Closed Bug 412726 Opened 17 years ago Closed 14 years ago

Flash stream to FF under XP uses 100% of cpu when Content-Length not specified (continuous NetStream Flush events)

Categories

(External Software Affecting Firefox Graveyard :: Flash (Adobe), defect)

x86
Windows XP
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: justin, Assigned: bugs)

Details

BitGravity is working on getting a live stream up and running, but is running into issues with a live flash stream *only* under Firefox *only* on WindowsXP.  Here is his description of the issue:

--
We've noticed that if you capture the stream (wget) to an FLV file, and then play that directly, FireFox doesn't have any problems.  The live stream isn't a continuous stream of packets, but a series of 50k-100k chunks that are sent once per second by the server.  Somehow this causes FireFox to chew CPU.
In this case, the server buffers ~10 seconds worth of data (1MB), sends it en masse as fast as it can, and starts buffering again.  If you look at the WinXP taskmgr, you'll see that the CPU is pegged at 100%, then dips for a few seconds, then jams back up to 100%.  When the CPU dips, the video playback no longer stutters.

If you try this in WinXP/IE, you'll see that it plays fine and cpu doesn't spike.

It almost looks like Flash is getting interrupted by FireFox a lot more for network events than by IE.  When there's no network traffic, the CPU dips.
--

I have private URLs I can test with and get any debug info you may need, but can't be posted to the bug as they are private for testing.  You don't see this under OS X.
Component: General → Plug-ins
Product: Firefox → Core
QA Contact: general → plugins
Version: 2.0 Branch → 1.8 Branch
I have verified that both test streams behave the same on WinXP with FireFox 3 beta2 -- chewing 100% CPU, brief dip, then back to 100%.
Hi - The Flash Player team is investigating and will follow-up whe nwe have more information to report.
Folks, we have been contacted by Adobe and asked for a "support tracking ID" to escalate this to the Flash team.  Do we have one associated with this bug?  Or is this already escalated properly?  Thanks in advance.
Here's a smoking gun (and work-around).  It turns out that the bug shows up if you don't specify a Content-Length HTTP header.

Background ...

If I set Content-Length to some arbitrary value, the video plays perfectly under WinXP/FireFox, and the CPU doesn't spike anymore.  The network traffic is virtually identical.  Remove the Content-Length header and the problem reappears.

In the Flash player ...

Without the Content-Length header we get a continuous stream of NetStream Flush events.  With the Content-Length header, we get just get one NetStream Full event.



Testcase ==> http://www.responsibilityproject.com/films/lighthouse/

The Flash video embedded in the above page uses 100% CPU on my system.

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.0.1pre) Gecko/2008070106 GranParadiso/3.0.1pre
Shockwave Flash 9.0 r124
Flash can use almost all CPU power. It is not even being played - only starting images are displayed. And the worse thing is no one of them are visible at the moment - they are located on inactive tab or outside the visible area of the page.

It is very serious bug because it makes using Mozilla impossible!
Component: Plug-ins → Flash (Adobe)
Product: Core → Plugins
QA Contact: plugins → adobe-flash
Version: 1.8 Branch → unspecified
Summary: Flash stream to FF under XP uses 100% of cpu → Flash stream to FF under XP uses 100% of cpu when Content-Length not specified (continuous NetStream Flush events)
The latest Flash player includes numerous fixes related to instance management and CPU usage. Please update this bug with your findings with this Player:
http://labs.adobe.com/downloads/flashplayer10.html
Assignee: nobody → jet
Marking resolved/fixed to get regression data.
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Product: External Software Affecting Firefox → External Software Affecting Firefox Graveyard
You need to log in before you can comment on or make changes to this bug.