ci: bound benchmark simulator setup and cleanup

[mrousavy]

Simulator creation and simctl boot can currently hang before the five-minute bootstatus timer starts. Put all three commands in one setup step with a five-minute Actions deadline, and give final simulator cleanup a one-minute deadline. Export the device ID immediately after creation so cleanup can run when boot fails or times out.

After successful setup, explicitly mark the newly created CI simulator as fresh. Skip preinstall terminate/uninstall calls for apps that cannot yet exist. Reused simulators keep that cleanup by default; each benchmark case still launches and terminates its own process.

Validation on c7576d288e314744bd3f52992924fe3a9cf74c2c: all GitHub checks passed, including TypeScript, lint, Nitrogen on both platforms, both native builds and both device measurement jobs. The iOS job completed all 92 measurements, created/booted the simulator in 38 seconds and deleted it in four seconds. The two expected missing-app termination errors seen in unchanged runs are absent.

Local tests cover fresh/reused installations, paired and same-SHA runs, invalid freshness flags and per-case process isolation. Workflow checks verify that create/boot/bootstatus share the setup deadline and final cleanup has its own deadline; GitHub Actions enforces the timeouts. TypeScript, actionlint, ShellCheck and diff checks pass. Rebase patch equivalence was verified with git range-diff.

Timing review:

Run	Entire iOS measurement job	Boot-ready to first app launch
Current PR	7m00s	6.61s
Unchanged iOS controller, #1628	7m18s	7.81s
Unchanged iOS controller, #1629	7m01s	8.46s
PR before rebase	Cancelled by rebase	24.53s

The current run and controls share the same base and iOS runner/toolchain. These are separate hosted runs, and app extraction moved to after boot in this PR. The latest run is slightly shorter, but startup variation prevents attributing a reliable overall speedup to the change. The demonstrated benefits are bounded setup/cleanup and removing unnecessary commands/errors. Leave open under the requested works-and-speeds-up-CI merge condition.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
nitro-docs	Ready	Preview	Sep 7, 2026 7:03pm UTC

[nitro-modules-bot]

Performance Report

⚠️ Advisory: Results do not fail this PR.

iOS

Benchmark	Before	After	Difference
C++ bounce(1 MiB)	232.7 ns	263.1 ns	🔴 +13.05% slower
C++ large array (1,024)	48.28 µs	45.62 µs	🟢 -5.51% faster
C++ deferred worker Promise (includes trigger call)	26.93 µs	24.80 µs	🟢 -7.9% faster
Swift copy(1 MiB)	59.03 µs	74.93 µs	🔴 +26.93% slower
Swift synchronousCallback()	242.6 ns	223.2 ns	🟢 -7.97% faster
Swift create()	26.41 µs	23.21 µs	🟢 -12.12% faster
Swift returnExisting()	232.1 ns	250.2 ns	🔴 +7.77% slower
Swift simpleFunc()	60.7 ns	67.8 ns	🔴 +11.77% slower
Swift deferred worker Promise (includes trigger call)	23.82 µs	28.37 µs	🔴 +19.14% slower
Swift nested Car struct	2.38 µs	2.52 µs	🔴 +5.85% slower

All Benchmarks

Benchmark	Before	After	Difference
JavaScript addNumbers()	22.0 ns	22.7 ns	🔴 +2.74% slower
C++ bounce(4 KiB)	240.0 ns	239.6 ns	🟢 -0.19% faster
C++ bounce native-owned buffer (1 MiB)	3.74 µs	3.73 µs	🟢 -0.43% faster
C++ bounce native-owned buffer (4 KiB)	397.0 ns	388.4 ns	🟢 -2.16% faster
C++ copy(1 MiB)	83.40 µs	85.30 µs	🔴 +2.29% slower
C++ copy(4 KiB)	1.10 µs	1.15 µs	🔴 +3.8% slower
C++ small array (16)	906.6 ns	894.2 ns	🟢 -1.37% faster
C++ synchronousCallback()	214.1 ns	204.7 ns	🟢 -4.4% faster
C++ create()	431.4 ns	423.0 ns	🟢 -1.93% faster
C++ returnExisting()	202.8 ns	206.5 ns	🔴 +1.8% slower
C++ typed map (8 entries)	1.67 µs	1.66 µs	🟢 -0.18% faster
C++ optional trailing string	147.7 ns	146.9 ns	🟢 -0.53% faster
C++ addNumbers()	84.9 ns	87.4 ns	🔴 +2.99% slower
C++ simpleFunc()	63.8 ns	65.0 ns	🔴 +1.94% slower
C++ immediatePromise()	902.0 ns	902.8 ns	🔴 +0.09% slower
C++ number property get/set	133.2 ns	134.5 ns	🔴 +0.98% slower
C++ short ASCII string	137.4 ns	135.4 ns	🟢 -1.43% faster
C++ Unicode string	284.4 ns	284.6 ns	🔴 +0.05% slower
C++ nested Car struct	2.24 µs	2.30 µs	🔴 +2.64% slower
C++ number | string variant	127.7 ns	126.9 ns	🟢 -0.56% faster
Swift bounce(1 MiB)	240.0 ns	237.1 ns	🟢 -1.19% faster
Swift bounce(4 KiB)	242.8 ns	242.0 ns	🟢 -0.29% faster
Swift bounce native-owned buffer (1 MiB)	3.87 µs	3.84 µs	🟢 -0.73% faster
Swift bounce native-owned buffer (4 KiB)	384.9 ns	398.5 ns	🔴 +3.53% slower
Swift copy(4 KiB)	1.22 µs	1.22 µs	🟢 -0.39% faster
Swift large array (1,024)	47.91 µs	48.04 µs	🔴 +0.26% slower
Swift small array (16)	994.0 ns	1.03 µs	🔴 +3.78% slower
Swift typed map (8 entries)	2.40 µs	2.33 µs	🟢 -2.76% faster
Swift optional trailing string	156.0 ns	152.9 ns	🟢 -1.95% faster
Swift addNumbers()	78.2 ns	77.9 ns	🟢 -0.36% faster
Swift immediatePromise()	1.29 µs	1.26 µs	🟢 -2.72% faster
Swift number property get/set	137.2 ns	138.0 ns	🔴 +0.59% slower
Swift short ASCII string	165.6 ns	159.0 ns	🟢 -4% faster
Swift Unicode string	347.3 ns	345.0 ns	🟢 -0.66% faster
Swift number | string variant	193.2 ns	187.3 ns	🟢 -3.02% faster
TurboModule addNumbers()	867.7 ns	906.7 ns	🔴 +4.49% slower

Android

Benchmark	Before	After	Difference
Kotlin bounce(1 MiB)	1.18 µs	1.03 µs	🟢 -12.64% faster
Kotlin copy(1 MiB)	303.66 µs	324.14 µs	🔴 +6.74% slower
Kotlin create()	1.55 µs	1.64 µs	🔴 +5.85% slower

All Benchmarks

Benchmark	Before	After	Difference
JavaScript addNumbers()	22.1 ns	21.2 ns	🟢 -4.09% faster
C++ bounce(1 MiB)	435.6 ns	435.9 ns	🔴 +0.05% slower
C++ bounce(4 KiB)	435.9 ns	432.2 ns	🟢 -0.85% faster
C++ bounce native-owned buffer (1 MiB)	3.42 µs	3.41 µs	🟢 -0.1% faster
C++ bounce native-owned buffer (4 KiB)	451.5 ns	444.6 ns	🟢 -1.53% faster
C++ copy(1 MiB)	19.19 µs	19.01 µs	🟢 -0.92% faster
C++ copy(4 KiB)	795.1 ns	770.1 ns	🟢 -3.15% faster
C++ large array (1,024)	34.48 µs	34.95 µs	🔴 +1.36% slower
C++ small array (16)	740.0 ns	752.0 ns	🔴 +1.63% slower
C++ synchronousCallback()	324.6 ns	312.3 ns	🟢 -3.8% faster
C++ create()	541.1 ns	532.8 ns	🟢 -1.53% faster
C++ returnExisting()	214.4 ns	217.5 ns	🔴 +1.4% slower
C++ typed map (8 entries)	1.64 µs	1.70 µs	🔴 +3.6% slower
C++ optional trailing string	142.0 ns	141.1 ns	🟢 -0.63% faster
C++ addNumbers()	86.6 ns	86.7 ns	🔴 +0.09% slower
C++ simpleFunc()	71.9 ns	71.5 ns	🟢 -0.57% faster
C++ deferred worker Promise (includes trigger call)	23.29 µs	23.71 µs	🔴 +1.78% slower
C++ immediatePromise()	826.8 ns	825.5 ns	🟢 -0.15% faster
C++ number property get/set	154.1 ns	153.4 ns	🟢 -0.43% faster
C++ short ASCII string	138.7 ns	136.9 ns	🟢 -1.28% faster
C++ Unicode string	251.0 ns	246.4 ns	🟢 -1.83% faster
C++ nested Car struct	2.08 µs	2.06 µs	🟢 -0.73% faster
C++ number | string variant	128.8 ns	129.2 ns	🔴 +0.29% slower
Kotlin bounce(4 KiB)	1.02 µs	1.04 µs	🔴 +1.08% slower
Kotlin bounce native-owned buffer (1 MiB)	4.05 µs	4.06 µs	🔴 +0.38% slower
Kotlin bounce native-owned buffer (4 KiB)	1.13 µs	1.14 µs	🔴 +0.26% slower
Kotlin copy(4 KiB)	5.18 µs	4.98 µs	🟢 -3.89% faster
Kotlin large array (1,024)	38.82 µs	39.67 µs	🔴 +2.18% slower
Kotlin small array (16)	990.2 ns	983.4 ns	🟢 -0.69% faster
Kotlin synchronousCallback()	1.14 µs	1.13 µs	🟢 -1.05% faster
Kotlin returnExisting()	565.1 ns	569.2 ns	🔴 +0.73% slower
Kotlin typed map (8 entries)	6.20 µs	6.23 µs	🔴 +0.52% slower
Kotlin optional trailing string	320.1 ns	325.6 ns	🔴 +1.73% slower
Kotlin addNumbers()	113.6 ns	114.9 ns	🔴 +1.19% slower
Kotlin simpleFunc()	93.4 ns	93.0 ns	🟢 -0.5% faster
Kotlin deferred worker Promise (includes trigger call)	41.69 µs	43.20 µs	🔴 +3.62% slower
Kotlin immediatePromise()	2.40 µs	2.33 µs	🟢 -2.98% faster
Kotlin number property get/set	222.5 ns	221.8 ns	🟢 -0.32% faster
Kotlin short ASCII string	439.2 ns	438.5 ns	🟢 -0.17% faster
Kotlin Unicode string	722.0 ns	725.9 ns	🔴 +0.54% slower
Kotlin nested Car struct	4.51 µs	4.42 µs	🟢 -2.05% faster
Kotlin number | string variant	386.6 ns	387.3 ns	🔴 +0.16% slower
TurboModule addNumbers()	226.0 ns	224.0 ns	🟢 -0.87% faster

Benchmarking Code Diff 2e815f49...c7576d28 (view raw output)

Raw measurements: performance-report-1 (JSON artifact). Run 34154047323, attempt 1. Download requires GitHub access.

Android: measurements, attempt 1, apps, attempt 1.
iOS: measurements, attempt 1, apps, attempt 1.