internal(pytest-mergify): honour an empty and a refused test selection

[AlexandreGaubert]

The engine is gaining two answers on the test-selection endpoint that no
published client knows: empty -- the predecessor's attempt of this job
already ran these tests and they passed, so run none of them -- and
refused -- several sessions share this job's name and collected the
same tests, so Mergify will not guess which one this run succeeds. Both
fall through the plugin's forward-compatibility guard today and quietly
run the full suite, which would make the feature's most visible outcome
never happen, and would leave an ambiguous job answered wrongly forever
with no symptom to notice.

Teach the plugin both. An empty answer empties the collection through
pytest's own deselection hook, so the session still finishes and still
uploads -- a job that legitimately ran nothing is the one a developer
comes asking about, and a shorter exit during collection would make it
the only one missing from Mergify's reporting -- and the run exits 0
instead of pytest's "no tests ran" code 5. A refused answer raises
pytest.UsageError, showing the server's own explanation of what it saw and
what to change -- the copy belongs there so it can be corrected without
publishing a client -- and marks the uploaded session errored so it cannot be
read as a clean empty one. Every other value
keeps degrading to a full run, which is what lets the engine grow
answers without breaking clients already published; the crate-level test
pinning that now uses a variant nothing implements, empty having
stopped being one.

pytest only. Vitest and Playwright share the Rust decoder but normalise
in their own TypeScript and have test selection disabled for the pilot
(MRGFY-8906), so wiring them now would be work on a path nobody runs.
They keep degrading to a full run on both answers, deliberately.

Fixes MRGFY-8931

[mergify]

Merge Protections

🔴 2 of 6 protections blocking · waiting on 👀 reviews

	Protection	Waiting on
🔴	👀 Review Requirements	👀 reviews
🔴	🔎 Reviews	👀 reviews
🟢	🤖 Continuous Integration	—
🟢	Enforce conventional commit	—
🟢	📕 PR description	—
🟢	🚦 Auto-queue	—

🔴 👀 Review Requirements

Waiting for

• #approved-reviews-by>=2

This rule is failing.

• any of:
  • #approved-reviews-by>=2
  • author = dependabot[bot]
  • author = mergify-ci-bot
  • author = renovate[bot]

🔴 🔎 Reviews

Waiting for

• #review-requested = 0

This rule is failing.

• #review-requested = 0
• #changes-requested-reviews-by = 0
• #review-threads-unresolved = 0

---

Show 4 satisfied protections

🟢 🤖 Continuous Integration

• all of:
  • check-success=ci-gate

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|internal|docs|style|refactor|perf|test|build|ci|chore|revert|ui)(?:\(.+\))?!?:

🟢 📕 PR description

• body ~= (?ms:.{48,})

🟢 🚦 Auto-queue

When all merge protections are satisfied, this pull request will be queued automatically.

[AlexandreGaubert]

Revision history

#	Type	Changes	Reason	Date
1	initial	cadc7ca		2026-09-01 15:49 UTC
2	content	cadc7ca → 2d73fdd (raw)	show the engine's own refusal message (real contract from monorepo#39711) instead of a hard-coded string	2026-09-01 15:49 UTC
3	content	2d73fdd → 739f042 (raw)	correct four comments the message-key delta made false (lens pass)	2026-09-01 15:59 UTC
4	content	739f042 → d7e6522 (raw)	rewrite the refusal copy for a CI user (Alexandre's review): what happened first, then why, then the fix with its docs link	2026-09-02 07:05 UTC