From Fedora Project Wiki

< CI

(Add Contribute sections)
(Add section Gating)
Line 132: Line 132:
  
 
Once the pull request is created CI Pipeline will detect it and execute tests. Once the test execution is finished you will see results of the testing on the pull request page. See the [[CI/Pipeline|Pipeline]] page for the list of active pipelines and result examples.
 
Once the pull request is created CI Pipeline will detect it and execute tests. Once the test execution is finished you will see results of the testing on the pull request page. See the [[CI/Pipeline|Pipeline]] page for the list of active pipelines and result examples.
 +
 +
== Gating ==
 +
 +
Currently gating the package on test results is an opt-in feature. In order to enable gating for you component create a `gating.yaml` file in the root of your component dist git repository. See [[CI/Gating|Gating]] for more details.

Revision as of 12:45, 22 August 2018

Introduction

Are you eager to try out how the Fedora CI tests work? Do you want to get a quick hands-on experience without having to read too much documentation? This quick introduction for the impatient will show you a minimal set of steps to execute existing tests as well as provide useful links to resources where you can learn more.

First Steps

Install the following essential packages on your system:

sudo dnf install fedpkg standard-test-roles

Use fedpkg to clone the package git repository. See the Package Maintenance Guide for more info about the tool.

fedpkg clone -a bash

Tests are defined according to the Standard Test Interface in the tests directory:

cd bash/tests/

Test coverage to be executed together with the basic set of metadata is described in the tests.yml playbook. Use ansible-playbook to run all available tests for the classic environment on the local host:

ansible-playbook --tags=classic tests.yml

From the ansible output you can directly see an overall summary of the testing. If you see failed=0 at the end of the log then all tests passed:

localhost: ok=29 changed=11 unreachable=0 failed=0

For more detailed test results check the test.log and other files in the artifacts directory:

vim artifacts/test.log

That's it! You just executed test coverage for the Bash package :)

Test Subjects

To execute tests against different test subjects we need to prepare the environment. Let's store the detailed test results in /tmp/artifacts, use dynamic inventory as defined by the Standard Test Roles and download the latest Atomic Host image.

export TEST_ARTIFACTS=/tmp/artifacts
export ANSIBLE_INVENTORY=/usr/share/ansible/inventory
curl -Lo /tmp/atomic.qcow2 https://getfedora.org/atomic_qcow2_latest

Now let's try to run tests against all supported test subjects.

Classic

Run tests against classic rpms installed on the system:

export TEST_SUBJECTS=
ansible-playbook --tags=classic tests.yml

See Classic for detailed docs.

Container

Run tests in a docker container:

export TEST_SUBJECTS=docker:docker.io/library/fedora:latest
ansible-playbook --tags=container tests.yml

See Container for detailed docs.

Atomic

Run tests against the Atomic Host:

export TEST_SUBJECTS=/tmp/atomic.qcow2
ansible-playbook --tags=atomic tests.yml

See Atomic for detailed docs.

Hints

Debug

Would you like to investigate why a test failed? Enable debugging to easilly connect to running Atomic or Container to investigate:

export TEST_DEBUG=1
ansible-playbook --tags=atomic tests.yml

See Debug for details about debugging.

Ignore

Use .gitignore to specify files that Git should ignore. Such files are created during tests run. Create a tests/.gitignore file with the following contents:

# Ignore tests runs/artefacts.
artifacts/**
**/*.retry

Contribute

Are you interested in contributing a new test coverage? You are most welcome! As you have seen Executing a test is quite easy. Writing a new test or Wrapping an existing one is quite simple as well. Here's a few recommendations for creating a new pull request.

Fork

Unless you are maintainer of the package, who has direct commit access, create a fork of the package git repository using the Fork button in Pagure web interface and add your private fork as a new remote. Create a branch for your new tests. For example:

git remote add fork ssh://psss@pkgs.fedoraproject.org/forks/psss/rpms/bash.git
git checkout -b tests

If you are not a Fedora packager, use fedpkg command to clone you fork and set up the git repo config so that you are able to push to it. See Pull Requests for more detailed info.

fedpkg clone -a forks/psss/rpms/bash
git checkout -b tests

Add

Create new test coverage under the tests directory, update the tests.yml file accorgingly or create a new one. Run tests and verify they are stable and working fine in all supported environments. Add files to git, commit and push:

git add tests.yml test1 test2 test3
git commit -m "Add CI tests using the Standard Test Interface"
git push fork tests:tests

It is a good idea to include more details and links in the commit message to make the pull request easier for review:

Add CI tests using the Standard Test Interface

Adding initial set of basic functionality tests for bash
according to the Standard Test Interface [1]. See Quick Start
Guide for brief introduction about how to run these tests [2].

[1] https://fedoraproject.org/wiki/CI/Standard_Test_Interface
[2] https://fedoraproject.org/wiki/CI/Quick_Start_Guide

Create a new pull request from your tests branch against the master branch in the Pagure web interface. You might want to include an additional info about the tests such as:

There are three tests available: smoke and func have been tested
across all environments (classic, container, atomic), login is
relevant for classic only (because of a missing dependency).
Please, merge the tests into all currently supported branches.

Results

Once the pull request is created CI Pipeline will detect it and execute tests. Once the test execution is finished you will see results of the testing on the pull request page. See the Pipeline page for the list of active pipelines and result examples.

Gating

Currently gating the package on test results is an opt-in feature. In order to enable gating for you component create a gating.yaml file in the root of your component dist git repository. See Gating for more details.