449324 - Take single-tile background path when there's a solid border

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fix

The default background-origin area is the element's padding rect, but the default background-clip area is the element's border rect. This means that any element with a repeating background which is the same size as the element's padding rect, but which has a solid opaque border, in theory should tile the background into 9 tiles even though all but the center tile are completely hidden by the border.

Since repeating is the default, this could be hurting us today when Web authors design a background image to fix a px-sized element. It hurts us even more when we start using SVG backgrounds or background-size or SVG paint servers to scale the background to the size of the element.

This patch detects solid opaque borders and essentially treats them as inducing background-clip:padding. I think this is a safe optimization.

Comment 1

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fix v2

Oops, that needed to be updated to trunk a bit more.

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Comment on attachment 332489 [details] [diff] [review]
fix v2

Yep, should be safe.

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Pushed 926b38cca720

Comment 4

[Benjamin Smedberg]

This checkin appears to have caused the mac tbox to turn orange with the following test failures:

REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/inline--float-left--01.xhtml
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/inline--float-right--01.xhtml
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/inline--position-absolute--01.xhtml
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/object--auto-auto--0-pct.html
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/object--auto-auto--0-px.html
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/object--auto-auto--pct-0.html
REFTEST TEST-UNEXPECTED-PASS | file:///builds/slave/trunk_darwin_mini01/build/layout/reftests/svg/sizing/object--auto-auto--px-0.html

Any clues?

Comment 5

[Bill Gianopoulos [:WG9s]]

Well, are these really failures?  They are all tests that now pass but previously did not.

Comment 6

[Simon Montagu :smontagu]

Attachment: Reenable the passing tests

Comment 7

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

Why not do this for ridge, groove, inset, and outset too?

Comment 8

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Comment on attachment 333567 [details] [diff] [review]
Reenable the passing tests

Sure, but would be good to know why they pass now.. I assume that the issue is some odd handling of OPERATOR_ADD on the borders (often 1-off on cocoa)?

Comment 9

[Simon Montagu :smontagu]

Yeah, on 3.0.1 the errors are 1-off drawing on the borders, but so it is on the tests that still fail, so more analysis is required.

Pushed 7bb9c0bb9663.

Comment 10

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Thanks everyone.

Comment 11

[Robert O'Callahan (:roc) (email my personal email if necessary)]

(In reply to comment #7)
> Why not do this for ridge, groove, inset, and outset too?

I dunno, we probably should.

Comment 12

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: patch for that

Comment 13

[Robert O'Callahan (:roc) (email my personal email if necessary)]

The extra patch here never landed.

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

http://hg.mozilla.org/mozilla-central/rev/0a45a7c62e1f