Android CI

Continous Integration (CI) for Android apps with an emulator

By using snapshots, the emulator in this image is able to boot in under 15 seconds. This Docker image contains the Android SDK, emulator packages and an AVD with a snapshot.

Image

image: yorickvanzweeden/android-ci:latest

Specifications

Build-tools: 28.0.2
Platform: 24
System-image: android-24;google_apis;x86_64

Sample GitLab usage

.gitlab-ci.yml

image: yorickvanzweeden/android-ci:latest

before_script:
    - export GRADLE_USER_HOME=`pwd`/.gradle
    - chmod +x ./gradlew

cache:
  key: "$CI_COMMIT_REF_NAME"
  paths:
     - .gradle/

stages:
  - build
  - test_instrumented

build:
  stage: build
  script:
     - ./gradlew assembleDebug
  artifacts:
    expire_in: 1 week
    paths:
      - app/build/outputs/apk/

instrumentedTests:
  stage: test_instrumented
  only:
    - scheduled
  script:
    # Running emulator
    - /sdk/emulator/emulator -avd ${AVD_NAME} -no-window -no-audio -snapshot ${SNAPSHOT_NAME} &

    # Wait for emulator
    - ./android-wait-for-emulator

    # Run instrumented Android tests
    - /sdk/platform-tools/adb shell input keyevent 82
    - ./gradlew cAT

How it works

In order to have an emulator running in a Docker container, the Docker container must run with a privileged flag. This flag is required for KVM (Linux kernel virtualization). Docker build does not contain a privileged flag and therefore docker-compose is used with a bash script. The bash script is allowed to run in privileged mode and start the emulator. In order to save a snapshot, an expect script is fired that sets up a telnet connection. In Gitlab/other CI tools, the emulator may start from the snapshot instead of a cold boot. This reduces the startup time of the emulator to about 15 seconds.

If you want to create your own image, you should adjust the .env file to your needs and run the build-image.sh script. The name of the AVD and snapshot used are stored in the variables AVD_NAME and SNAPSHOT_NAME.