Integration in memory tests overhaul

[kalverra]

Big Changes

• Greatly simplifies the Integration In-Memory Tests, extracting core logic away from the legacy workflows to run only what we need.
  • Also includes various caching, speed, and cost improvements now possible thanks to the simplification.
• Changes core/store/ to use embedded go string instead of looking for a file. This enables some CI optimizations, and makes sense to me, but could change behavior/intent in ways I'm not sure of. Calling in @jmank88 on this.

Integration In-Memory Tests Improvements

These would run for a long time, were very complex, and burned a lot of money. They've been greatly simplified, made faster, and cheaper.

• Restructured them to remove legacy e2e-tests workflow that was unnecessarily complex
• Implemented caching for plugin installations and builds
• Fixed Go caching and added to general caching flow of the repo
• Pre-compile + warm caches in a single step to save a little time + a ton of money.

Comparison

Metric	Before	After	Delta	% Delta
Duration	28m20s	14m47s	-13m33s	-47.8%
Cost	$4.43	$0.80	-$3.62	-81.8%

Small Changes

• Update golangci-lint action + version in CI
• Linting in core/store/

[github-actions]

✅ No conflicts with other open PRs targeting develop

[trunk-io]

   

View Full Report ↗︎ ⋅ Docs

[Copilot]

Pull request overview

Copilot reviewed 21 out of 21 changed files in this pull request and generated no new comments.

Suppressed comments (9)

core/store/models/common_test.go:80

• t.Parallel() inside the subtest uses the range variable test from the outer loop without capturing it. Because subtests can start running after the loop continues, this can cause the wrong test case data to be used and lead to flaky tests.

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			t.Parallel()

core/store/models/common_test.go:141

• Same loop-variable capture issue with t.Parallel(): rebind test := test inside the loop before starting parallel subtests to avoid races/flakiness.

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			t.Parallel()

core/store/models/common_test.go:226

• Same loop-variable capture issue with t.Parallel() in subtests: rebind the loop variable before starting the parallel subtest to avoid non-deterministic behavior.

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			t.Parallel()

core/store/models/common_test.go:117

• Same loop-variable capture issue with t.Parallel(): the subtest closure reads test from the surrounding range loop without rebinding it, which can make the test non-deterministic under parallel execution.

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			t.Parallel()

core/store/models/common_test.go:206

• Same loop-variable capture issue with t.Parallel() in subtests: the closure uses test from the outer range loop without rebinding it, which can lead to flaky results.

	for _, test := range tests {
		t.Run(test.name, func(t *testing.T) {
			t.Parallel()

core/store/database_ddl.go:33

• Removing the gosec suppression here is likely to reintroduce a gosec warning for SQL string formatting (even though the identifier is validated/escaped). This may break CI linting; keep a targeted //nolint:gosec with justification.

	// PostgreSQL does not support bound parameters for database identifiers.

	_, err = db.ExecContext(ctx, fmt.Sprintf(dropDatabaseSQLFmt, quoted))

core/store/database_ddl.go:44

• Same as above: fmt.Sprintf-based SQL construction tends to trigger gosec; without a justified //nolint:gosec this can fail linting even though quotePostgresDBName validates/escapes the identifier.

	// PostgreSQL does not support bound parameters for database identifiers.

	_, err = db.ExecContext(ctx, fmt.Sprintf(createDatabaseSQLFmt, quoted))

.github/workflows/integration-in-memory-tests.yml:24

• The workflow_call input is named/described as a Slack channel ID, but the default fallback uses a channel name (#ccip-testing) and callers may pass #channel as well. Consider clarifying in the description that this accepts either a channel ID or a #channel name, to avoid confusing/misleading callers.

      slack_channel_id:
        description: "Slack channel ID to notify test results"

core/store/store_test.go:15

• Minor typo in comment: "without depending on on-disk paths" has a duplicated "on".

// insertFixtures must execute the given fixture SQL without depending on
// on-disk paths derived from runtime.Caller, so that pre-compiled binaries

[cl-sonarqube-production]

Quality Gate failed

Failed conditions
5 Security Hotspots
C Security Rating on New Code (required ≥ A)

See analysis details on SonarQube

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE