Java Jacoco Coverage Backfill

Java JaCoCo Coverage Backfill (Universal)

When to use

• User asks to add/fix Java tests.
• User asks to improve coverage.
• User asks for branch-gap based supplementary tests.
• Project is Maven or Gradle based.

Goals

• Use JaCoCo XML as the coverage source of truth.
• Track four coverage metrics consistently: CLASS, METHOD, LINE, BRANCH.
• Prioritize classes with worst BRANCH coverage for test backfill.
• Raise branch quality without causing regression on other coverage dimensions.

Coverage Model

• CLASS: covered classes / total classes
• METHOD: covered methods / total methods
• LINE: covered lines / total lines
• BRANCH: covered branches / total branches

• Primary backfill target metric: BRANCH
• Secondary guardrail metrics: CLASS, METHOD, LINE

Inputs And Defaults

• Class selection: 10 lowest BRANCH-coverage classes with branch counters.
• Default threshold: BRANCH >= 80% for each selected class.
• Secondary constraint: no unexplained regression in CLASS/METHOD/LINE for selected classes.
• If user provides different thresholds/scope/metric priorities, follow user input.

Important: Where Files Go

| Location | What Goes There | | ------------------------------------------------------------ | --------------- | | Skill directory (${HOME}/.codex/skills/java-jacoco-coverage-backfill) | scripts |

Bundled Scripts (Use First)

Scripts are under scripts/ and should be preferred over ad-hoc one-off shell snippets.

1. scripts/detect_build_tool.sh

• Detect Maven/Gradle and resolve effective build root.

1. scripts/generate_jacoco_report.sh [--ignore-test-failures]

• Generate JaCoCo XML/HTML reports for Maven or Gradle.

1. scripts/find_jacoco_reports.sh

• Locate all JaCoCo XML report files in mono-repo/multi-module layouts.

1. scripts/analyze_jacoco_xml.py --xml --top 10 [--out ] [--selected ]

• Produce class ranking (BRANCH asc), module metrics, and optional snapshot JSON for diffing.

1. scripts/compare_jacoco_snapshots.py --before --after [--selected ]

• Show metric delta at module and class granularity.

Required workflow

1) Discover project layout

1. Detect build system and build root using script:

scripts/detect_build_tool.sh 

1. Detect module(s) to analyze:

• Prefer module(s) touched by the task.
• If unspecified, start with the primary module containing Java production code and tests.

2) Ensure JaCoCo is configured (mandatory)

If JaCoCo is missing, add it first.

Maven (if missing)

• Ensure org.jacoco:jacoco-maven-plugin exists in ``.
• Required executions:
• prepare-agent
• report (bound to test phase or invoked via jacoco:report)

Reference snippet:

  org.jacoco
  jacoco-maven-plugin
  0.8.12
  
    
      
        prepare-agent
      
    
    
      report
      test
      
        report
      
    
  

Gradle (if missing)

• Apply plugin: id 'jacoco'.
• Ensure XML report is enabled.
• Ensure test finalizes jacocoTestReport.

Reference snippet (Groovy DSL):

plugins {
  id 'java'
  id 'jacoco'
}

test {
  finalizedBy jacocoTestReport
}

jacocoTestReport {
  dependsOn test
  reports {
    xml.required = true
    html.required = true
  }
}

3) Generate baseline JaCoCo XML

Use script:

scripts/generate_jacoco_report.sh 

4) Locate JaCoCo XML report(s)

Find candidate XML files:

scripts/find_jacoco_reports.sh 

Selection rules:

1. Prefer module-specific report for requested scope.
2. If multiple are relevant, analyze per module.
3. If only one exists, use it.

5) Analyze coverage with jacoco_reporter_server

For each selected XML report:

1. Call jacoco_reporter_server on that file (authoritative source for missed line/branch positions).
2. Use script to compute ranked metrics and baseline snapshot:

python3 scripts/analyze_jacoco_xml.py \
  --xml  \
  --top 10 \
  --out .coverage/baseline.json

1. Rank classes primarily by BRANCH ascending:

• primary: BRANCH% ascending
• tie-breaker 1: missed branches descending
• tie-breaker 2: LINE% ascending

1. Select top 10 classes with branch counters (or fewer if not available).
2. Extract missed branch lines from jacoco_reporter_server or analyze_jacoco_xml.py output.
3. Preserve baseline snapshot for delta comparison.

6) Design targeted tests

For each selected class:

1. Map missed branch lines to concrete conditions.
2. Add tests for missing branch directions (true/false, success/error, allow/deny, found/not-found).
3. Prefer unit tests first; use integration tests only when branch behavior depends on framework/runtime boundaries.
4. Avoid tests that only increase line hits but still miss branch directions.

7) Run tests and iterate

1. Run targeted tests first.
2. Regenerate JaCoCo XML.
3. Re-run jacoco_reporter_server.
4. Generate post snapshot and compare:

python3 scripts/analyze_jacoco_xml.py \
  --xml  \
  --top 10 \
  --out .coverage/post.json

python3 scripts/compare_jacoco_snapshots.py \
  --before .coverage/baseline.json \
  --after .coverage/post.json

1. Repeat until completion criteria are met or blockers are confirmed.

Test design rules

• One test should target one explicit missed-branch intent.
• Use deterministic inputs to force both branch directions.
• Include failure paths (validation, unauthorized, conflict, not found) where relevant.
• Avoid redundant tests that hit same branch direction.
• Keep tests local, readable, and consistent with project conventions.

Multi-module rule

• Do not merge unrelated module coverage into one ranking unless user requests aggregate mode.
• Default: report baseline and post-change per module.

Completion criteria

• For selected class set(s):
• each class reaches BRANCH >= 80% (or user-defined threshold), and
• CLASS/METHOD/LINE do not regress without explanation.
• If any selected class remains below target, report blockers:
• class name
• current CLASS/METHOD/LINE/BRANCH percentages
• missed branch lines
• why not coverable now
• exact next action needed

Output format (every run)

1. Project/build detection summary (Maven/Gradle, module scope).
2. JaCoCo setup status:

• already present, or
• added/updated files and config changes.

1. Baseline table (selected classes):

• class name
• CLASS%, METHOD%, LINE%, BRANCH%
• covered/missed branch counts

1. Added/updated test files list.
2. Validation commands and pass/fail status.
3. Post-change table for same selected classes with delta columns.
4. Module-level summary before/after for CLASS/METHOD/LINE/BRANCH.
5. Remaining gaps and concrete next tests (if not complete).

Script-First Commands

scripts/detect_build_tool.sh .
scripts/generate_jacoco_report.sh .
scripts/find_jacoco_reports.sh .
python3 scripts/analyze_jacoco_xml.py --xml  --top 10 --out .coverage/baseline.json
python3 scripts/analyze_jacoco_xml.py --xml  --top 10 --out .coverage/post.json
python3 scripts/compare_jacoco_snapshots.py --before .coverage/baseline.json --after .coverage/post.json

Notes

• Do not modify production code only to increase coverage unless user explicitly requests it.
• Coverage decisions must be based on JaCoCo XML + jacoco_reporter_server output.
• If build policy blocks adding JaCoCo, report blocker and required approval explicitly.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: changheng10000
• Source: changheng10000/java-jacoco-coverage-backfill
• License: MIT
• Homepage: https://github.com/changheng10000/java-jacoco-coverage-backfill

Install and usage instructions live in the source repository linked above.