cmux-next agent pane: don't snap a small scroll-up back to the latest row

[teamleaderleo]

If a reader at the latest row of the agent pane scrolls up slightly while a row below finishes drawing shorter, they now stay where they scrolled. Until now the transcript could snap them back to the latest row.

The anchor effect detects when the browser has clamped the scroll offset after the content shrank. Before this change it treated any offset within 1px of the layout's fractional end as clamped, because the browser rounds scrollHeight. That 1px also caught a real scroll-up of U when the shrink S satisfied S < U ≤ S + 1. A clamped offset lands exactly on the scroller's own end, scrollHeight - clientHeight, so the check now compares against that with a 0.5px allowance. This is a follow-up from the #16491 review.

Stacked on #16497 (same layout effect). Once that merges, this branch takes feat-cmux-next and the diff narrows to the two commits below.

Validation

cd webviews && bun test src/agent-session/

• Red at 4db8286: a small scroll-up at the latest row survives a row below drawing shorter fails. The reader scrolls up 10.75px, then the latest row draws 10px shorter. The view snaps to the new end, 8474, instead of staying at 8473.25. The fake viewport now reports scrollHeight.
• Green at 469359e: 147 of 147 pass. bunx tsc --noEmit -p . and oxlint --deny-warnings are clean, and build-agent-pane-web.sh --check says the bundle is current.

Not checked in a running app. The test's fake viewport assumes a clamp lands exactly on scrollHeight - clientHeight. WebKit keeps element scroll positions on whole pixels, so that should hold, but nobody has checked it live.

Changelog

none (cmux-next is unreleased)

🤖 Generated with Claude Code

https://claude.ai/code/session_01RYQHfug1ZVQDp4eWgwVUtD

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

---

Summary by cubic

Fixes the agent pane snapping a small scroll-up back to the latest row when a row below draws shorter, and keeps the latest row in view when the viewport shrinks.

Bug Fixes

• A small scroll-up at the latest row no longer snaps back when a row below draws shorter; the clamp check now uses the scroller's own end rather than a fractional layout height.
• Shrinking the viewport (e.g., a permission card appearing) keeps the latest row in view; an offset that hasn't moved keeps its recorded at-latest state.
• Merged the latest feat-cmux-next changes to keep this branch current.

^{Written for commit e72c736. Summary will update on new commits.}

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository: manaflow-ai/cmux/.coderabbit.yaml

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 2eb942da-7043-44a5-9ee4-595a601d105d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Autopilot is currently an internal CodeRabbit preview.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[cubic-dev-ai]

No issues found across 3 files

_{Re-trigger cubic}

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.

[teamleaderleo]

Subagent review at e72c736: LGTM.

• WebKit clamps an overflow scroller to scrollableContentsSize - visibleSize. That's the same rounded integer scrollHeight returns, and overflow scroll positions are whole pixels even at DPR 2. So a real clamp lands exactly on scrollHeight - clientHeight, and the 0.5 allowance catches it.
• scrollHeight is the browser's real scroll extent. It's more accurate than the model's totalHeight.
• Reading scrollTop first already forces layout, so scrollHeight reflects the spacer that was just committed.
• Residual risk, not checked on a device and not new: a clamp is missed only if WebKit's visible height rounded differently from clientHeight. The old -1 tolerance on an unrounded total had a similar edge.
• Test nits: the repro scrolls up a fractional 10.75px, and the fake scrollHeight isn't rounded. The logic is still modelled correctly.
• 147/147 bun test, tsc and oxlint clean, bundle current. The new test fails on base App.tsx (8474 against 8473.25) and passes at the head.