1787623 - backdrop-filter: url(#svg-filter) causes the element to not render

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, defect)

Description

[Kyza]

Attachment: firefox_bug.png

Steps to reproduce:

1. Create an SVG filter.
2. Apply the SVG filter to the backdrop-filter of an element outside of the SVG.

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:104.0) Gecko/20100101 Firefox/104.0

Issue Example: https://codepen.io/Kyza/pen/WNzqjML

Actual results:

The element does not render. Similar to opacity: 0.

Expected results:

The element renders with the appropriate filter from the SVG.

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'Core::Graphics: WebRender' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[Gregory Pappas [:gregp]]

With WebRender and backdrop-filter force-enabled, I got this range with mozregression. Suspects include bug 1578777, bug 1581307. Botond, Sotaro, what do you think?

Comment 3

[Glenn Watson [:gw]]

It's possible that we hit a path where we don't support that SVG filter on a backdrop-filter, and we should just drop the filter but render the rest of the content for now, at least. Any ideas if something like that might be happening Emilio?

Comment 4

[Sotaro Ikeda [:sotaro]]

(In reply to Gregory Pappas from comment #2)

With WebRender and backdrop-filter force-enabled, I got this range with mozregression. Suspects include bug 1578777, bug 1581307. Botond, Sotaro, what do you think?

bug 1581307 does not change a behavior. It seems not related to this problem.

Comment 5

[Emilio Cobos Álvarez (:emilio)]

Attachment: Test-case from reporter.

Comment 6

[Emilio Cobos Álvarez (:emilio)]

So, that's already what's happening for some unsupported backdrop-filters:

• https://searchfox.org/mozilla-central/rev/380fc5571b039fd453b45bbb64ed13146fe9b066/layout/painting/nsDisplayList.cpp#8382

However in this case we hit the !initialized code-path below:

• https://searchfox.org/mozilla-central/rev/380fc5571b039fd453b45bbb64ed13146fe9b066/layout/painting/nsDisplayList.cpp#8387

Because we hit this code-path in SVGFilterInstance:

• https://searchfox.org/mozilla-central/rev/380fc5571b039fd453b45bbb64ed13146fe9b066/layout/svg/SVGFilterInstance.cpp#342-344

Same happens if we replace backdrop-filter with just filter.

So there are two issues:

• Invalid filters should probably not paint nothing, that's also true of masks, see bug 1483225. That's what Chrome does in the test-case if you use filter rather than backdrop-filter.
• Not sure how we should treat the BackgroundImage / BackgroundAlpha filter input, maybe we should treat it as SourceImage / SourceAlpha when using backdrop-filter? Markus seems familiar with this code, do you know (when you have the time / are back from PTO)?

Comment 7

[Botond Ballo [:botond]]

(In reply to Gregory Pappas from comment #2)

With WebRender and backdrop-filter force-enabled, I got this range with mozregression. Suspects include bug 1578777, bug 1581307. Botond, Sotaro, what do you think?

This does seem to be a regression from bug 1578777.

The culprit appears to be https://phabricator.services.mozilla.com/D46060, which was intended to be just a refactoring, but accidentally introduced a functional change: it factored out a helper function to be shared by nsDisplayFilters::CanCreateWebRenderCommands and nsDisplayBackdropFilters::CanCreateWebRenderCommands, without realizing that the first was looking at StyleEffects()->mFilters and the second was looking at StyleEffects()->mBackdropFilters. The helper function looked at StyleEffects()->mFilters, thereby giving an incorrect answer when called from nsDisplayBackdropFilters::CanCreateWebRenderCommands.

The effect of this function was to gate the creation of an nsDisplayBackdropFilters display item.

That condition was removed in bug 1423746, whose commit message includes:

Remove the BackdropFilters::CanCreateWebrenderCommands call because it
was testing for StyleSVGEffects::mFilters rather than mBackdropFilters,
so it was doing nothing.

I suspect that the correct implementation returned false at the time, thereby inhibiting the creation of an nsDisplayBackdropFilters display item, and resulting in rendering the element without the backdrop filter.

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1578777

Comment 9

[Botond Ballo [:botond]]

Attachment: Bug 1787623 - Do not create nsDisplayBackdropFilters display item if we cannot create WebRender commands for it. r=emilio

This fixes a regression introduced inadvertently in
https://phabricator.services.mozilla.com/D46060.

Comment 10

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1787623 - Make invalid and unsupported filters paint the unfiltered frame. r=botond

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/b134d587690d
Make invalid and unsupported filters paint the unfiltered frame. r=botond

Comment 12

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Created web-platform-tests PR https://github.com/web-platform-tests/wpt/pull/35775 for changes under testing/web-platform/tests

Comment 13

[Marian-Vasile Laza]

Backed out changeset b134d587690d (Bug 1787623) for causing reftest failures on filter-basic-01.svg.
Backout link
Push with failures <--> Rs7
Failure Log

Comment 14

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Upstream PR was closed without merging

Comment 15

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/c3d6bd944022
Make invalid and unsupported filters paint the unfiltered frame. r=botond

Comment 16

[Cristina Cozmuta (:CrissCozmuta)]

Backed out for causing mochitest failures on test_filter_crossorigin.html . CLOSED TREE
Backout link
Push with failures
Backfill range
Link to failure log
Failure line : TEST-UNEXPECTED-FAIL | layout/svg/tests/test_filter_crossorigin.html | reftest comparison: == Testcase loaded from mochi.test Reference: black/yellow

Comment 17

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Upstream PR was closed without merging

Comment 18

[Emilio Cobos Álvarez (:emilio)]

Third time's the charm.

Comment 19

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/06480f550b00
Make invalid and unsupported filters paint the unfiltered frame. r=botond

Comment 20

[Iulian Moraru]

https://hg.mozilla.org/mozilla-central/rev/06480f550b00

Comment 21

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Upstream PR merged by moz-wptsync-bot

Comment 22

[BugBot [:suhaib / :marco/ :calixte]]

The patch landed in nightly and beta is affected.
:emilio, is this bug important enough to require an uplift?

• If yes, please nominate the patch for beta approval.
• If no, please set status-firefox105 to wontfix.

For more information, please visit auto_nag documentation.

Comment 23

[Emilio Cobos Álvarez (:emilio)]

It's non-trivial and not a regression so probably not, but please feel free to request uplift if you disagree.