fix(presentation-editor): bail on clicks in inter-page gap (SD-2749)

[tupizz]

Summary

• Fixes SD-2749: clicking in the vertical gap between two pages caused the viewer to jump to a different page.
• One guard line added to EditorInputManager.#handlePointerDown that bails when no .superdoc-page is under the cursor (matches the existing in-margin guard from SD-2356).
• Defensive layout?.pages?.[pageIndex] access in isOutsidePageBodyContent to prevent a latent crash when layout.pages is missing.

Root cause

PointerNormalization uses elementsFromPoint to find the .superdoc-page under the cursor. In the inter-page gap there is none, so it returns pageIndex: undefined. The existing isOutsidePageBodyContent guard from SD-2356 short-circuits to false when pageIndex is not finite, so the pointer flow fell through to the geometry fallback (clickToPositionGeometry → snapToNearestFragment), which always returns some nearest fragment. TextSelection.create + ProseMirror scrollIntoView then jumped the page.

clickToPositionDom        → null (no DOM hit)
findPageElement           → null (no page under cursor)
clickToPositionGeometry   → hitTestPage returns nearest page by center distance
  → hitTestFragment       → null
  → snapToNearestFragment → returns fragment closest to the click
→ TextSelection.create    → editor.view.dispatch
→ ProseMirror scrollIntoView → page jumps

The fix is symmetric with SD-2356: clicking in the gap is conceptually the same as clicking in a page margin — preserve selection and scroll.

Files changed

File	Change
EditorInputManager.ts	Extended pointerdown guard to bail when pageIndex is undefined; added defensive layout?.pages?.[pageIndex]
EditorInputManager.pageMarginClick.test.ts	New regression test for SD-2749 (RED → GREEN)
EditorInputManager.{activeCommentClick,dragAutoScroll,footnoteClick,structuredContent}.test.ts	Mocks updated to return realistic { x, y, pageIndex, pageLocalY } matching the production contract
PresentationEditor.test.ts	Two integration tests stub document.elementsFromPoint so happy-dom returns the simulated .superdoc-page under the click point

Test plan

• pnpm --filter super-editor test — 18,421/18,421 pass (1,240 files)
• Targeted: EditorInputManager.pageMarginClick.test.ts (3 tests pass; new SD-2749 test fails on main, passes on this branch)
• Browser end-to-end with IT-923 Headers.docx (7 pages):
  • Gap click at (600, 366) between pages 0 & 1: scrollTop and selection.from UNCHANGED
  • In-page text click at (600, 500): selection moves to clicked text, no scroll jump
• Manual confirmation in Word-style customer doc (NVCA-Model-IRA.docx)
• Manual confirmation that the existing SD-2356 fix (margin clicks) still works

[linear]

SD-2749 BUG: Clicking Near Page Border Jumps Documents to Different Page

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[tupizz]

Follow-up: extended fix for the header→body click variant of SD-2749.

Repro confirmed on origin/main (and on previous commit on this branch):

• dblclick header → enters editing session
• type some text in header
• scroll down a few pages
• click in middle of any body paragraph
• → scrollTop jumps from 3500 → 97 (page snaps back to where the header is)

Root cause: #handlePointerDown captured sessionMode, useActiveSurfaceHitTest, and editor locals at function start. When #handleClickInHeaderFooterMode exits the session for a bodyContent click and returns false, those locals stayed stale. Selection then got dispatched on the header editor → ProseMirror's scrollIntoView pulled the viewport back to the (now off-screen) header.

Fix (commit 57c46e5): convert the three locals to let and re-derive them from a fresh getHeaderFooterSession() read after the H/F handler returns. If the session is now 'body' and we're not in a note session, point editor at bodyEditor and recompute useActiveSurfaceHitTest.

Tests added EditorInputManager.headerFooterBodyClick.test.ts with one regression test asserting that body view.dispatch is called and header view.dispatch is NOT called for the post-H/F body click.

Verified end-to-end in the browser:

Scenario	Before fix	After fix
Dblclick header → edit → scroll → body click	scrollTop 3500 → 97 (-3403)	scrollTop unchanged (delta 0)
Gap click between pages (original SD-2749)	(already fixed)	scrollTop unchanged ✅
Normal in-page text click	works	selection moves correctly ✅

Full pnpm --filter super-editor test: 18,422 / 18,422 passing.

[chatgpt-codex-connector]

💡 Codex Review

superdoc/packages/super-editor/src/editors/v1/core/presentation-editor/tests/PresentationEditor.test.ts

Lines 2327 to 2329 in 57c46e5

	(document as unknown as { elementsFromPoint: (x: number, y: number) => Element[] }).elementsFromPoint = () => [
	fakePage,
	];

Restore document.elementsFromPoint after stubbing

This test mutates the global document.elementsFromPoint and never restores it, so all later tests in this file inherit a fake hit chain even after fakePage is removed from the DOM. Any subsequent test that relies on pointer normalization or page hit detection can get a fabricated pageIndex/page-local coordinate and pass or fail for the wrong reason. Please save the original function and restore it in cleanup (or use a scoped spy/mock) to avoid cross-test contamination.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[tupizz]

Addressed both P2 review comments in 8ddc44e.

1. [P2] Refresh story session after exiting H/F — correct catch. exitHeaderFooterMode synchronously clears the backing StoryPresentationSessionManager, so activeStorySession becomes stale right when I recomputed useActiveSurfaceHitTest from it. In production every H/F session is story-backed, so the old variable was always truthy → useActiveSurfaceHitTest = true → my SD-2749 page-gap guard was bypassed for the first click. Now I also re-read getActiveStorySession() after the H/F handler returns.

Strengthened the existing unit test to mock a story-backed session, and added a second test that exercises the inter-page gap path (pageIndex undefined) while an H/F session was just exited. The new test fails on the prior commit and passes after the fix, so the regression is now actually pinned.

2. [P2] Restore document.elementsFromPoint after stubbing — correct, my bad. The two PresentationEditor.test.ts tests that simulated the page-under-cursor were leaking the stub onto subsequent tests in the same file. Wrapped both in try/finally that restores the original (or deletes the stub if there wasn't one — happy-dom doesn't ship the API).

Verified end-to-end:

• pnpm --filter super-editor test — 18,423 / 18,423 passing (one more test than before)
• Browser repro (header dblclick → edit → scroll → body click) → scrollTop delta = 0

[caio-pizzol]

@tupizz good call re-reading the story session after H/F exit, that path was easy to miss :)

i pushed a sibling regression test for the note-session × gap path, and ran render-parity against the merge-base (no rendering delta).

lgtm.

[superdoc-bot]

🎉 This PR is included in @superdoc-dev/mcp v0.3.0-next.63

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in @superdoc-dev/react v1.2.0-next.105

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in superdoc-cli v0.8.0-next.79

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in vscode-ext v2.3.0-next.107

[superdoc-bot]

🎉 This PR is included in superdoc v1.30.0-next.61

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in superdoc-sdk v1.8.0-next.61

[superdoc-bot]

🎉 This PR is included in superdoc-cli v0.9.0

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in superdoc v1.32.0

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in @superdoc-dev/mcp v0.4.0

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in @superdoc-dev/react v1.3.0

The release is available on GitHub release

[superdoc-bot]

🎉 This PR is included in vscode-ext v2.4.0