fix(cursor): seed default cursor with enabled=false to prevent overlay process leak by chxrro · Pull Request #1900 · trycua/cua

chxrro · 2026-06-15T16:53:54Z

Problem

Every anonymous (non-session) CUA tool call spawns a cursor overlay process
that is never cleaned up, causing 30+ zombie processes after moderate use.

Root cause: the default cursor is seeded in three places with a hardcoded
assumption that it should always render:

CursorRegistry::new() seeds "default" with enabled: true
overlay::init() seeds default render state from the launch config
RenderStateCore::new() hardcodes visible: true, ignoring
cfg.enabled entirely — the documented contract says "without a
session, actions run cursor-less" but the code always renders

The "default" key is explicitly guarded against removal (both in
CursorRegistry::remove() and overlay::apply_msg()), so these
overlays accumulate until daemon restart.

Fix (3 files, 24 lines)

platform-macos/src/cursor/state.rs — seed default cursor
registry entry with enabled: false
platform-macos/src/cursor/overlay.rs — seed default render
state with enabled: false
cursor-overlay/src/render_state.rs — derive visible from
cfg.enabled instead of hardcoding true

The overlay NSWindow still exists (AppKit lifecycle unchanged) but the
default cursor draws nothing — paint_cursor early-returns on
!visible, and animate_cursor_to skips on !cfg.enabled.
Session cursors still work normally via set_agent_cursor_enabled /
start_session.

Verification

Updated test: default_cursor_starts_disabled verifies the seed
Existing cursor tests pass unchanged for session-based paths
Manual: HermesBar + Hermes agent, 20+ anonymous CUA calls → 1 daemon
process (vs 30+ before)

Closes #1777

Summary by CodeRabbit

Bug Fixes

Cursor overlay visibility now properly respects launch-time configuration settings instead of defaulting to always enabled.
Default cursor overlay initializes to disabled state for predictable, configuration-based control.
Updated tests to verify correct initialization and configuration behavior.

…y leak The default cursor was seeded with enabled=true in three places: - CursorRegistry::new() (platform-macos/src/cursor/state.rs) - overlay::init() (platform-macos/src/cursor/overlay.rs) - RenderStateCore::new() hardcoded visible=true (cursor-overlay/src/render_state.rs) This caused every anonymous (non-session) tool call to render a cursor overlay that was never cleaned up — the "default" key is explicitly guarded against removal. The documented contract is "without a session, actions run cursor-less," but the hardcoded visible=true bypassed the config entirely. Fixes: 1. state.rs: seed default with enabled=false 2. overlay.rs: seed default render state with enabled=false 3. render_state.rs: derive visible from cfg.enabled instead of hardcoding true Without a declared session, the overlay window still exists (AppKit NSWindow lifecycle) but the default cursor draws nothing — paint_cursor early-returns on !visible, and animate_cursor_to skips on !cfg.enabled. Session cursors work normally via set_agent_cursor_enabled / start_session. Closes #1777

vercel · 2026-06-15T16:53:59Z

Someone is attempting to deploy a commit to the Cua Team on Vercel.

A member of the Team first needs to authorize it.

coderabbitai · 2026-06-15T16:54:14Z

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: c5f31d6a-c291-4118-810c-f6ab683b2cf5

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

📝 Walkthrough

Walkthrough

Three files are updated to ensure the seeded "default" anonymous cursor starts disabled: CursorRegistry::new explicitly sets enabled: false for the default entry, overlay.rs init clones the incoming config with enabled forced to false before inserting the default cursor, and RenderStateCore::new derives visible from cfg.enabled instead of hardcoding true. The unit test for the default cursor is updated to match.

Changes

Default cursor disabled at initialization

Layer / File(s)	Summary
Disable default cursor in registry, overlay, and render state `libs/cua-driver/rust/crates/platform-macos/src/cursor/state.rs`, `libs/cua-driver/rust/crates/platform-macos/src/cursor/overlay.rs`, `libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs`	`CursorRegistry::new` seeds `"default"` with `enabled: false`; overlay `init` clones the incoming config with `enabled` forced to `false` for the `"default"` entry; `RenderStateCore::new` sets `visible` from `cfg.enabled` instead of `true`. Unit test updated to assert the default cursor starts disabled and transitions to enabled via `set_enabled`.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

trycua/cua#1787: Directly overlaps with cursor-overlay logic that conditionally seeds and animates the session cursor based on enablement, coupling the behavior of cursor visibility initialization between the two PRs.

Poem

🐇 A cursor once blinked on without asking leave,
Now it waits, politely hidden up its sleeve.
enabled: false — a shy little dot,
Until set_enabled wakes it from its cot.
Default is silent, the registry agrees,
No phantom cursor blinking as it please!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: seeding the default cursor with enabled=false to fix the overlay process leak.
Linked Issues check	✅ Passed	The PR directly addresses objective 1 from `#1777` by implementing cursor session-scoping and preventing anonymous cursor overlay accumulation through the enabled=false seeding approach.
Out of Scope Changes check	✅ Passed	All changes are directly related to cursor session-scoping as specified in `#1777`; no unrelated modifications to other features or systems were introduced.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs`:
- Around line 99-100: The issue is that visible and cfg.enabled can diverge
after initialization. While RenderStateCore::new initializes visible from
cfg.enabled, the OverlayCommand::SetEnabled handler only mutates visible without
updating cfg.enabled. This causes the animation gating check in
animate_cursor_to (which uses cfg.enabled) to become stale after toggles. Fix
this by ensuring that whenever the enabled state is toggled in the SetEnabled
handler at line 115, both visible and cfg.enabled are updated together to keep
them synchronized.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 19234aa1-ae73-4e06-aee0-9666d9572c4c

📥 Commits

Reviewing files that changed from the base of the PR and between edbcf3b and 547e9da.

📒 Files selected for processing (3)

libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs
libs/cua-driver/rust/crates/platform-macos/src/cursor/overlay.rs
libs/cua-driver/rust/crates/platform-macos/src/cursor/state.rs

coderabbitai · 2026-06-15T16:57:37Z

+        let visible = cfg.enabled;
        Self {


⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Keep cfg.enabled and visible in sync to avoid stale enablement checks.

RenderStateCore::new now initializes visible from cfg.enabled, but runtime toggles (OverlayCommand::SetEnabled) still mutate only visible. Since animate_cursor_to gates on rs.core.cfg.enabled in platform-macos/src/cursor/overlay.rs (Line 308), enable/disable state can diverge and break animation gating after toggles.

Suggested fix

diff --git a/libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs b/libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs @@ OverlayCommand::SetEnabled(v) => { self.visible = v; + self.cfg.enabled = v; true }

Also applies to: 115-115

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@libs/cua-driver/rust/crates/cursor-overlay/src/render_state.rs` around lines 99 - 100, The issue is that visible and cfg.enabled can diverge after initialization. While RenderStateCore::new initializes visible from cfg.enabled, the OverlayCommand::SetEnabled handler only mutates visible without updating cfg.enabled. This causes the animation gating check in animate_cursor_to (which uses cfg.enabled) to become stale after toggles. Fix this by ensuring that whenever the enabled state is toggled in the SetEnabled handler at line 115, both visible and cfg.enabled are updated together to keep them synchronized.

injaneity · 2026-08-10T11:45:56Z

closing as outdated, as this contradicts the deliberate default-on cursor behavior documented and retained by #1787, while #1779 already merged session-scoped cursor cleanup, crash detection, and resurrection guards

coderabbitai Bot reviewed Jun 15, 2026

View reviewed changes

Chirag Tripathi and others added 2 commits June 15, 2026 17:03

fix(cursor): update get_config test for default enabled=false

ada00cb

fix(cursor): keep cfg.enabled in sync with visible in SetEnabled (#1900)

6b96958

f-trycua requested review from ddupont808, f-trycua and r33drichards as code owners July 16, 2026 11:21

injaneity closed this Aug 10, 2026

Navigation Menu

Uh oh!

Uh oh!

fix(cursor): seed default cursor with enabled=false to prevent overlay process leak - #1900

fix(cursor): seed default cursor with enabled=false to prevent overlay process leak#1900
chxrro wants to merge 3 commits into
trycua:mainfrom
chxrro:fix/default-cursor-anonymous-leak-1777

chxrro commented Jun 15, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

vercel Bot commented Jun 15, 2026

Uh oh!

coderabbitai Bot commented Jun 15, 2026 •

edited

Loading

Review skipped

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Poem

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Jun 15, 2026

Uh oh!

injaneity commented Aug 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

chxrro commented Jun 15, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix (3 files, 24 lines)

Verification

Summary by CodeRabbit

Bug Fixes

Uh oh!

vercel Bot commented Jun 15, 2026

Uh oh!

coderabbitai Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Poem

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Jun 15, 2026

Choose a reason for hiding this comment

Uh oh!

injaneity commented Aug 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

chxrro commented Jun 15, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Jun 15, 2026 •

edited

Loading