Beautiful Testing

What Have We Learned So Far?

We know how to generate caller–callee dependencies among the functions in the source code of the executable using dynamic binary analysis, with the help of the Valgrind tool suite.

We also have identified the test case to source files mapping from the Callgrind tool. To do this, we need a debug or optimized build where build symbols are preserved for generating useful call graphs.

We have identified the source, sync, and pipe functions in the source code using the KCachegrind tool.

We have generated an instrumented executable using gcov.

We also have generated the extent of cumulative code coverage after executing all tests at once on the executable to identify covered areas of code and code gaps.

From the source code control system, we have identified the list of changed files in the last 24 hours, as well as details of the changed code.

Armed with this data, we can generate an effective change-centric test suite that has the highest probability of testing the changes in the code base.

Example 4

Source files file1.cpp, file2.cpp, and file3.cpp are modified in the last 24 hours, and the changes are checked into the source code control system.

The batch file/cron job, which runs on a daily basis, runs a diff command on the source tree with respect to the revision that was used yesterday, and would generate a text report that not only shows the list of source files that are changed, but also the lines of source that have changes.

From the code coverage data, we ...