Add Support for the ContainerTiming PaintMetric by jasonwilliams · Pull Request #46966 · servo/servo

jasonwilliams · 2026-08-03T18:43:41Z

Hi, I'm a spec champion for the Container Timing feature and I wanted to add it to Servo (behind a flag).

The Container Timing API enables monitoring when annotated sections of the DOM are displayed on screen and have finished their initial paint. I am one of the spec authors, and i've been helping this land in both Chromium and Firefox

This PR implements the Container Timing API, following the same layout > paint > constellation pipeline already used for LCP.

Steps for calculating Container Timing:

When an element has the containertiming attribute, a flag is propagated down through its descendants (set on the Node/fragment) so layout can cheaply tell "is this inside a container-timing root" without walking back up the tree on every fragment.
While building the display list, each painted text/image fragment that belongs to a container-timing root reports its viewport-clipped size and rect to a PaintTimingHandler, similar to how LCP candidates are collected.
PaintTimingHandler keeps one running record per container root (keyed by a stable node ID, since the containertiming value itself is optional and not unique), accumulating size/rect across paints, and tracks which containers were actually touched since the last hand-off.
Layout hands the finished per-container records to script as part of the reflow result
once webrender confirms that epoch composited, it sends PaintMetricEvent::ContainerTiming(paint_time, ids) to script, which stamps the already-stored records with the real paint/first-render time and queues the finished PerformanceContainerTiming entries.

Testing:

There are plenty of Web Platform Tests for Container Timing. I have been testing the build against these as i have been going along, 19/29 tests pass.
There are tests for some components, such as paintedRegion

Follow ups & investigation:

lastPaintedElement selection picks by traversal order, not largest area
[Tidy Up] The effective_visual_size check was originally written for LCP, but is useful for Container Timing as well, we should make this API generic

jdm · 2026-08-19T23:53:47Z

@jasonwilliams I've been ignoring this because it's marked as a draft. Are you waiting for someone to review it?

jasonwilliams · 2026-08-20T11:16:45Z

@jasonwilliams I've been ignoring this because it's marked as a draft. Are you waiting for someone to review it?

Thanks for reaching out @jdm, no not yet, it’s still a WIP.

Ill mark it as review and ping you when I need reviews.

shubhamg13 · 2026-09-22T03:10:58Z

Yeah, Nice.

Please put a separate PR to enable WPT for container-timing/ in include.ini. So, that we can see the visible effects.

Edit: + With WPT (Failing) Expectations

jasonwilliams · 2026-09-22T10:43:54Z

Yeah, Nice.

Please put a separate PR to enable WPT for container-timing/ in include.ini. So, that we can see the visible effects.

Done: #48290

github-actions · 2026-10-05T10:41:28Z

Test results for linux-wpt from try job (#37294387426):

Flaky unexpected result (21)

OK /IndexedDB/idbtransaction-oncomplete.any.html (#42845)

FAIL [expected PASS] subtest: IDBTransaction - complete event

assert_array_equals: lengths differ, expected array ["upgradeneeded", "complete", "success", "opencursor"] length 4, got ["upgradeneeded", "complete", "success"] length 3

CRASH [expected OK] /_webgl/conformance2/wasm/readpixels-2gb-in-4gb-wasm-memory.html
OK /beacon/beacon-basic.https.window.html (#41723)
- FAIL [expected PASS] subtest: Payload size restriction should be accumulated: type = string
```
assert_false: expected false got true
```
FAIL [expected PASS] /css/css-image-animation/image-animation-img-running-to-normal.html (#46567)
OK /dom/nodes/moveBefore/iframe-document-preserve.window.html (#43152)
- FAIL [expected PASS] subtest: moveBefore(): cross-origin iframe is preserved: remove new parent
```
assert_equals: iframe does not fire a second load event expected 1 but got 0
```
- FAIL [expected PASS] subtest: moveBefore(): cross-origin iframe is preserved: remove self via innerHTML
```
assert_equals: iframe does not fire a second load event expected 1 but got 0
```
CRASH [expected TIMEOUT] /html/browsers/history/the-location-interface/location_replace_session_history.html (#41896)
TIMEOUT [expected OK] /html/browsers/windows/consume-user-activation/window-open-multi-global.html (#47399)
- TIMEOUT [expected FAIL] subtest: window.open() and consuming user activation with multiple globals in play
```
Test timed out
```
TIMEOUT [expected OK] /html/semantics/embedded-content/media-elements/autoplay-allowed-by-permissions-policy.https.sub.html (#46803)
- TIMEOUT [expected PASS] subtest: Permissions-Policy header: autoplay * allows same-origin iframes.
```
Test timed out
```
TIMEOUT [expected OK] /html/semantics/embedded-content/media-elements/src_object_blob.html (#40340)
- TIMEOUT [expected PASS] subtest: HTMLMediaElement.srcObject blob
```
Test timed out
```
OK [expected CRASH] /html/semantics/embedded-content/the-iframe-element/iframe-loading-lazy-reload-location-reload.html (#32595)
- FAIL [expected PASS] subtest: Reloading iframe loading='lazy' before it is loaded: location.reload
```
uncaught exception: Error: assert_equals: expected "http://web-platform.test:8000/html/semantics/embedded-content/the-iframe-element/support/blank.htm?src" but got "about:blank"
```
OK /html/semantics/forms/form-submission-0/jsurl-form-submit.tentative.html (#36489)
- PASS [expected FAIL] subtest: Verifies that form submissions scheduled inside javascript: urls take precedence over the javascript: url's return value.
CRASH [expected ERROR] /html/semantics/forms/the-select-element/customizable-select/select-appearance-button-after-span.html

OK /html/semantics/scripting-1/the-script-element/execution-timing/077.html (#22139)

FAIL [expected PASS] subtest: adding several types of scripts through the DOM and removing some of them confuses scheduler

assert_array_equals: expected property 1 to be "Script #1 ran" but got "Script #3 ran" (expected array ["Script #2 ran", "Script #1 ran", "Script #3 ran", "Script #4 ran"] got ["Script #2 ran", "Script #3 ran", "Script #4 ran", "Script #1 ran"])

TIMEOUT [expected OK] /html/user-activation/navigation-state-reset-sameorigin.html (#45151)
- TIMEOUT [expected FAIL] subtest: Post-navigation state reset.
```
Test timed out
```
OK /html/webappapis/user-prompts/print-during-unload.html (#35944)
- PASS [expected FAIL] subtest: print() during unload
ERROR [expected OK] /intersection-observer/v2/animated-opacity.html (#40827)
OK /preload/link-header-preload-delay-onload.html (#39622)
- FAIL [expected PASS] subtest: Makes sure that Link headers preload resources and block window.onload after resource discovery
```
assert_true: expected true got false
```
OK /resource-timing/test_resource_timing.https.html (#25216)
- PASS [expected FAIL] subtest: PerformanceEntry has correct name, initiatorType, startTime, and duration (img)
- FAIL [expected PASS] subtest: PerformanceEntry has correct name, initiatorType, startTime, and duration (link)
```
assert_equals: expected 8.600000000000001 but got 8.6
```
OK /upgrade-insecure-requests/gen/sharedworker-classic-data.http-rp/upgrade/fetch.https.html
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to cross-http-downgrade origin and downgrade redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to cross-http-downgrade origin and no-redirect redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to cross-https origin and downgrade redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to same-http-downgrade origin and downgrade redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to same-http-downgrade origin and no-redirect redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for fetch to same-https origin and downgrade redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
OK /upgrade-insecure-requests/gen/sharedworker-classic-data.meta/upgrade/websocket.https.html
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for websocket to cross-ws-downgrade origin and no-redirect redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
- FAIL [expected PASS] subtest: Upgrade-Insecure-Requests: Expects allowed for websocket to same-ws-downgrade origin and no-redirect redirection from https context.
```
assert_equals: The resource request should be 'allowed'. expected "allowed" but got "blocked"
```
OK /visual-viewport/resize-event-order.html (#41981)
- PASS [expected FAIL] subtest: Popup: DOMWindow resize fired before VisualViewport.

Stable unexpected results that are known to be intermittent (11)

TIMEOUT [expected OK] /_mozilla/mozilla/window-resize-event.html (#36741)
- TIMEOUT [expected PASS] subtest: Popup onresize event fires after resizeTo
```
Test timed out
```
CRASH [expected OK] /html/browsers/browsing-the-web/navigating-across-documents/005.html (#27062)
TIMEOUT [expected OK] /html/browsers/browsing-the-web/navigating-across-documents/cross-origin-top-navigation-with-user-activation-in-parent.window.html (#48205)
- TIMEOUT [expected FAIL] subtest: Cross-origin top navigation is blocked without user activation, even if the parent has user activation
```
Test timed out
```
OK /html/browsers/browsing-the-web/navigating-across-documents/refresh/same-document-refresh.html (#34597)
- PASS [expected FAIL] subtest: Same-Document Referrer from Refresh
TIMEOUT /html/semantics/embedded-content/media-elements/autoplay-default-permissions-policy.https.sub.html (#46805)
- TIMEOUT [expected PASS] subtest: Default "autoplay" feature policy ["self"] allows the top-level document.
```
Test timed out
```
CRASH [expected OK] /html/semantics/embedded-content/the-iframe-element/iframe-loading-lazy-nav-window-open.html (#32596)
OK /preload/prefetch-document.html (#37210)
- FAIL [expected PASS] subtest: different-site document prefetch with 'as=document' should not be consumed
```
assert_equals: expected 2 but got 1
```
OK /resource-timing/buffer-full-add-then-clear.html (#40819)
- PASS [expected FAIL] subtest: Test that if the buffer is cleared after entries were added to the secondary buffer, those entries make it into the primary one
OK /resource-timing/buffer-full-then-increased.html (#44408)
- FAIL [expected PASS] subtest: Test that overflowing the buffer and immediately increasing its limit does not trigger the resourcetimingbufferfull event
```
assert_equals: Number of entries does not match the expected value. expected 3 but got 1
```
OK /touch-events/single-tap-when-touchend-listener-use-sync-xhr.html (#41175)
- PASS [expected FAIL] subtest: Click event should be fired when touchend opens synchronous XHR
OK [expected TIMEOUT] /trusted-types/trusted-types-navigation.html?16-20 (#44835)
- PASS [expected TIMEOUT] subtest: Navigate a frame via area with javascript:-urls in report-only mode.
- PASS [expected NOTRUN] subtest: Navigate a frame via area with javascript:-urls w/ default policy in report-only mode.

github-actions · 2026-10-05T10:41:38Z

✨ Try run (#37294387426) succeeded.

This enables testing for the experimental [container timing](https://wpt.fyi/results/container-timing/tentative?label=experimental&label=master&aligned) within WPT Part of servo#46966 Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Initial start by setting experimental runtime flags and support in Node and Base fragments. This is to reduce the cost of traversal to discover if container timing is needed Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Signed-off-by: jasonwilliams <jase.williams@gmail.com>

- Update paint_regions within a pending container, then merge into the permenent container record when the display list has finished building - Fix up some WPT's by checking the node itself for the container root before starting at the parent Signed-off-by: jasonwilliams <jase.williams@gmail.com>

`container_id` was always `root_element.id()`. Key by the root element's `OpaqueNode` instead. Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Instead of walking up the DOM trying to find container roots, use the traversalState to pass roots down instead. LCP has been doing this with text decorations etc. We have a ContainerTimingRoots linked list which starts at the nearest root then follows on until None. This allows us to implement containertimingignore and pass more tests. This should also have much improved performance, because there is no "lookup" now which iterates every single node up the tree, we know ahead of time where the roots are. The identifier is now read later on in script a we have access to the root element here anyway, we check again if the attribute is there, and if not we skip. Signed-off-by: jasonwilliams <jase.williams@gmail.com>

jdm

Very nice!

shubhamg13 · 2026-10-10T13:04:00Z

@servo-highfive monthly update

servo-highfive · 2026-10-10T13:04:06Z

Someone thinks this change could be added to the monthly blog post! To help with this, we need someone to answer the following questions: 😄

Who is most impacted by this change: users, Servo developers, embedders, or some other group?
What observable difference does this change make?
What preferences (if any) need to be enabled to observe this difference?
What (if any) specific URLs are affected?

If this change is part of a broader feature/project please make sure the PR description contains a Fixes: #12345 or Part of: #12345 issue reference.

Please add @servo-highfive monthly update answer when answering these questions so the bot notices your answer (or just quote this comment).

Thanks for helping us prepare the monthly blog post! ❤️

This enables testing for the experimental [container timing](https://wpt.fyi/results/container-timing/tentative?label=experimental&label=master&aligned) within WPT Part of servo#46966 Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Hi, I'm a spec champion for the Container Timing feature and I wanted to add it to Servo (behind a flag). The Container Timing API enables monitoring when annotated sections of the DOM are displayed on screen and have finished their initial paint. I am one of the spec authors, and i've been helping this land in both [Chromium](https://developer.chrome.com/blog/container-timing-origin-trial) and [Firefox](https://phabricator.services.mozilla.com/D279373) This PR implements the [Container Timing API](https://github.com/WICG/container-timing), following the same layout > paint > constellation pipeline already used for LCP. Steps for calculating Container Timing: 1. When an element has the `containertiming` attribute, a [flag is propagated down through its descendants](https://github.com/jasonwilliams/servo/blob/403908fa99f47da3e18faaca88d2abdb01498cf7/components/script/dom/node/node.rs#L378-L387) (set on the Node/fragment) so layout can cheaply tell "is this inside a container-timing root" without walking back up the tree on every fragment. 2. While building the display list, each painted text/image fragment that belongs to a container-timing root [reports its viewport-clipped size and rect to a PaintTimingHandler](https://github.com/jasonwilliams/servo/blob/403908fa99f47da3e18faaca88d2abdb01498cf7/components/layout/display_list/mod.rs#L706-L718), similar to how LCP candidates are collected. 3. PaintTimingHandler keeps one running record per container root (keyed by a stable node ID, since the containertiming value itself is optional and not unique), accumulating size/rect across paints, and tracks which containers were actually touched since the last hand-off. 4. Layout [hands the finished per-container records to script](https://github.com/jasonwilliams/servo/blob/403908fa99f47da3e18faaca88d2abdb01498cf7/components/layout/layout_impl.rs#L1561-L1567) as part of the reflow result 5. once webrender confirms that epoch composited, [it sends PaintMetricEvent::ContainerTiming(paint_time, ids)](https://github.com/jasonwilliams/servo/blob/403908fa99f47da3e18faaca88d2abdb01498cf7/components/paint/painter.rs#L585-L589) to script, which stamps the already-stored records with the real paint/first-render time and queues the finished PerformanceContainerTiming entries. Testing: - There are plenty of [Web Platform Tests](https://wpt.fyi/results/container-timing/tentative?label=master&label=experimental&aligned&q=container%20timing) for Container Timing. I have been testing the build against these as i have been going along, 19/29 tests pass. - There are tests for some components, such as paintedRegion Follow ups & investigation: - `lastPaintedElement` selection picks by traversal order, not largest area - [Tidy Up] The `effective_visual_size` check was [originally written for LCP](servo#43378), but is useful for Container Timing as well, we should make this API generic --------- Signed-off-by: jasonwilliams <jase.williams@gmail.com>

jasonwilliams · 2026-10-16T16:49:53Z

@servo-highfive monthly update answer

Who is most impacted? Web developers, mostly people who measure site performance and build real-user monitoring (RUM) tools. It's an experimental API, so it doesn't change anything for regular users yet.
What observable difference does this change make? Servo now has an experimental version of the Container Timing API. If a page adds the containertiming attribute to an element, it can use a PerformanceObserver with type: "container" to get PerformanceContainerTiming entries. Each entry says when that part of the DOM, including the text and images inside it, was painted to the screen and how much area it covers. Entries are only reported after the frame has actually been composited. It works like Largest Contentful Paint, but for any part of the page the author picks instead of just the largest element. Chromium (origin trial)
and Firefox are also implementing it.
What preferences need to be enabled? container_timing_enabled, e.g. --pref container_timing_enabled.
What specific URLs are affected? None in particular. It affects any page that uses the containertiming attribute. Most of the WPT tests in container-timing/tentative now pass.

jasonwilliams force-pushed the container-timing branch 6 times, most recently from 4f18c08 to bd2d948 Compare August 7, 2026 11:07

jasonwilliams force-pushed the container-timing branch 9 times, most recently from 0282ab5 to 83e2dc0 Compare September 21, 2026 16:00

jasonwilliams marked this pull request as ready for review September 21, 2026 16:07

jasonwilliams requested review from Loirooriol, gterzian and mrobinson as code owners September 21, 2026 16:07

servo-highfive added the S-awaiting-review There is new code that needs to be reviewed. label Sep 21, 2026

jasonwilliams force-pushed the container-timing branch 2 times, most recently from 88ae8e7 to f93db82 Compare September 22, 2026 10:38

jasonwilliams mentioned this pull request Sep 22, 2026

Enable container timing in WPT #48290

Merged

shubhamg13 mentioned this pull request Sep 29, 2026

Feature: ContainerTiming #48490

Open

shubhamg13 approved these changes Sep 29, 2026

View reviewed changes

servo-highfive added the S-needs-rebase There are merge conflict errors. label Sep 29, 2026

servo-highfive added the S-needs-rebase There are merge conflict errors. label Oct 8, 2026

jasonwilliams force-pushed the container-timing branch from 90d9f7e to bcfaa24 Compare October 8, 2026 11:54

servo-highfive removed the S-needs-rebase There are merge conflict errors. label Oct 8, 2026

jasonwilliams force-pushed the container-timing branch from bcfaa24 to 0a2c3f6 Compare October 8, 2026 17:32

servo-highfive added the S-needs-rebase There are merge conflict errors. label Oct 8, 2026

jasonwilliams added 7 commits October 8, 2026 18:36

Container Timing: Add initial flags and support

8ca14a4

Initial start by setting experimental runtime flags and support in Node and Base fragments. This is to reduce the cost of traversal to discover if container timing is needed Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Add Container Timing records and PaintedRegion

d922b9d

Signed-off-by: jasonwilliams <jase.williams@gmail.com>

paint/script: Report Container Timing entries once a frame is composited

d67b13e

Signed-off-by: jasonwilliams <jase.williams@gmail.com>

layout: Remove redundant ContainerTimingRecord::container_id

3dc7bc8

`container_id` was always `root_element.id()`. Key by the root element's `OpaqueNode` instead. Signed-off-by: jasonwilliams <jase.williams@gmail.com>

Update WPT expectations

b300f93

Signed-off-by: jasonwilliams <jase.williams@gmail.com>

jasonwilliams force-pushed the container-timing branch from 0a2c3f6 to a2f50e5 Compare October 8, 2026 17:45

servo-highfive removed the S-needs-rebase There are merge conflict errors. label Oct 8, 2026

jdm approved these changes Oct 9, 2026

View reviewed changes

servo-highfive removed the S-awaiting-review There is new code that needs to be reviewed. label Oct 9, 2026

jdm added this pull request to the merge queue Oct 9, 2026

servo-highfive added the S-awaiting-merge The PR is in the process of compiling and running tests on the automated CI. label Oct 9, 2026

Merged via the queue into servo:main with commit e1fe321 Oct 9, 2026
33 checks passed

servo-highfive removed the S-awaiting-merge The PR is in the process of compiling and running tests on the automated CI. label Oct 9, 2026

servo-highfive added the monthly update label Oct 10, 2026

Navigation Menu

Repository navigation

Add Support for the ContainerTiming PaintMetric - #46966

Add Support for the ContainerTiming PaintMetric#46966
jdm merged 7 commits into
servo:mainfrom
jasonwilliams:container-timing

jasonwilliams commented Aug 3, 2026 •

edited

Loading

Uh oh!

jdm commented Aug 19, 2026

Uh oh!

jasonwilliams commented Aug 20, 2026

Uh oh!

shubhamg13 commented Sep 22, 2026 •

edited

Loading

Uh oh!

jasonwilliams commented Sep 22, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

jdm left a comment

Uh oh!

Uh oh!

shubhamg13 commented Oct 10, 2026

Uh oh!

servo-highfive commented Oct 10, 2026

Uh oh!

jasonwilliams commented Oct 16, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

jasonwilliams commented Aug 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jdm commented Aug 19, 2026

Uh oh!

jasonwilliams commented Aug 20, 2026

Uh oh!

shubhamg13 commented Sep 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jasonwilliams commented Sep 22, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

github-actions Bot commented Oct 5, 2026

Uh oh!

jdm left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

shubhamg13 commented Oct 10, 2026

Uh oh!

servo-highfive commented Oct 10, 2026

Uh oh!

jasonwilliams commented Oct 16, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jasonwilliams commented Aug 3, 2026 •

edited

Loading

shubhamg13 commented Sep 22, 2026 •

edited

Loading