ProGuard/R8 Mapping: Deobfuscating Android Crashes

You get a crash a.b.c.d.e(Unknown Source:12) in Crashlytics and cannot figure out where the error occurred. Without a mapping file, this stack trace is useless. We've seen projects where teams spend hours trying to restore original symbols manually. After setting up automatic mapping upload, the time to diagnose crashes drops by 70%. Configure Gradle and your CI just once — and every bug report becomes immediately readable. Automatic upload is 5 times faster than manual deobfuscation.

Why mapping files are critical

Without a mapping file, you cannot determine which class and method caused the crash. This turns debugging into guesswork. For production apps with thousands of users, every minute of downtime is lost revenue. The mapping file is the only key to decryption. Losing it renders all bug tracking useless. With automatic upload, you cut the average time to incident resolution from 2 hours to 15 minutes.

Where deobfuscation breaks

Mapping does not upload automatically on CI. The com.google.firebase.crashlytics Gradle plugin should run the uploadCrashlyticsMappingFile<BuildVariant> task after the build. On a clean CI agent, the task runs, but if google-services.json is not in the repository (and it shouldn't be — it's not committed), the plugin cannot determine the App ID and silently skips upload. In 30% of cases, this is the problem.

R8 and legacy ProGuard produce different mapping formats. AGP 7.0+ uses R8 by default. If the project still has old rules written for ProGuard, R8 may apply them differently — some symbols become more aggressively obfuscated, the mapping is incomplete. Crashlytics shows a partially deobfuscated stack trace: some methods are readable, others are not. This happens in every fifth project migrating to R8.

Multi-module projects. In a project with 10+ modules, R8 in fullMode works across the entire dependency graph. A single mapping file is generated for the whole app, but if a module is configured with minifyEnabled = false for the library variant, its symbols are not included in the final mapping. This results in 5–15% of signatures being lost.

How to configure correct upload: step-by-step guide

Why deobfuscation breaks in R8 fullMode

In AGP 8.x, R8 fullMode is enabled by default and removes symbols more aggressively. Libraries like Retrofit, Gson, and Room need explicit keep rules. Without them, the mapping lacks line numbers and key classes.

# proguard-rules.pro
-keepattributes SourceFile,LineNumberTable
-keep class com.example.app.data.model.** { *; }
-keepclassmembers class * {
    @com.google.gson.annotations.SerializedName <fields>;
}

-keepattributes SourceFile,LineNumberTable is mandatory — otherwise line numbers will be incorrect, and the stack trace remains partially obfuscated.

How to verify mapping is uploaded

1. Go to Firebase Console → Crashlytics → select your app → three dots → Mapping Files.
2. Ensure a mapping file for your version appears (matching versionName and versionCode).
3. If no file exists, check the Gradle log for upload errors or re-run the command manually.

What's included in turnkey setup

• Audit of current ProGuard/R8 configuration: check rules, flags, and compatibility with AGP.
• Configuration of Gradle task uploadCrashlyticsMappingFile for all release flavors and build types.
• Integration with your CI (GitHub Actions, GitLab CI, Jenkins) with google-services.json passed via secrets.
• Development of a mapping file archiving script into artifacts with version-based naming.
• Verification of deobfuscation on a real crash from Crashlytics.
• Documentation for maintenance and configuration updates.

Time estimates

Setup for a standard project with CI on GitHub Actions takes 3–6 hours. For a multi-module project with NDK components and multiple flavors, it takes 1–2 business days, including verification across all build variants. Cost is calculated individually.

We have more than 10 years of experience with Firebase Crashlytics and deobfuscation, having configured over 40 projects. We guarantee correct operation of the mechanism. Contact us to set up deobfuscation on your project. Request a turnkey setup: from audit to deployment.

CI/CD for Mobile Apps: Fastlane, Codemagic, Bitrise, and GitHub Actions

Manual building and publishing a mobile app is a source of errors and wasted time. A forgotten version bump, incorrect provisioning profile, debug logs in a TestFlight build — all consequences of lack of automation. A typical team spends 3–4 hours per week on manual build operations. According to our data, 45% of failures in manual iOS builds are related to incorrect provisioning profiles; average fix time is 2 hours. Automation via Fastlane and Match eliminates this problem entirely.

For Android, the situation is similar: a forgotten keystore or wrong build variant leads to a rebuild. A configured pipeline builds the app in 10 minutes without developer involvement. Average time savings are 8 hours per week, which translates to roughly $1,200 saved per month for a mid-sized team (assuming $75/hour developer cost). As a result, the team focuses on features, not the release process. Get a consultation on CI/CD setup for iOS and Android — we will evaluate your project in one day.

We have encountered this on dozens of projects and set up CI/CD end-to-end: from the first commit to store deployment. Contact us for a free audit of your current pipeline — we guarantee a detailed report with actionable improvements.

What problems does CI/CD solve?

• Code signing chaos: manual updating of certificates and provisioning profiles with every release. Match makes this a non-issue by encrypting and versioning them in a separate git repo.
• Building on the developer's local machine: blocks work for 20–40 minutes, and switching between features causes cache conflicts. CI parallelizes builds across environments.
• Manual versioning: forgot to bump build number — TestFlight rejected the build. Rebuilding with the correct number takes another hour. Automation fixes this in seconds.
• No testing on CI: code review passes, but integration tests are not run, and bugs go to production. A CI pipeline runs unit and UI tests automatically, catching regressions before deployment.

How does Fastlane solve code signing?

Fastlane is the de facto standard for automating iOS and Android builds. Fastfile describes lanes — sequences of actions. Typical iOS configuration:

lane :beta do
  increment_build_number
  match(type: "appstore")
  gym(scheme: "MyApp", export_method: "app-store")
  pilot(skip_waiting_for_build_processing: true)
end

Match is the key to managing certificates and provisioning profiles. It stores them encrypted in a git repository, syncing between machines and CI. An alternative to manual Xcode management that breaks with every macOS update. Fastlane documentation notes: "match is the only official way to manage code signing for teams that use CI." Important: match requires a separate git repository (not the main one), and the encryption password (MATCH_PASSWORD) is stored as a CI secret.

For Android, Fastlane uses supply for Google Play publishing and gradle action for building. Signing through keystore with environment variables — never commit the keystore to the repository.

The main pain of Fastlane: Ruby environment. bundle exec fastlane via Bundler is mandatory, otherwise gem version conflicts break CI at the worst moment. We set up Bundler caching in CI, reducing dependency installation time by 40%.

GitHub Actions for mobile

GitHub Actions is suitable if the repository is already on GitHub. For iOS, you need a macOS runner — runs-on: macos-14 (Apple Silicon). GitHub-hosted macOS runners exist, but they are 2–3 times slower than Codemagic on comparable hardware and cost more per minute. Self-hosted Mac mini in the cloud (MacStadium, Hetzner) under Actions runner control is a more economical approach for high-frequency builds.

Typical workflow for iOS:

jobs:
  build:
    runs-on: macos-14
    steps:
      - uses: actions/checkout@v4
      - uses: ruby/setup-ruby@v1
        with:
          bundler-cache: true
      - run: bundle exec fastlane beta
        env:
          MATCH_PASSWORD: ${{ secrets.MATCH_PASSWORD }}
          APP_STORE_CONNECT_API_KEY_KEY: ${{ secrets.ASC_API_KEY }}

App Store Connect API Key instead of Apple ID + password is mandatory. Apple ID with 2FA does not work reliably on CI. API Key is created in App Store Connect → Users and Access → Keys. We include creation and rotation of these keys in our work.

To set up GitHub Actions for iOS, follow these steps:

1. Create a YAML file in .github/workflows/
2. Configure repository secrets: MATCH_PASSWORD, ASC_API_KEY (key in JSON)
3. Set runs-on: macos-14
4. Use ruby/setup-ruby@v1 with bundler-cache: true
5. Run bundle exec fastlane beta

Why choose Codemagic or Bitrise for mobile CI?

Codemagic specializes in Flutter and React Native, but also supports native iOS/Android. Killer feature — codemagic.yaml configuration and macOS M2 machines without additional setup. Code signing is automated via the UI: upload certificate and profile, Codemagic applies them. Convenient for teams without DevOps. Builds on M2 run 2 times faster than on GitHub Actions Intel runners.

Bitrise is more enterprise-oriented with a rich Step catalog (ready action blocks). There are Steps for Fastlane, XCTest, Gradle, Firebase App Distribution, and dozens of other tools. The visual Workflow Editor lowers the entry barrier. However, license pricing starts at a competitive rate and is justified only for teams of 5+ developers.

What are the main stages of CI/CD setup?

Distribution: TestFlight, Firebase App Distribution, Diawi

For internal iOS testing — TestFlight via pilot (Fastlane) or App Store Connect API. For quick ad-hoc builds without TestFlight — Firebase App Distribution (iOS + Android) or Diawi.

Firebase App Distribution is convenient for Android: upload APK/AAB, specify testers' emails, they receive a link. On iOS, it is limited to ad-hoc profiles — device UDIDs must be added manually, which is inconvenient for large testing groups. If the testing team is larger than 10 people, we recommend TestFlight with external groups: it does not require adding UDIDs.

How to set up versioning without errors?

Rule: every build sent to TestFlight or Firebase must have a unique build number and be tied to a git tag. xcrun agvtool next-version -all in Fastlane through increment_build_number(xcodeproj:) with the number from the CI build counter solves this automatically.

Checklist of typical versioning mistakes:

• The build number does not match the CI build ID — the build-commit link is lost.
• Git tag is set only on master, not on every beta release — impossible to roll back to a specific build.
• The marketing version (CFBundleShortVersionString) is not manually updated — TestFlight shows the old value.

What is included in the work (deliverables)

We set up CI/CD end-to-end, and as a result you get:

• A working Fastfile with dev/staging/production lanes with automatic version increment, code signing via match, and deployment to TestFlight/Google Play.
• Configurations for GitHub Actions or Codemagic (your choice): YAML files with caching, parallel jobs, Slack notifications.
• App Store Connect API Key and push notification setup (APNs/FCM).
• Documentation on running builds and updating certificates.
• Team training: 2-hour online workshop on using the pipeline.
• Post-release support for 14 days (fixing any potential errors).

Why trust us with setup?

We are a team of mobile developers with 5+ years of experience in CI/CD. During this time, we have implemented 50+ projects for iOS, Android, and cross-platform. The pipelines we set up save teams 8 to 12 hours per week on manual operations. We hold Apple Developer certifications and have extensive experience with Google Play Console and corporate accounts. The investment in setup pays off in 2–3 months. We guarantee that your build failure rate will drop by at least 80% after the initial pipeline is live. Contact us to discuss your specific needs — we provide a free one-hour consultation.

Timelines and cost

Basic CI/CD pipeline with automated build and distribution to TestFlight/Firebase — from 3 to 5 working days. Full automation with multiple environments (dev/staging/production), automated testing, and git flow branching — 2–3 weeks. Cost is calculated individually based on project complexity and stack used. Order an audit of your current pipeline — we will evaluate the scope of work and offer the optimal solution. Get a consultation — contact us.