Documentation
Assuring FORTE Code Quality
Reducing the number of bugs and keep a high code quality is a constant struggle. For FORTE we apply several means and tools which we summarize here as we think they can be helpful for users planing to extend FORTE.
Static Code Analysis
Static code analyzers as the name implies perform an off-line analysis of your code and try to identify potential issues. For FORTE we are currently using the following two open source static code analyzers:
- CPP Check: is a tool you can directly use to check your code. However for Eclipse there also exists an Eclipse plugin you can install from the Eclipse market place. Which makes its use much more convenient. CPP Check helps for example to find if correct pairs of new/delete are used or also on potential memory leaks.
- Clang's scan-build: scan-build is a tool that comes with the C++ Compiler clang. It uses the clang compiler to perform the static analysis during compile time.
To use it for checking FORTE you first need to change the C to clang's ccc-analyzer and the C++ compiler to clang's c++-analyzer during the CMake configuration of FORTE. Therefore it is recommended to use a separate build directory.
Furthermore instead of running
make allyou need to invoke the comman.d
scan-build -o ./static makefor building FORTE. The nice thing about scan-build is that it creates nice html files with explanations about the found problems.
Dynamic Analysis
Dynamic analysis tools help to find issues during the runtime. Typical issues are memory issues like memory leaks, out of bound access, stack over/underflows, or the use after memory is freed.
- Valgrind: Valgrind is definitely the most commonly known dynamic checker. It checks mainly for memory issues. Currently it is only available for Linux and there only for Intel platforms. The main drawback of Valgrind is that it tremendously slows down the execution of the application under test.
- Sanitizers: Sanitizers are means for runtime checking that the complier adds to your application. They can be activated with compiler switches and have the great advantage that they are currently much faster than Valgrind. However they are only available for Clang and gcc. For the latter only partially. Currently we have worked with the following sanitizers
- Address: Checks for memory and heap issues. Can be activated by adding
-fsanitize=addressto the CMake option CMAKE_CXX_FLAGS.
- Undefined Behavior: Checks for undefined behavior in your code (e.g, overflows, div by zero). Can be activated by adding
-fsanitize=undefinedto the CMake option CMAKE_CXX_FLAGS.
- Thread: To be tested
- Address: Checks for memory and heap issues. Can be activated by adding
Unit Tests
We applied the unit test framework provided by the Boost-library called Boost Test. You can find the current set of available unit test in the directory test. New unit tests are always more then welcome.
Install Boost Test
To use execute unit tests at least boost 1.59 is needed. Download it and unzip it.under Linux
under Windows
For detailed instructions on installation under Windows have a look at this or this Blog. Otherwise follow this summary for cygwin:- open a cygwin command line and go to your unziped folder
- Start bootstrap.bat and specify your toolset.
Supported toolsets are: borland, como, gcc, gcc-nocygwin, intel-win32, metrowerks, mingw, msvc, vc7, vc8, vc9, vc10, vc11.
Use the following command for cygwin
./bootstrap.sh gcc
- define an installation directory and specify a toolset.
Toolsets here are a little bit different from the ones we used before.
The following command creates two folders, include and lib.
Both folders should contain files afterwards, which takes a lot of time.
./b2 install --prefix=c:/installation/path toolset=gcc
Set CMake of the 4diac-rte
Set the following configuration in CMake:The make target "all" generates a fortetest executable which executes all Tests. In case you want code coverage results also check FORTE_TEST_CODE_COVERAGE_ANALYSIS and create a make target in eclipse called TestCoverage the make target TestCoverage generates a folder with html files containing the coverage results. In case an error occures, uncomment
Function Block Tests
Based on Boost Test we developed a helper class which allows you to write unit test for Function Blocks in C++. For examples how to use it have a look on the test cases in the test/stdfblib directory. Write positive and negative test cases for your unit tests.
Where to go from here?
Go back to Development index:
If you want to go back to the Start Here page, we leave you here a fast access
Or Go to top