1697558 - Corruption with masks and isolate.

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, defect)

Description

[Emilio Cobos Álvarez [:emilio]]

Attachment: Test-case.

With the attached test-case I see corrupted masks, only on WebRender.

Comment 1

[Glenn Watson [:gw]]

We've backed out the regressing patch for now, while I investigate this further.

Comment 2

[Glenn Watson [:gw]]

Notes to self, for more investigation tomorrow:

• Seems specific to mask - changing to border-radii (which results in the same task graph) doesn't repro.
• Readback seems correct.
• Seems related either to reading of the source stacking context (but UVs / target rects seem correct), or that the source surface itself is incorrect.

Comment 3

[Glenn Watson [:gw]]

The bug is somewhere around [1].

In this test case, it assumes that the entire clip mask will be covered (so doesn't need to clear it), but then the loop to add the tiles calculates incorrect local sub-rects, resulting in only part of the clip mask being written to.

I confirmed that forcing clip_is_axis_aligned to be false here makes the test case render correctly. That's not a proper fix (it means we overdraw for each tile), but it confirms that the issue is related to the tile sub-rect coords created here.

[1] https://searchfox.org/mozilla-central/source/gfx/wr/webrender/src/batch.rs#3583

Comment 4

[Glenn Watson [:gw]]

Attachment: Bug 1697558 - Fix mix-blend readbacks from raster roots.

Picture cache code was using final device-space rect instead of the
actual raster rect when determining the readback coordinates for
mix-blend-mode effects. These are equivalent in most cases, but can
be different if the picture cache is also establishing a raster root.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/e19685e641eb
Fix mix-blend readbacks from raster roots. r=gfx-reviewers,lsalzman

Comment 6

[Marian-Vasile Laza]

https://hg.mozilla.org/mozilla-central/rev/e19685e641eb