Overview of Code Coverage and Differential Coverage Methods Using JaCoCo

Code Coverage Overview – Code coverage is defined as the number of executed code elements divided by the total code elements of the target under test. It is triggered by functional and interface testing and can be measured at various dimensions such as function, statement, branch, condition, and path coverage.

Measuring Test Coverage – There are two main approaches: requirement‑based coverage, which evaluates test case execution against requirement test points, and code‑based coverage, which assesses execution at the code level.

Limitations of Full‑Code Coverage – Redundant, duplicate, or abandoned code inflates coverage numbers and makes precise assessment difficult.

Advantages of Differential Coverage – Provides precise assessment for small changes, monitors developer self‑testing, and allows setting thresholds as pass/fail criteria.

Significance of Code Coverage – Analyzing uncovered code helps verify test case completeness, clarify requirement understanding, and detect dead code, thereby improving overall code quality.

Statistical Method for Differential Coverage

Based on the Beetle management platform and test‑server agents.

Utilizes a code‑diff feature to present both full and differential coverage on a web page.

Core principle: count executed code lines and collect execution trace information.

JaCoCo Integration – JaCoCo (Java Code Coverage) uses bytecode instrumentation with two modes: On‑The‑Fly (runtime agent, no source changes) and Offline (pre‑instrumented classes, no agent required).

On‑The‑Fly workflow: the JVM is started with -javaagent:jacocoagent.jar, the agent inserts a boolean probe array into each class before loading, and probes are set to true when bytecode instructions execute.

Full Coverage vs. Differential Coverage

Full coverage reflects the current deployed version only, using JaCoCo’s complete report with multiple metrics.

Differential coverage focuses on lines changed in the current development branch, calculated as covered diff lines divided by total diff lines, and can accumulate across versions.

JaCoCo Report Interpretation

Provides multi‑dimensional counters such as instruction (C0), branch (C1), cyclomatic complexity, line, method, and class coverage.

Branch coverage is visualized with red (none), yellow (partial), and green (full) diamonds; line coverage uses red, yellow, and green backgrounds.

Conflict Resolution with SCF Service Code

Instrumentation adds a static variable $jacocoData and method $jacocoInit() to classes, which can cause errors in entity classes lacking getters/setters and in code that does not recognize the added variable.

Solution: filter out affected classes from instrumentation, modify JaCoCo source to apply matching rules (class annotations like SCFSerializable, Table, Data; method signatures; class name patterns) and manage these rules via a dynamic configuration file pushed to test servers.