1879957 - Generalize "percent-height quirk" to apply to block axis (or root element's block axis)

Status: ASSIGNED

(Core :: Layout, defect)

Description

[Daniel Holbert [:dholbert]]

[Spinning this off from bug 1687557 comment 14 and associated comments.]

We've got a quirks-mode behavior where percentage heights resolve against the height of the root element, or something to that effect.

This is specced here:
https://quirks.spec.whatwg.org/#the-percentage-height-calculation-quirk

In Chromium and WebKit, it looks like this applies in the body element's block axis, I think. We should probably align with them, for interoperability, since there's no strong reason this quirk should be vertically-biased in the absence of vertical writing-modes.

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: testcase 1 (writing-mode: vertical-rl, width:90%)

Here's testcase to demonstrate. Chrome/WebKit honor the 90% block-size in this case, Firefox does not.

(We do honor it if you remove the vertical writing-mode, making the block-size a height rather than a width.)

Comment 2

[Daniel Holbert [:dholbert]]

Attachment: testcase 2: the element used as a CB isn't the root

This quirk isn't specific to the root, actually; here's a testcase where it's a div that provides the concrete size to resolve percentages against.

Chrome/WebKit resolve the percent in this case (drawing a bunch of lime area). Firefox draws no lime (resolving the percent to zero).

Comment 3

[Daniel Holbert [:dholbert]]

Attachment: testcase 3: percent height resolving through opposite-block-flow writing modes

This is a variant of testcase 2, with vertical-lr nested inside of vertical-rl. This block-flow direction-flip doesn't serve as a blocker for Chromium/WebKit when they're tunneling up to find a percent basis.

Comment 4

[Daniel Holbert [:dholbert]]

Attachment: testcase 4: bunch of cases

Here's a testcase with a grid of scenarios. In Chromium/WebKit, the center-right, bottom-left, and bottom-right examples are blank (I think because the span with an orthogonal writing-mode from its parent ends up forming a box that blocks the traversal).

Comment 5

[Simon Pieters [:zcorpan]]

What should the spec say?