release/v2.2.1#22
Merged
Merged
Conversation
Add `--runInBand` to `test:typeorm:mysql` so jest runs all spec files in a single worker process. The failing v2.2.0 c.basic-crud compound-key route test points at parallel-worker contention against the shared MySQL container — another spec in a sibling worker mutates the User row mid-suite, so the `WHERE companyId=1 AND profileId=5` lookup returns 0 rows. Apply the same isolation only to MySQL because release.yml's sequential `yarn test:all` runs typeorm:postgres FIRST (clean DB) then typeorm:mysql SECOND (where the contention manifests). Postgres run never had the bug because mutations across workers don't cross-pollute when each adapter run gets its own DB-prepare drop+seed and is the first sequential pass. Un-skips the c.basic-crud compound-key test that v2.2.0 shipped with `it.skip` to unblock publish.
Pre-request: dump registered Express routes filtered to /users4 + total route count. Post-failure: dump full response shape (status/body/text/headers). The test passes locally but returns 404 in CI MySQL with zero SQL emitted — need to determine whether the route is registered at all in CI before applying any further fix attempt. Diagnostic is no-op on success path.
Route registration is confirmed (count: 7 incl. GET /users4/:companyId/:profileId). Response shape confirms 404 from CrudService NotFoundException, not from NestJS routing. So the SELECT WHERE companyId=1 AND profileId=5 is returning 0 rows. Add a raw SQL probe via the app's DataSource right before the request: - All rows where id=5 OR profileId=5 - Targeted lookup matching the route's WHERE clause Will reveal whether user 5 was deleted, soft-deleted, or had its companyId/ profileId mutated mid-suite by an earlier spec.
…reAll Earliest jest-side observation point. Will reveal whether seed actually inserted user 5 in MySQL or whether it vanished during seed itself.
cursor.spec.ts cell 9 ("soft-delete still applies in cursor mode")
selected `targetId` as the last id of the first cursor page. The
/users-cursor controller is configured with `limit: 5`, so under
`?sort=id,ASC` the first page returns ids [1..5] and `targetId` lands
on user 5.
User 5 is part of the shared seed (companyId=1, profileId=5) and is
the row exercised by c.basic-crud's compound-key route test
`GET /users4/1/5`. The /users-cursor controller has no softDelete
config so the DELETE is a hard delete via TypeORM remove, dropping
the row from the shared `users` table.
Whether the bug surfaces depends on jest's file-execution order.
Locally jest tends to run c.basic-crud first (largest file in
sequencer cache) and the bug stays hidden. release.yml's MySQL
runner sequences cursor.spec before c.basic-crud and the compound-key
test then sees user 5 gone — `[diag user5 rows] []`, status 404.
Replace the seed-row target with a POST throwaway user (matching the
shape used in cell 5), capture the returned id, and DELETE that id.
The seed range is now untouched and execution order no longer matters.
Post-v2.2.0 fix for CI MySQL c.basic-crud compound-key flake. Root cause: cursor.spec cell 9 hard-deleted seeded user 5 from the shared `users` table when jest sequenced cursor.spec before c.basic-crud (jest default file-order varies by environment; CI hit the failing order, local did not). Includes: - d3a91ac: --runInBand on test:typeorm:mysql + un-skip compound-key - 47cdeb4: revert 5 diagnostic instrumentation commits - 7e5fbab: fix(test) — cursor.spec cell 9 uses POST throwaway target
The @zmotivat0r/mrepo orchestrator was replaced with native lerna + TypeScript composite project references in v2.2.0 (see CHANGELOG [Unreleased]). The orphan mrepo.json config file was left behind and is no longer read by any tooling.
v2.2.0 shipped 2026-05-03 (npm + tag v2.2.0 + GitHub Release) but the CHANGELOG sections were never promoted from [Unreleased]. Renames the section header in: - CHANGELOG.md - packages/core/CHANGELOG.md - packages/drizzle/CHANGELOG.md - packages/mikro-orm/CHANGELOG.md - packages/prisma/CHANGELOG.md - packages/request/CHANGELOG.md - packages/typeorm/CHANGELOG.md Adds a fresh [2.2.0] — 2026-05-03 stub to packages/util/CHANGELOG.md (version-only republish — bumped in lockstep with the rest of the monorepo, no package-specific source changes). Date pulled from `gh release view v2.2.0`.
knip 5 auto-detects per-package `src/index.ts` (via package.json `main`) and `jest.config.js` (via the jest plugin), so the explicit entries in each workspace block were redundant and surfaced as configuration hints on every run. Also drops: - `@mikro-orm/knex` from root `ignoreDependencies` — was a peer of @nestjs-crud/mikro-orm in v2.0.0 and removed in v2.0.1; nothing in the root references it any more. - `lerna` from root `ignoreDependencies` — now actually used by the release workflow and `yarn run rebuild` after the v2.2.0 mrepo→lerna swap, so it correctly resolves as a binary. - `src/crud/swagger.helper.ts` from packages/core `ignore` — knip can reach the file through normal import graph traversal now that the v3-gate dead-code paths have been removed. Adds `optionalPeerDependencies: "off"` so the @nestjs/swagger optional peer (declared optional in 2.1.1, intentionally referenced via safeRequire in swagger.helper.ts) does not flip the exit code. `yarn knip` now exits 0 with no findings.
Migrates the supertest done(callback) pattern to direct await across
five spec files:
- packages/typeorm/test/b.query-params.spec.ts (~23 tests)
- packages/typeorm/test/c.basic-crud.spec.ts (~32 tests)
- packages/drizzle/test/real-db-smoke.spec.ts (5 tests)
- packages/mikro-orm/test/real-db-smoke.spec.ts (5 tests)
- packages/prisma/test/real-db-smoke.spec.ts (5 tests)
Why: under shared-DB stress (CI MySQL container, parallel-worker
contention) the supertest .end((_, res) => { ...; done() }) callback
can fail to fire and the test hangs to jest's 30s timeout instead of
producing a Promise rejection with a clean stack. supertest >= 7
returns a thenable from request(server).get(...), so direct `await`
removes the manual callback bridge and surfaces failures with the
expected supertest assertion.
Behavior unchanged. All assertions preserved.
Verified locally:
- yarn test:typeorm:mysql 15/15 suites, 275 passed + 1 skipped
- yarn test:drizzle:postgres 10/10 suites, 91 passed
- yarn test:mikro-orm:postgres 12/12 suites, 131 passed
- yarn test:prisma:postgres 13/13 suites, 149 passed + 1 todo
packages/mikro-orm/test/users-repo.spec.ts (4 done() refs) is excluded
— that file is user-owned per the working-tree preserve rule.
Coordinated lockstep bump across all 7 publishable packages plus the root package.json (which lerna leaves untouched — it's outside the `packages/*` glob). lerna.json mirrors the new floor. Maintenance release. No consumer source changes. - Closes the v2.2.0 release-notes "known issue" on the c.basic-crud compound-primary-key route test (cursor.spec cell 9 was hard-deleting the seeded user 5 row that compound-key relies on). - Test-suite reliability hardening — done(callback) → async/await across typeorm/drizzle/mikro-orm/prisma specs; --runInBand on test:typeorm:mysql to dodge parallel-worker DB contention. - Removes orphan mrepo.json (the v2.2.0 mrepo→lerna swap left the config file behind; nothing reads it any more).
Force-bumps two transitive dev-only deps to their patched versions: - tmp 0.0.33 → ^0.2.4 (closes GHSA-52f5-9888-hmc6, low) - @conventional-changelog/git-client 1.0.1 → ^2.0.0 (closes GHSA-vh25-5764-9wcr, medium) Both surface only in yarn.lock — no publishable package.json declares them. CHANGELOG entries deliberately skipped per the v2.1.1 dev-tree advisory convention; the lockfile diff plus this commit message is the audit trail. uuid 11 → 14 (GHSA-w5hq-g745-h8pq, medium) is intentionally NOT resolved. uuid 14 ships as `"type": "module"` (pure ESM, no CJS exports). Loading it under the typeorm and core ts-jest CJS preset breaks every spec with `SyntaxError: Unexpected token 'export'` (verified). Fix waits on either lerna 9.x and typeorm 0.3.x publishing transitive bumps OR a coordinated jest-config ESM transform pass that touches all five jest configs. Carried concern; no consumer impact. Verified locally: - yarn build exit 0 - yarn test:typeorm:postgres 15/15 - yarn test:typeorm:mysql 15/15 - yarn test:drizzle:postgres 10/10 - yarn test:mikro-orm:postgres 12/12 - yarn test:prisma:postgres 13/13 - yarn test:parity 1/1
The single `img/logo.svg` rendered the wordmark and two of the four CRUD-grid squares in slate `#1E293B`, which is invisible on GitHub's dark theme — the README hero collapsed to a red-only fragment for any dark-mode reader. Splits the asset into a light/dark pair (slate text on light, frost text on dark). Same CRUD 2x2 mark, identical stroke and corner radius. Wordmark viewBox tightened from 440x80 to 310x80 (drops ~130px of trailing whitespace). NestJS red `#E0234E` is preserved as the constant accent across both variants. - img/logo-light.svg, img/logo-dark.svg (wordmark, 310x80) - img/logo-icon-light.svg, img/logo-icon-dark.svg (icon, 128x128) - img/logo.svg, img/logo-icon.svg (removed)
|
Packages with version v2.2.1 have been released |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
v2.2.1 — maintenance release
No consumer source changes. Resolves the v2.2.0 release-notes "known issue" on the
c.basic-crudcompound-primary-key route test, plus a round of test-suite reliability hardening, dev-tree advisory closures, and a brand refresh for GitHub dark-mode readers.Fixed
cursor.speccell 9 (soft-delete still applies in cursor mode) was selectingtargetIdas the last id of the first cursor page. The/users-cursorcontroller is configured withlimit: 5, so under?sort=id,ASCthe first page returned ids[1..5]andtargetIdlanded on user 5 — the row exercised byc.basic-crud'sGET /users4/1/5compound-key route test. The cell now POSTs a throwaway user, captures the returned id, and DELETEs that id; the seed range is no longer touched and execution order no longer matters. Test-fixture-only — no consumer code is affected.Changed
test:typeorm:mysqlscript now passes--runInBandto jest. Single-worker execution avoids parallel-worker contention against the shared MySQL container that v2.2.0 hit.done(callback)pattern is migrated to directawaitacrossb.query-params.spec.ts,c.basic-crud.spec.ts, and the threereal-db-smoke.spec.tsfiles (drizzle, mikro-orm, prisma). Failures now surface as Promise rejections with a clean stack instead of hanging to jest's 30s timeout.Removed
mrepo.jsonfile at the repo root is removed (the v2.2.0 mrepo→lerna swap left the config file behind; nothing reads it any more).Internal
tmp 0.0.33 → ^0.2.4(closes GHSA-52f5-9888-hmc6, low) and@conventional-changelog/git-client 1.0.1 → ^2.0.0(closes GHSA-vh25-5764-9wcr, medium) force-pinned via rootresolutions.uuid 11 → 14(GHSA-w5hq-g745-h8pq, medium) intentionally NOT resolved — uuid 14 ships as"type": "module", breaks ts-jest CJS preset across all five jest configs. Carried concern; no consumer impact.knip.jsonreduced to the entries knip 5 cannot auto-detect (16 hints → 0).[Unreleased]sections promoted to[2.2.0] — 2026-05-03across the 8 CHANGELOGs (the v2.2.0 publish itself missed the rename — process gap closed for next cycle).img/logo.svgsplit intoimg/logo-{light,dark}.svg+ matching icon pair;README.mdhero now uses<picture>withprefers-color-scheme: darkso the wordmark is visible in either GitHub theme.Verified locally
yarn build→ exit 0yarn test:typeorm:postgres→ 15/15 suitesyarn test:typeorm:mysql→ 15/15 suites (compound-key un-skipped, all 276 tests green)yarn test:drizzle:postgres→ 10/10 suitesyarn test:mikro-orm:postgres→ 12/12 suitesyarn test:prisma:postgres→ 13/13 suitesyarn test:parity→ 1/1 suite (72 tests)yarn knip→ exit 0