matlab-test-execution（matlab 测试执行）

MATLAB Test Execution

Must-Follow Rules

• Execute MATLAB via MCP — If the MATLAB MCP core server is available, use its evaluate_matlab_code tool to run MATLAB commands. Fall back to matlab -batch only if the MCP server is not available.
• Use buildtool for CI — When running tests in CI/CD, always use buildtool with a buildfile.m. Do not use manual runtests or TestRunner scripts in CI.
• Do not guess — If the test folder location, source folder, or desired output format is unclear, ask.

Running Tests

results = runtests('tests');                          % All tests in a directory
results = runtests('myFunctionTest');                  % Specific test file
results = runtests('myFunctionTest/testAddition');     % Specific test method

Filtering

results = runtests('tests', 'Name', '*Calculator*');   % By name pattern
results = runtests('tests', 'Tag', 'Unit');            % By tag
results = runtests('tests', 'ExcludeTag', 'Slow');     % Exclude tag
results = runtests('tests', 'ProcedureName', 'testX'); % By procedure name
results = runtests('tests', 'UseParallel', true);      % Parallel execution
results = runtests('tests', 'Strict', true);            % Treat warnings as failures
results = runtests('tests', 'Debug', true);             % Enter debugger on failure
results = runtests('tests', 'OutputDetail', 'Detailed');% Verbose diagnostics

Parallel requirements: Tests must be independent — no shared state, no order dependence, no shared file system artifacts.

For advanced control (custom plugins, programmatic suite manipulation), use matlab.unittest.TestRunner with testsuite and runner.run(suite). See the Code Coverage section for an example.

Analyzing Test Results

results = runtests('tests');
disp(results)

for r = results([results.Failed])
    fprintf('\nFAILED: %s\n', r.Name);
    disp(r.Details.DiagnosticRecord.Report);
end

for r = results([results.Incomplete])
    fprintf('\nINCOMPLETE: %s\n', r.Name);
    disp(r.Details.DiagnosticRecord.Report);
end

Code Coverage

If code coverage is required (explicitly requested or clearly implied), follow these steps: Collect and display → Identify gaps (only when user asks to generate tests for missing coverage).

1. Collect and Display

Run tests with coverage. Include CoverageResult (programmatic) and CoverageReport (HTML). Add CoberturaFormat for CI. Use the highest MetricLevel available — "mcdc" if "MATLAB Test" is installed, otherwise omit it.

import matlab.unittest.TestRunner
import matlab.unittest.plugins.CodeCoveragePlugin
import matlab.unittest.plugins.codecoverage.CoverageResult
import matlab.unittest.plugins.codecoverage.CoverageReport

runner = TestRunner.withTextOutput;
covFormat = CoverageResult;
runner.addPlugin(CodeCoveragePlugin.forFolder('src', ... % use forFile or forNamespace if needed
    'Producing', [covFormat, CoverageReport('coverage-report')], ...
    'MetricLevel', 'mcdc'));          % omit MetricLevel if MATLAB Test is unavailable
results = runner.run(testsuite('tests'));

covResults = covFormat.Result;
disp(covResults)              % aggregated summary
for i = 1:numel(covResults)   % per-file breakdown
    disp(covResults(i))
end

2. Identify gaps — only when generating tests for missing coverage

Read and run scripts/printCoverageGaps.m. It expects covResults from step 1 and prints uncovered items. Include tiers up to the MetricLevel used (the script has comments marking where to cut).

3. Act on gaps if requested by the user

Use the uncovered items from step 2 to target test generation. Defer to the MATLAB test generation skill for writing tests.

CI/CD Integration

Always use buildtool with a buildfile.m for CI.

Example buildfile.m

function plan = buildfile
    plan = buildplan(localfunctions);

    plan("clean") = matlab.buildtool.tasks.CleanTask;

    plan("check") = matlab.buildtool.tasks.CodeIssuesTask("src");

    plan("test") = matlab.buildtool.tasks.TestTask("tests", ...
        SourceFiles = "src", ...
        ReportFormat = ["html", "cobertura"], ...
        OutputDirectory = "reports");

    plan("package") = matlab.buildtool.tasks.PackageTask("toolbox.prj");

    plan("ci") = matlab.buildtool.Task( ...
        Description = "Full CI pipeline", ...
        Dependencies = ["check", "test", "package"]);

    plan.DefaultTasks = "test";
end

Running buildtool

buildtool              % Run default task
buildtool test         % Run specific task

CI Pipeline Configurations

GitHub Actions

# .github/workflows/matlab.yml
name: MATLAB Tests
on: [push, pull_request]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: matlab-actions/setup-matlab@v2
      - uses: matlab-actions/run-build@v2

Azure DevOps

# azure-pipelines.yml
trigger: [main]
pool:
  vmImage: 'ubuntu-latest'
steps:
  - task: InstallMATLAB@1
  - task: RunMATLABBuild@1

GitLab CI example

# .gitlab-ci.yml
test:
  image: mathworks/matlab:r2024a
  script:
    - matlab -batch "buildtool"