Repository navigation
Transcript: a followed reply moves only by what is drawn - #20697
Merged
Merged
Conversation
Base flake on loaded CI runners (#20644, #20633): "keeps the view at its end" saw the text sink a few pixels, and "scrolls smoothly" saw a 24px step. The streaming tests now run with Chromium's CPU throttled 4x, the load the runners had, so every run reproduces the first. A smoothness step is measured over the time its frame took, so a loaded page that paints fewer frames is not read as a jump. red: bun test test/agent-pane-chat-column.test.ts (browser lane, builder) -> 'keeps the view at its end' fails 2/2 at movedDown Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two causes of the chat-column flake on loaded runners: - A streaming reply's text draws after the version that holds it, so the row was placed by an estimate of the new version (a line or more ahead of the text), then by its drawn height a frame later: the view scrolled ahead and back at every chunk. A row now keeps its last drawn height until its new version is drawn; growth comes from what is drawn. - The queue cleared --acpmux-queue-space at every change before setting it again; the transcript's layout pass could run between the two, read the taller view and scroll the followed reply back by a row. The space now stays while the list redraws and goes with the list. red: bun test test/agent-pane-chat-column.test.ts (loaded, builder) -> 'keeps the view at its end' fails 2/2 green: same command -> 7 pass, 0 fail, 3/3 runs; related transcript and queue browser tests 31/31 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
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. Comment |
…pt/glide-retarget # Conflicts: # webviews/test/agent-pane-chat-column.test.ts
…y frame rate From the #20697 review: dividing a step by its frame's time let a snap through on a slow frame, and a fixed px-per-painted-frame bound depends on frame timing (a loaded frame holds two lines of growth). The check now reads, per frame, how far the followed row moved under the thread's glide and the glides the thread started since the frame before: a move the glides do not account for is a step drawn at once. Checked: with glide() disabled the test fails at snapped(); with it, the file passes 8/8 three times on a 4x-throttled page (builder). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bugbot needs on-demand usage enabledBugbot uses usage-based billing for this team and requires on-demand usage to be enabled. A team admin can enable on-demand usage in the Cursor dashboard. |
…n is drawn Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Collaborator
Author
|
Bead |
Contributor
|
all web/size checks green |
Contributor
|
Passes: CI passes on CI passes on Written by |
Contributor
|
|
Contributor
Gallery25 states changed: agent-pane.transcript/long-code, agent-pane.transcript/shell-grouped-commands, agent-pane.transcript/copy-last-code-block, agent-pane.transcript/archive-chat-key, agent-pane.transcript/mark-unread-menu, ... · 1 play check failed Diff page and renders (the run's
19 more on the diff page. Live preview (follows this branch; start it with
Play checks failed (0 px anchor movement, 0 layout shift, no frame over 33 ms unless the entry gives a reason):
77 nondeterministic states differed from a second render of the same head (a clock or fixture leak) and are not counted as changes. |
…pt/glide-retarget
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
…red) The #20467 test 'a pane resized at every frame keeps its last line at the bottom' failed on a loaded CI runner (#20697 react-apps-check, https://github.com/manaflow-ai/cmux/actions/runs/38614942014/job/116065309559: one frame with the last line 341px off). The chat-column tests now all run at 6x CPU throttle. red: bun test test/agent-pane-chat-column.test.ts -t 'pane resized' (builder) -> 16 of 16 fail Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… before its paint The chat-column watcher samples each frame in a task after its animation frame. On a loaded page a viewport resize can land between that frame and the sample: the read forces a layout at the new width (the reply rewrapped 341px taller, the spacer and offset still placed for the old width) that the page never painted, and the next frame shows it settled. The rendering steps apply a resize before animation-frame callbacks and no task runs between them and the paint, so a frame paints at the viewport its callback saw; a sample whose viewport changed since is not a frame and is skipped. The assertions are unchanged. red: bun test test/agent-pane-chat-column.test.ts -t 'pane resized' (6x, builder) -> 16 of 16 fail green: same -> 0 of 16 fail; whole file 8/8 at 6x (5 runs); bun run check clean Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…pt/glide-retarget
![cmux-autofix[bot]](https://github.githubassets.com/images/gravatars/gravatar-user-420.png){kind=avatar}
- vp fmt webviews/src/pages/home-channels/HomeChannelsPage.tsx Autofix-Run: https://github.com/manaflow-ai/cmux/actions/runs/38637927193
Bugbot needs on-demand usage enabledBugbot uses usage-based billing for this team and requires on-demand usage to be enabled. A team admin can enable on-demand usage in the Cursor dashboard. |
…pt/glide-retarget # Conflicts: # webviews/src/agent-session/acpmux/ComposerQueue.tsx
…transcript/glide-retarget
…pt/glide-retarget
- vp fmt webviews/src/pages/home-channels/cloudHome.ts Autofix-Run: https://github.com/manaflow-ai/cmux/actions/runs/38647133983
Bugbot needs on-demand usage enabledBugbot uses usage-based billing for this team and requires on-demand usage to be enabled. A team admin can enable on-demand usage in the Cursor dashboard. |
teamleaderleo
added a commit
that referenced
this pull request
Oct 14, 2026
… option fades Leo on #20610 (2026-10-14): - The transcript's rows are placed under the thread's 16px top padding, but its height (the spacer, and the end the view follows to) left the padding out: the last row ended 16px under the view, and with its own 8px padding the action row's time, Copy and Fork sat 8px under the composer. The height now counts the padding and 12px of room, so the action row clears the composer by 20px. - The fade option only faded the step rows that appeared; the rows the fold moved jumped, and the steps that left faded out under them, so nothing visibly moved. Every row that changes place now fades in at its new place over 120-180 ms (Reduce Motion keeps its 100 ms cap), and the leaving copies are dropped in that mode. - With the true end, a streaming reply's estimate overshoot (fixed the same way in #20697) moved the followed text back each chunk; a row keeps its last drawn height until its new version is drawn. green: bun test test/agent-pane-chat-column.test.ts -t 'never under the composer|each fold motion' -> 2 pass agent-pane-*, transcript-*, chat-find*, queue-* browser tests: 127 pass; the 3 locale first-paint fails need a built Resources/agent-pane bundle (absent on the builder) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
cmux-next tour of Path-selected interactions
Agent pane (before): 3 ok
Screenshots (3)1. Launch
Screenshots (3)1. Launch
Screenshots (11)1. Launch
Screenshots (11)1. Launch Tours are data in |
This was referenced Oct 15, 2026
teamleaderleo
added a commit
that referenced
this pull request
Oct 15, 2026
…6fbz) (#20733) * test: a webviews test file can declare a longer limit (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: editor-page declares a 240 s limit; the runner honors per-file limits (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: browser tests warm their Vite server before the first page load (cx-v6fbz) Vite prints its port before it has transformed a module, so a test's first goto paid for the whole module graph inside its 30 s navigation timeout. On a busy CI runner that failed chat-find and text-field-focus (#20697), and the gallery-frame and agent-pane browser warm-ups hit the same wait (#20681). warmDevServer requests the page and every module it imports first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: diff-virtualization clicks a tree row only while it is still under the pointer (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * ci: react-apps-check job cap 30 -> 40 min (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: the composer browser tests new on base warm their Vite server too (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * check: a test that starts a Vite dev server must warm it (cx-v6fbz) The browser-lane check now fails a test file that starts a dev server (vite.listen() or a vp dev spawn) without warmViteServer/warmDevServer. On base it flags 80 of 240 test files; on this head none. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: a declared webviews file limit never shortens the default (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: the longer of a declared and the default webviews file limit wins (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: review follow-ups: settings-scroll warms after its readiness poll; the diff row check climbs out of shadow roots (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: agent-pane-bottom-stack (new on base) warms its Vite server (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: text-field-focus declares a 240 s limit; its gallery warm-up and 22 loads took about 115 s (cx-v6fbz) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * ci: move the react-apps-check cap raise to its own PR The ci-web.yml edit widens this PR's web scope to the web/ unit shards, which are red on base. The cap raise lands separately. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test: give dev-server files 240 s and settle abandoned warm-up visits Review follow-up: the spawned-server files and menu-focus warm the whole gallery graph before their browser warm-up, which can pass the default 120 s file limit on a loaded runner. A failed warm-up visit no longer leaves the other in-flight visits as unhandled rejections. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes the base flake in
webviews/test/agent-pane-chat-column.test.ts: "scrolls smoothly: its last line holds still" (24 > 23) and "a streaming reply keeps the view at its end, through a growing command and a queue". Both failed under CI load (seen on #20644 and #20633) and passed locally.The load exposed two real scroll jitters:
App.tsx,heightAt).The queue let the followed reply slip back a row.Superseded: base Agent chat: one bottom stack, so the queue never covers the PR dock #20521 (one bottom stack) removed--acpmux-queue-space, so that part of the fix was dropped in the base merge.The test now loads the page itself with
Emulation.setCPUThrottlingRateat 6×, so a local run reproduces what CI sees. The smoothness test no longer counts pixels per painted frame. That bound depends on frame timing: one loaded frame can hold two lines of growth. Instead it checks, for each frame, that the followed row moved exactly as far as the glides the thread started since the frame before. A step drawn at once fails at any frame rate. Withglide()disabled, the test fails on that check. It still checks that the text never moves down or back.The live-resize test (#20467) flaked too, but in the test, not the app. The watcher samples each frame in a task after its animation frame. On a loaded page a viewport resize could land between that frame and the sample. The sample's forced layout then read the new width (the reply rewrapped 341 px taller, the spacer and offset still placed for the old width), which the page never painted; the next frame was settled. A frame paints at the viewport its animation-frame callback saw, so a sample whose viewport changed since is skipped. The assertions are unchanged.
Testing
All runs were on the Linux builder, with the throttled page.
bun test test/agent-pane-chat-column.test.tsbun test test/agent-pane-chat-column.test.ts -t 'pane resized'Also run on the fix commit:
bun run check: clean.Changelog
🤖 Generated with Claude Code