CI/Metadata

Requirements

In order to use the Flexible Metadata Format effectively for the CI testing we need to agree on the essential set of attributes to be used. For each attribute we need to standardize:

• Name ... unique, well chosen, possibly with a prefix
• Type ... expected content type: string, number, list, dictionary
• Purpose ... description of the attribute purpose

For names we should probably consider using namespace prefix (such as test-description, requirement-description) to prevent future collisions with other attributes. Each attribute definition should contain at least one apt example of the usage. Or better, a set of user stories to be covered.

Attributes

In this section there are attributes proposed so far. Material for discussion. Nothing final for now.

Summary

In order to efficiently collaborate on test maintenance it's crucial to have a short summary of what the test does.

• Name ... summary
• Type ... string (one line, up to 50 characters)
• Purpose ... concise summary of what the test does

User stories:

• As a developer reviewing 10 failed tests I would like to get quickly an idea of what my change broke.

Notes:

• Shall we recommend 50 characters or less? Like there is for commits? Yes.

Example:

summary: Test wget recursive download options

Description

For complex tests it makes sense to provide more detailed description to better clarify what is covered by the test.

• Name ... description
• Type ... string (multi line, plain text)
• Purpose ... detailed description of what the test does

User stories:

• As a tester I come to a test code I wrote 10 years ago (so I have absolutely no idea about it) and would like to quickly understand what it does.
• As a developer I review existing test coverage for my component and would like to get an overall idea what is covered without having to read the whole test code.

Example:

description: |
  This test checks all available wget options related to
  downloading files recursively. First a tree directory
  structure is created for testing. Then a file download
  is performed for different recursion depth specified by
  the "--level=depth" option.

Tags

Throughout the years, free-form tags proved to be useful for many, many scenarios. Primarily to provide an easy way how to select a subset of objects.

• Name: tags
• Type: list
• Purpose: free-form tags for easy filtering

Notes:

• Tags are case-sensitive.
• Using lowercase is recommended.

User stories:

• A a developer/tester I would like to run only a subset of available tests.

Example:

tags: [Tier1, fast]

Test

This is the key content attribute defining how the test is to be executed.

• Name: test
• Type: string
• Purpose: shell command which executes the test

User stories:

• As a developer/tester I want to easily execute all available tests with just one command.
• As a test writer I want to run a single test script in multiple ways (e.g. providing different parameters)

Example:

test: ./runtest.sh

Path

As the object hierarchy does not need to copy the filesystem structure (e.g. virtual test cases) we need a way how to define where the test is located.

• Name: path
• Type: string
• Purpose: filesystem directory to be entered before executing the test

User stories:

• As a test writer I define two virtual test cases, both using the same script for executing. See also the Virtual Tests example.

Example:

path: wget/recursion

Environment

Test scripts might require certains environment variable to be set. Although this can be done on the shell command line as part of the Test attribute it makes sense to have a dedicated field for this, especially when the number of parameters grows. This might be useful for virtual test cases as well.

• Name: environment
• Type: dictionary
• Purpose: environment variables to be set before running the test

User stories:

• As a tester I need to pass environment variables to my test script to properly execute the desired test scenario.
• As a tester I'm using a single test script for testing different Python implementations specified by environment variable PYTHON.

Example:

environment:
   PACKAGE: python37
   PYTHON: python3.7

Duration

In order to prevent stuck tests consuming resources we should be able to define a maximum time for test execution.

• Name: duration
• Type: string
• Purpose: maximum time for test execution after which a running test is killed by the test harness

Notes:

• Let's use the same format as the sleep command. For example: 3m, 2h, 1d.

User stories:

• As a deverloper/tester I want to prevent resource wasting by stuck tests.
• As a test harness I need to know after how long time I should kill test if it is still running.

Example:

duration: 5m

Relevancy

Sometimes a test case is only relevant for specific environment. Test Case Relevancy allows to filter irrelevant test cases out.

• Name: relevancy
• Type: list
• Purpose: Test Case Relevancy rules used for filtering relevant test cases for given environment.

User stories:

• As a tester I want to skip execution of a particular test case in given test environment.

Notes:

Environment is defined by one or more environment dimensions such as product, distro, collection, variant, arch, component. Relevancy consists of a set of rules of the form condition: decision. For more details see the Test Case Relevancy documentation.

Example:

relevancy:
   - "distro < f-28: False"
   - "distro = rhel-7 & arch = ppc64: False"

Contact

When there are several people collaborating on tests it's useful to have a way how find who is responsible for what.

• Name ... contact
• Type ... string (name with email address)
• Purpose ... person maintaining the test

User stories:

• As a developer reviewing a complex failed test I would like to contact the person who maintains the code and understands it well.

Example:

   contact: Name Surname <email@address.org>

Component

It's useful to be able to easily select all tests relevant for given component or package. As they do not always have to be stored in the same repository and because many tests cover multiple components a dedicated field is needed.

• Name ... component
• Type ... list of strings
• Purpose ... relevant fedora/rhel source package names for which test should be executed

User stories:

• As a SELinux tester testing the checkpolicy component I want to run Tier1 tests for all SELinux components plus all checkpolicy tests.

Example:

   component: [libselinux, checkpolicy]

Notes:

The following fmf command can be used to select test set described by the user story above:

   fmf --key test --filter 'tags:Tier1 | component:checkpolicy'

Tier

It's quite common to organize tests into "tiers" based on their importance, stability, duration and other aspects. For this tags have been used quite often as there was not corresponding attribute available. It might make sense to have a dedicated field for this functionality as well.

• Name ... tier
• Type ... string
• Purpose ... name of the tier set this test belongs to

User stories:

• As a tester testing a security advisory I want to run the stable set of important tests which cover the most essential functionality and can provide test results in a short time.

Example:

   tier: 1

Provision

In some cases tests have special requirements for the environment in order to run successfully. For now just simple qemu options for the standard-inventory-qcow2 provisioner are supported.

• Name ... provision
• Type ... dictionary
• Purpose ... set of environment requirements

User stories:

• As a tester I want to specify amount of the memory which needs to be available for the test.
• As a tester I want to specify network interface card to be used in qemu.

Example:

provision:
  standard-inventory-qcow2:
    qemu:
      m: 3G
      net_nic:
        model: e1000

Memory size is specified in megabytes. Optionally, a suffix of “M” or “G”. Use qemu-system-x86_64 -net nic,model=help for a list of available devices. See also real life example usage.

Examples

Below you can find some basic examples using the metadata definen above. Separate Examples page illustrates integration with Standard Test Roles on some real-life components.

BeakerLib Tests

Three beakerlib tests, each in it's own directory:

main.fmf

 test: ./runtest.sh

 /one:
     path: tests/one
 /two:
     path: tests/two
 /three:
     path: tests/three

fmf

 tests/one
 path: tests/one
 test: ./runtest.sh

 tests/two
 path: tests/two
 test: ./runtest.sh

 tests/three
 path: tests/three
 test: ./runtest.sh

Three Scripts

Three different script residing in a single directory:

main.fmf

path: tests

/one:
    test: ./one
/two:
    test: ./two
/three:
    test: ./three

fmf

tests/one
path: tests
test: ./one

tests/two
path: tests
test: ./two

tests/three
path: tests
test: ./three

Virtual Tests

Thre virtual test cases based on a single test script:

main.fmf

path: tests/virtual

/one:
    test: ./script --one
/two:
    test: ./script --two
/three:
    test: ./script --three

fmf

tests/one
path: tests
test: ./script --one

tests/two
path: tests
test: ./script --two

tests/three
path: tests
test: ./script --three