Tests, coverage & benchmarks

The Go stack runs tests before it builds. A suite that fails should not be followed by minutes of compiling and packaging something nobody will ship.

Detection

test is auto by default: run them if the module has any *_test.go, and say so if it does not.

test
auto	run when tests exist; emit a notice and continue when they do not
true	require tests — a module that should have them and does not fails
false	skip

The middle setting is the useful one for a repository that has decided tests are not optional. A silent skip and a pass look identical in a green tick; true makes the difference visible.

Coverage

Off by default, because a profile costs time and most builds do not read it.

- uses: dAppCore/build@v4
  with:
    build-name: mytool
    coverage: true
    coverage-threshold: 70

coverage-threshold fails the job below that percentage. 0, the default, measures without judging. The percentage is exposed as the COVERAGE output either way.

Codecov

- uses: dAppCore/build@v4
  with:
    build-name: mytool
    codecov: true
    codecov-token: ${{ secrets.CODECOV_TOKEN }}

codecov: true implies coverage: true — there is nothing to upload otherwise. Public repositories can upload without a token; private ones need one. An upload failure does not fail the build, because a reporting outage is not a defect in your code.

Benchmarks

- uses: dAppCore/build@v4
  with:
    build-name: mytool
    benchmarks: true

Runs with -run='^$' so no test executes twice — benchmarks are a separate reading, and rerunning the suite would add noise and time. Results land in the job summary and in benchmarks.txt.

Note what this does not do: compare against a previous run, or fail on a regression. Runner hardware varies between jobs enough to make that a source of false alarms rather than a signal. The numbers are for a human to read.

The race detector

- uses: dAppCore/build@v4
  with:
    build-name: mytool
    race: true