meshai/work/tests/test_satpass_persisted_timer.py
malice 41178831e4
chore(central-ripout 2b): relocate satellite code to env/satellite/ (#162)
* chore(central-ripout 2b): create env/satellite package, move pass_predictor

pass_predictor.py was 100% live (no dead entrypoint) — SGP4 pass
computation used by both the native satpass adapter and the on-demand
!satpass command. Straight move, no code changes: meshai.central.pass_predictor
-> meshai.env.satellite.pass_predictor. Owner directive: satellite code gets
its own folder under the feed adapters, separate from env.satpass (the
adapter) to avoid colliding with env/satpass.py.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(central-ripout 2b): split satpass_handler.py -> env/satellite/pass_format.py

satpass_handler.py was a split file: live wire-formatting/gate logic plus
dead Central-envelope ingest machinery whose only caller was the
already-deleted central/consumer.py NATS bridge.

Moved (live, verified via rg — external callers in env/satpass.py and
commands/satpass_cmd.py, or transitively called by them):
  gate_consolidated_pass, format_pass, _check_rate_cap, _upsert_satpass,
  _attach_commit, _map_severity, _canonical_id, _azimuth_to_compass,
  _short_sat_name, _collapse_compass, _region_paren, _is_synthetic_observer,
  _format_time_12h/24h, _format_ampm, _tz_abbr, _date_label,
  plus the _SHORT_SAT_NAMES/_SHORT_NAME_SUBSTR/_SYNTHETIC_OBSERVERS tables.

Dropped (dead — zero callers outside the already-deleted consumer.py and
handle_satpass/consolidate_satpass_pending themselves; verified with rg):
  handle_satpass, consolidate_satpass_pending, _cleanup_pending,
  load_pending_schedule, _log_event_returning_id, _coerce_float,
  _coerce_int, _parse_iso_epoch, _now, CONSOLIDATION_DELAY,
  _pending_consolidation_ids, drain_pending_consolidation_ids,
  _elevation_bucket (already-orphaned pre-ripout: superseded by numeric
  "max NN°" wire format, zero callers anywhere but its own tests),
  SCHEMA_SATPASS_EVENTS/SCHEMA_SATPASS_PENDING (unused string constants —
  actual schema lives in persistence/migrations/*.sql, never imported).
  Also dropped now-unused `json`/`time`/`Any` imports.

Straight code move otherwise — no logic changes to any moved function.
Two docstrings updated for accuracy (module docstring, and
gate_consolidated_pass's docstring which referenced the now-deleted
Central consumer path).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(central-ripout 2b): split tle_handler.py -> env/satellite/tle_store.py

tle_handler.py was a split file: live storage helpers plus a dead
Central-envelope ingest entrypoint whose only caller was the
already-deleted central/consumer.py NATS bridge.

Moved (live — used by env.tle_fetch, env.satpass, commands.satpass_cmd,
verified via rg): upsert_tle, get_fresh_tles, get_tle_by_norad,
search_tle_by_name.

Dropped (dead — handle_tle's only callers were tests and the deleted
consumer.py; verified with rg): handle_tle.

Straight code move otherwise — no logic changes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(central-ripout 2b): repoint satellite consumers at env.satellite

Rewire the three production consumers (including their lazy/function-body
imports, not just module-top ones) to the new location:
  - env/satpass.py: central.tle_handler -> env.satellite.tle_store,
    central.pass_predictor -> env.satellite.pass_predictor,
    central.satpass_handler -> env.satellite.pass_format
  - env/tle_fetch.py: central.tle_handler.upsert_tle -> env.satellite.tle_store
  - commands/satpass_cmd.py: all three, same mapping

Also refreshed docstrings that pointed at the old module paths or described
the now-fully-deleted Central consolidation path
(consolidate_satpass_pending / satpass_pending buffer) as a live
alternative, and updated central/__init__.py's module docstring to stop
listing the three relocated modules among central's remaining contents.

No behavior changes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(central-ripout 2b): update satpass/tle test suite for the relocation

Repoints every remaining test import at the new env.satellite.* modules
and removes/adapts coverage for the Central envelope-ingest path deleted
in this pass (handle_satpass, consolidate_satpass_pending, handle_tle, and
the filtering/coercion/staleness logic that lived only inside them):

- test_satpass_native.py, test_tle_fetch.py, test_satpass_command.py:
  import-path updates only (pass_predictor, tle_store). Also dropped
  test_satpass_command.py's TestTLEUpsert.test_returns_none_always
  (handle_tle-specific contract, no longer applicable) and rewrote its two
  latest-wins tests to call upsert_tle directly — same behavior under test,
  now exercised through the still-live primitive instead of the dead
  wrapper.
- test_satpass_native.py: deleted test_central_consolidate_feeds_shared_gate_merged
  (spied on consolidate_satpass_pending, which no longer exists). The
  merge-across-observers logic it guarded is native-side (_consolidate)
  and already covered by test_two_observers_consolidate_to_one_broadcast.
- test_satpass_handler.py: gutted to the one test that calls format_pass
  directly (test_format_pass_worst_case_fits_140); the rest exercised
  handle_satpass's observer/norad/elevation filters, which have no live
  equivalent (the native adapter filters at the config level, not
  per-envelope) and is redundant with test_satpass_native.py's dedup/wire
  coverage via the real SatpassAdapter path.
- test_satpass_broadcast_safety.py: kept every test that calls format_pass
  or gate_consolidated_pass-adjacent REGISTRY checks directly (wire format,
  clean-format rules, REGISTRY defaults); deleted TestNoradIdTypeCoercion
  and TestStalenessGuard (handle_satpass-only logic, no live equivalent)
  and the 6 _elevation_bucket tests (_elevation_bucket itself was dead
  before this pass too — zero callers anywhere but its own tests, already
  superseded by the numeric "max NN°" wire format per its own docstring).
- test_satpass_persisted_timer.py: dropped test_due_at_persisted_on_normal_ingest
  (handle_satpass-only); kept the two schema/migration tests, which don't
  touch satpass_handler.
- Deleted outright (tested ONLY the dead Central envelope-ingest path, no
  live equivalent to port to): test_satpass_event_path.py,
  test_satpass_compass_fallback.py, test_satpass_wire_fields.py.

Full suite: 2059 passed, 0 failed (was 0 failed on main pre-change).
Satpass/TLE subset (99 tests across 8 files) verified green in isolation.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Matt Johnson <mj@k7zvx.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-17 15:46:36 -06:00

58 lines
2.9 KiB
Python

"""Tests for the satpass persisted-timer `due_at` column (schema only).
Pending satellite-pass consolidations used to be scheduled only as in-memory
asyncio TimerHandles, so a restart orphaned any satpass_pending rows: the row
survived but its timer did not, and it was never consolidated/broadcast. The
fix persisted a durable `due_at` on each pending row and added a startup
sweep (`CentralConsumer._sweep_pending_satpass`) that reconstructed a timer
for every pending consolidated_id off its persisted due_at.
The Central NATS consumer (and `_sweep_pending_satpass` with it) was retired
2026-07 -- the sweep's tests are gone with it. The native satpass path
(env/satpass.py) never used the satpass_pending buffer or this sweep in the
first place (it consolidates in-memory within a single tick), so nothing
live is affected.
The ingest path that wrote `due_at` (`central.satpass_handler.handle_satpass`)
was itself dead -- its only caller was the already-retired Central consumer
-- and was deleted in the same pass that relocated the still-live satellite
code to `env.satellite` (2026-07). Its due_at-persistence test is gone with
it; the `satpass_pending` table (and its due_at/peak_compass columns) are
now unused by any live code path, but the migrations that created them are
left in place (schema changes are out of scope for that pass). What remains
here:
- SCHEMA_VERSION == 26 and the v22 migration (which added the due_at
column) still applies cleanly on a fresh DB
"""
from __future__ import annotations
import pytest
from meshai.persistence import get_db, init_db, SCHEMA_VERSION
# ── schema / migration ───────────────────────────────────────────────
def test_schema_version_is_current():
# SCHEMA_VERSION is derived from the highest vN.sql in migrations/, so
# this just guards against the derivation returning something bogus
# (e.g. 0, which would mean the migrations dir wasn't found). The v22
# migration (this file's focus) must always be <= the current version.
assert SCHEMA_VERSION >= 22
def test_v22_migration_applies_and_adds_due_at_column(tmp_path, monkeypatch):
"""Fresh DB migrates cleanly and satpass_pending has due_at (v22 column)."""
from meshai.persistence import close_thread_connection
from meshai.persistence import db as persistence_db
db = str(tmp_path / "fresh-v22.sqlite")
monkeypatch.setenv("MESHAI_DB_PATH", db)
persistence_db._initialised.clear()
close_thread_connection()
conn = init_db()
row = conn.execute("SELECT value FROM schema_meta WHERE key='version'").fetchone()
assert int(row["value"]) == SCHEMA_VERSION
cols = {r["name"] for r in conn.execute("PRAGMA table_info(satpass_pending)")}
assert "due_at" in cols
close_thread_connection()
persistence_db._initialised.discard(db)