Proposal Details

I propose a new helper for running x/tools/go/analysis/passes over the files in a module, and reporting the results as a test pass/fail.

// Test runs the given analyzers against the given package patterns.
// Each package matching the pattern and analyzer is reported as a sub test,
// in the format "<analyzer.Name>/<package path>".
// It fails the subtest if there are any findings.
func Test(t *testing.T, analyzers []*analysis.Analyzer, patterns []string)

Example usage:

import (
    "testing"

    "golang.org/x/tools/go/analysis"
    "golang.org/x/tools/go/analysis/passes/appends"
    "golang.org/x/tools/go/analysis/passes/asmdecl"
    "golang.org/x/tools/go/analysis/testchecker"
)

func TestLint(t *testint.T) {
    analyzers := []*analysis.Analyzer{
        appends.Analyzer,
        asmdecl.Analyzer,
    }
    testchecker.Test(t, analyzers, "./...")
}

Go has a rich ecosystem of static analysis, much of which is built on top of x/tools/go/analysis.
While these can aide developers while authoring code, like when integrated into gopls https://github.com/golang/tools/blob/master/gopls/doc/analyzers.md , many projects also enforce a common set of checks during CI.

Today, there are a variety of options to do that, for example, a random sampling:

• this repo
  • has a main package built using multichecker
  • builds and runs the analyzer in CI
  • (I think it's part of some online course, there are many copies of the same name / approach)
• this repo runs analysis binary in a test
• this repo installs and runs an analyzer directly (shadow)
• this repo runs a meta linter in git hooks and again as a ci jon
• this repo runs it in a makefile

Go already runs a set of vet checks by default during go test, these are high confidence checks that apply to everyone. Individual projects can usually be stricter and enforce more checks by default.
I think a helper to directly integrate linters into tests can ease the sprawl of complexity in ensuring projects continue to adhere to their desired quality gates.