fix(shader-transitions): pin cloned scene roots to composition dimensions

[vanceingalls]

What

Pins cloned scene roots in the page-side shader compositor (engineModePageComposite.ts) to the exact box each measures via getBoundingClientRect() while still live in the document — not the composition's full pixel size. New pure exported helper clonePinStyleFor(rect) computes the style values from that measured rect.

Why

Wild report: page-side shader clones collapsed to 0x0 inside the canvas layoutsubtree when scenes relied only on position:absolute; inset:0 sizing (no containing-block dimensions exist there).

The live document already resolves inset:0 (and any authored explicit width/height) correctly against the real ancestor chain — measuring the source element's rect there and reapplying that exact box to the clone fixes the collapse without ever overriding an author's own sizing, unlike the original fix which unconditionally hardcoded every clone to the composition's full width/height regardless of the source scene's own dimensions.

Validation

New unit tests for clonePinStyleFor: proves a 0x0-collapse-prone inset:0 scene measures/reproduces the full frame box, AND proves an authored explicit width/height/offset (e.g. a picture-in-picture panel) survives verbatim rather than being overridden. 9/9 tests pass. shader-transitions package build clean.

Review history

Addressed review feedback: the previous version clobbered authored explicit dimensions. Now measures and reapplies the live-document box instead of hardcoding composition size.

🤖 Generated with Claude Code

[vanceingalls]

• fix(cli): surface video frame-injection failures in snapshot output #2255
• fix(shader-transitions): pin cloned scene roots to composition dimensions #2254 👈 (View in Graphite)
• fix(cli): fold canvas pixel hashes into the frozen-sweep fingerprint #2253
• chore(producer): resync stale committed puppeteer-core symlinks to lockfile version #2259
• fix(producer): disk-headroom gate before hdr raw frame pre-extraction #2256
• feat(cli): preview --browser-no-gpu for gpu-unstable hosts #2252
• fix(cli): settle post-seek font subset loads before audit screenshots #2251
• fix(engine): densify drawelement self-verify with parallel worker count #2250
• fix(producer): withhold de parallel router below a machine ram floor #2249
• fix(producer,engine): gate drawelement on ancestor background-image + tail verify sample #2247
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[miga-heygen]

#2254 — pin cloned scene roots to composition dimensions

Verdict: LGTM 🟢

11 lines. Cloned scene roots in the page-side shader compositor get explicit dimensions (width: ${width}px; height: ${height}px) so scenes sized only by position:absolute; inset:0 don't resolve to 0x0 inside the staging canvas's layout subtree. Scenes with explicit dimensions are unaffected (the explicit values win).

Minimal, correct fix for a measured failure (blank transition textures).

Review by Miga

[miguel-heygen]

Blocker: the new clone pinning unconditionally assigns clone.style.width/height = composition dimensions (engineModePageComposite.ts). That overrides authored scene-root layout, contrary to the PR contract and its comment that explicit dimensions are unaffected. Please preserve authored dimensions (only pin when the root relies on auto/zero sizing, or otherwise establish the containing block without overwriting explicit author styles) and add a regression test proving an authored width/height survives.

[james-russo-rames-d-jusso]

Reviewed at 1d2aca6.

Right diagnosis: position:absolute; inset:0 needs a containing-block with dimensions, and the staging canvas's layout subtree doesn't provide one. Pinning to the composition's pixel size unconditionally fixes the collapse in both transitions.

One small clarification worth adding to the docstring:

Nit — "scenes that already carry explicit dimensions are unaffected". The block at packages/shader-transitions/src/engineModePageComposite.ts:250-256 overwrites clone.style.width/height/position/left/top unconditionally, so a scene with style="width: 500px; height: 500px" inline WILL be pinned to composition dims. In practice shader-transition scene roots are full-canvas so the visual output is unchanged for correctly-authored scenes, but the docstring wording implies pin-then-preserve semantics that the code doesn't actually implement. Either soften to "scenes that already match the composition dimensions render identically" or add a if (!clone.style.width) … guard if the intent was actually preservation.

Not blocking — happy to leave the code and just tweak the comment.

LGTM from my side.

— Review by Rames D Jusso

[miguel-heygen]

Re-reviewed current head 01350bf. The fix now measures each live scene root and pins the clone to that measured box; authored dimensions/offsets are preserved and regression tests cover both collapsed inset roots and explicit sizing. No required-check failures are present (Graphite mergeability pending only). APPROVE.

[jrusso1020]

Approving per the channel stamp-flow: Magi re-reviewed the fix and approved at 01350bf (authored-dim blocker resolved via measured-rect clone + regression tests), Rames-D had cleared it, CI clean.

The base branch was changed.

The base branch was changed.