1841316 - WPT failures in css/css-grid/grid-items/replaced-element-[011-013].html

Status: RESOLVED DUPLICATE of bug 1944423

(Core :: Layout: Grid, defect)

Description

[Thomas Wisniewski [:twisniewski]]

wpt.fyi:

• https://wpt.fyi/results/css/css-grid/grid-items/replaced-element-011.html
• https://wpt.fyi/results/css/css-grid/grid-items/replaced-element-012.html
• https://wpt.fyi/results/css/css-grid/grid-items/replaced-element-013.html

All of these seem to rely on a <canvas> being placed/sized as expected in a grid, so I suspect there is a common underlying reason for them to fail.

Comment 1

[Daniel Holbert [:dholbert]]

We're failing all three of these tests because we render the grid as being 10px wide, when the test expects it to be 100px wide. This seems to be a case with a sort-of circular relationship.

When we lay out this testcase in Firefox, we resolve the intrinsic width of the grid based on the intrinsic width of the canvas (its sole grid item), which has width=10.

But the test seems to be expecting that our intrinsic sizing algorithm for the grid will take into account the fact that we already know that the canvas will ultimately be 100px tall (from its 100% and the grid's grid-template-rows declaration), and we'll translate that height through the canvas's aspect ratio to instead produce an intrinsic width contribution of 100px from the canvas.

In this case that's easy to intuitively see and reason about, but I'm not sure this is doable or sensible in general.... This would require us to do grid layout (to know the precise cell placement of grid items, and the sizing of those cells) before we can resolve the intrinsic width of the grid, which is a bit circular.

Comment 2

[Ting-Yu Lin [:TYLin] (PDT, UTC-7)]

replaced-element-[011-013].html are all fixed by bug 1944423.