Developer/Tests

Introduction#

We use TextTest to test whether the software still behaves as expected. TextTest compares an application's file output, including the output to stdout and stderr with predefined outputs from files.

There are some advantages and disadvantages of this approach. On the one hand, you can guarantee that the application is doing what is wanted by comparing all outputs with files you think are right. But this is also the problem with this approach: you do not have the assurance that the files you are comparing the current outputs to are right - you have to prepare them once and update/verify them again every time the expected results change.

At the moment all our [Developer/Nightly_Build tests run each night] with the results on this summary webpage.

Setup#

We use TextTest as our testing environment which is Python based and currently available for Python 2 only. To install it, you can open a terminal / Windows console and type:

pip install texttest

on all platforms, but it will not pull in the dependencies, so you need to follow the TextTest installation instructions and the additional info here.

Windows Setup#

You will need the following software (at least for a setup with GUI)

Visual guide:

Python path shouldn't include whitespaces

Python requieres a full installation

PyGTK must be installed in the same folder of Python

TkDiff path shouldn't include whitespaces

Check that python works typing "Python" in CMD.

Type "quit()" to exit Type pip install texttest to install TextTest automatically from Python repository

Every value of environment variable "PATH" must be separated with ";"

If Installation was successfully, this window will appear after execution of "runAllTests.bat"

Linux setup#

This is much easier because at least with recent distributions there is probably everything included. With openSUSE 10.2 you need python-gtk and python-cairo as well as tkdiff before installing TextTest. If you don't need the GUI you can probably even skip these.

MacOS setup#

  1. Follow the steps here
  2. If you have multiple python versions (2.x/3.x) installed (see here), make sure that your pip points to version 2.x
  3. Install TextTest via pip: pip install texttest

Customize configuration#

Common options added to the personal config file $HOME/.texttest/config are the used diff-viewer, editor, and the flag to collapse the static test suites on program start. E.g.:

   diff_program:meld
   view_program:geany
   static_collapse_suites:1

Running the Tests#

Within the <SUMO_HOME>/tests - folder you can find batch-files which start TextTest with our test suites. "runAllTests.bat" starts TextTest for testing all applications located in the folder, "runNetconvertTests.bat" will only show tests for NETCONVERT, "runDuarouterTests.bat" only those for DUAROUTER etc.

Adding a Test to an existing Application#

The easiest method is probably to copy an existing test using the TextTest GUI context menu functions (copy/paste). Then you can modify the copied input files, run the test once and save the expected results. Note, that input files may be picked up along the folder hierarchy so you may have to copy additional files to your test folder manually if these inherited input files must be changed as well.

Adding an Application Test Suite#

To add a test suite for a new application, you have to perform the following steps. For the examples below we'll use "polyconvert" as the example application.

  • go to <SUMO_HOME>/tests
  • copy one of the run...Tests.bat-files and rename it properly (runPolyconvertTests.bat in our case); change the name of the application within it. In our case the resulting file will look as this:
call %~dp0\testEnv.bat %1
start %TEXTTESTPY% -a polyconvert
  • add the application to the list of applications that are tested each night by
    • adding it to runTests.sh; in our case, the following line was added:
export POLYCONVERT_BINARY="$SUMO_BINDIR/polyconvert"
  • adding it to testEnv.bat; in our case, the following line was added:
set POLYCONVERT_BINARY=%~dp0\..\bin\polyconvert%1.exe
  • build a test folder for the application, named as the application itself (without the ".exe" extension), in our case the folder is named polyconvert
  • go the folder
  • build a configuration file; its name is "config", the extension is the application's to test name, so in our case it's config.polyconvert. Please consult TextTest documentation about the content, nonetheless, here are some notes
    • do not forget the import of the main config file (config_all)
    • name the binary correct
    • name the file name properly in output
    • In all files that are collated, Version should be ignored - the tests should be working along all versions

The initial file looks as following:

import_config_file:../config_all
binary:$POLYCONVERT_BINARY
copy_test_path:input_net.net.xml
[collate_file]
config:config.cfg
log:log.txt
[run_dependent_text]
output:polyconvert.exe{REPLACE polyconvert}
net:Version
  • build a top-level testsuite file; its name is "testsuite", the extension is the application's to test name, so in our case it's testsuite.polyconvert
    • I suppose, it is a good idea to start with tests of meta-output; Simply copy them from another application and patch the file names...

Example of use#

The main function of existing test is to verify that a modification of a function does not affect the final output, comparing the output of our modification in SUMO with a with an previously tested output control file. Note that SUMO must be compiled in release mode for testing. To show an example of use, the MSDevice_Battery.cpp file will be modified to include an error in the output file battery.xml. Once modified the line, run executable runSumoTests.bat and find the test sumo/extended/electric/braunschweig, and then right click to run the test

Manual error

Running test

After execution can be observed that the battery.sumo file is marked in red. That means that the execution of our SUMO modification generated a different battery output than expected. If the file marked in red was errors.sumo, it means that our modification of SUMO has caused a runtime error, and finally if the file marked in red is output.sumo, it means that SUMO generated some warning during execution. Double-clicking on the red box battery.sumo opens ''TkDiffv automatically, and displays the differences of our battery output with respect to the original battery output control file.

Error in battery.sumo

Detail of the error comparing both battery.sumo

At this point, we undone the modification in MSDevice_battery.cpp which caused the error, and compile. Once compiled, we will again right click over the failed text and press rerun. A new dialog will be opened, and then click over Accept, and test will be repeated. After the new test run no error is displayed.

Undo error

Repeating test

After clicking on the check box next to the green message Succeeded can be observed that all output shown in green, and at the bottom appear a message indicating that the test has been succeeded.

Test succeeded

Details of the test succeeded

Organization of Tests#

Our TextTest tests can be grouped into the categories described below

Application Tests#

These tests are for the executable files in the bin folder such as NETCONVERT, DUAROUTER and SUMO. They work by running their application once with a predefined set of input files and compare against expected output files. The tests for sumo run once with the GUI version and once without GUI..

Tool Tests#

These tests are for the python tools in the tools folder such as randomTrips.py and flowRouter.py. Instead of defining a new application for each tool (as is the standard for TextTest) they all call the tests/toolrunner.py script and select the actual python tool to test via options. Be aware that the script to use is always listed last in the options file (this is due to TextTest trying to reorder options cleverly).

Complex Tests#

These tests go beyond calling an application once. Each tests defines a runner.py script which performs arbitrary application calls. This is used to call netconvert multiple times to perform back-and-forth format conversions or to test tutorials where first the network is generated and afterwards simulation is performed. They are also used for testing the traci python client (runner.py is a traci script in this case).

Interactive GUI Tests#

These tests are currently only active for NETEDIT and are described in more detail on the page Developer/GUI_Testing.

Extra Tests#

These tests run one or multiple of the test suites above with additional options or a different python interpreter and are usually not run using a batch file but the runExtraTests.py. To enable python3 tests you need to have an executable named python3 in your PATH (which is usually there on Linux). On Windows you will need to add it by copying the existing python.exe in your python3 installation to python3.exe (in the same directory) and adding that directory to the end of your PATH (if it is not already there).