jd sez: Summary is he had a recurring intermittent video problem, but only on one machine in a workgroup. The difference was not isolated and identified, although this machine also had problems installing CS3.]
    I’ve been on a project where I’ve been dealing with this bug since at least February of this year. It is only on my WinXP desktop machine, but not on my XP laptop with nearly identical configuration. And it is reproducible, in so far as it sticks around for a few weeks, maybe a month, and then goes away for a while, then comes back for no apparent reason. But while the bug is affecting my system, it seems to be on all browsers, and only for streaming video. It affects streaming video from a local installation FMS 2&3, remote FMIS3 and two CDNs I tested.
    I tracked another related bug where the Camera permissions dialog freezes the Flash player, of sorts — the buttons could not be clicked to re-enable to the SWF application. Happens to Flash and Flex-compiled projects, some of my old AS2 projects, and other open source projects where recording is enabled. I say related cause it only has ever happened on my desktop, across all browsers, just like the streaming video issue, waxing and waning with the same incident frequency of the streaming video bug. I finally nailed down that issue by discovering that it only ever happened when a NetStream connection was attempted before the user had actually given permission to the camera. After hacking around a bit I was a able to find a reasonable workaround by delaying the NetStream connection until the user had given permission, or a certain time delay had gone by. Not a fix, but good enough.
    But never found a fix nor any workaround for the streaming playback issue. Due to the elusive nature of this bug, it was with considerable embarrassment that I finally resolved the issue by coincidentally switching to another machine and completing the development of the video module. I thought I was going insane at one point — HOW CAN STREAMING VIDEO JUST NOT WORK FOR NO REASON, ON ANY BROWSER!! ARRGHHH!!! πŸ™‚ So I can perfectly understand how Adobe is having such a hard time reproducing this bug. Because it came up intermittently, it took me nearly four months to figure out that switching to another machine solved the problem, that it wasn’t my code, or some quirk in the flash player I was unaware of, or my firewall or antispyware settings, or the alignment of the planets or the price of oranges in China.
    And for the trolls in this space, if you believe Adobe is purposefully not putting in the manpower to fix this, you’ve been spending too many solitary hours in your cave. I’d no sooner believe that than fly off into conspiracy-land and chalk it up to a secret plot to overthrow Flash. We’re all in this together, so stop wasting everyone’s time.
    Consequently, and perhaps related, this desktop is the same machine that resisted all attempts to install Adobe CS3 Web Premium and AIR without an MS-CONFIG installation hack. When the problem has come up, I have not tried playing or recording video in Admin/MS-CONFIG mode, which would be interesting to try.
    I don’t have time to get into debugging log files and trading specs at the moment, but I will be in touch with the team at Adobe in the coming weeks and see if together we can’t figure this thing out.
    At least now I know I’m not completely crazy, that other people been experiencing the same thing on this issue. Thanks for shedding some light on this John.

  17. Michael says:

    See http://bugs.adobe.com/jira/browse/FP-211
    Note the need for a bunch of instances. This reliably reproduces the problem on my machine (XP, FF3), and installing beta 10 ‘fixes’ the issue.
    [jd sez: Thanks… if there’s a recipe in the bugbase, then engineers can follow it to make the problem happen too. (Me, I just usually muse on weblog strangenesses, other people do the real work…. πŸ˜‰ ]

  18. Joeflash says:

    I don’t think it’s the same issue, though it may well have the same cause. The “video streaming stops working after two seconds” issue occurs on my one machine with just one instance playing, and it’s not confined to Firefox, it’s all browsers. My MP3 player(s) do not experience this problem with streaming media.

  19. Dave Glover says:

    Is this an allied bug? [jd sez: Not sure, I’m still trying to ID the first report.] I migrated to FF3 from FF2, had some runaround with uninstalling google desktop and reinstalling it because it wouldn’t run in an XP home SP3 admin account. When up and running I ran a flash:
    http://www.brightcove.tv/title.jsp?title=1438490562
    which displayed the behaviour described – a few seconds play then back to start. This was just after trying to listen to:
    http://www.myspace.com/bishimusic which simply did not work on the audio stream at all.
    Both the above worked in IE (with Flash 9 installed)and FF3 could play other myspace streams. I thought the audio stream problem was reproducible initially because it occurred in a fresh instance of FF3 but I have just opened 20 tabs, 2 running paused instances of the video stream, 1 a playing instance of it, 1 webform (this) and a WORKING INSTANCE of the audio stream – all in the instance of FF3 that originally wouldn’t run the audio when first opened…. this isn’t going to be easy! good luck!

  20. Lori says:

    As a developer, I just wanted to say thanks to jd for helping users understand what they need to do to help us fix bugs. In order to FIX a problem, you have to be able to REPRODUCE a problem at will. If you can’t reproduce reliably, you can’t verify that whatever changes you made have fixed the problem.
    [jd sez: And Lori’s been on the bitter end of that deal before, so she knowsw…. πŸ˜‰ ]

  21. patriotboy says:

    I’m getting the same problem with Google Chrome. It occurs on both Firefox and chrome but not on IE in the same session. Haven’t checked the memory–I’ll do that when I get home–but it occured right after reboot yesterday and I have 4 gig of ram, so I doubt that’s it.