Directly use the taskcat/taskcat Docker container #20

ShahradR · 2020-07-11T02:00:30Z

Eliminate the need for a local Dockerfile—we can directly use the taskcat/taskcat:latest container image on Docker Hub. Custom scripts can be executed without the need to define a Dockerfile by using the pre-entrypoint and post-entrypoint.

The text was updated successfully, but these errors were encountered:

Add a test to verify the success scenario, where taskcat is provided with valid AWS credentials and deploys a CloudFormation template. Because the mock layer has not yet been implemented, we are actually hitting AWS to run these tests. The AWS credentials are provided to the test via Maven Surefire parameters passed through the `mvn` command from the workflow. The credentials themselves are stored as repository secrets. We implement the success scenario ahead of the changes to the Docker container to ensure that the action works both before and after we use the taskcat/taskcat container instead of the local Dockerfile. Associated issue: #20

When running the JUnit tests through GitHub Actions, we often end up with "java.lang.IllegalThreadStateExceptions" being thrown due to asserts being ran before act has finished executing. By adding the Process.waitFor() call before the assertions, we wait until the act process has finished before running the rest of the tests. Associated issue: #20

ShahradR · 2020-07-12T01:33:41Z

Because both the entrypoint and args both expect lists, we can't call taskcat directly without a script to tokenize the string and create a CMD array for Docker to consume. While there is some string manipulation available from yaml when running the tests (not sure if it's provided by YAML itself, act, or if it's even available from GitHub Actions), it probably isn't enough to make it work without the need for a Dockerfile and an entrypoint script to split the command into tokens.

Using the `.split()` function to return a list

A .split() function actually works against stings. Unfortunately, the array is cast back into a string when passed to the Docker container and taskcat still throws an error.

`action.yaml`

runs:
  using: docker
  image: docker://taskcat/taskcat:latest
  entrypoint:
    - taskcat
  args:
    - ${{ inputs.commands.split(" ") }}

`.github/workflow`

steps:
  - uses: actions/checkout@v2
  - name: taskcat
    uses: ./../../../../
    with:
      commands: test run

act output

[main.yaml/taskcat] ⭐  Run taskcat
[main.yaml/taskcat]   🐳  docker run image=taskcat/taskcat:latest entrypoint=["taskcat"] cmd=["test,run"]
[main.yaml/taskcat]   |  _            _             _   
[main.yaml/taskcat]   | | |_ __ _ ___| | _____ __ _| |_ 
[main.yaml/taskcat]   | | __/ _` / __| |/ / __/ _` | __|
[main.yaml/taskcat]   | | || (_| \__ \   < (_| (_| | |_ 
[main.yaml/taskcat]   |  \__\__,_|___/_|\_\___\__,_|\__|
[main.yaml/taskcat]   |                                 
[main.yaml/taskcat]   | 
[main.yaml/taskcat]   | 
[main.yaml/taskcat]   | version 0.9.18
[main.yaml/taskcat]   | usage: taskcat [args] <command> [args] [subcommand] [args] 
[main.yaml/taskcat]   | taskcat: error: argument : invalid choice: 'test,run' (choose from 'delete', 'deploy', 'lint', 'list', 'package', 'test', 'update-ami', 'upload')
[main.yaml/taskcat]   ❌  Failure - taskcat

Passing list values using the index

Because the split() function actually returns an array, we can get individual items by referencing its index. However, because we can't predict how many arguments will be passed, this is useless without the ability to iterate and dynamically build the args list.

`action.yaml`

runs:
  using: docker
  image: docker://taskcat/taskcat:latest
  entrypoint:
    - taskcat
  args:
    - ${{ inputs.commands.split(" ")[0] }}

`.github/workflow`

steps:
  - uses: actions/checkout@v2
  - name: taskcat
    uses: ./../../../../
    with:
      commands: test run

act output

[main.yaml/taskcat] ⭐  Run taskcat
[main.yaml/taskcat]   🐳  docker run image=taskcat/taskcat:latest entrypoint=["taskcat"] cmd=["test"]
[main.yaml/taskcat]   |  _            _             _   
[main.yaml/taskcat]   | | |_ __ _ ___| | _____ __ _| |_ 
[main.yaml/taskcat]   | | __/ _` / __| |/ / __/ _` | __|
[main.yaml/taskcat]   | | || (_| \__ \   < (_| (_| | |_ 
[main.yaml/taskcat]   |  \__\__,_|___/_|\_\___\__,_|\__|
[main.yaml/taskcat]   |                                 
[main.yaml/taskcat]   | 
[main.yaml/taskcat]   | 
[main.yaml/taskcat]   | version 0.9.18
[main.yaml/taskcat]   | usage: taskcat [args] test [args] [subcommand] [args] 
[main.yaml/taskcat]   | taskcat [args] <command> [args] [subcommand] [args] test: error: the following arguments are required: 
[main.yaml/taskcat]   ❌  Failure - taskcat

Add a test that prints taskcat's help output by calling taskcat with no commands. This test is designed to act as a starting point for the implementation of a generic `commands` input to the GitHub Action, allowing users the flexibility of running arbitrary tasckat commands, rather than the hard-coded `taskcat test run`. Associated issue: #20

Allow users of this action to run any supported taskcat command, rather than the previously hard-coded `taskcat test run`. This commit also removes the dependency on the local Dockerfile, in favor of the official taskcat/tasckat container hosted on Docker Hub. Commands are passed to the container by overriding the entrypoint through the action.yaml file. The `taskcat` invocation itself is also handled by the action. Users only need to specify the command, subcommands, and arguments to be passed to run their tests. Associated issue: #20

Add a test to verify the success scenario, where taskcat is provided with valid AWS credentials and deploys a CloudFormation template. Because the mock layer has not yet been implemented, we are actually hitting AWS to run these tests. The AWS credentials are provided to the test via Maven Surefire parameters passed through the `mvn` command from the workflow. The credentials themselves are stored as repository secrets. We implement the success scenario ahead of the changes to the Docker container to ensure that the action works both before and after we use the taskcat/taskcat container instead of the local Dockerfile. Associated issue: #20

When running the JUnit tests through GitHub Actions, we often end up with "java.lang.IllegalThreadStateExceptions" being thrown due to asserts being ran before act has finished executing. By adding the Process.waitFor() call before the assertions, we wait until the act process has finished before running the rest of the tests. Associated issue: #20

Add a test that prints taskcat's help output by calling taskcat with no commands. This test is designed to act as a starting point for the implementation of a generic `commands` input to the GitHub Action, allowing users the flexibility of running arbitrary tasckat commands, rather than the hard-coded `taskcat test run`. Associated issue: #20

Allow users of this action to run any supported taskcat command, rather than the previously hard-coded `taskcat test run`. This commit also removes the dependency on the local Dockerfile, in favor of the official taskcat/tasckat container hosted on Docker Hub. Commands are passed to the container by overriding the entrypoint through the action.yaml file. The `taskcat` invocation itself is also handled by the action. Users only need to specify the command, subcommands, and arguments to be passed to run their tests. Associated issue: #20

github-actions · 2020-07-12T10:54:50Z

🎉 This issue has been resolved in version 1.0.0-develop.2 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀