Mask backdrop-filter content until all mask-image layers settle [chromium/src : main]

[Philip Rogers (Gerrit)]

Attention needed from Helmut Januschka and Wangsong Jin

Philip Rogers added 1 comment

Patchset-level comments

File-level comment, Patchset 4 (Latest):

Philip Rogers . unresolved

Wangsong, can you review this?

Helmut, can you please put this change behind a RuntimeEnabledFeature flag so we can disable it in the unlikely case that it causes regressions?

Open in Gerrit

Related details

Attention is currently required from:

• Helmut Januschka
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review
• No-Unresolved-Comments
• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 4

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Helmut Januschka <hel...@januschka.com>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Mon, 02 Mar 2026 21:56:10 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

[Helmut Januschka (Gerrit)]

Attention needed from Philip Rogers and Wangsong Jin

Helmut Januschka added 1 comment

Patchset-level comments

File-level comment, Patchset 4:

Philip Rogers . unresolved

Wangsong, can you review this?

Helmut, can you please put this change behind a RuntimeEnabledFeature flag so we can disable it in the unlikely case that it causes regressions?

Helmut Januschka

added killswitch!

Open in Gerrit

Related details

Attention is currently required from:

• Philip Rogers
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review
• No-Unresolved-Comments
• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 5

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Tue, 03 Mar 2026 00:05:34 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

Comment-In-Reply-To: Philip Rogers <p...@chromium.org>

[Wangsong Jin (Gerrit)]

Attention needed from Helmut Januschka and Philip Rogers

Wangsong Jin added 1 comment

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1897, Patchset 5 (Latest): if (style.HasMask() && !style.BackdropFilter().IsEmpty() &&

Wangsong Jin . unresolved

Thanks for taking a look!
Is there a plan to fix the same loading issue for the regular mask image?
This change only updates the backdrop‑filter mask case, and I’m a bit concerned this may be confusing for developers since it’s not aligned with the regular mask behavior anymore (though it's not expected). And that's the reason I didn’t update this change in my original fix with FillLayer::AllImagesAreInvalid().

Ideally, we could fix both cases together, though I’m not sure about the effort required for the regular mask image. If it’s not a minor change, it might make sense to ship them separately. CC: @pdr

Open in Gerrit

Related details

Attention is currently required from:

• Helmut Januschka
• Philip Rogers

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review
• No-Unresolved-Comments
• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 5

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Helmut Januschka <hel...@januschka.com>

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Comment-Date: Tue, 03 Mar 2026 01:01:46 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

[Helmut Januschka (Gerrit)]

Attention needed from Philip Rogers and Wangsong Jin

Helmut Januschka added 2 comments

Patchset-level comments

File-level comment, Patchset 4:

Philip Rogers . resolved

Wangsong, can you review this?

Helmut, can you please put this change behind a RuntimeEnabledFeature flag so we can disable it in the unlikely case that it causes regressions?

Helmut Januschka

added killswitch!

Helmut Januschka

Done

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1897, Patchset 5: if (style.HasMask() && !style.BackdropFilter().IsEmpty() &&

Wangsong Jin . resolved

Thanks for taking a look!
Is there a plan to fix the same loading issue for the regular mask image?
This change only updates the backdrop‑filter mask case, and I’m a bit concerned this may be confusing for developers since it’s not aligned with the regular mask behavior anymore (though it's not expected). And that's the reason I didn’t update this change in my original fix with FillLayer::AllImagesAreInvalid().

Ideally, we could fix both cases together, though I’m not sure about the effort required for the regular mask image. If it’s not a minor change, it might make sense to ship them separately. CC: @pdr

Helmut Januschka

agree we should align regular mask loading behavior.
updated this CL to do that.

Open in Gerrit

Related details

Attention is currently required from:

• Philip Rogers
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review

• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 6

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Tue, 03 Mar 2026 22:31:50 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

Comment-In-Reply-To: Helmut Januschka <hel...@januschka.com>

Comment-In-Reply-To: Philip Rogers <p...@chromium.org>

Comment-In-Reply-To: Wangsong Jin <wangs...@microsoft.com>

[Wangsong Jin (Gerrit)]

Attention needed from Helmut Januschka and Philip Rogers

Wangsong Jin added 2 comments

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1898, Patchset 8 (Latest): style.MaskLayers().HasLoadedImage()) {

Wangsong Jin . unresolved

I’m not entirely sure what happens when all mask images are still in the loading phase. From this condition, it appears that we don’t explicitly hide the content, but just tested it nothing is rendered. I think there're some logic elsewhere that handles regular case. The existing fix sets `state.opacity = 0.f` specifically for the backdrop-filter + invalid case; extending this approach to regular masks feels a bit hacky to me. But I’m not an expert on the mask pipeline, so I’d suggest having someone with deeper expertise in this area take a look.

Line 1900, Patchset 8 (Latest): }

Wangsong Jin . unresolved

This mask loading change may not be appropriate for WPT, as I don’t see an explicit mention in the spec. But it may still be worth adding a few web tests to cover it.

Open in Gerrit

Related details

Attention is currently required from:

• Helmut Januschka
• Philip Rogers

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review

• No-Unresolved-Comments

• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 8

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Helmut Januschka <hel...@januschka.com>

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Comment-Date: Wed, 04 Mar 2026 22:38:49 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

[Helmut Januschka (Gerrit)]

Attention needed from Philip Rogers and Wangsong Jin

Helmut Januschka added 2 comments

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1898, Patchset 8: style.MaskLayers().HasLoadedImage()) {

Wangsong Jin . resolved

I’m not entirely sure what happens when all mask images are still in the loading phase. From this condition, it appears that we don’t explicitly hide the content, but just tested it nothing is rendered. I think there're some logic elsewhere that handles regular case. The existing fix sets `state.opacity = 0.f` specifically for the backdrop-filter + invalid case; extending this approach to regular masks feels a bit hacky to me. But I’m not an expert on the mask pipeline, so I’d suggest having someone with deeper expertise in this area take a look.

Helmut Januschka

Thanks for the pushback. I narrowed this CL so opacity gating is only applied to the backdrop-filter + mask case (`style.HasMask() && !style.BackdropFilter().IsEmpty() && ...`) and removed the regular-mask test that encoded broader behavior.

Line 1900, Patchset 8: }

Wangsong Jin . resolved

This mask loading change may not be appropriate for WPT, as I don’t see an explicit mention in the spec. But it may still be worth adding a few web tests to cover it.

Helmut Januschka

Done

Open in Gerrit

Related details

Attention is currently required from:

• Philip Rogers
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review

• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 9

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Fredrik Söderquist <f...@opera.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Thu, 05 Mar 2026 22:59:42 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

Comment-In-Reply-To: Wangsong Jin <wangs...@microsoft.com>

[Helmut Januschka (Gerrit)]

Attention needed from Philip Rogers and Wangsong Jin

Helmut Januschka added 1 comment

Patchset-level comments

File-level comment, Patchset 9 (Latest):

Helmut Januschka . resolved

@wangs...@microsoft.com feedback addressed

Gerrit-Comment-Date: Fri, 06 Mar 2026 12:54:38 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

[Wangsong Jin (Gerrit)]

Attention needed from Helmut Januschka and Philip Rogers

Wangsong Jin added 1 comment

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1898, Patchset 8: style.MaskLayers().HasLoadedImage()) {

Wangsong Jin . resolved

I’m not entirely sure what happens when all mask images are still in the loading phase. From this condition, it appears that we don’t explicitly hide the content, but just tested it nothing is rendered. I think there're some logic elsewhere that handles regular case. The existing fix sets `state.opacity = 0.f` specifically for the backdrop-filter + invalid case; extending this approach to regular masks feels a bit hacky to me. But I’m not an expert on the mask pipeline, so I’d suggest having someone with deeper expertise in this area take a look.

Helmut Januschka

Thanks for the pushback. I narrowed this CL so opacity gating is only applied to the backdrop-filter + mask case (`style.HasMask() && !style.BackdropFilter().IsEmpty() && ...`) and removed the regular-mask test that encoded broader behavior.

Wangsong Jin

To clarify, my earlier comment was about the case where all mask images are still in the pending state. Based on the HasUnloadedImage() && HasLoadedImage() logic, we don’t set the opacity to 0, which means the content should not be hidden.
For the backdrop‑filter mask case, the content is hidden because we additionally check AllImagesAreInvalid(). What surprised me is that the content also appears hidden for the regular mask case, which suggests there must be some other logic involved. If we could handle the regular mask through the same flow, that would be ideal. I’m not sure how much work that would be, though, and it’s possible that the current "hacky" solution ends up being the pragmatic choice.

The changes related to the backdrop‑filter mask image loading look good to me. The open question is whether we’re good to update the loading behavior only for the backdrop‑filter mask case. Honestly, I don’t have an answer to that. cc: @pdr for suggestions.

Open in Gerrit

Related details

Attention is currently required from:

• Helmut Januschka
• Philip Rogers

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review
• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 9

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Fredrik Söderquist <f...@opera.com>

Gerrit-CC: Philip Rogers <p...@chromium.org>

Gerrit-CC: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Helmut Januschka <hel...@januschka.com>

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Comment-Date: Fri, 06 Mar 2026 20:17:12 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

Comment-In-Reply-To: Helmut Januschka <hel...@januschka.com>

Comment-In-Reply-To: Wangsong Jin <wangs...@microsoft.com>

[Philip Rogers (Gerrit)]

Attention needed from Helmut Januschka and Wangsong Jin

Philip Rogers added 2 comments

File third_party/blink/renderer/core/paint/compositing/compositing_test.cc

Line 3942, Patchset 9 (Parent):TEST_P(CompositingSimTest, CanvasDrawElementLayersWithWillChange) {

Philip Rogers . unresolved

These shouldn't be removed

File third_party/blink/renderer/core/paint/paint_property_tree_builder.cc

Line 1898, Patchset 8: style.MaskLayers().HasLoadedImage()) {

Wangsong Jin . resolved

I’m not entirely sure what happens when all mask images are still in the loading phase. From this condition, it appears that we don’t explicitly hide the content, but just tested it nothing is rendered. I think there're some logic elsewhere that handles regular case. The existing fix sets `state.opacity = 0.f` specifically for the backdrop-filter + invalid case; extending this approach to regular masks feels a bit hacky to me. But I’m not an expert on the mask pipeline, so I’d suggest having someone with deeper expertise in this area take a look.

Helmut Januschka

Thanks for the pushback. I narrowed this CL so opacity gating is only applied to the backdrop-filter + mask case (`style.HasMask() && !style.BackdropFilter().IsEmpty() && ...`) and removed the regular-mask test that encoded broader behavior.

Wangsong Jin

To clarify, my earlier comment was about the case where all mask images are still in the pending state. Based on the HasUnloadedImage() && HasLoadedImage() logic, we don’t set the opacity to 0, which means the content should not be hidden.
For the backdrop‑filter mask case, the content is hidden because we additionally check AllImagesAreInvalid(). What surprised me is that the content also appears hidden for the regular mask case, which suggests there must be some other logic involved. If we could handle the regular mask through the same flow, that would be ideal. I’m not sure how much work that would be, though, and it’s possible that the current "hacky" solution ends up being the pragmatic choice.

The changes related to the backdrop‑filter mask image loading look good to me. The open question is whether we’re good to update the loading behavior only for the backdrop‑filter mask case. Honestly, I don’t have an answer to that. cc: @pdr for suggestions.

Philip Rogers

The bug mentions 2 cases: multiple masks images and multiple mask images with backdrop filter. The bug is about switching to safari's behavior for both, but the bug is also not super clear (e.g., starts with "Consider...").

Maybe we could make a table of:
multiple loading masks:
chrome, firefox, safari, chrome-with-fix

and a table of:
multiple loading masks with backdrop-filter:
chrome, firefox, safari, chrome-with-fix

The tests in this patch have just one loading image, but I think multiple simultaneously loading images is the interesting case?

Open in Gerrit

Related details

Attention is currently required from:

• Helmut Januschka
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review

• No-Unresolved-Comments

• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 9

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Philip Rogers <p...@chromium.org>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Fredrik Söderquist <f...@opera.com>

Gerrit-CC: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Helmut Januschka <hel...@januschka.com>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Sat, 07 Mar 2026 02:19:21 +0000

[Helmut Januschka (Gerrit)]

Attention needed from Philip Rogers and Wangsong Jin

Helmut Januschka added 1 comment

File third_party/blink/renderer/core/paint/compositing/compositing_test.cc

Line 3942, Patchset 9 (Parent):TEST_P(CompositingSimTest, CanvasDrawElementLayersWithWillChange) {

Philip Rogers . resolved

These shouldn't be removed

Helmut Januschka

Done

Open in Gerrit

Related details

Attention is currently required from:

• Philip Rogers
• Wangsong Jin

Submit Requirements:

• Code-Coverage
• Code-Owners
• Code-Review

• Review-Enforcement

Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings.

Gerrit

Gerrit-MessageType: comment

Gerrit-Project: chromium/src

Gerrit-Branch: main

Gerrit-Change-Id: I31e212600b42fe899e2ae42046ac26bc6eaf977c

Gerrit-Change-Number: 7620503

Gerrit-PatchSet: 10

Gerrit-Owner: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Helmut Januschka <hel...@januschka.com>

Gerrit-Reviewer: Philip Rogers <p...@chromium.org>

Gerrit-CC: AyeAye <android-bu...@system.gserviceaccount.com>

Gerrit-CC: Chromium LUCI CQ <chromiu...@luci-project-accounts.iam.gserviceaccount.com>

Gerrit-CC: Fredrik Söderquist <f...@opera.com>

Gerrit-CC: Wangsong Jin <wangs...@microsoft.com>

Gerrit-CC: chromium...@chromium.org

Gerrit-Attention: Philip Rogers <p...@chromium.org>

Gerrit-Attention: Wangsong Jin <wangs...@microsoft.com>

Gerrit-Comment-Date: Sat, 07 Mar 2026 19:53:46 +0000

Gerrit-HasComments: Yes

Gerrit-Has-Labels: No

Comment-In-Reply-To: Philip Rogers <p...@chromium.org>