mdoucha / os-autoinst

OS-level test automation

os-autoinst

Table of Contents

• Getting started
• How to run test cases
• Develop test modules
• Verifying a runtime environment
• How to contribute
  • Rules for commits
  • Deprecation approach
• Build instructions
  • Installing dependencies
  • Conducting the build
• Standalone tool for image search
• History of os-autoinst
• Further notes

The OS-autoinst project aims at providing a means to run fully automated tests. Especially to run tests of basic and low-level operating system components such as bootloader, kernel, installer and upgrade, which can not easily and safely be tested with other automated testing frameworks. However, it can just as well be used to test firefox and openoffice operation on top of a newly installed OS.

os-autoinst can be executed alone, but is currently designed to be executed together with openQA, the web user interface that allows to run more than one os-autoinst instance at the same time.

More information on os-autoinst and openQA can be found on http://os-autoinst.github.io/openQA/

Getting started

Under openSUSE the os-autoinst package can be installed from the official repository or from our devel repository. For further details, have a look at the openQA documentation.

For building os-autoinst manually checkout the build instructions below.

The main executable isotovideo can read test parameters from the command line or read test parameters from a file named vars.json. This file stores the values of the different variables that will configure the behavior of the test execution.

A container is provided and can be pulled and the main execution can be called in one step, for example using the podman container engine for tests defined in the current directory on x86_64 if your environment supports KVM virtualization acceleration:

podman run --rm -it -v .:/tests registry.opensuse.org/devel/openqa/containers/isotovideo:qemu-kvm casedir=/tests

Use the image variant ending with qemu-x86 on x86_64 if no KVM support is available.

Take a look on https://registry.opensuse.org/cgi-bin/cooverview?srch_term=project%3Ddevel%3AopenQA for all available container images.

Additional test variables can be supplied on the command line. There are some variables used by os-autoinst itself and other that are used by the tests. A minimal command line can look like this:

isotovideo distri=opensuse casedir=/full/path/for/tests iso=/full/path/for/iso

As alternative or completementary a corresponding vars.json with additional parameters could be:

{
   "DISTRI" :      "opensuse",
   "CASEDIR" :     "/full/path/for/tests",
   "NAME" :        "test-name",
   "ISO" :         "/full/path/for/iso",
   "VNC" :         "91",
   "BACKEND" :     "qemu",
   "DESKTOP" :     "kde"
}

Be advised that the file vars.json is also modified by os-autoinst so make sure to backup handcrafted versions of this file.

For more concrete instructions read on in the "How to run test cases" section below. Find sections about "How to contribute" or "Build instructions" further below.

How to run test cases

This following instructions shows how to run test cases. First one needs to clone the test distribution. Checkout os-autoinst-distri-example for an example of a minimal test distribution.

Example for openSUSE’s tests:

mkdir distri && cd distri
git clone [email protected]:os-autoinst/os-autoinst-distri-opensuse.git opensuse
cd opensuse/products/opensuse
git clone [email protected]:os-autoinst/os-autoinst-needles-opensuse.git needles

Example for openQA’s self-tests ("openQA-in-openQA" test):

mkdir distri && cd distri
git clone [email protected]:os-autoinst/os-autoinst-distri-openQA.git openqa
cd openqa
git clone [email protected]:os-autoinst/os-autoinst-needles-openQA.git needles

Then create a working directory for the test execution, e.g.:

mkdir /tmp/os-autoinst-run && cd /tmp/os-autoinst-run

Create a minimal vars.json config file within that directory, e.g.:

vars.json

{
   "ARCH" : "x86_64",
   "BACKEND" : "qemu",
   "CASEDIR" : "/path/to/os-autoinst-distri-opensuse",
   "DESKTOP" : "gnome",
   "DISTRI" : "opensuse",
   "ISO" : "/path/to/openSUSE-Tumbleweed-DVD-x86_64-Snapshot20160715-Media.iso",
   "PRODUCTDIR" : "/path/to/os-autoinst-distri-opensuse/products/opensuse",
   "VNC" : 90,
}

You will need to correct the file paths to point to real locations. Some of the variables you can use are listed here. Test case specific variables are listed in the distri directories e.g. os-autoinst-distri-opensuse/variables.

Then you can run the isotovideo script within the created working directory. When doing a manual build, that script can be found at the top-level of the os-autoinst Git checkout.

All of these examples were using the QEMU backend which is usually the easiest backend to handle and therefore recommended. If you need to develop and test other backends, have a look at the backend-specific documentation.

When using the QEMU backend it is possible to access the system under test via VNC:

vncviewer localhost:91 -ViewOnly -Shared

Run isotovideo with the environment variable RUN_VNCVIEWER set to autostart a VNC viewer on the right port.

Run isotovideo with the environment variable RUN_DEBUGVIEWER to start the internal debug screenshot viewer updated with an always recent screenshot of the test run.

Develop test modules

Individual test modules are written with one test module per file using the test API in Perl code. Experimental support for test modules in the Python programming language is provided.

Find more details about how to write tests on http://open.qa/docs/#_how_to_write_tests

Verifying a runtime environment

To check if your hardware is able to successfully execute os-autoinst based tests one can execute openQA tests, all the development tests or simply call something like

podman run --pull=always --rm -it --entrypoint '' registry.opensuse.org/devel/openqa/containers/os-autoinst_dev:latest /bin/sh -c 'git -C /opt clone --depth 1 https://github.com/os-autoinst/os-autoinst && make -C /opt/os-autoinst/ test-perl-testsuite TESTS=t/99-full-stack.t'

which only requires the container runtime environment "podman" and will run a container based os-autoinst full-stack test, here without KVM hardware accelerated virtualization support.

How to contribute

If you want to contribute to this project, please clone and send pull requests via https://github.com/os-autoinst/os-autoinst.

More information on the contribution can be found on http://os-autoinst.github.io/openQA/contact/, too.

Issues are tracked on https://progress.opensuse.org/projects/openqav3/.

For an overview of the architecture, see doc/architecture.md.

Rules for commits

• Every commit is checked by our CI system as soon as you create a pull request but you should run the os-autoinst tests locally. Checkout the build instructions for further details.

• For git commit messages use the rules stated on How to Write a Git Commit Message as a reference

• Every pull request is reviewed in a peer review to give feedback on possible implications and how we can help each other to improve

If this is too much hassle for you feel free to provide incomplete pull requests for consideration or create an issue with a code change proposal.

Deprecation approach

In case you want to deprecate functionality consider the use of the function backend::baseclass::handle_deprecate_backend.

Build instructions

Installing dependencies

On openSUSE one can install the package os-autoinst-devel which provides all the dependencies to build and run os-autoinst for the corresponding version of the sources. To build a current version of os-autoinst it is recommended to install os-autoinst-devel from devel:openQA as the distribution-provided packages might be too old or miss dependencies. This is particularly true for openSUSE Leap. Also see the openQA docs.

The required dependencies are also declared in dependencies.yaml. (The names listed within that file are specific to openSUSE but can be easily transferred to other distributions.)

Conducting the build

Simply call

make

in the top folder which automatically creates a build directory and builds the complete project.

Call

make help

to list all available targets.

The above commands use a convenience Makefile calling cmake. For packaging, when using an IDE or to conduct the steps manually it is suggested to use CMake directly and do the following: Create a build directory outside of the source directory. The following commands need to be invoked within that directory.

Configure build:

cmake $path_to_os_autoinst_checkout

You can specify any of the standard CMake variables, e.g. -DCMAKE_BUILD_TYPE=Debug and -DCMAKE_INSTALL_PREFIX=/custom/install/prefix.

The following examples assume that GNU Make is used. It is possible to generate for a different build tool by adding e.g. -G Ninja to the CMake arguments.

Build executables and libraries:

make symlinks

This target also creates symlinks of the built executables and libraries within the source directory so isotovideo can find them.

Run all tests:

make check

By default CTest is invoked in verbose mode because prove already provides condensed output. Add -DVERBOSE_CTEST=OFF to the CMake arguments to avoid that.

Run all Perl tests (*.t files found within the t and xt directories):

make test-perl-testsuite

Run individual tests by specifying them explicitly:

make test-perl-testsuite TESTS="t/15-logging.t t/28-signalblocker.t"

Notice that the user needs to include the test directory for each test (either t for normal or xt for developer-centric tests) when specifying individual tests.

Add additional arguments to the prove invocation, e.g. enable verbose output:

make test-perl-testsuite PROVE_ARGS=-v

Gather coverage data while running tests:

make test-perl-testsuite WITH_COVER_OPTIONS=1

Generate a coverage report from the gathered coverage data:

make coverage

If no coverage data has been gathered so far the coverage target will invoke the testsuite automatically.

Reset gathered coverage data:

make coverage-reset

Install files for packaging:

make install DESTDIR=…

Further notes:

• When using the test-perl-testsuite target, ctest is not used (and therefore ctest specific tweaks have no effect).

• One can always run Perl tests manually via prove after the build has been conducted with make symlinks. Note that some tests need to be invoked within the t directory. An invocation like prove -vI.. -I../external/os-autoinst-common/lib 28-signalblocker.t is supposed to work.

• It is also possible to run ctest within the build directory directly instead of using the mentioned targets.

• All mentioned variables to influence the test execution (TESTS, WITH_COVER_OPTIONS, …) can be combined and can also be used with the coverage target.

Standalone tool for image search

The script imgsearch in the repository’s top-level allows to use the fuzzy image comparison independently of the normal test execution. Invoke the script with no parameters to show its usage. There is also an example file showing what output you can expect. There is one key for each file to be searched. The best matching image to be found will show up as match and the other images under candidates. If no image matches well enough, match will be null.

To use the script the previously shown build instructions need to be executed (including the invocation of the symlinks target).

History of os-autoinst

At a time Bernhard M. Wiedemann who later joined was on the openSUSE testing team and was assigned the task of testing the installer. Which meant tedious and dull work of waiting for 4GB ISO files to download when it’s not even clear if those things even boot. And as the Perl founder Larry Wall states, important traits of programmers are lazyness, impatience and hybris. Which quickly led to developing os-autoinst to automate installations ;) See https://lizards.opensuse.org/2010/04/29/making-of-the-opensuse-install-video/ and https://lizards.opensuse.org/2010/05/25/automated-opensuse-testing/ for Bernhard’s blog posts.

Further notes

When using the QEMU backend, also ensure your user running os-autoinst has access to /dev/kvm.

modprobe kvm-intel || modprobe kvm-amd
chgrp kvm /dev/kvm ; chmod g+rw /dev/kvm # maybe redundant
# optionally use a new user; just to keep things separate
useradd -m USERNAME -G kvm
passwd USERNAME # and/or add ~USERNAME/.ssh/authorized_keys