662347 - scrolling on pages with Flash is slower in FF7 Nightly build than in FF4

Status: RESOLVED WORKSFORME

(Core Graveyard :: Plug-ins, defect)

Description

[Josh Aas]

1. Load YouTube.com main page. I get a big Flash-based Duke Nukem Forever ad on top.
2. Put your mouse over the page but on the side with no actual content above or below the mouse on the page (to avoid mouseovers).
3. Scroll up and down rapidly with the scroll wheel.

This is really fast, nearly flawless with Firefox 4. It is noticeably glitchy (little pauses) with the latest Firefox 7 Nightly build.

Comment 1

[Benoit Girard (:BenWa)]

The tentative cause is bug 598425 but I tried the nightly from June 5th and it shows the same problem. It's a bit hard to measure so if others want to compare the June 5th/6th nightly with a flash ad on the page such as the front page of youtube typically does.

I will look at this once I finish my current bugs, in the meantime it would be fantastic if someone can pin-point the regression window.

Comment 2

[Steven Michaud [:smichaud] (Retired)]

It's *not* a good idea to have an ad on a particular web page be the
only way to reproduce a bug -- the ad might be gone in a few hours

Here's a "trailer" from the Duke Nukem Forever site:

http://www.dukenukemforever.com/html/us/media.html

Can this bug also be reproduced with it?

Comment 3

[Steven Michaud [:smichaud] (Retired)]

I can't reproduce this bug with either the Duke Nukem Forever ad at
the top of the YouTube home page (http://youtube.com/, which by the
way is still there) or the "trailer" from comment #2.  I tested with a
Mighty Mouse on OS X 10.6.7 (running on an approximately 2.5-year-old
MacBook Pro).

I couldn't see any difference between Firefox 4.0.1 and today's
mozilla-central nightly.  I scrolled the page rapidly up and down
using both the mouse's scroll-wheel and the page's scroll bar.

Comment 4

[Benoit Girard (:BenWa)]

The regression may only be related to the Core Graphics rendering model, and I suspect this page is using Core Animation.

Comment 5

[Steven Michaud [:smichaud] (Retired)]

I should add that I tested with the current version of the Flash plugin (which as of today is 10.3.181.22).

Comment 7

[Benoit Girard (:BenWa)]

Test case: http://middaybrew.com/

Comment 8

[(no longer active)]

Actually, this is a really bad regression, and I think we should solve it for Firefox 7.

Comment 9

[Benoit Girard (:BenWa)]

I get the same performance on both ff4 or recent builds. There doesn't appear to be any performance regression on my end, however scrolling performance does not improve with the async patch which would be expect. I believe it would be worthwhile to investigate why.

(Responding to Bug 662843 Comment #2)
Steven do you notice that scrolling performance is worse when you start the first two videos (regardless of the build)?

Comment 10

[Steven Michaud [:smichaud] (Retired)]

> (Responding to Bug 662843 Comment #2)
> Steven do you notice that scrolling performance is worse when you
> start the first two videos (regardless of the build)?

Even playing *all* the videos seems to make no difference to scrolling
performance, in any of the builds.

Though (of course) doing this causes other weirdness (in all builds)
-- the audio seems to switch randomly between the playing videos, and
many of the videos "pause while loading".

Comment 11

[Johnathan Nightingale [:johnath]]

(In reply to comment #8)
> Actually, this is a really bad regression, and I think we should solve it
> for Firefox 7.

Well, we're tracking it, but 7 is now in Aurora and this hasn't had activity for more than a month - is it still an active issue?

Comment 12

[Benoit Girard (:BenWa)]

This has been pref-ed off for firefox 7 so there shouldn't be in regression in 7. Keeping the bug to confirm that this doesn't re-appear in 8 with the async plugin pref-ed on.

Comment 13

[Benoit Girard (:BenWa)]

Bug 663259 landed yesterday, can anyone reproduced the bug with today's nightly?

Comment 14

[(no longer active)]

I couldn't, but after our talk about this on IRC last time, I wouldn't know what should be considered a regression and what shouldn't... :/

Comment 15

[Benoit Girard (:BenWa)]

Closing per Comment 13. I still can't reproduce the issue. Please let us know if you run into this.