383351 - A hang or very high CPU usage on this page

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Ria Klaassen (not reading all bugmail)]

Attachment: zipped page (hangs without internet connection)

When I go to the URL in the URL field I get a hang or a high CPU usage.
Before bug 300030 it was about 30-35% and after bug 300030 85-100%.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Would be nice to have a minimal-ish testcase...

Comment 2

[Martijn Wargers (dead)]

Attachment: unminimized testcase

Thanks for attaching the page, Ria.
I've minimized that page to this, thus far.
I still see a cpu increase when testing between 2006-12-07 and 2006-12-08 builds. (unfortunately, I have difficulty with quantifying it with a timeout, that would make minimizing easier)

Comment 3

[Boris Zbarsky [:bzbarsky]]

So is this Windows-only?  I don't see any CPU usage on Linux on that testcase once it finishes loading...

Comment 4

[Martijn Wargers (dead)]

Sorry, should have said it, you need to click on the runscroller button to begin testing.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Hmm.  Yeah, a profile doesn't show anything unexpected... about half the time spent in reflow, half elsewhere...

Comment 6

[dolphinling]

Attachment: Still unminimized testcase

Cut out a bunch of the JS, didn't touch the HTML

Comment 7

[Arjan]

Attachment: Minimized testcase

This is a somewhat minimized testcase. I didn't try to change the javascript. Press the "runscroller"-button for the test.

The CPU usage doesn't jump extremely high when you remove one of the following:
- the 'height="585"' of the <td>
- the 'style="width: 760px;"' of the <div id="ticker">
- move the entire '<div id="ticker">' out of the <table>

Comment 8

[Boris Zbarsky [:bzbarsky]]

I just tried profiling this on Mac, and we spend all our time painting.  Removing the height from the table cell drops CPU usage from 40% to 20% or so.

I wonder whether the height matters just because we end up invalidating a bigger area or something?  I still wouldn't expect the repaint here to be all that expensive here...

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

...could also be triggering special height reflow, requiring an extra pass (and perhaps additional invalidation resulting from the oscillation between the two results).

Comment 10

[Wayne Mery (:wsmwk)]

(In reply to comment #7)
> Created an attachment (id=292949) [details]
> Minimized testcase
> 
> This is a somewhat minimized testcase. I didn't try to change the javascript.
> Press the "runscroller"-button for the test.
> 
> The CPU usage doesn't jump extremely high when you remove one of the following:
> - the 'height="585"' of the <td>
> - the 'style="width: 760px;"' of the <div id="ticker">
> - move the entire '<div id="ticker">' out of the <table>

ranges 5-15% cpu for me with
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2a1pre) Gecko/20090621 Minefield/3.6a1pre (.NET CLR 3.5.30729)

Comment 11

[Mihai Morar, (:MihaiMorar)]

Works for me using testcases that had been attached here using Lastest Nighty 25:

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:25.0) Gecko/20130730 Firefox/25.0