Skip to main content

safe-code-change

After a code change, find affected tests, update them to match new behavior, then guide the user to run validation. Use when the user has made or asked for a code change and wants to make sure nothing is broken.

Informations de source

Dépôt
openshift/lightspeed-operator
Dernière activité de la source
18 mars 2026 à 11:47
Langue détectée de SKILL.md
anglais
Étoiles
17
Forks
59

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

Affichage de SKILL.md

SKILL.md
Instructions source · Aperçu en lecture seule
name
safe-code-change
description
After a code change, find affected tests, update them to match new behavior, then guide the user to run validation. Use when the user has made or asked for a code change and wants to make sure nothing is broken.
disable-model-invocation
true
# Safe Code Change After a code change is made, find and fix affected tests before running validation. ## Rules - The code change is already done. Do not modify production code. - Only update tests to match the new behavior, not the other way around. - Do not reformat or lint-fix during test updates. Save that for validation. - If a test change is ambiguous (unclear what the new expected behavior is), ask the user. ## Step 1: Identify What Changed ```bash git diff --name-only git diff --stat ``` List the modified production files (ignore test files, configs, docs). ## Step 2: Find Affected Tests Search for imports and uses of changed functions/types across all test files: ```bash # Find test files that import the changed package rg "github.com/openshift/lightspeed-operator/<changed_package>" --type go -g '*_test.go' # Find direct function/type references rg "<ChangedFunctionOrType>" --type go -g '*_test.go' ``` For controller changes, also check: - `internal/controller/suite_test.go` (shared test setup) - `test/e2e/` (E2E tests) ## Step 3: Analyze Impact on Tests For each affected test file, check whether the change breaks existing tests: 1. **Signature changes** — function renamed, parameters added/removed/reordered. 2. **Behavior changes** — return value, error messages, side effects differ. 3. **Removed code** — tests for deleted functions/types need removal. 4. **New code** — consider whether new tests are needed (ask user if unclear). 5. **Interface changes** — mock implementations need updating. ## Step 4: Update Tests Apply minimal fixes to each affected test: ### For Ginkgo Tests (most controller tests): - Update `Expect()` assertions to match new return values - Update mock return values in test fixtures - Add/remove parameters in function calls - Update error message checks - Adjust `Eventually()` timeouts if reconciliation logic changed ### For Standard Go Tests (utils, helpers): - Update table-driven test cases with new expected values - Update mock implementations - Add/remove parameters in function calls - Update error assertions ### Common Fixes: - **Error constant renamed**: Update all `Expect(err).To(MatchError(ContainSubstring(oldName)))` → `newName` - **Function signature changed**: Update all call sites in tests - **Resource structure changed**: Update test fixtures and expected values - **Owner reference logic changed**: Update assertions that check `OwnerReferences` ## Step 5: Verify Test File Syntax Before telling the user tests are ready, verify Go syntax: ```bash go fmt <modified_test_file> ``` If formatting changes the file significantly, there may be syntax errors. ## Step 6: Report List all test files updated and what was changed in each: 1. File name 2. What was updated (function calls, expectations, mocks, fixtures) 3. Number of changes Then guide the user to run validation: ``` Tests are updated. Run validation with: make test # Run all unit tests make test-e2e # Run E2E tests (requires cluster) make lint # Check code style If tests fail, review the specific failures and adjust expectations. ``` Do not run `make test` automatically — let the user control when tests run.
Voir sur GitHub