Revision as of 13:22, 10 July 2009

This page describes the Akonadi test and benchmark facilities.

Tip

Further Reading: Akonadi Development Tools.

Akonadi Testrunner

Igor's GSoC project, found in kdepimlibs/akonadi/tests/testrunner. The Akonadi Testrunner sets up an isolated Akonadi server (which implies a separated D-Bus server) based on an environment configuration file.

Creating Testrunner Environments

A testrunner environment consists of two components: a set of configuration and data files and a XML description file of the environment.

Here is an example listing based on the environment used for the libakonadi unittests:

unittestenv/ unittestenv/config.xml unittestenv/xdglocal unittestenv/kdehome unittestenv/kdehome/share unittestenv/kdehome/share/config unittestenv/kdehome/share/config/akonadi-firstrunrc unittestenv/kdehome/share/config/akonadi_knut_resource_0rc unittestenv/kdehome/testdata.xml unittestenv/xdgconfig

The environment description file (config.xml)looks like this:

<config>

 <kdehome>kdehome</kdehome>
 <confighome>xdgconfig</confighome>
 <datahome>xdgdata</datahome>
 <agent synchronize="true">akonadi_knut_resource</agent>

</config>

The first three elements define the relevant paths inside the environment data, relative to the config.xml file. The <agent> element can be used to create instances of the specified agent (multiple such elements are allowed). If the agent is a resource, it can also be synced initially by adding the synchronize="true" attribute. Tests will not be launched before the syncing has been complete in this case.

Agents set up in this way can be configured by simply providing the corresponding configuration file in $KDEHOME, such as akonadi_knut_resource_0rc in our example.

Global configuration files can be provided in the same way, akonadi-firstrunrc as shown below is in particular useful to avoid the Akonadi default setup mechanism interfering with the test:

[ProcessedDefaults] defaultaddressbook=done defaultcalendar=done

Same for kdedrc which allows to disable kbuilsycoca4. That can greatly speed up tests.

[General] CheckSycoca=false CheckFileStamps=false

Using the Testrunner

Interactive Use

For manual usage, the testrunner provides an interactive mode in which it sets up the environment and provides a way to "switch" into it.

First, start the testrunner:

$ akonaditest -c config.xml &

Once the setup is complete, it creates a shell script containing the necessary environment variable changes to switch into the test environment:

$ source testenvironment.sh

The environment variables of the current shell are then changed to point to the test environment (eg. KDEHOME, DBUS_*, etc.). Every Akonadi application run in that shell operates on the Akonadi server of the test environment.

To terminate and cleanup the test environment, run:

$ shutdown-testenvironment

Note that your shell afterwards still points to the (now no longer existing) test environment and might not work as expected anymore.

Non-Interactive Use

kdepimlibs/akonadi/tests uses the Akonadi Testrunner to run unittests in an isolated environment. For automated usage, the testrunner can be used in a non-interactive way:

$ akonaditest -c config.xml <comand> <params>

The testrunner will run command params within the isolated environment and terminate afterwards.

This can be used from within CMake (example based on kdepimlibs/akonadi/tests):

macro(add_akonadi_isolated_test _source)

 set(_test ${_source})
 get_filename_component(_name ${_source} NAME_WE)
 kde4_add_executable(${_name} TEST ${_test})
 target_link_libraries(${_name}
   akonadi-kde akonadi-kmime ${QT_QTTEST_LIBRARY}
   ${QT_QTGUI_LIBRARY} ${KDE4_KDECORE_LIBS}
 )

 # based on kde4_add_unit_test
 if (WIN32)
   get_target_property( _loc ${_name} LOCATION )
   set(_executable ${_loc}.bat)
 else (WIN32)
   set(_executable ${EXECUTABLE_OUTPUT_PATH}/${_name})
 endif (WIN32)
 if (UNIX)
   set(_executable ${_executable}.shell)
 endif (UNIX)

 find_program(_testrunner akonaditest)

 add_test( libakonadi-${_name} 
   ${_testrunner} -c
   ${CMAKE_CURRENT_SOURCE_DIR}/unittestenv/config.xml
   ${_executable} 
 )

endmacro(add_akonadi_isolated_test)

Using QtTest unittests with KDE extensions (QTEST_KDEMAIN) together with the testrunner is problematic as they modify some of the environment variables set by the testrunner. Instead, use the following:

include <qtest_akonadi.h>

QTEST_AKONADIMAIN( MyTest, NoGUI )

KNUT Test Data Resource

In kdepim/akonadi/resources, fully featured resource that operates on a single XML file. File format is decribed in knut.xsd and follows closely the internal structure of Akonadi. New files can be created in eg. Akonadiconsole by creating a resource and specifying a non-existing file.

Akonadi Benchmarker

In kdepimlibs/akonadi/test, part of Robert's thesis. It is a set of test that show the time to process many item/collection operations.

Warning

This section needs improvements: Please help us to

cleanup confusing sections and fix sections which contain a todo

Unittests

Akonadi Server

Usable without installation, run with ctest/make test as usual.

kdepimlbs/akonadi

These tests use the Akonadi Testrunner, the test environment is found in kdepimlibs/akonadi/tests/unittestenv.

Setup

The tests do not yet completely work without having certain components installed, namely:

Akonadi Server
KNUT resource

Running the tests

The tests can be run automatically using ctest/make test as usual. To run a single test manually, it needs to be executed using the Akonadi testrunner:

$ cd kdepimlibs/akonadi/test $ akonaditest -c unittest/config.xml $BUILDDIR/test_executable.shell

kdepim/akonadi

Are there any?

Warning

This section needs improvements: Please help us to

cleanup confusing sections and fix sections which contain a todo

Resource Testing

Tools to automatically test Akonadi resources are currently in development in playground/pim/akonaditest/resourcetester. There are two basic modes of operation, read tests and write tests. The resourcetester tool provides convenience methods for common operations needed to perform those tests.

Read Tests

To verify the read code in a resource works correctly we need to read pre-defined test data from the resource and compare that with independently provided reference data.

Write Tests

Once the reading code is verified we can use that to verify the writing code. This is done by writing a change to the resource, re-creating it to ensure the change was persistent and finally comparing the re-read change with the expected result.

Warning

This section needs improvements: Please help us to

cleanup confusing sections and fix sections which contain a todo