test(server): pin the queen export route and bundle helper - createQueenExportRoute (Closes #1676)

[dmitrii-f-t27]

Closes #1676

What changed

One new file: trios/agent-server/apps/server/tests/api/queen-export.test.ts (the path the issue's Success Criteria names). The subject queen-export.ts is untouched.

The single exported route factory and the exported bundleOfBranch helper are exercised (7 it( blocks) against a real throwaway git repository steered through the WORKSPACE_DIR / TRIOS_REPO_URL / TRIOS_REPO_REF environment overrides (restored in an afterAll):

• bundleOfBranch (3 tests): 404 for a branch this checkout does not hold (pinned error text); 409 for a branch level with the base; and the ok case - branch, base, the commit list (sha, author, date, subject), the file list, and bytes that are a REAL bundle proven by git bundle verify against the same checkout (the singular wording measured on this git).
• The route (4 tests): GET / lists only the branches ahead of the base (the level branch and the foreign name excluded), sorted by issue; GET /:issue serves the branch as JSON with the commit list, file list and base64 bundle bytes (measured: the base64 decodes to a bundle whose header starts with # v); GET /abc answers 400 with the pinned message; GET /999 answers 404 (the stored-bundle fallback is inert without DATABASE_URL - measured).

The suite registers no mock.module at all and touches nothing outside its mkdtempSync tree, so it cannot affect sibling suites. No network, no database, no container.

How it was verified (FR-004)

• Passing run against the unmodified subject: 7 pass / 0 fail / 30 expect() calls, Ran 7 tests across 1 file (~0.5 s).
• Deliberately broken copy (the no-commits check widened to -1; the 404 error reworded; the listing sort reversed; the issue validation loosened to allow 0): 4 fail out of 7, quoted raw above. The subject was restored byte-identical (git checkout) and the suite passes again.
• Three draft corrections came from measurement, not guessing: this git's bundle-verify wording (The bundle contains this ref:), the branch topology (a branch cut from another work branch counts both commits), and a subarray length. The file is biome ci-clean and contributes no typecheck error (local monorepo typecheck: base set, 0 mentions of the new file).

Not verified

• Nothing beyond the quoted local runs; the CI Tests / server-api job re-runs the suite on this SHA.

Dmitrii Fedorov (@dmitrii-f-t27)

[tri-claw]

Summary by t27-bees

This PR adds a test suite that pins the behavior of createQueenExportRoute and its bundleOfBranch helper by exercising them against a real throwaway git repository. The tests verify bundle creation, route listing, and error responses without mocks, databases, or network.

Worth a look

• Global process.env mutation (lines 54–59, 61–67): The test overwrites WORKSPACE_DIR, TRIOS_REPO_URL, TRIOS_REPO_REF and deletes DATABASE_URL/RAILWAY_SSOT_URL at module level, restoring them only in afterAll. If the test crashes or is interrupted, these leak into other suites running in the same process.

• Temp directory cleanup only in afterAll (lines 44, 61–67): mkdtempSync creates parent; rmSync(parent, …) runs only in afterAll. A test failure before afterAll leaves the temp tree on disk.

• Git output string matching (line 118): The test asserts verified.includes('The bundle contains this ref:'), which the description notes was "measured on this git". Different git versions or locales may phrase this differently, causing false failures.

• Shell interpolation in git helper (line 48): execSync(git -C "${repo}" ${args}) interpolates args directly into the shell command. While current calls use hardcoded strings, the pattern invites injection if reused with dynamic input.

• Branch sorting assumption (line 146): The test expects body.branches.map(b => b.issue) to equal [45, 123], relying on the implementation sorting by issue number. If the sort order changes (e.g., by date), the test fails even though the contract ("sorted by issue") is unchanged.

---

_{Read: 1 files, +192 -0, the whole diff. Head dcf2cc1. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}

[github-actions]

✅ Tests passed — 3560/3630

Suite	Passed	Failed	Skipped
✅ agent	188/188	0	0
✅ build	9/9	0	0
✅ cdp-protocol	5/5	0	0
✅ eval	93/93	0	0
✅ server-agent	280/280	0	0
✅ server-api	2159/2228	0	69
✅ server-browser	6/6	0	0
✅ server-integration	10/11	0	1
✅ server-lib	279/279	0	0
✅ server-pglive	166/166	0	0
✅ server-root	68/68	0	0
✅ server-sim	8/8	0	0
✅ server-skills	31/31	0	0
✅ server-tools	244/244	0	0
✅ shared	14/14	0	0

View workflow run

[dmitrii-f-t27]

Checks note, before merge:

• runner / Biome and runner / Typecheck are red on the base branch queen with pre-existing failures (reference base run https://github.com/gHashTag/trios/actions/runs/38087839871 on dfb6d97). On this PR's SHA: the Typecheck log has the 19 base error TS lines and 0 mentions of the new queen-export.test.ts; the Biome log has 0 diagnostics for the new file. All Tests / * suites including server-pglive are green on this SHA, and the suite registers no mock.module at all (a real throwaway git repository steered by environment overrides), so it cannot affect sibling suites.

Dmitrii Fedorov (@dmitrii-f-t27)