[build-tools] disable apsd in iOS simulator

[gwdp]

Why

apsd (the push daemon) can't get a push token in the Simulator — a fresh Simulator has no push identity — so it retries forever and floods the system log. That keeps diagnosticd busy for the whole life of the Simulator, eating about a CPU core. Simulators can't get real pushes anyway.

Seen on Xcode 16.4, 26.2 and 26.4.

How

Disable apsd as soon as the Simulator boots, before we wait for it to be ready:

xcrun simctl spawn <udid> launchctl disable system/com.apple.apsd
xcrun simctl spawn <udid> launchctl bootout system/com.apple.apsd

If it fails, we warn and keep going. Same for cloned Simulators.

Results

Idle Simulator, Xcode 26.4:

	apsd on	apsd off
apsd logs/min	~100k	~0
diagnosticd CPU	~30-50%	~0%
Simulator CPU	~75%	~0%

App launch time was the same either way (~210ms).

Test Plan

• Unit test for disableApsdAsync.
• Before/after on a build (Xcode 26.4): https://expo.dev/accounts/gwdp/projects/coin-flip/workflows/019e66ad-11c5-7ae1-a5a8-f4f590b2b8f8
• Watched it stay idle for 3 min with apsd off: https://expo.dev/accounts/gwdp/projects/coin-flip/workflows/019e67e4-9a80-7416-b631-25de268b9b15
• yarn typecheck, oxlint, tests pass.

Notes

Still not sure why iOS 26.0 didn't storm. There apsd connected once and went quiet, while 16.4, 26.2 and 26.4 all kept spamming — same missing token, same network. Haven't worked out what's different about 26.0.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 57.39%. Comparing base (86a891b) to head (19400d3).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3795      +/-   ##
==========================================
+ Coverage   57.32%   57.39%   +0.08%     
==========================================
  Files         905      905              
  Lines       39314    39324      +10     
  Branches     8242     8242              
==========================================
+ Hits        22532    22568      +36     
+ Misses      15315    15291      -24     
+ Partials     1467     1465       -2     

☔ View full report in Codecov by Sentry.
📢 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.

[github-actions]

✅ Thank you for adding the changelog entry!

[github-actions]

Subscribed to pull request

File Patterns	Mentions
**/*	@douglowder

Generated by CodeMention

[hSATAC]

LGTM

[sjchmiela]

are we certain apsd absolutely does not work in simulators or is not needed? see https://exponent-internal.slack.com/archives/C01D24DMP98/p1778175913170849, https://developer.apple.com/forums/thread/795433?answerId=881732022#881732022

[gwdp]

are we certain apsd absolutely does not work in simulators or is not needed? see https://exponent-internal.slack.com/archives/C01D24DMP98/p1778175913170849, https://developer.apple.com/forums/thread/795433?answerId=881732022#881732022

​Thank you for the review and bubbling some of your the concerns @sjchmiela 🙏

APNS token registration in the simulator is bound to the host Mac (per the Xcode 14 release notes), and EAS workers aren't signed into any Apple ID, so apsd can't get a token regardless of the runtime. Measured through "no token found in keychain" ~130/min on iOS 18.x/26.0/26.2/26.4. Enabling the registration path would mean signing the CI workers into an Apple ID, which we currently don't.

In the other hand,simctl push doesn't go through apsd (verified — succeeds with the same output whether apsd is running or launchctl bootout'd), and that's already how Maestro delivers notifications, so disabling doesn't break anything we use. snorkling session documented at notion

[sjchmiela]

AWESOME