Skip to content

platform-changes.md

Latest commit

 

History

History
96 lines (75 loc) · 5.34 KB

platform-changes.md

File metadata and controls

96 lines (75 loc) · 5.34 KB

← Back to main index | ← Back to folder

13. BONUS: Platform Changes & Security (2026)

Android 16 Breaking Changes (Migration Guide)

Change Impact Action Testing Strategy
Adaptive apps (sw600dp+) Orientation/resize ignored on tablets/foldables. Declared orientation forced. Test multi-size layouts. material3.adaptive for responsive UI. Run on physical tablet or resizable-emulator. Check portrait ↔ landscape.
Predictive back Peek animation default, cannot opt-out. User sees animation before back triggers. Use OnBackPressedCallback for custom back logic. Coordinate animations with peek. Test back gesture visually. Ensure animations don't conflict with peek.
16KB page size Native libraries MUST align to 16KB pages (was 4KB). Misaligned libs fail to load. Recompile NDK code with latest NDK. Update CMakeLists.txt. Verify .so alignment. Test on API 36 device/emulator. Check adb logcat for "Failed to dlopen" errors.
Edge-to-edge Status/nav bar background colors ignored. System draws colors; app draws behind. Use WindowInsets APIs to inset content (padding). WindowCompat.setDecorFitsSystemWindows(false). Test with light/dark status bar. Check padding on notch devices.
Safer Intents Implicit intents must match exported intent filters. Unmatched intents throw ActivityNotFoundException. Audit all startActivity() calls. Use explicit intents or add matching <intent-filter>. Run adb logcat, search for "ActivityNotFoundException". Test intent-receiving apps.

Migration checklist for existing apps:

  1. Update compileSdk = 36
  2. Run lint: ./gradlew lint → fix hardcoded widths for tablets
  3. Test on API 36+ emulator (create multi-size AVD)
  4. Update native build if using NDK
  5. Audit startActivity() calls → add <intent-filter> or use explicit Intents
  6. Test predictive back gesture (pull from edge, release to cancel)

Security Quick Hits

Topic Key Detail
Credential Manager Standard auth. CredentialManager.getCredential()
EncryptedSharedPreferences security-crypto for tokens
Network Security Config Cert pinning via XML
R8 full mode Default AGP 8+. Know -keep rules.
exported Required since API 31. Default false.

Common Performance Pitfalls (& Fixes)

Pitfall Impact Symptom Fix
Too many recompositions 60fps → 15fps (jank) Frames drop during scroll Use derivedStateOf, stable params, remember for expensive computations
Large image unoptimized Memory spikes, OOM crash OutOfMemoryError on load Use Coil/Glide with .size(200, 200) sampling
Blocking I/O on Main ANR (App Not Responding) crash after 5s "App not responding" dialog Move to Dispatchers.IO via withContext
Allocating in scroll callback GC pauses every frame Jank during fast scroll Pre-allocate objects, reuse lists, avoid new in callbacks
Database queries N+1 100 queries instead of 1. 500ms → 50ms Load screen slow, list items sluggish Use SQL joins, @Relation, Room .with()
Flow doesn't unsubscribe Memory leak, coroutine runs forever Battery drain, memory growth Use repeatOnLifecycle(STARTED) or collectAsStateWithLifecycle()

Common N+1 query example:

💻 Code Example 
// ❌ BAD: Fetch users, then loop to fetch posts (100 queries for 100 users)
val users = db.getAllUsers()  // 1 query
for (user in users) {
    user.posts = db.getPostsByUserId(user.id)  // 100 queries!
}

// ✅ GOOD: Fetch in one query
@Entity
data class UserWithPosts(
    @Embedded val user: User,
    @Relation(parentColumn = "id", entityColumn = "userId") val posts: List<Post>
)

val usersWithPosts = db.getUsersWithPosts()  // 1-2 queries total

Performance Tooling

Tool Purpose How to Read
Studio Profiler CPU, memory, network (live) Look for sustained high CPU (>30%), memory spikes. Allocations = garbage collection trigger.
Perfetto (System Trace) Frame-level trace. Shows which functions run per frame. Green = on-time frame (60fps). Red = dropped frame. Drill into to see which function took time.
Baseline Profiles AOT compilation → 30-40% faster startup, 20% less jank Add via android.baselineProfileRules. Run on real device to generate .txt profile.
Macrobenchmark Measure startup time, scroll FPS, frame compositing time Baseline against previous version. Target: startup <1s, scroll ≥55fps.
Compose Compiler Metrics Stability/skippability per composable. restartable, skippable, scheme restartable = can recompose alone. skippable = does if params stable. Find unstable params → cause cascade recompositions.

Profiling workflow:

  1. Identify jank: Feel sluggish? Run Perfetto → see which frame drops
  2. Profile CPU: Studio Profiler → CPU tab → record 5s → see top functions
  3. Check memory: Memory tab → Dump heap → analyze for leaks (retained objects)
  4. Baseline: Macrobenchmark on release build → startup time, scrolling FPS

Example: Finding recomposition hotspot:

💻 Code Example 
./gradlew debugComposeCompilerMetrics
# Generates build/compose-metrics/ComposableMetrics.txt
# Look for `unstable` → find why (e.g., mutableListOf passed as param)