[teamleaderleo]

Leo, 2026-11-03: "im still getting these weird layout shift things when im already scrolled up due to idk tool calls shifting/messages coming in maybe that should not happen / likewise i don't want to see flickers of the session info or anything when there's random re-renders EVER"

This PR is the Session Info half: an update that changes nothing Session Info shows (a streamed token, a tool call that makes nothing, a usage tick) draws nothing in it. Keeping the transcript in place while scrolled up is another lane's work.

Root causes

• SessionInfo.tsx SessionInfoPresence: every acpmux snapshot is made of new objects, so the lists it derives (summary, made, links, documents, agents, transcriptTasks, processes, waiting, snapshot.sessions, peerKinds, grants, the started chats and workflows) were new on every token, and the <SessionInfo> sections, which were not memoized, all rendered again on every update. They now keep their identity while they hold the same data (new session-info/sameValue.ts, useSameValue), the sections render through a memoized SessionInfoSections, and the App's callbacks go through useStableCallback.
• The same component computed the chat column with chatColumnOf(...), which returns a new object when the display sets its own width or the page is zoomed. The placement layout effect depends on it, so on every render it rewrote the shell's --acpmux-session-info-column and --acpmux-header-room and dispatched CHAT_LAYOUT_EVENT. The column is now kept by value.
• The rows given to the sections (for the chat's gallery) stay the same until a row is added, the gallery count changes or the panel lists another agent. The tool-call ids a remote chat's background task row looks up are read in Presence, so a call added to an existing row still shows up (this was the review's BLOCK at d922a51).

Tests

• webviews/test/session-info-no-flicker.test.tsx (jsdom): mounts the pane with Session Info open, then applies 50 updates: a token at a time, a tool call every fifth, and a usage tick with new objects holding the same values. It asserts no DOM mutation inside the panel (MutationObserver, takeRecords included), the same section elements, an unchanged frame style, and zero renders of OutputsSection/EnvironmentSection after mount (both spies count at mount as a positive control). Red at 95aa733 (all 50 updates rendered OutputsSection), green here. A second test (a remote chat's task row gets its Show call when the call lands in the existing activity row) is red at d922a51 and green here.
• webviews/test/session-info-no-flicker-browser.test.ts + test/browser/session-info-no-flicker.harness.tsx (Chromium): 50 updates at 1400/1000/600 px (float, push, overlay). It asserts no mutation in the panel, no layout-shift entry with a source inside it, and one bounding rect throughout.

Performance

metric	before	after	delta	harness or run link
Session Info sections rendered per streamed update (jsdom, 50 updates)	50 / 50	0 / 50	-100%	test/session-info-no-flicker.test.tsx
ui.session_info.cls (open play, Chromium)	0.0000	see the follow-up comment		bun bench/ui-stability.bench.ts 3 --only session_info, base run https://github.com/manaflow-ai/cmux/actions/runs/41593125846

ui.session_info.cls was already 0.0000 at this PR's base (the 0.028 reading predates it). That bench measures the open play, not streaming, so this change is not expected to move it.

No new timer, observer or animation loop.

Observed / Not run

• Ran locally: bun test on every test/session-info*.test.tsx, background-task-arrival, native-session-info-toggle and chat-header-every-tool-acts, all green except session-info-dead-clicks ("every line of the tray…"), which fails the same way at base. scripts/web-lint.sh is clean except check:design-tokens on cloudStart.css, a base red.
• The browser test ran on the fleet at 95aa733: green at 1400 px. At 1000/600 px it recorded one style write on the card, caused by the harness itself: its dock's live line first appeared inside the observed window. The harness now starts that line before it observes. A fleet run of the browser test and the bench at f8af7fb is in progress.
• Native panel: unchanged. It already sends its model only when the JSON changes.

Agent session: agent session

🤖 Generated with Claude Code

[github-actions]

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

Review f8af7fb: 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=pr33203 -f ref=lane/si-no-flicker -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4160373381601 (d485585f77)
• cmux NEXT nightly 1.0.0-nightly.4160614853601 (0e8eb9bf99)

[blacksmith-sh]

Found 7 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmp39hc0tdv/summary-missing.json (live):/summary-missing.json (live):	View Logs
/tmp/tmpviba5geh/summary-fail.json (live):/summary-fail.json (live):	View Logs
[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows/ [new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows	View Logs
[new_workspaces] the window draws no frames/[new_workspaces] the window draws no frames	View Logs
[tabs] debug.tab_switch timed 0 of 40 switches/ [tabs] debug.tab_switch timed 0 of 40 switches	View Logs
[tabs] the window draws no frames/[tabs] the window draws no frames	View Logs
build_only must be a conjunctive exclusion on the helper, signing, and publication jobs , and must not gate the unsigned app build/ build_only must be a conjunctive exclusion on the helper, signing, and publication jobs , and must not gate the unsigned app build	View Logs

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

[teamleaderleo]

Review (claude-opus-5-5): SHIP at f8af7fb

Review (subagent claude-opus-5-5): SHIP at f8af7fb

Visual verdict: unchanged by this diff (no UI drawn differently; Session Info only stops redrawing for updates that change nothing it shows).

Checked the fix for the earlier BLOCK at d922a51 (a remote chat's background task row lost Show call when a call landed in the existing activity row): SessionInfoPresence now reads the call ids itself through useSameValue and hands them to the sections; the remote-call test is red at d922a51 and green here. Also checked the structural-equality helper, the memoized sections and the column kept by value. session-info-dead-clicks fails the same way on base.

Follow-ups:

• [follow-up] read the remote tool-call ids only when a background command has a toolCallId.
• [follow-up] refresh the kept gallery rows when an image changes in place (same row count and gallery count).
• [follow-up] make the remaining callback props to the sections stable, or test that they are.