ref(macrobenchmark): Trim comments and drop the cache encryption key

The retrieval rationale these comments carried is documented in the module
README, so removing the duplicated prose loses nothing.

Also drops cache-encryption-key from the Setup Gradle step, leaving this
workflow without the Gradle configuration-cache encryption the other workflows
still configure.

Author: runningcode

.github/workflows/integration-tests-macrobenchmark.yml

@@ -2,15 +2,6 @@ name: 'Integration Tests - Macrobenchmark'
 # Runs the sentry-uitest-android-macrobenchmark cold-start benchmark on a Sauce Labs real
 # device and recovers timeToInitialDisplay from the device log.
 #
-# Getting the numbers back is the awkward part. saucectl only downloads assets Sauce hosts
-# for a job (device.log, junit.xml, video.mp4, ...) and never reads the device filesystem,
-# so Macrobenchmark's benchmarkData.json cannot be fetched directly. The Real Device Access
-# API can pull device files, but POST /sessions accepts only private devices
-# (deviceClasses=[PRIVATE_DEVICE]) and we run on the public cloud. So the benchmark echoes
-# its own results into logcat, and scripts/parse-macrobenchmark-log.py reassembles them.
-#
-# Manual trigger only -- this reports numbers, it does not gate PRs. Cloud devices have
-# unlocked CPU clocks, so run-to-run spread is far wider than most SDK-init changes.
 on:
   workflow_dispatch:
   # Temporary scaffolding: workflow_dispatch cannot target a workflow that does not exist on
@@ -45,8 +36,6 @@ jobs:
 
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@9c971963bec38e04b3d30dcc455b5382be2fdbfb # v6.3.0
-        with:
-          cache-encryption-key: ${{ secrets.GRADLE_ENCRYPTION_KEY }}
 
       - name: Assemble target app and Macrobenchmark apk
         if: env.SAUCE_USERNAME != null

.sauce/sentry-uitest-android-macrobenchmark.yml

@@ -14,10 +14,7 @@ defaults:
   timeout: 40m
 
 espresso:
-  # Target app under test: Macrobenchmark cold-starts sentry-samples-android. It must be
-  # release-like; the release build type is signed with the debug key so it installs on Sauce.
   app: ./sentry-samples/sentry-samples-android/build/outputs/apk/release/sentry-samples-android-release.apk
-  # Instrumentation APK: the self-instrumenting com.android.test macrobenchmark module.
   testApp: ./sentry-android-integration-tests/sentry-uitest-android-macrobenchmark/build/outputs/apk/benchmark/sentry-uitest-android-macrobenchmark-benchmark.apk
 
 suites:
@@ -26,21 +23,9 @@ suites:
     # No test orchestrator and no clearPackageData: Macrobenchmark manages its own process
     # restarts and AOT compilation, and StartupMode.COLD intentionally keeps app data and
     # permissions (it force-stops rather than `pm clear`).
-    #
-    # Older hardware would give a better signal -- it spends longer in SentryAndroid.init, so a
-    # regression is larger against the same relative noise -- but Google_Pixel_4_13_real_us could
-    # not be allocated in three attempts, each queueing ~61 minutes before Sauce gave up with a
-    # concurrency retry-timeout. This device allocates reliably.
-    #
-    # Keep this to one device: two would produce two device logs, which the parser rejects rather
-    # than splice together -- see scripts/parse-macrobenchmark-log.py.
     devices:
       - id: Google_Pixel_9_Pro_XL_15_real_sjc1 # Google Pixel 9 Pro XL - api 35 (15) - high end
 
-# The device log carries the actual results: Sauce only returns assets it produces itself, so it
-# cannot pull Macrobenchmark's `<pkg>-benchmarkData.json` off the device. SentryStartupBenchmark
-# echoes that JSON into logcat instead, and scripts/parse-macrobenchmark-log.py reassembles it.
-# The log also holds Macrobenchmark's device guard warnings (unlocked clocks, low battery, ...).
 artifacts:
   download:
     when: always

sentry-android-integration-tests/sentry-uitest-android-macrobenchmark/src/main/java/io/sentry/uitest/android/macrobenchmark/SentryStartupBenchmark.kt

@@ -27,10 +27,6 @@ import org.junit.runner.RunWith
  *   [android.os.Trace] section the SDK emits, isolating SDK-init cost from the rest of the start.
  *
  * [CompilationMode.Full] pins ART AOT compilation so dexopt state does not drift between runs.
- * Iterations are capped at 12: on an unthrottled Pixel 3, back-to-back cold starts hit thermal
- * throttling after ~14 iterations, which inflates the tail of longer runs. This is NOT a CI test;
- * it requires a connected device. To A/B an SDK change, see README.md (build the app twice, once
- * per SDK variant, in interleaved rounds).
  */
 @OptIn(ExperimentalMetricApi::class)
 @RunWith(AndroidJUnit4::class)
@@ -63,10 +59,7 @@ class SentryStartupBenchmark {
     /** Kept in sync with `scripts/parse-macrobenchmark-log.py`. */
     private const val LOG_TAG = "SentryBenchmarkData"
 
-    /**
-     * Well under logcat's ~4 KB per-message cap, so a chunk is never silently truncated, while
-     * still keeping the whole document to a handful of messages.
-     */
+    /** Well under logcat's ~4 KB per-message cap, so a chunk is never silently truncated. */
     private const val CHUNK_LENGTH = 2000
 
     /**
@@ -102,11 +95,8 @@ class SentryStartupBenchmark {
 
     private fun findBenchmarkData(): File? {
       val context = InstrumentationRegistry.getInstrumentation().targetContext
-      // Deprecated since API 30 in favour of MediaStore, which hands back content URIs rather
-      // than the filesystem path androidx.benchmark writes its File to -- so there is nothing to
-      // migrate to. Suppressed on this call alone; externalCacheDir below is not deprecated.
+      // This is where Macrobenchmark writes to for some reason.
       @Suppress("DEPRECATION") val mediaDirs = context.externalMediaDirs.toList()
-      // Outputs uses the media dir from API 29 on, and externalCacheDir on API 24-28.
       return (mediaDirs + context.externalCacheDir).filterNotNull().firstNotNullOfOrNull { dir ->
         dir.listFiles()?.firstOrNull { it.name.endsWith(BENCHMARK_DATA_SUFFIX) }
       }