Yocto 1.5 Overall Test Plan

From Yocto Project
Jump to navigationJump to search

Reversion history

Version Modifier Comments
1.0 Alexandru Georgescu Initial Version

Introduction

This article is the overall test plan for version 1.5 of the Yocto Project. It contains an overview of the testing process, such as testing areas, types, cycles and reports, along with a summary and objective for each of the conducted validation activities. Further on, sections may be linked to other articles that contain further details or information related to them. As a note, the information provided in this article, or one that's linked here, is subject to changes at any time, as to reflect the actual activities held for the current version of the project.

Objectives

The test process is mainly focused to track and review the quality and performance of the Yocto Project, along with its reference system and internal projects. The plan also includes identifying and tracking areas subject to improvement, regression, validation of enhancements and bugs, development of testing methods with emphasis on automated testing. Documentation and licensing status is not included in the scope of the testing process, unless otherwise noted e.g. as part of the process of verifying new features.

Test Areas

Each internal project, under Yocto Project, is an area subject to a testing process. Areas are grouped by the nature of their functionality, as follows:

Build System

The build engine and the surrounding components, that provide the means to build an image or bake a bit of software. In this area the build-time tests are executed.

BitBake

Functional testing of BitBake, as a build engine, with all its features and components, against various configuration and scenarios.

HOB

Functional and usability testing of HOB as a graphical user interface for BitBake.

Metadata

Testing the core metadata of the Yocto Project is mainly covered in the overall testing process, through other #Test Areas like #BitBake and #HOB mentioned above.

Build Performance

The performance of the build system is tracked, with regards to time spent on passing through a build process, in multiple, commonly used, configurations.

Tool
http://git.yoctoproject.org/cgit/cgit.cgi/poky/tree/scripts/contrib/build-perf-test.sh
Results page
Performance Test

Target System

Area focused on a target operating system or an application that comes with it, as the output of a build process. In this area the run-time tests are executed.

QEMU Image

Covering all core, QEMU, machine definitions:

  • qemuarm
  • qemumips
  • qemuppc
  • qemux86
  • qemux86-64

BSP Image

Covering BSPs included in the Yocto Project:

  • atom-pc
  • beagleboard
  • mpc8315e-rdb
  • routerstationpro

Compliance Testing

Compliance test suites / frameworks used:

  • LSB tests
  • POSIX tests
  • LTP tests

Stress Testing

Stress tests are run on Beagleboard and Jasperforest BSPs. Details as follow:

  • Beagleboard core-image-sato-sdk image is tested using LTP and Crashme stress tests
  • Jasperforest core-image-lsb-sdk image is tested using Helltest and Crashme stress tests

System Performance

The run-time performance metrics will be measured using the following methods:

  • measure boot time for SystemD and SysVinit features
  • record image size from buildhistory to track regression
  • use Phoronix Test Suite tests to benchmark Yocto images
  • run Piglit test suite

Developer Tools

Application Development Toolkit

ADT testing includes tests for ADT-installer, meta-toolchain-sdk and user build sdk. It will be covered in Weekly and Fullpass testing.

  • Cross-toolchain install&compiling Test
  • relocatable SDK
  • ADT installer
  • toolchain tarballs
  • yocto build tree

Eclipse IDE Plugin

Eclipse plugin tests will cover the basic functionalities. This includes installation, configuring Yocto Project ADT settings, Yocto BSP, Bitbake project and project compiling and deployment to the target. Based on the features that will be implemented, new test cases will be added, to support Windows and Mac support. This will be more detailed in the Features section.

  • headless build
  • C/C++ project creation
  • debug/deploy
  • user space tools
  • Bitbake project

Build Appliance

The basic functionality of the Build appliance will be tested. The tests consists on building successfully a build-appliance-image, launch HOB.

Distribution Support

The most recent previous version of one of the 4 supported distributions Ubuntu, Fedora, CentOS, OpenSUSE, and the latest beta version of those are the target for these tests. Basic functionalities of ADT, toolchaion, HOB2 and Core Build System ar testested on these targets.

The following coverage matrix will reprezent the supported distributions:

Distribution Ubuntu Fedora CentOS OpenSUSE
Version 13.04 13.10 18 19 6.3 6.4 12.3 13.1
RC Supported Yes Yes Yes Yes
1.5 Supported Yes Yes Yes Yes Yes Yes Yes Yes

Test Cycle

Test execution cycle
#Sanity Test #Weekly Test #Full Pass Test #Release Test
Build type Daily (M.U.T.) yes
Weekly yes yes
Release Candidate yes yes yes
Release yes yes yes yes
#Test Areas #BitBake yes yes yes
#HOB yes yes
#Build Performance yes
#QEMU Image yes yes yes
#BSP Image yes yes
#Compliance Testing yes
#Stress Testing yes
#System Performance
#Application Development Toolkit yes yes
#Eclipse IDE Plugin yes yes yes
#Build Appliance yes

Note:

  • On weekly test, the basic BSP will be covered: Atom-PC, Beagleboard, Routerstationpro, MPC8315e-rdb, P1022ds-rdb, and all the QEMU images (arm, x86, x86_64, ppc, mips)
  • as mentioned in the Weekly test section, the weekly tests will cover the basic functionalities of the component.
  • Compliance Test on Beagleboard and routerstation will only cover LTP and POSIX Test, since LSB test suite does not support ARM and MIPS.
  • Jasperforest is selected for stress test, because it is powerful with 12 CPUs and 2GB memory. Moreover, it is a server, which needs more stress tests.

The following matrix represents Full pass and Weekly coverage of BSP testing.

Test levels Atom-PC NUC Emenlow Sugarbay Jasperforest Crownbay FRI2 Beagleboard RouterstationPro MPC8315e-rdb P1022ds-rdb
Weekly Yes Yes Yes Yes Yes
Full Pass Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Sanity Test

Brief and quick automated tests, with execution time of maximum 10 minutes.

  • Objective
    • Build finished with no errors;
    • Check basic QEMU image functionality, e.g. boot, network, package manager, etc.;
    • Establish if testing cycle can continue, depending on the build type.

Weekly Test

  • Scope
    • Images built weekly and released through the distribution team.
    • Passed #Sanity Test
  • Objective
    • Functionality test on most areas with minimum sets of tests;
    • Regression test with high probability to find bugs.

Full Pass Test

  • Scope
    • Images built as candidates for milestone or final release;
    • Passed #Weekly Test
  • Objective
    • Ensure functionality of all Yocto Project components.

Release Test

  • Objective
    • All scheduled features are covered, or rescheduled;
    • All relevant bugs are fixed and verified.
  • Coverage
    • Stress test on RC
    • Compliance test on RC
    • Distribution test on RC

Test Automation

  • Objectives
    • Reduce effort with manual testing, by automating current tests;
    • Improve build-time and run-time testing.

Validation

  • Objective
    • Verify the correct functionality of new changes introduced in version 1.5 of the Yocto Project.
  • Entry criteria
    • The change is tracked and prioritized in Bugzilla;
    • Bugzilla entry has a target milestone within version 1.5;
    • The change is documented or pointed out when no documentation is necessary;
    • Bug status is set to RESOLVED.
  • Exit criteria
    • The change is well documented for writing test case, where applicable;
    • Planned test case has passed;
    • Bug status is set to VERIFIED.

Test report