We present the TSW system (TestSystem Web), a web-based environment currently developed at the Rome 1 University, for the delivery of C programming exercises and their automatic correction.
The core of the correction system automatically tests the student’s programs by applying unit-tests and/or by comparing the behaviour of the student’s code to a reference implementation. Care is taken to avoid error propagation from a function to other functionally depending procedures by redirecting the failing calls to the corresponding reference implementation. The system “instruments” the student’s code by using a code analyser and rewriter to handle instruction tracing and function calls redirection. The rewriter can be easily extended to develop other analysis instruments. As an example, we have developed: a
code coverage tool
that reports how much of the student’s code has been visited during the test, a
cyclomatic complexity evaluator
to compare the number of different logic paths in the code, a
tracker for stack depth usage
to check for proper implementation of recursive functions, a
function/loop execution counter
for the evaluation of the execution complexity. Additional care is taken to capture disruptive errors that would abort the program: “segmentation faults” caused by wrong pointer dereferentiation, and time-out caused by run-away processes.
With these tools, the teacher can write rich unit tests that can either compare the behaviour of the function under analysis with a reference implementation (e.g. by generating random input and comparing the results), or by submitting well-crafted special inputs to elicit special cases or by comparing the complexity and/or stack depth counters. Each test applied will then explain to the student what was the problem found.
TSW is the core component of a future larger social knowledge project, in which students will cooperatively/competitively participate to the definition and test of each-other’s programs, sharing ideas and learning from each other.