1105878 - background-size: 100%; flickers and lags wildly

Status: RESOLVED DUPLICATE of bug 1060200

(Core :: General, defect)

Description

[blubban]

Attachment: Test page

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:33.0) Gecko/20100101 Firefox/33.0
Build ID: 20141013200257

Steps to reproduce:

Just look at the attached page. (The original had messy indentation and three-level table nesting, but programmers like you always like reduced testcases.)


Actual results:

The background image jumps around; it looks like it's switching back and forth between nearest-neighbor scaling to an integer pixel count, and something more smooth, as fast as it can.
While doing this, Firefox uses ~15% of my CPU - and Xorg uses ~25%, with most of the idle time being spent in radeon_fence_wait_seq. It also gives me about one second of input lag on everything on my entire computer, both Firefox itself and text editors.

The time taken for each size switch is linear in the number of <td>s on the page; 40 of them gives me about five switches per second. If you want to adjust the speed, just add or remove some of them.


Expected results:

A page without scripts and animations should not move unless the user is interacting with it. Nor should Firefox suck a large amount of CPU for viewing a static page. (Nor should it lag my computer, but I'd say that's more of a bug in X than Firefox.)

The bug exists on Windows 7 64bit too. (I only tested the original page I found it on in W7, but it would surprise me greatly if the same bug appears on multiple OSes but under different circumstances.)

There seems to be a certain amount of randomness in this bug; using Firebug to view the offending image often (but not always) makes it stand still, and it once (out of ~50 tries) didn't dance like that after a refresh.

Comment 1

[Alice0775 White]

Seems to be duplication of Bug 1060200

Comment 2

[blubban]

Seems plausible; it stands still in 35.0a2, and CPU use is zero where it belongs, even when I increased it to 5000 <td>s. (34 gets rid of the width change and makes it stand mostly still, but it still switches between nearest and smooth and lags my computer.)

Bug 1060200 seems to me like it's about making the rescaling fast (or act fast, at least); I'm not sure if it stops the scaling method from changing in the first place. If it still does that, that's also a bug, though a minor one - probably not worth caring about, there are more important bugs around here.

I'll keep in mind that new Firefoxes should be tested before reporting bugs. Thanks for the pointer.