Bug 3237 – CSS2: Background repeats when it shouldn't

• Component: CSS
• Status: VERIFIED FIXED

[Dave Hyatt]

2/13/05 10:44 PM Randal Hansen:
Summary:
When told to position a background image taller than its container below the container top and not to 
repeat it, Safari does repeat the image.  This is fiddly to describe but easy to see.

Example:
I have a simple test case at:

http://sonofhans.net/css/snippets/background_position_safari.html

This test page passes W3C XHTML and CSS validators, and works correctly in Mozilla 1.7 and Windows 
Internet Explorer 6.

[Dave Hyatt]

Apple Bug: rdar://4005553/

[Nate Cook]

Created attachment 2221 [details]
Test of x and y positioning

The same thing also happens if the image is wider & positioned to the right of
0. This test case demonstrates that.

[Nate Cook]

Created attachment 2222 [details]
Proposed fix

This makes it render properly.

[Darin Adler]

Looks good to me, but I think Dave should review.

Style-wise, I think the code is a tiny bit inelegant. The existing code always sets sx in all the code paths, 
even though it's initialized at the top of the function. The new code introduces code paths that rely on the 
default value of 0 for sx and sy. I'd prefer a patch that goes further and gets rid of the "sx = 0" cases by 
changing the if structure and just doing nothing in the "pixh == 0" cases.

This is an example of something we can't test with a dumped render tree, so we probably shouldn't add 
this to the layout-tests subdirectory. While I don't want to go to a design where we dump the pixmap of 
every test result, it would be good to figure out a way to make a regression test for this bug some day.

[Nate Cook]

Created attachment 2231 [details]
Proposed fix #2

Good points. I was trying to edit as little as possible.

For testing, maybe we could extend the layoutTestController so that individual
tests could force a pixmap dump?

[Dave Hyatt]

Fixed.  I added a new directory to the layout tests... fast/backgrounds, that can at least be visually 
inspected for correctness until we come up with something better.  Additional bg tests are encouraged.

[Joost de Valk (AlthA)]

Fixed in 10.4.2