Travis: docker push at deploy instead of after_success?

[junaruga]

I found a tip for docker push. This tip can be used for not only multiarch/fedora, but also other multiarch/*.

Right now our .travis.yml's docker push logic is like this.

after_success:
    - |
      if [[ $TRAVIS_BRANCH == 'master' && $TRAVIS_PULL_REQUEST == 'false' ]]; then
          docker login -u="$DOCKER_USERNAME" -p="$DOCKER_PASSWORD" "$DOCKER_SERVER" && \
              docker push $DOCKER_REPO
      fi

This logic 's challenge is

• We cat not catch an script failure in the process of after_success. If the error happens in it, the Travis CI still succeeds.

I assume when using deploy: for docker login and docker push, we can catch an error in the deploy, also can write conditional statement like "if branch == master" simply.

https://medium.com/mobileforgood/5-tips-to-optimise-your-travis-ci-file-eced09d2d74e
3. Docker image caching

.travis.yml

before_deploy:
    - echo -n "$REGISTRY_PASS" | docker login -u "$REGISTRY_USER" --password-stdin
deploy:
    provider: script
    script: docker push "$IMAGE_TAG"
    on:
        branch: master

https://docs.travis-ci.com/user/docker/#branch-based-registry-pushes
https://docs.travis-ci.com/user/deployment