fix: avoid false idle reports from powerMonitor on macOS (naturalBreaks churn)

[kyson-zhu-888]

What

On macOS 26 (Tahoe / Darwin 25), powerMonitor.getSystemIdleTime() reports false idle values — it oscillates between a few seconds and values above naturalBreaksInactivityResetTime, even during active use. Because NaturalBreaksManager.idleTime trusts powerMonitor.getSystemIdleTime() first and only falls back to the native node-desktop-idle-v2 when it returns 0, the inflated (non-zero) value means the accurate native source is never reached.

The oscillation makes _checkIdleTime() fire clearBreakScheduler (idle > reset time) and then naturalBreakFinished (idle < 20s) in rapid succession, every 1–2 minutes. Each naturalBreakFinished calls reset() on the planner, restarting the break countdown. Net effect: with naturalBreaks enabled, breaks never fire.

Fixes #1797.

Change

Use the native node-desktop-idle-v2 (IOHID) source directly on darwin; keep the existing behaviour on every other platform.

get idleTime () {
  if (this.usingNaturalBreaks) {
    if (process.platform === 'darwin') {
      return desktopIdle.getIdleTime() * 1000
    }
    return (powerMonitor.getSystemIdleTime() || desktopIdle.getIdleTime()) * 1000
  } else {
    return 0
  }
}

Why the native source is trustworthy

node-desktop-idle-v2 reads IOHID. Querying the same IOHID source directly is accurate:

$ ioreg -c IOHIDSystem | grep HIDIdleTime   # no input for 4s
109.899 s
$ sleep 4; ioreg -c IOHIDSystem | grep HIDIdleTime
113.979 s     # +4s, monotonic and correct

So only powerMonitor.getSystemIdleTime() misbehaves on this OS — likely a macOS 26 / Electron regression in the CGEventSource-backed idle API.

Testing

• standard passes on the changed file.
• Existing test/naturalBreaksManager.js still passes (it only asserts idleTime === 0 when natural breaks are off).
• Patched into a local 1.21.0 build, naturalBreaks=true: monitored main.log for 5 minutes during active use → 0 idle-churn events (vs. one pause/resume pair every 1–2 min before). Breaks now schedule and fire correctly, and genuine idle periods still reset the countdown as intended.

Scope

Single-file, single-expression change, guarded by process.platform === 'darwin'. Non-macOS behaviour is unchanged. The powerMonitor event subscriptions in main.js (suspend/lock-screen/resume/unlock-screen) are unrelated and untouched.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 68.31%. Comparing base (888f1f1) to head (9e6c7af).
⚠️ Report is 15 commits behind head on trunk.

Files with missing lines	Patch %	Lines
app/utils/naturalBreaksManager.js	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##            trunk    #1798      +/-   ##
==========================================
+ Coverage   67.39%   68.31%   +0.91%     
==========================================
  Files          16       17       +1     
  Lines         506      527      +21     
==========================================
+ Hits          341      360      +19     
- Misses        165      167       +2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.