cmux-next: diff viewer paints big diffs fast and gets the cmux NEXT look by teamleaderleo · Pull Request #23154 · manaflow-ai/cmux

teamleaderleo · 2026-10-21T07:40:02Z

The diff viewer (the chat header's Changes page and every other diff-page.html host) now draws its first rows of a big diff 2.5-5x sooner. It gets the cmux NEXT chrome: a glass toolbar, accent-tinted open menus, a lifted file list and a per-file change bar. Word-level highlights are now on by default.

What was slow

I measured from the moment the host's patch was ready, in headless Chromium on huge-blue. The page was test/browser/diff-virtualization.html, which runs the same mountDiffSurface path as the app. I took a CPU profile for each fixture.

A big file made the stream quadratic. createStreamingPatchFileSplitter appended every 256 KB chunk to its search buffer, and then searched it for the next diff --git. That flattened the whole V8 rope again on each chunk. On a 60 MB generated bundle, nextGitBoundaryIndex alone took 4.4 s before anything drew.
The highlight pool started late. The WorkerPoolContextProvider mounted only with the first files, and each worker then compiles Shiki's WASM (300-450 ms). Until every worker is up, @pierre/diffs draws file headers with no rows.
One long parse task. The first batch parsed on the main thread in one task, so worker replies and the first paint queued behind it.

What changed

diff-stream.ts: text that cannot contain a file boundary is held aside. It joins the search buffer only when a diff --git shows up (checked across the chunk seam) or at finish(). Each file is now flattened once. Parsing yields to the event loop every 12 ms through a MessageChannel post. scheduler.yield() measured worse, because its continuation runs ahead of the worker replies.
App.tsx / useRenderDiff.ts: the worker pool and the theme registration start as soon as the page knows a diff is coming, not when the first files arrive. The workers boot while the patch streams.
WorkerRenderOptionsSync.tsx (drawPlainBeforeWorkers): while the workers boot, the pool gets a Shiki highlighter on the JavaScript regex engine with no grammars. That highlighter draws plain rows in the right theme with no WASM. Once the workers are up they highlight as before. A changed @pierre/diffs that no longer has the field this reads reports unsupported, and the perf test fails on that.
Restyle, using existing tokens only (--cmux-accent, --surface-blur, --cmux-hairline, fg/bg mixes; no new literal colors):
- The toolbar is glass (--surface-blur over a 3% fg lift).
- The pills and round buttons have a hairline rim, a separate hover, and an accent tint while their menu is open.
- The files sidebar sits on its own lifted surface, with an accent-tinted selection.
- Each file header shows a 5-block added/deleted bar next to its counts, in --text-detail.
Word diffs default to on (diff.wordDiffs in DiffViewerSettingsSchema.swift and the webview default). They render as word-alt. No Debug toggle: the setting already exists in Settings and in the toolbar's options menu (Word diffs), so turning it off is one click. The settings exports (settings-schema.json, its source stamp, docs/mdm) were regenerated with scripts/cmux-next/regenerate-settings-exports.sh on cmux16, because the PR autofix's Mac job did not regenerate them on recent heads and its push job is skipped. The defaults snapshot is left to the base autofix (bot-only).
Review fixes (from an independent review; it also fuzzed the new splitter against the old one, 50,000 cases, no difference):
- Parse yields share one MessageChannel for the page, with a queue of waiters. A channel per yield stayed alive (a started, connected port is never collected), hundreds per big patch.
- The plain-rows Shiki instance is disposed once the pool replaces or clears it (the pool's next stats broadcast). Before, the pool overwrote it without disposing it.

Numbers

All numbers are in ms from the patch being ready, on huge-blue, headless Chromium.

Production build (vite build of the harness, best of 3):

fixture	first row, before -> after	first highlight, before -> after
default: 2,000 files	710 -> 253	788 -> 674
`mixed`: 2,000 files, every fifth line changed, then a 1,500-line rewrite (~10 MB)	1,181 -> 387	1,360 -> 1,046
`huge`: one 1,000,000-line bundle (~60 MB) ahead of 50 files	6,185 -> 1,195	6,250 -> 1,350

The new perf test itself (dev server, best of 2): base 7a4b464b375 vs this branch.

fixture	first row	first highlight	first row as a share of the full stream
`mixed`	1,129 -> 456	1,414 -> 1,086	78% -> 21%
`huge`	5,896 -> 1,149	5,959 -> 1,366

At 940faca2aad (with the review fixes, word diffs on), the same test on huge-blue: mixed first row 398 ms, first highlight 962 ms, complete 1,505 ms (first row at 26% of the stream); huge first row 1,216 ms, first highlight 1,471 ms.

Streaming the whole of mixed takes slightly longer (about 1.6-1.8 s against 1.3-1.9 s), because parsing now yields so the rows can paint.

Perf test and budget

webviews/test/diff-paint-perf.test.ts is new, uses requireBrowserLane, and is Chromium only. The webviews runner picks it up from the git index; there is no registry for webviews tests. A probe script marks, each frame, the first diff row inside the viewer, the first row with Shiki token colors, and the end of the stream. The test runs each fixture twice, keeps the best run, and asserts:

the budgets: mixed row < 1,500 and highlight < 3,300; huge row < 3,500 and highlight < 4,500. They are about 3x this branch's dev-server numbers, which leaves room for a slower CI runner. Base misses the huge ones by 2,400 ms or more.
on mixed, that the first row is drawn before 60% of the stream. This check does not depend on machine speed (base 78-86%, this branch 21-34%).
that the plain-rows-before-workers path actually ran (drawn), so a @pierre/diffs upgrade cannot silently undo it.

Base fails the test and this branch passes it, on huge-blue.

webviews/test/diff-stat-blocks.test.ts is also new and covers the header bar.

webviews/test/diff-stream-splitter.test.ts is also new. It checks that the splitter returns the same files for every chunking, including a diff --git marker split across chunks. It also checks that a big file comes back once and whole, the no-git-header fallback, and format-patch commit headers.

Gallery

There are new pages.diff variants:

word-diff: unified, a config file with changes inside lines
word-diff-split
big-review: 120 files at 760 px, for the file list and the change bars
collapsed-files: a play step that clicks a header to collapse it

The existing variants show the new chrome. To compare: https://cmux-lawrences-mac-mini.tail137216.ts.net:18796/compare/cmux-diffviewer-fast/?entry=pages.diff

No capture: the only native-app change is one default value (diff.wordDiffs in DiffViewerSettingsSchema.swift); everything visible is the webview diff page, shown by the gallery stills below.

Gallery stills from the gallery dev server on huge-blue (Chromium, 1000 px). Base small-split, then this branch:

Word changes inside lines, dark and light (an even +6 -6 split draws two green blocks, two red, one neutral):

120 files: the file list on its own surface, and the change bar in every header:

A header clicked to collapse its file:

Changelog

Changed: big diffs in the diff viewer show their first rows several times sooner, the viewer has the cmux NEXT look (glass toolbar, file list, per-file change bars), and word-level highlights are on by default.

Observed / Not run

Observed:

On huge-blue (Linux, Chromium):
- diff-paint-perf.test.ts, diff-stat-blocks.test.ts and diff-stream-splitter.test.ts pass at 940faca2aad (numbers above).
- diff-paint-perf.test.ts fails on base 7a4b464b375.
- diff-virtualization.test.ts passes at 940faca2aad (10/10, Chromium and WebKit). Run in one bun test process right after the perf file it timed out in setup (a cold dev-server warm-up); alone it passes.
- The production-build measurements above.
- The gallery screenshots.
On cmux16: regenerate-settings-exports.sh for the committed exports; defaults_guard.py check-stamp passes.
On cosmic-gray: diff-stream-splitter.test.ts and diff-stat-blocks.test.ts, tsc --noEmit, lint and prepush-gate.py.

Not run:

WebKit timing. The perf test is Chromium only; the app's diff page runs in WKWebView, where the same fixes apply but were not timed.
The native app and any Swift build. The Swift change is one default value.
The full webviews suite.

Agent session: cosmic-gray

🤖 Generated with Claude Code

…cmux NEXT chrome Measured on huge-blue (headless Chromium, production build of the virtualization harness, from the patch being ready): - 2,000 files / 10 MB: first row 1,180-1,330 ms -> 390-510 ms - one 1,000,000-line file: first row 6,190-6,740 ms -> 1,195-1,260 ms What was slow and what changed: - The patch splitter appended every 256 KB chunk to its buffer and searched it, flattening the whole rope each time: quadratic inside one big file (4.4 s of copying). Text with no file boundary now waits outside the buffer. - The highlight pool was created with the first files; it now mounts with the viewer, so its workers start while the patch streams. - Rows drew only after every worker compiled Shiki's WASM; the pool now gets a JavaScript-engine instance at once and draws plain rows until its colors arrive. - Between batches the parse yields every 12 ms, so the first files' rows and worker replies are not held behind a 100 ms parse task. Look: word changes are on by default (word-alt marks), a hairline rim and an accent open state on toolbar chips, a translucent toolbar, the files panel on one lift with an accent selection, and a five-block change bar per file. A Chromium perf test (diff-paint-perf.test.ts) holds the first row and the first highlight to a budget; a unit test covers the splitter. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

teamleaderleo · 2026-10-21T07:40:25Z

Owned by projects-ba on cosmic-gray · session

teamleaderleo · 2026-10-21T07:40:25Z

Owned by projects-ba on cosmic-gray · session

teamleaderleo · 2026-10-21T07:40:34Z

Bead cx-77aso created for this PR by the cmux coordinator.

github-actions · 2026-10-21T07:40:48Z

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

Review 19ecab0: 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=pr23154 -f ref=cmux/diffviewer-fast -f platform=macos -f backend_mode=local

Unclear: web / react-apps-check failed: no known signature; failed step: Check webview test shards.

CI failed on 19ecab035c (run 39598170651 attempt 2): 1 code, 1 unknown.
macOS jobs did not run: compile and app tests were skipped: Fast static checks skipped.

Job	Verdict	Why
web / react-apps-check (3)	code	a web test failed
web / react-apps-check	unknown	no known signature; failed step: `Check webview test shards`

Failures:

seen on other PRs agent-pane-stream-jitter.test.ts webkit: a queued prompt sent as the next turn moves the view up only: failed (job)

Matched log lines

web / react-apps-check (3): (fail) webkit: a queued prompt sent as the next turn moves the view up only [5852.92ms]

Not re-run automatically: web / react-apps-check (3), web / react-apps-check are not machine failures.

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.

red: web / Web status; NEW: green on the base 0284c98; do not merge until green; see lane-rules
red: web / react-apps-check; NEW: green on the base 0284c98; do not merge until green; see lane-rules
red: web / react-apps-check (3); NEW: green on the base 0284c98; do not merge until green; see lane-rules

CI fast guards failed on 19ecab035c (https://github.com/manaflow-ai/cmux/actions/runs/39598170651). 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.

Gallery before/after for 19ecab0: 16 states rendered, 16 images differ, 1420 more touched states not rendered. Images are in the run's artifact, not on this public PR.

gh run download 39598170674 -R manaflow-ai/cmux -n gallery-advisory-23154-19ecab035c3681461e81e986a504df38260ff1b2 (then read gallery-review.json)

States

State	Theme	Status	Diff	Picked by
`diff.find-bar/empty`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/empty`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/matches`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/matches`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/next`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/next`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/previous`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/previous`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/keyboard`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/keyboard`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/refocus`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/refocus`	Apple System Colors	unchanged	0%	own `styles.css`
`diff.find-bar/no-matches`	Apple System Colors Light	unchanged	0%	own `styles.css`
`diff.find-bar/no-matches`	Apple System Colors	unchanged	0%	own `styles.css`
`pages.diff/word-diff`	Apple System Colors Light	new	0%	own `App.tsx`
`pages.diff/word-diff`	Apple System Colors	new	0%	own `App.tsx`
`pages.diff/word-diff-split`	Apple System Colors Light	new	0%	own `App.tsx`
`pages.diff/word-diff-split`	Apple System Colors	new	0%	own `App.tsx`
`pages.diff/big-review`	Apple System Colors Light	new	0%	own `App.tsx`
`pages.diff/big-review`	Apple System Colors	new	0%	own `App.tsx`
`pages.diff/collapsed-files`	Apple System Colors Light	new	0%	own `App.tsx`
`pages.diff/collapsed-files`	Apple System Colors	new	0%	own `App.tsx`
`pages.diff/small-split`	Apple System Colors Light	changed	4.19%	own `App.tsx`
`pages.diff/small-split`	Apple System Colors	changed	4.03%	own `App.tsx`
`pages.diff/unified`	Apple System Colors Light	changed	4.19%	own `App.tsx`
`pages.diff/unified`	Apple System Colors	changed	4.03%	own `App.tsx`
`pages.diff/deleted`	Apple System Colors Light	changed	2.16%	own `App.tsx`
`pages.diff/deleted`	Apple System Colors	changed	2.14%	own `App.tsx`
`pages.diff/many-files`	Apple System Colors Light	changed	3.17%	own `App.tsx`
`pages.diff/many-files`	Apple System Colors	changed	3.13%	own `App.tsx`
`agent-pane.add-project-panel/environment`	Apple System Colors Light	unchanged	0%	names `styles.css`
`agent-pane.add-project-panel/environment`	Apple System Colors	unchanged	0%	names `styles.css`

_{Pick other states with a Gallery: entry.id, entry.id/variant line in the PR body.}

blacksmith-sh · 2026-10-21T07:41:55Z

Found 2 test failures on Blacksmith runners:

Failures

Test	View Logs
`test/agent-pane-stream-jitter.test.ts/agent-pane-stream-jitter.test`	View Logs
`test/agent-pane-stream-jitter.test.ts/agent-pane-stream-jitter.test`	View Logs

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

…t row mid-stream The mixed fixture also asserts the first row draws before 60% of the stream, which holds on any machine speed and fails on the old path (78-86%). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

+6 -6 drew two green blocks and three red; ties now drop a block from both sides and leave a neutral one, as GitHub's bar does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

The settings exports regenerated by the PR autofix (its push job leaves them in the pr-autofix-patch artifact), applied with scripts/cmux-next/pr-autofix-apply.sh: diff.wordDiffs now defaults to true. The defaults snapshot stays with the base autofix (bot-only). Autofix-Run: https://github.com/manaflow-ai/cmux/actions/runs/39566105569 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

The settings exports take the base's copies (GitHub merges them by line and called the pull request conflicting); the PR autofix regenerates them with this branch's diff.wordDiffs default on top. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

…highlighter once replaced Review findings: a MessageChannel per parse yield stayed alive (a started, connected port is never collected), hundreds per big patch; and the pool overwrites the plain-rows Shiki instance without disposing it. The yield now shares one channel with a FIFO of waiters, and the plain highlighter is disposed on the pool's first stats broadcast after it is replaced or cleared. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

…40faca scripts/cmux-next/regenerate-settings-exports.sh on cmux16 (the PR autofix's Mac job did not regenerate them on the last two heads, and its push job is skipped). The export run's SettingsSchemaExportTests reports a base gap (notifications.attention.flash keys have no agent policy) and still writes the exports; defaults_guard.py check-stamp passes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

The settings exports take the base's copies; the next commit regenerates them on this merge with diff.wordDiffs on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

cursor · 2026-10-21T10:50:47Z

Bugbot needs on-demand usage enabled

Bugbot 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.

scripts/cmux-next/regenerate-settings-exports.sh on cmux16 at this branch's merge with feat-cmux-next cca5809. Three base test files that do not compile there (PageEnginePolicyTests, ChiefAppAcpmuxTests, BrowserProfileNewTabMenuTests; the exports do not read them) were left out of that scratch build only. defaults_guard.py check-stamp passes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

The settings exports take the base's copies (#23239 regenerated them); the next commit regenerates them on this merge with diff.wordDiffs on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

…ations of two base tests belong to the base) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

The settings exports the PR autofix regenerated on its Mac job (its push job is skipped; artifact pr-autofix-patch), applied with scripts/cmux-next/pr-autofix-apply.sh: diff.wordDiffs defaults to true. The defaults snapshot stays with the base autofix (bot-only). Autofix-Run: https://github.com/manaflow-ai/cmux/actions/runs/39590411725 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo · 2026-10-21T14:17:35Z

Review (independent subagent, claude-opus-5-5): SHIP at 19ecab0

No blocking findings. Checked: streaming parse gives the same files as a whole-patch parse at every chunk size; cancellation drops batches from stale streams; plain rows before workers start can't use the wrong language and leak no worker; word diffs are capped on huge diffs (lines over 1000 chars, plain diffs over 1000 lines); the toolbar blur doesn't change how menus stack; the settings export sha matches.

cmux16, base → head: mixed first row ≈330 → ≈93 ms, huge ≈585 → ≈220 ms. The perf test fails on base, as intended.

Non-blocking:

The temporary highlighter takes its themes from the first-mount options, not from the pool's theme setting.
A highlighter created after the pool shuts down isn't disposed (garbage collection frees it).
Total stream time on mixed is about 15% longer, in exchange for the first row 3.5x sooner.
The wordDiffs: true default also applies to the classic app.
plans/cmux-next/customization.md:183 still says diff.wordDiffs: bool = false.

Reds at this head are base: PaletteMenuVocabularyTests.swift:74 compile failure (fixed on feat-cmux-next) and the agent-pane-stream-jitter flake in react-apps-check (3).

github-actions · 2026-10-21T14:18:33Z

Default changes

This pull request changes 1 default at 19ecab035c36 (what everyone gets without configuring anything):

setting.diff.wordDiffs: false -> true

Needs a call. It passes only with an approving review on this commit from @lawrencecchen or @austinywang (the call owners in .github/defaults-call-owners.json, never the author). The needs a call label alone does not pass, and a new push needs a new approval.

Merge receipt: merged as fc82e05c2263 with these defaults changed, WITHOUT an approving call owner review on its head.

Regenerate the snapshot with python3 scripts/ci/defaults_guard.py snapshot.

…3480) #23154 added diff-paint-perf with a probe that read layout in a requestAnimationFrame callback; #23374's frame-read rule (check:browser-lanes) landed meanwhile, so webviews lint went red on feat-cmux-next for every PR. The probe now reads in window.cmuxAfterPaint (test/support/paintedFrames.ts) and marks each milestone at the painted frame's time. Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…ghlighter uses the pool's themes (#23465) * test: diff viewer word changes default per host, plain-rows highlighter themes from the pool (red) The classic viewer (embedded config) should start with word changes off and cmux NEXT (page host, transport `page`) with them on, NEXT's defaults matching the exported settings schema. The plain-rows highlighter should load the pool's own themes and not be handed to a pool shut down while it was built. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray * cmux-next diff viewer: word changes default on in cmux NEXT only; plain-rows highlighter takes the pool's themes #23154 turned word changes on in the shared webview default, which also changed the classic viewer. The default now follows the host: on cmux NEXT's page host (config transport `page`) the display defaults are the `diff.*` settings' (DiffViewerSettingsSchema.swift; diff-option-defaults.test.ts holds them to the exported schema), word changes on; the classic viewer and the dev server keep word changes off. The virtualization harness seeds word changes on so diff-paint-perf keeps measuring cmux NEXT's default. drawPlainBeforeWorkers loaded its themes from the viewer options captured at first mount; it now reads the pool's (getDiffRenderOptions().theme), the ones the pool draws plain rows with. A pool terminated while the instance was built no longer gets it (it is disposed; outcome `stopped`). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray * test: diff-option-defaults installs the DOM globals @pierre/diffs reads at import state.ts reaches @pierre/diffs, whose web components module reads customElements when it loads. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?harness=claude&session=01a15d24-cf0a-7373-a03a-75491e4d724a&host=cosmic-gray --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

github-actions · 2026-10-21T16:06:05Z

Shipped in:

cmux NEXT nightly 1.0.0-nightly.3962509665801 (ed547f0e27)
cmux NEXT nightly 1.0.0-nightly.3962960774501 (913f2e6d53)

teamleaderleo added the risk:normal Risk tier (docs/ci/risk-tiers.md) label Oct 21, 2026

teamleaderleo and others added 3 commits October 21, 2026 00:49

teamleaderleo temporarily deployed to cmux-tui-merge-tree October 21, 2026 08:30 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to sentry-dsyms October 21, 2026 08:53 — with GitHub Actions Inactive

teamleaderleo had a problem deploying to artifacts October 21, 2026 08:53 — with GitHub Actions Failure

teamleaderleo and others added 3 commits October 21, 2026 03:04

teamleaderleo temporarily deployed to cmux-tui-merge-tree October 21, 2026 11:10 — with GitHub Actions Inactive

teamleaderleo and others added 3 commits October 21, 2026 04:13

teamleaderleo temporarily deployed to cmux-tui-merge-tree October 21, 2026 11:35 — with GitHub Actions Inactive

github-actions Bot temporarily deployed to ci-routing-agent October 21, 2026 12:04 Inactive

teamleaderleo temporarily deployed to artifacts October 21, 2026 12:04 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to sentry-dsyms October 21, 2026 12:04 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to artifacts October 21, 2026 12:08 — with GitHub Actions Inactive

Merge feat-cmux-next into cmux/diffviewer-fast

19ecab0

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo temporarily deployed to cmux-tui-merge-tree October 21, 2026 12:15 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to sentry-dsyms October 21, 2026 12:47 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to artifacts October 21, 2026 12:47 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to artifacts October 21, 2026 12:51 — with GitHub Actions Inactive

github-actions Bot temporarily deployed to ci-routing-agent October 21, 2026 12:54 Inactive

teamleaderleo merged commit fc82e05 into feat-cmux-next Oct 21, 2026

teamleaderleo deleted the cmux/diffviewer-fast branch October 21, 2026 14:17

This was referenced Oct 21, 2026

test(webviews): the chat-open perf gate says its margins out loud #23483

Merged

test(webviews): U1 waits for conditions, not fixed times (114 waits) #23416

Merged

teamleaderleo mentioned this pull request Oct 21, 2026

cmux-next: Change Font… with live preview, fonts in their own face, recent fonts first #22762

Merged

Navigation Menu

Repository navigation

cmux-next: diff viewer paints big diffs fast and gets the cmux NEXT look - #23154

cmux-next: diff viewer paints big diffs fast and gets the cmux NEXT look#23154
teamleaderleo merged 13 commits into
feat-cmux-nextfrom
cmux/diffviewer-fast

teamleaderleo commented Oct 21, 2026 •

edited

Loading

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026 •

edited

Loading

Uh oh!

blacksmith-sh Bot commented Oct 21, 2026 •

edited

Loading

Uh oh!

cursor Bot commented Oct 21, 2026

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026 •

edited

Loading

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

teamleaderleo commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What was slow

What changed

Numbers

Perf test and budget

Gallery

Changelog

Observed / Not run

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blacksmith-sh Bot commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Failures

Uh oh!

cursor Bot commented Oct 21, 2026

Bugbot needs on-demand usage enabled

Uh oh!

teamleaderleo commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026

Default changes

Uh oh!

github-actions Bot commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

teamleaderleo commented Oct 21, 2026 •

edited

Loading

github-actions Bot commented Oct 21, 2026 •

edited

Loading

blacksmith-sh Bot commented Oct 21, 2026 •

edited

Loading

github-actions Bot commented Oct 21, 2026 •

edited

Loading