Scoped View Transitions · Chromium baseline

Source: chromium/src@main Build commit: 553843c3bb5b Confirmed against origin/main: 8910eb28f6aa Suite pinned to wpt e8ca14556268 (2026-11-20) Build: content_shell, debug/component (reused) css-view-transitions-2 Status: shipped no flag

Baseline reference for the scoped view transitions cross-browser project. No patches on top of chromium/src@main, no feature flag toggled. Element.startViewTransition() and the scoped pseudo-tree have shipped in Chromium, so Before and After here are the same captures: there is no "unpatched" build to diff against. This page exists so Firefox and WebKit samplers in the same project have a stable reference point for the identical four cases and the same nested-clip video.

CSS View Transitions Level 2 adds Element.startViewTransition(): a view transition scoped to one element instead of the whole document. The pseudo-element tree is generated on the scope element, only its subtree participates, and the rest of the page keeps running -- including other scoped transitions at the same time.

const transition = card.startViewTransition(() => updateCard());
await transition.finished;
90/93scoped tests pass
at wpt e8ca14556268 (2026-11-20)
3reftests fail, all expected
(non-HTML-block scope roots)
0regressions, 0 flakes, 0 crashes
0flags, shipped

Run via the virtual/threaded/ prefix (the only virtual suite in VirtualTestSuites that owns external/wpt/css/css-view-transitions exclusively, so the physical path is always skipped on its own). Suite source is pinned to wpt e8ca14556268 (2026-11-20): on the build this baseline was measured on (chromium/src@553843c3bb5b), the scoped/ dir was rsync'd --ignore-existing against that upstream SHA so the suite runs at the same 93-test revision Firefox and WebKit are measured on. The import on chromium/src@origin/main (8910eb28f6aa) ships the identical 93-file scoped directory and encodes the three failures as crbug.com/561596305 [ Failure ], confirming this is the current chromium baseline, not an import-lag artefact. Numbers come straight from out/Default/layout-test-results/full_results.json: num_passes: 90, num_flaky: 0, num_failures_by_type: { PASS: 90, FAIL: 3, CRASH: 0, TIMEOUT: 0, SKIP: 0 }.

Overlay reftest breakdown · the group Firefox/WebKit are tracking

The user-facing value of this baseline card: which of the "non-HTML-block scope root" and "shadow roots tentative" overlay reftests already pass on Chromium main, so Firefox and WebKit reviewers can see which cases are upstream behavior and which are still pending across the board.

reftestChromium mainwhat it checks
scope-svg-root.htmlFAIL Scope is an <svg> root; overlay pseudo tree must still anchor under the SVG frame. Same ProcessChildren-style bug as Gecko/WebKit.
scope-mathml-root.htmlFAIL Scope is a MathML root. Same shape of failure as above.
scope-replaced-elements.htmlFAIL Scope is a replaced element (img, canvas, video, iframe, embed, audio variants); replaced frames do not accept children, so the overlay never gets built.
shadow-roots-auto-and-match-element.tentative.htmlPASS Tentative shadowRoots option interacting with match-element name masking.
shadow-roots-classes-groups-scopes.tentative.htmlPASS Tentative shadowRoots cross-tree group/scope classification.
shadow-roots-cross-tree-morph.tentative.htmlPASS Tentative shadowRoots morph across the host/shadow tree boundary.
shadow-roots-discovery.tentative.htmlPASS Tentative shadowRoots participant discovery.
shadow-roots-flat-tree-descendants.tentative.htmlPASS Tentative shadowRoots flat-tree descendant walk.
shadow-roots-nested-groups.tentative.htmlPASS Tentative shadowRoots nested groups and ::view-transition-group-children().
shadow-roots-options.tentative.htmlPASS Tentative shadowRoots option surface area.
shadow-roots-pairing-and-collisions.tentative.htmlPASS Tentative shadowRoots name collision handling.
hit-test-overflow-clipped-scope.htmlPASS Hit-testing through the overlay of an overflow:hidden-clipped scope.
occlusion-zero-opacity-scope.htmlPASS Overlay occlusion when the scope itself has opacity: 0.
scope-becomes-document-element.htmlPASS Scope element is promoted to the document element during the transition.

Chromium main passes all 8 tentative shadowRoots reftests (the block that grew the suite from 81 -> 93), and the three new non-shadow upstream tests (hit-test-overflow-clipped-scope, occlusion-zero-opacity-scope, scope-becomes-document-element). It fails exactly the three scope-{svg,mathml,replaced}-* reftests that Firefox and WebKit are also tracking -- the same "overlay pseudo tree never anchors under non-block scope frames" class of bug. Not hiding it: this baseline is 90, not 93.

Before / after

Same build, same demo page. The Chromium baseline has no flag to toggle, so Before and After are identical copies of the same capture. The point of the pair is to keep the four-tab layout aligned with the Firefox and WebKit samplers in this project, where before/after do differ.

before -- scope morph, baseline capturebaseline
four tiles re-grid inside the scope
after -- same build, identical capturebaseline
four tiles re-grid inside the scope
before -- clipping/overflow demo, baseline capturebaseline
scope escapes its own clip, ancestor clip holds
after -- same build, identical capturebaseline
scope escapes its own clip, ancestor clip holds

Video

Live motion of the same two demos, recorded with ffmpeg -f x11grab against content_shell inside Xvfb (800x600 window, 10 s each, VP9 + H.264). Baseline, so Before and After are byte-for-byte the same file. The clip video is mirrored to nested-clip-before/after.{webm,mp4} for the vt/ overview page.

before -- scoped morph, baseline capturebaseline
after -- same build, identical capturebaseline

A second capture page exercises the clipping and overflow rules: the pseudo tree escapes the scope's own bounds but stays clipped by an ancestor overflow: hidden, and a scope with overflow: auto gains no scroll range from the overlay.

before -- clip demo, baseline capturebaseline
after -- same build, identical capturebaseline

Live sampler

Element-scoped morph · card.startViewTransition(mutate)
A
B
C
D
before
after
Four named tiles re-grid inside the scope: the old and new snapshots cross-fade and the groups morph between the two layouts. No flag to toggle on this build, so the Rendered tab animates by default.
Scope inside a shadow root · adopted stylesheet styles the pseudos
before
after
The scope lives inside a shadow root and its view-transition-name comes from an adopted stylesheet. The dynamic ::view-transition styles resolve in the shadow root's style scope, so the part morphs from square to circle across the card.
Scope clip escaped, ancestor clip applied
VT
scope
before
after
The tile flies beyond the green scope bounds -- the pseudo tree escapes the scope's own clip -- but the dashed amber ancestor with overflow: hidden still cuts it at the edge. Mid-flight the circle is visibly outside the scope and clipped by the ancestor.
Overlay adds no scrollable overflow to the scope
VT
scroll 0,0
before
after
The scope has overflow: auto with styled (non-overlay) scrollbars. The flying group leaves the content box during the morph, yet no scrollbars appear and scrollLeft/scrollTop stay 0,0: the overlay is excluded from the scope's scrollable overflow.

What this baseline covers

The Chromium shipping implementation of scoped view transitions, served as a stable peer to the Firefox (bug 2075158) and WebKit (PR #81741) samplers. 90 of 93 tests in external/wpt/css/css-view-transitions/scoped at wpt e8ca14556268 (2026-11-20) pass under virtual/threaded/ on this build, with zero regressions, zero flakes, zero unexpected results -- the 3 failures are already encoded as FAIL SKIP expectations and match the group Firefox and WebKit are tracking. There is no patch to toggle: the feature is behind no flag.

Result

baseline (this build)
scoped WPT suite (virtual/threaded/.../scoped, 93 tests at wpt e8ca14556268) 90 of 93 pass; 0 flakes, 0 crashes, 0 timeouts; 3 failures encoded as [ Failure ] in chromium/src@origin/main under crbug.com/561596305
expected FAIL reftests 3: scope-svg-root, scope-mathml-root, scope-replaced-elements
feature flag requirednone; shipped
Firefox implementation (bug 2075158) 77 of 93 at wpt e8ca14556268; see firefox sampler
WebKit implementation (PR #81741) 87 of 93 at wpt e8ca14556268; see webkit sampler
Commands behind the numbers and captures
# pin the suite to a named wpt sha so the denominator is auditable
git clone --depth=1 https://github.com/web-platform-tests/wpt.git /tmp/wpt
#   -> /tmp/wpt HEAD = e8ca14556268 (2026-11-20)

SDIR=third_party/blink/web_tests/external/wpt/css/css-view-transitions/scoped
rsync -av --ignore-existing /tmp/wpt/css/css-view-transitions/scoped/ $SDIR/
diff -rq $SDIR /tmp/wpt/css/css-view-transitions/scoped/   # must be empty
third_party/wpt_tools/wpt/wpt manifest \
    --path third_party/blink/web_tests/external/wpt/MANIFEST.json \
    --tests-root third_party/blink/web_tests/external/wpt

third_party/blink/tools/run_web_tests.py -t Default \
    --skipped=ignore --no-retry-failures --jobs=8 \
    virtual/threaded/external/wpt/css/css-view-transitions/scoped
# num_failures_by_type: { SKIP: 0, PASS: 90, FAIL: 3, CRASH: 0, TIMEOUT: 0 }
# num_passes: 90  num_flaky: 0  (no retries, full directory, one job set)

# per-case still captures (chromium headless, same binary as the videos)
/usr/bin/chromium --headless=new --disable-gpu --hide-scrollbars --no-sandbox \
    --window-size=WxH --screenshot=before-X.png file://$PWD/capture-X.html
cp before-X.png after-X.png     # baseline = no patch to diff against

# video pair (Xvfb + content_shell + ffmpeg x11grab, 10s each)
Xvfb :77 -screen 0 800x600x24 &
DISPLAY=:77 out/Default/content_shell --no-sandbox --disable-gpu \
    --content-shell-host-window-size=800x600 file://$PWD/nested-clip.html &
ffmpeg -y -video_size 800x600 -framerate 30 -f x11grab -i :77.0 -t 10 \
    -c:v libvpx-vp9 -b:v 1200k -pix_fmt yuv420p -an before-clip.webm
ffmpeg -y -i before-clip.webm -c:v libx264 -pix_fmt yuv420p \
    -movflags +faststart before-clip.mp4
cp before-clip.webm after-clip.webm   # same build, no toggle
cp before-clip.mp4  after-clip.mp4