[teamleaderleo]

In progress: CI running. Independent review passed; its one real finding (inline frame waits followed by reads in pr-dock and tab-row) is fixed in the second commit.

What

A requestAnimationFrame callback runs before the frame's style, layout, ResizeObserver callbacks and paint. A test that reads layout there can see a state the frame never paints, such as the transcript's last row 30 px low on an opened chat's first frame. This PR makes reading after paint the one way frame tests sample.

• webviews/test/support/paintedFrames.ts (new): installPaintedFrames(page) defines two page globals:

  • window.cmuxAfterPaint(read, drive?) calls read(frame) after each painted frame until it returns false. drive(frame) runs in the frame's callbacks first. It returns a stop function.
  • window.cmuxPaintFrame(drive?) resolves with the frame time once that frame has painted.

  A read runs in a MessageChannel task posted from the frame callback, so it runs after the frame renders.

• Migrated every browser test that sampled in frame callbacks: agent-pane-stream-jitter, -chat-column, -history-paging, -queue-collapse, -reading-place, -steer, -switch-frames, chat-open-perf, code-block-highlight-height, hover-row-sweep, new-tab-composer-still, new-tab-first-frame-project, new-tab-warm-first-frame, reopened-chat-history, settings-open-frames, transcript-fling, transcript-fold-anchor-browser, plus agent-pane-pr-dock and agent-pane-tab-row, which awaited two frames inside the page and then read layout.

• Guard: webviews/scripts/check-frame-reads.ts runs from check-browser-test-lanes. It fails a browser test that does any of these inside a requestAnimationFrame callback, unless a // raf-read: <reason> comment says why:

  • reads layout there;
  • reschedules a sampler from there;
  • wraps a frame in a named awaiter (const frame = () => new Promise(r => requestAnimationFrame(r)));
  • awaits a frame inline and then reads layout in the same function.

  Unit tests are in test/check-frame-reads.test.ts.

• Pending (FRAME_READS_PENDING): these wait for other lanes' open PRs, and the guard fails once one of them reads after paint so the entry gets removed:

  • agent-pane-scrollbar-full-height (Agent chat: Jump to latest glides to the end after the composer grew #23257)
  • thread-minimap-feel and thread-minimap-motion (cmux NEXT: the New Tab page keeps up with a resize, and animations share one short scale #23092); I move these in a follow-up once cmux NEXT: the New Tab page keeps up with a resize, and animations share one short scale #23092 lands.

• New regression tests on the helper:

  • A chat with rich history opens at its latest line and holds it (agent-pane-stream-jitter).
  • A long chat opened at its end holds its latest line on every painted frame while history pages in above it (agent-pane-history-paging).
  • Thread switching is already covered by chat-open-perf, which now samples post-paint.

Verification

• Every changed browser test ran locally on cmux16 with CMUX_BROWSER_TESTS=1 and all of them pass, except settings-open-frames. Its dev server doesn't start on cmux16, and base fails the same way here, so CI is the judge for that file.
• agent-pane-chat-column failed 1 of 13 once ("a streaming reply keeps the view at its end…", resting 12 px off). It passed 13/13 on the rerun, and base passed 13/13. I'm watching it in CI.
• Guard, unit tests and vp check are clean.

Agent session: cmux16

🤖 Generated with Claude Code

[teamleaderleo]

Owned by scroll-jitter-2 on cmux16 · session

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review e3a2b63: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr23374 -f ref=scroll-jitter-2/post-paint-frames -f platform=macos -f backend_mode=local

Passes: CI passes on e3a2b63a15.

CI passes on e3a2b63a15 (run 39601299709 attempt 1).

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

web and size checks: green

CI fast guards failed on 3561e42c50 (https://github.com/manaflow-ai/cmux/actions/runs/39597389773). It does not block the merge; a red guard merged into main breaks it for every open PR.
The log named no failed step; see the run.

Agents: python3 scripts/ci/guard_attribution.py fix applies the mechanical fixes locally.

Tour of e3a2b63a15 (run): no media. The tour did not run: only webview code changed (23 path(s)), which the gallery advisory run checks in Linux Chromium on the pull request (its artifact holds any captures); add the media-tour-mac label to tour it on a Mac.

[teamleaderleo]

Bead cx-o97ov created for this PR by the cmux coordinator.

[blacksmith-sh]

Found 2 test failures on Blacksmith runners:

Failures

Test	View Logs
test/thread-minimap-feel.test.ts/thread-minimap-feel.test	View Logs
test/thread-minimap-feel.test.ts/thread-minimap-feel.test	View Logs

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

[teamleaderleo]

base-red-watch: cmux-next web / web / react-apps-check (2) went red on feat-cmux-next in https://github.com/manaflow-ai/cmux/actions/runs/39601369539 (first red at 8083ae2, last green at d6dcbed). Suspect: #23374 (commit 8083ae2, which touches webviews/test/chat-open-perf.test.ts). If it is yours, fix forward on feat-cmux-next.

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3960833477001 (4482e41d31)
• cmux NEXT nightly 1.0.0-nightly.3960894798201 (0b1a93e9c0)
• cmux NEXT nightly 1.0.0-nightly.3961101575901 (101cc33db2)