tests: realism migration regression coverage

Four gaps from the realism migration plan, plus one flaky test
fixed.

Added:

- tests/deploy/test_orchestrator_unit.py — replaces the dead
  test_emailgen_unit.py. Asserts:
  * decnet-orchestrator.service.j2 carries the DECNET_REALISM_*
    env block (LLM, MODEL, TIMEOUT, PERSONAS) so per-host tuning
    works without editing the .j2.
  * Legacy DECNET_EMAILGEN_* vars are NOT referenced — clean break
    contract from stage 5.
  * decnet.target wants orchestrator + canary, does NOT want
    decnet-emailgen.service. Anti-regression for service-collapse.
  * deploy/decnet-emailgen.service.j2 stays deleted.

- tests/orchestrator/test_worker_integration.py — new
  test_one_tick_email_branch_records_orchestrator_email. Pins the
  action-roll to email, seeds a topology with an IMAP mail decky +
  two personas, stubs LLM + docker-exec write paths, verifies an
  orchestrator_emails row + bus event land. Restores end-to-end
  email coverage that was lost when the pre-collapse
  test_worker_integration.py was deleted.

- tests/realism/test_synthetic_files_truncation.py — pins the 64KB
  last_body cap on create + edit, and documents the consequence:
  edit candidates carry a truncated snapshot of files that exceeded
  the cap. If a future change lifts the cap, _LIMIT in the test
  must lift with it.

Fixed flaky:

- tests/orchestrator/test_scheduler.py — two pick_file tests
  pinned to random.Random(1). Without a seed, the 3% canary gate
  (stage 7) and 10% leave-alone roll occasionally flaked the
  assertions because the _FakeRepo doesn't carry a
  create_canary_token method.

Note: the existing
test_realism_subprocess_import_personas_rejects_in_agent_mode
already covers agent-mode rejection of decnet realism
import-personas; no new gating test needed.
This commit is contained in:
2026-04-27 17:29:25 -04:00
parent a07fb3fe08
commit b86129e35e
5 changed files with 382 additions and 93 deletions

View File

@@ -0,0 +1,109 @@
"""``synthetic_files.last_body`` is capped at 64 KB.
The orchestrator caps the persisted body at 64 KB on every write
(create + edit) so the table doesn't bloat with large blobs. This
introduces a real edge: an EditAction whose ``previous_body`` is
sourced from the cap (not the file on disk) sees truncated bytes.
Today the realism templates produce well under 64 KB, so the edge
isn't reachable from the planted-content path. But a future change
that lifts the cap, an LLM that returns a long body, or a
``honeydoc_pdf`` body cultivated through the realism path could all
hit it. These tests pin the contract so a regression that drops the
cap or applies it inconsistently fails loudly.
"""
from __future__ import annotations
from datetime import datetime, timezone
import pytest
import pytest_asyncio
from decnet.web.db.sqlite.repository import SQLiteRepository
_LIMIT = 65536 # decnet/orchestrator/worker.py uses [:65536]
@pytest_asyncio.fixture
async def repo(tmp_path):
r = SQLiteRepository(db_path=str(tmp_path / "decnet.db"))
await r.initialize()
yield r
await r.engine.dispose()
def _row(body: str) -> dict:
import hashlib
now = datetime.now(timezone.utc)
return {
"decky_uuid": "d1",
"path": "/home/admin/notes.txt",
"persona": "admin",
"content_class": "note",
"created_at": now,
"last_modified": now,
"edit_count": 0,
# The hash is over the *full* body in the orchestrator's write
# path; if the body comes from a row that was already truncated,
# the hash reflects the truncation. Tests check both paths.
"content_hash": hashlib.sha256(body.encode("utf-8")).hexdigest(),
"last_body": body[:_LIMIT],
}
@pytest.mark.asyncio
async def test_oversized_body_is_truncated_at_write(repo):
body = "A" * (_LIMIT * 2)
uuid = await repo.record_synthetic_file(_row(body))
rows = await repo.list_synthetic_files(decky_uuid="d1")
assert len(rows) == 1
stored = rows[0]
assert stored["uuid"] == uuid
assert len(stored["last_body"]) == _LIMIT
@pytest.mark.asyncio
async def test_body_at_exact_limit_is_preserved(repo):
"""Boundary: a body of exactly 64 KB must not be silently
truncated. Off-by-one regression target."""
body = "B" * _LIMIT
await repo.record_synthetic_file(_row(body))
rows = await repo.list_synthetic_files(decky_uuid="d1")
assert len(rows[0]["last_body"]) == _LIMIT
@pytest.mark.asyncio
async def test_pick_for_edit_returns_truncated_body(repo):
"""Stage 3b contract: the edit candidate carries the *stored*
last_body — necessarily truncated when the original exceeded the
cap. Document the consequence so a future test author doesn't
expect the full body to round-trip."""
body = "C" * (_LIMIT * 3)
await repo.record_synthetic_file(_row(body))
candidate = await repo.pick_random_synthetic_file_for_edit("d1")
assert candidate is not None
assert len(candidate["last_body"]) == _LIMIT
# The edit driver mutates this body via realism.bodies.next_iteration,
# so callers must accept they're editing a truncated snapshot of
# the file that's actually on the decky. This is documented
# behaviour pre-v1; if the cap rises, lift _LIMIT here too.
@pytest.mark.asyncio
async def test_edit_path_keeps_cap(repo):
"""An update_synthetic_file call that tries to write a >cap body
must clip to the cap on the way in. Mirrors the orchestrator
worker's ``last_body=body[:65536]`` line."""
uuid = await repo.record_synthetic_file(_row("seed"))
big = "D" * (_LIMIT * 4)
await repo.update_synthetic_file(
uuid,
{
"last_modified": datetime.now(timezone.utc),
"edit_count": 1,
"last_body": big[:_LIMIT], # caller is responsible for clipping
},
)
rows = await repo.list_synthetic_files(decky_uuid="d1")
assert len(rows[0]["last_body"]) == _LIMIT