Fix stale Chromium composer page during pane holds by teamleaderleo · Pull Request #28002 · manaflow-ai/cmux

teamleaderleo · 2026-10-30T05:18:56Z

Summary

Leo (2026-10-29): “Dude, I hate how the fing composer acts like this... Number one: the fing rendering is stupid as hell.”

The stale second composer came from a native Chromium page window surviving a paint hold. PanePaintHold hid the app view with alphaValue, but Chromium renders in a separate child window and AppKit does not propagate ancestor alpha to that window. The old page therefore stayed composited over the live chat. Launch images also remained valid across a pane resize even though their captured width was stale.

This PR synchronizes Chromium visibility with ancestor alpha during paint holds, release, and page attach/hide/unhide; queues host visibility when a CEF tab is concealed; and clears launch images when the pane content size changes or the live page paints. Focused regressions cover Chromium visibility, CEF conceal ordering, and launch-image resize invalidation.

Validation

python3 scripts/cmux-next/prepush-gate.py --base origin/feat-cmux-next --head HEAD — no new violations (existing warnings only).
Linux Swift package checks passed for portable packages; CmuxNext is macOS-only.
Focused CmuxNext suites are dispatched through the fleet route: PageWebViewChromiumTests, CEFDeferredTabShowTests, and LaunchFirstFrameRevealTests.

Proof

Before: the native hold made the app view transparent while its separate Chromium child window stayed visible, allowing an idle stale chat to cover the live composer.

After: hold transitions explicitly hide the Chromium page window, release re-shows only the incoming page, conceal queues the host hide, and a resize drops the old-size launch snapshot.

No capture: the native Chromium child-window lifecycle requires a fleet macOS build; the focused fleet tests provide the regression evidence.

Agent session: agent session

github-actions · 2026-10-30T05:24:15Z

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

Review 3740ab4: 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=pr28002 -f ref=lane/composer-native-ghost -f platform=macos -f backend_mode=local

web and size checks: green

blacksmith-sh · 2026-10-30T05:24:56Z

Found 2 test failures on Blacksmith runners:

Failures

Test	View Logs
`test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt` `epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)/` `test_every_listed_hook_is_one_glaeda_released_with_the_check (main.SigningHooksInSt` `epWithGlaedaTests.test_every_listed_hook_is_one_glaeda_released_with_the_check)`	View Logs
`test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl` `ights_validate)/` `test_the_repository_highlights_validate (main.DigestTests.test_the_repository_highl` `ights_validate)`	View Logs

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

cursor · 2026-10-30T06:52:09Z

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.

cursor · 2026-10-30T07:14:25Z

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.

Agent-Session: https://cmux.com/deeplink/agent?id=aid-17e6cbfcbccf435755d0

teamleaderleo · 2026-10-30T09:52:47Z

Review (codex gpt-6-astra high, next-review): BLOCK at 799d1e3

Packages/macOS/CmuxNext/Sources/CmuxNextPages/PageWebView+Chromium.swift:16-19 — Chromium visibility is sampled when setChromiumVisibility is called, but the agent loading flow changes PageWebView.alphaValue from 0 to 1 without calling it. If Chromium attaches while the loading cover is active, it receives setVisible(false) and can remain hidden after the page is revealed; a fresh pane has no paint-hold callback to restore it.
Packages/macOS/CmuxNext/Sources/CmuxNextApp/AgentTabContent.swift:84-87 — PageWebView.whenPainted fires on the generic document paint probe before the agent’s pane.painted handshake/chat-ready signal. This clears the saved launch image while the agent page is still intentionally transparent, and the clear callback re-enables the loading state, causing a visible loading/blank gap during slow attach.

Agent-Session: https://cmux.com/deeplink/agent?id=aid-17e6cbfcbccf435755d0

teamleaderleo · 2026-10-30T11:24:49Z

SHIP: independent Astra review at exact head d1df487. Reviewed Chromium visibility during paint holds, CEF conceal ordering, launch reveal/resize lifecycle, and regression coverage; prior P1s fixed in this head.

PageWebViewChromiumTests' new alpha-visibility test builds an NSWindow; the file imported only Foundation, so the CmuxNextPages test target failed to compile. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Exercise window-owned PanePaintHold, loading-state reveal, move covers and handoff through FakeChromiumPage visibility. Fleet red/green verification follows on the pushed commits; no tests run on this host. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-35aad81d5881fa12f16a

Exercise window-owned PanePaintHold, loading-state reveal and move covers through FakeChromiumPage visibility. Fleet red/green verification follows on the pushed commits; no tests run on this host. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-35aad81d5881fa12f16a

The shared PageWebView subtree sync now serves PanePaintHold begin, end, released outgoing and handoff, plus loading and move-cover alpha transitions. PaneContentView uses the shared traversal. Add a handoff detach/reuse regression. Alpha audit of CmuxNextAgentPane and CmuxNextApp: loading/page and move-cover transitions and PanePaintHold are covered. NewTabSpareParking sets alpha zero before children exist; page attach/start derives visibility from that ancestor. Loading cover, toolbar/titlebar glass, labels and drag ghost views contain no PageWebView. Popup/window alpha is separate; WindowCloseMotion already propagates it to child windows. Static Swift syntax and prepush-gate passed. No builds or tests on this host; focused fleet verification follows at the pushed SHA. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-35aad81d5881fa12f16a

Fleet job cbe2ca3ee493a09cef3072dc found AgentTabContent reading internal AgentPaneView.page, introduced before the alpha repair. Keep the page private to its target and provide a package-scoped paint callback with the same behavior so the focused suite can compile. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-35aad81d5881fa12f16a

…er-native-ghost

…cument paint whenPagePainted forwarded to the page document probe, which fires before the chat is read, so the launch image went while the page was still clear and the empty loading state showed. It now waits on AgentPaneModel.whenPainted, the gate the page is revealed on. The Chromium visibility window test keeps its window (isReleasedWhenClosed = false), as other window tests do. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo · 2026-10-30T17:24:02Z

Before/after stills of the agent pane (the agent-pane tour; the composer scene fails to seed its agent turn both before and after this PR). Settled frames match, one composer each: the doubled composer only appeared during a pane or window hold, which the new AgentPaneChromiumVisibilityTests and PageWebViewChromiumTests cover.

After aeb3a6de053c

Light	Dark

Before aead767cca64

Light	Dark

github-actions · 2026-10-30T18:29:41Z

Shipped in:

cmux NEXT nightly 1.0.0-nightly.4110084161401 (92b6fe2118)
cmux NEXT nightly 1.0.0-nightly.4110335000301 (abe139dc5a)
cmux NEXT nightly 1.0.0-nightly.4110516029701 (b7196082e9)
cmux NEXT nightly 1.0.0-nightly.4110699040101 (723153fb7a)

… frame (#28461) * fix(pages): keep a Chromium page drawing while its pane waits for its frame #28002 hid the Chromium page whenever any view up to it, itself included, was transparent. The agent pane clears the page itself while it waits for the first frame (loading state) or the next one (move cover), and a hidden Chromium page draws no frame, so a new chat showed the loading spinner until the 8-12 s give-up instead of its composer. Only ancestors alpha (pane and window paint holds, the ghost-composer case) now hides the page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(pages): skip the page itself without an optional sequence start Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…e paints) (#28980) * test(new-tab): a parked Chromium spare keeps drawing in its off-screen parking Red on base: the parking's alpha hides the Chromium page (#28002), so the spare never paints before Cmd-T adopts it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-d750cf10922c785efa64 * perf(new-tab): let a parked Chromium spare paint before Cmd-T adopts it #28002 hides a Chromium page whenever an ancestor is transparent. The New Tab spare's parking view is transparent (and 100000 pt off screen) only to stay out of sight, so a Chromium spare never painted while parked. Cmd-T then adopted an unpainted page, the pane's paint hold hid it again, and the New Tab page showed only at the hold's 500 ms limit. The parking now opts out of that rule (PageChromiumOffscreenParking); paint holds still hide. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-d750cf10922c785efa64 --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

… drawing while it is held off screen (#32448) * test(agent pane): a hidden adopted page still reports its paint (fails: it waits for a frame) The page's pane.painted report moves to paintReport.ts unchanged (one frame, then a task). paint-report.test.ts fails on it: a hidden document runs no animation frames, so the report never comes (HQ#5914). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-14c3a892e82255726cbb * cmux NEXT: an adopted spare page reports its paint while hidden, and existing chats take the spare again The parked spare can be hidden (Chromium marks the page hidden), and a hidden document runs no animation frames. The page's pane.painted waited for one, and the host keeps an adopted page away and covered until that report: reopens held the cover for the ~15 s give-up (HQ#5914; diag bench at 7d8c57e: visibilityState hidden, last frame before the report was asked). A hidden page now reports on a task, at once or when it goes hidden while it waits. The chat-open trace counts an adopted page as drawn once its pane painted: its first frame (page_painted) came before the trace, as New Tab, so every adopted open read page_empty. Turns adoption back on (#31486 switched it off). paint-report.test.ts: red on the previous commit (2 fail), green here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-14c3a892e82255726cbb * test(agent pane): an adopted page opens an unlisted chat with no list read beside its attach (fails) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-14c3a892e82255726cbb * cmux NEXT: the adopted page reads the chat list again only when its attach fails adoptSession read the whole list beside the attach, on the same socket; a busy acpmux paces that list for seconds, and adopted reopens drew later than a cold page (HQ#5914 bench). The list now decides only a failed attach, and only a list that was read says the chat is missing (#31033's review follow-up). spare-session-adopt.test.ts: 1 fail on the previous commit, 7 pass here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-14c3a892e82255726cbb * test(agent pane): an adopted page held outside the window keeps drawing under the pane's paint hold (fails) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-23bb1f8ee2895ef522be * cmux NEXT: an adopted page held outside the window keeps drawing under the pane's paint hold The pane's paint hold makes its incoming view transparent, and a transparent ancestor hides a Chromium page (#28002). An adopted New Tab spare waits 100 000 pt outside the window until it drew the chat (adoptSession), so the hold hid a page nobody could see: Chromium throttled it, and it drew the chat only once the hold gave up at 500 ms. A page laid out wholly outside its window's content now stays visible to Chromium whatever its ancestors' alpha; in place, a hold still hides it. Drops #31619's hidden-page paint report (on a task, without a frame): with the held page drawing, the report waits for a real frame again, and a report with no frame drawn showed the page's previous frame when the host revealed it. AgentPaneChromiumVisibilityTests.anAdoptedPageHeldAwayKeepsDrawingUnderAPaintHold: red on the previous commit, green here. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-23bb1f8ee2895ef522be * test: cover selection after a failed spare adoption Agent-Session: https://cmux.com/deeplink/agent?id=aid-81082f85640c1fa29ca9 * fix: retain a newer selection after spare adoption fails Agent-Session: https://cmux.com/deeplink/agent?id=aid-81082f85640c1fa29ca9 * Agent pane: an adopted spare turns its cover on before its loading state begins The chat-open bench counted a flash on 7-10 of 10 adopted opens: its trace recorded the chat's saved frame ("image"), then the empty cover, then the chat. No such frame was drawn. beginLoadingState binds the trace and samples the surface at once, while the New Tab page's cover was still hidden; adoptSession turned the cover on only after that, in the same main turn, so the first sample read a state no frame showed. The cover now goes on first, so the pane is in its loading state from the first sample. Diagnosed on the fleet with a probe build that recorded every cover change after the adoption; it found none, which left the trace's own first sample. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-758f5f82cc10b26ecf3c --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 05:19 — with GitHub Actions Inactive

github-actions Bot added the risk:normal Risk tier (docs/ci/risk-tiers.md) label Oct 30, 2026

teamleaderleo force-pushed the lane/composer-native-ghost branch from 429d734 to 164084e Compare October 30, 2026 06:52

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 06:52 — with GitHub Actions Inactive

teamleaderleo force-pushed the lane/composer-native-ghost branch from 164084e to 3726e92 Compare October 30, 2026 07:14

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 07:14 — with GitHub Actions Inactive

Fix stale Chromium page windows during pane holds

799d1e3

Agent-Session: https://cmux.com/deeplink/agent?id=aid-17e6cbfcbccf435755d0

teamleaderleo force-pushed the lane/composer-native-ghost branch from 3726e92 to 799d1e3 Compare October 30, 2026 07:30

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 07:30 — with GitHub Actions Inactive

fix: preserve move cover and reveal Chromium pages

996075d

Agent-Session: https://cmux.com/deeplink/agent?id=aid-17e6cbfcbccf435755d0

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 10:30 — with GitHub Actions Inactive

fix: keep launch loading fallback on resize

d1df487

Agent-Session: https://cmux.com/deeplink/agent?id=aid-17e6cbfcbccf435755d0

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 10:45 — with GitHub Actions Inactive

test(pages): import AppKit for the Chromium visibility window test

97616e6

PageWebViewChromiumTests' new alpha-visibility test builds an NSWindow; the file imported only Foundation, so the CmuxNextPages test target failed to compile. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 12:25 — with GitHub Actions Inactive

teamleaderleo and others added 2 commits October 30, 2026 06:19

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 13:22 — with GitHub Actions Inactive

teamleaderleo had a problem deploying to ci-routing-agent October 30, 2026 13:22 — with GitHub Actions Error

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 13:24 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 13:25 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 13:56 — with GitHub Actions Inactive

Merge remote-tracking branch 'origin/feat-cmux-next' into lane/compos…

03ee17f

…er-native-ghost

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 15:05 — with GitHub Actions Inactive

teamleaderleo temporarily deployed to ci-routing-agent October 30, 2026 15:24 — with GitHub Actions Inactive

teamleaderleo merged commit aeb3a6d into feat-cmux-next Oct 30, 2026
147 of 151 checks passed

teamleaderleo deleted the lane/composer-native-ghost branch October 30, 2026 16:54

teamleaderleo mentioned this pull request Oct 30, 2026

fix(pages): keep a Chromium page drawing while its pane waits for its frame #28461

Merged

teamleaderleo mentioned this pull request Oct 31, 2026

New Tab: Cmd-T shows the page at once again (the parked Chromium spare paints) #28980

Merged

This was referenced Nov 2, 2026

cmux NEXT: an existing chat opens on the warm spare page, which keeps drawing while it is held off screen #32448

Merged

cmux NEXT: opening a chat on the prewarmed page draws it first time (a hidden adopted page reports its paint) #31619

Open

Navigation Menu

Repository navigation

Fix stale Chromium composer page during pane holds - #28002

Fix stale Chromium composer page during pane holds#28002
teamleaderleo merged 10 commits into
feat-cmux-nextfrom
lane/composer-native-ghost

teamleaderleo commented Oct 30, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 30, 2026 •

edited

Loading

Uh oh!

blacksmith-sh Bot commented Oct 30, 2026 •

edited

Loading

Uh oh!

cursor Bot commented Oct 30, 2026

Uh oh!

cursor Bot commented Oct 30, 2026

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

github-actions Bot commented Oct 30, 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 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Validation

Proof

Uh oh!

github-actions Bot commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

blacksmith-sh Bot commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Failures

Uh oh!

cursor Bot commented Oct 30, 2026

Bugbot needs on-demand usage enabled

Uh oh!

cursor Bot commented Oct 30, 2026

Bugbot needs on-demand usage enabled

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

Uh oh!

teamleaderleo commented Oct 30, 2026

Uh oh!

github-actions Bot commented Oct 30, 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 30, 2026 •

edited

Loading

github-actions Bot commented Oct 30, 2026 •

edited

Loading

blacksmith-sh Bot commented Oct 30, 2026 •

edited

Loading

github-actions Bot commented Oct 30, 2026 •

edited

Loading