test: add corpus golden report snapshots

Rakefile

@@ -9,4 +9,15 @@ require "rubocop/rake_task"
 
 RuboCop::RakeTask.new
 
+namespace :corpus do
+  desc "Regenerate committed golden reports for the vendored GitLab corpus"
+  task :regenerate_golden_reports do
+    require_relative "lib/code_keeper"
+    require_relative "spec/support/corpus_report_runner"
+    require_relative "spec/support/corpus_golden_report"
+
+    CorpusGoldenReport.regenerate_all
+  end
+end
+
 task default: %i[spec rubocop]

spec/code_keeper/golden_report_spec.rb

@@ -0,0 +1,45 @@
+# frozen_string_literal: true
+
+require "spec_helper"
+require "support/corpus_report_runner"
+require "support/corpus_golden_report"
+
+RSpec.describe "GitLab corpus golden reports" do
+  it "has one snapshot per corpus file" do
+    expect(Dir[File.join(CorpusGoldenReport::ROOT, "**/*.json")].sort).to eq CorpusGoldenReport.snapshot_paths.sort
+  end
+
+  it "matches the current CodeKeeper report for every corpus file" do
+    CorpusReportRunner.file_paths.each do |path|
+      aggregate_failures path do
+        expect(File.read(CorpusGoldenReport.snapshot_path_for(path))).to eq CorpusGoldenReport.serialized_report_for(path)
+      end
+    end
+  end
+
+  it "stores repository-relative paths in snapshots" do
+    paths = CorpusGoldenReport.snapshot_paths.flat_map do |snapshot_path|
+      paths_in(JSON.parse(File.read(snapshot_path)))
+    end
+
+    expect(paths).to all(match(%r{\Aspec/fixtures/corpus/gitlab/}))
+  end
+
+  it "detects stale snapshots outside the corpus file list" do
+    stale_path = File.join(CorpusGoldenReport::ROOT, "obsolete.rb.json")
+    existing_paths = CorpusGoldenReport.snapshot_paths + [stale_path]
+
+    expect(CorpusGoldenReport.stale_snapshot_paths(existing_paths: existing_paths)).to eq [stale_path]
+  end
+
+  def paths_in(value)
+    case value
+    when Array
+      value.flat_map { |item| paths_in(item) }
+    when Hash
+      value.flat_map { |key, item| key == "path" ? [item] : paths_in(item) }
+    else
+      []
+    end
+  end
+end

spec/fixtures/corpus/gitlab/README.md

@@ -36,3 +36,18 @@ MIT Expat license notice included here.
 - `app/models/concerns/from_set_operator.rb`
 - `app/models/concerns/ci/partitionable.rb`
 - `spec/models/ability_spec.rb`
+
+## Golden Reports
+
+Committed golden reports live under `spec/fixtures/golden_reports/gitlab/`.
+There is one pretty-printed JSON report per corpus file, and report paths are
+recorded with repository-relative corpus paths.
+
+Regenerate them only as an explicit out-of-band maintenance task:
+
+```bash
+bundle exec rake corpus:regenerate_golden_reports
+```
+
+This regeneration task is not part of CI. Review the resulting JSON diff before
+committing it.