loomio: enable by default; idempotent add (never re-invite) + targeted remove

The guard direction from 2026-11-04 is corrected: the co-op DOES use Loomio and
wants every new member added. Removing members/notifications is not the goal —
not duplicating them is. The feature is switched back ON and the sync is made
idempotent instead.

Enablement
  * LOOMIO_SYNC_ENABLED now defaults to ENABLED (kill switch: set =0 to pause).
    Restoring LOOMIO_API_KEY and deploying is all that is needed again.

Non-duplicating ADD (the real fix), two layers:
  * (a) live pre-check against Loomio group membership (b3 index, ordered and
    paginated stably). An address already in the group is skipped SILENTLY —
    it is never placed in an invite payload, so no API call about it can fire
    a notification or email.
  * (b) members.loomio_invited_at: an address we have handed over before is
    never sent again, even if it is not (yet) visible in the live roster.
  * FAIL CLOSED if the live roster cannot be read.
  * remove_absent is never sent.

Removal reconciliation preserved (deactivation must still propagate):
  * A member in the group who is no longer active is removed via a TARGETED
    b3 DELETE by membership id — it names only the member being removed, so it
    cannot re-invite or notify anyone staying. Operator allowlist
    (LOOMIO_ALWAYS_KEEP) is never removed. 404 (already gone) counts as
    success, so a re-run neither errors nor churns.
  * The keep-set is computed from the UNFILTERED active list: an active member
    whose Cloudron account is not yet activated is not removed (only their
    invite is deferred).

Tests: tests/test_loomio_guard.py (26/26) — add/dedupe, silent skip, record
guard, fail-closed, remove-once, allowlist-never-removed, targeted removal
(no present member in any payload), double-run no-op for both phases, 404
tolerance, kill switch.
This commit is contained in:
inference-bot committed 2026-11-04 19:25:45 +00:00
1 parent 17329c5d6f
commit f7bbec8a0a
2 files changed
+282 -145

No files matched your search

+140 -67
View File
@@ -1,20 +1,30 @@
#!/usr/bin/env python3
"""Regression test: the Loomio sync can never invite an existing member twice.
"""Regression test: the Loomio sync is idempotent and non-duplicating.
WHY THIS EXISTS
On 2026-11-04 a live run of sync_loomio_memberships() emailed a Loomio invite
notification to (effectively) the whole membership. The guards that must hold
are proven here on a FIXTURE db with the network stubbed:
are proven here on a FIXTURE db with the network STUBBED:
1. OFF by default — no Loomio call happens unless LOOMIO_SYNC_ENABLED is on,
so restoring LOOMIO_API_KEY alone cannot re-arm the integration.
2. POSITIVE EXCLUSION — an address is only ever sent if it is neither in
Loomio's current member list NOR recorded as previously invited
(members.loomio_invited_at). This matters because the invite-record
column starts NULL for the ~89 members ALREADY in Loomio.
3. FAIL CLOSED — if Loomio's current member list can't be read, no invite is
sent.
4. `remove_absent` is never sent (no evict-then-re-add re-invite path).
ADD / dedupe
1. A member already in the Loomio group is skipped SILENTLY — no invite
payload ever contains their address.
2. members.loomio_invited_at is a second guard: an address we have handed
over before is never sent again, even if it is not (yet) visible in the
live roster.
3. `remove_absent` is never sent (no evict-then-re-add re-invite path).
4. FAIL CLOSED — if the live roster cannot be read, nothing is sent.
5. Running the sync twice adds nobody a second time.
REMOVE / reconciliation (a deactivation MUST still propagate to Loomio)
6. A member in the group who is no longer active is removed exactly once.
7. The operator allowlist is never removed, even when not active.
8. Removal is targeted — it never re-sends a present member's address, so it
cannot re-invite or notify anyone who is staying.
9. A second run removes nobody again (the roster no longer lists them).
10. remove_loomio_membership treats 404 as success (idempotent, no error).
The sync is ON by default (kill switch LOOMIO_SYNC_ENABLED=0 pauses it).
Runnable: python3 tests/test_loomio_guard.py (from the repo root)
"""
@@ -24,36 +34,37 @@ import tempfile
import asyncio
os.environ["DB_PATH"] = os.path.join(tempfile.mkdtemp(), "fixture.db")
os.environ["LOOMIO_API_KEY"] = "test-key"
os.environ["LOOMIO_API_KEY"] = "test-user-key"
os.environ["LOOMIO_B3_KEY"] = "test-server-key"
os.environ["LOOMIO_GROUP_ID"] = "1"
os.environ["LOOMIO_SYNC_ENABLED"] = "off"
# Leave LOOMIO_SYNC_ENABLED unset -> the sync must default to ENABLED.
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
import app.main as m # noqa: E402
m.DB_PATH = os.environ["DB_PATH"]
m.LOOMIO_ALWAYS_KEEP = [] # keep the assertion about membership emails isolated
# Keep the membership assertions isolated: no extra keep accounts, and no
# username-fix pass (it makes its own /api/b3/users calls).
m.LOOMIO_ALWAYS_KEEP = ["keep@x.org"]
sent = [] # every b2/memberships POST payload
GET_FAILS = False # toggled to exercise fail-closed
posts = [] # every b2/memberships POST payload
deletes = [] # every b3 membership DELETE url
ROSTER = {} # email -> membership id currently in the group
GET_FAILS = False # exercise fail-closed
DELETE_STATUS = 200 # toggle to prove 404 tolerance
class FakePostResp:
status_code = 200
def json(self):
return {"added_emails": ["new@x.org"]}
class FakeGetResp:
def __init__(self):
self.status_code = 200
class FakeResp:
def __init__(self, status=200, payload=None):
self.status_code = status
self._payload = payload or {}
self.text = str(payload or {})
def raise_for_status(self):
pass
if self.status_code >= 400:
raise RuntimeError(f"HTTP {self.status_code}")
def json(self):
# member@x.org is ALREADY in the Loomio group.
return {"memberships": [{"user_email": "Member@x.org"}]}
return self._payload
class FakeClient:
@@ -68,19 +79,27 @@ class FakeClient:
async def get(self, url, headers=None, params=None):
if GET_FAILS:
raise RuntimeError("simulated Loomio read failure")
return FakeGetResp()
raise RuntimeError("simulated Loomio roster read failure")
offset = (params or {}).get("offset", 0)
limit = (params or {}).get("limit", 50)
rows = [{"id": i, "user_email": e} for e, i in ROSTER.items()]
page = rows[offset:offset + limit]
return FakeResp(200, {"memberships": page, "meta": {"total": len(rows)}})
async def post(self, url, headers=None, json=None):
sent.append({"url": url, "json": json or {}})
return FakePostResp()
posts.append({"url": url, "json": json or {}})
return FakeResp(200, {"added_emails": (json or {}).get("emails", [])})
async def delete(self, url, headers=None):
deletes.append(url)
return FakeResp(DELETE_STATUS)
async def patch(self, *a, **k):
return FakePostResp()
return FakeResp(200, {})
async def fake_active():
# new@ never seen; old@ previously invited (record); member@ already in Loomio
# active members: new@ / old@ are NOT in the group; member@ IS.
return ["new@x.org", "old@x.org", "member@x.org"]
@@ -88,6 +107,10 @@ async def fake_activated(emails):
return emails
async def noop_fix(_client):
return {}
results = []
@@ -96,7 +119,7 @@ def check(name, ok):
async def main():
global GET_FAILS
global GET_FAILS, ROSTER, DELETE_STATUS
m.get_db().close() # create + migrate the fixture schema
conn = m.get_db()
for e in ("new@x.org", "old@x.org", "member@x.org"):
@@ -111,53 +134,103 @@ async def main():
m.get_active_member_emails = fake_active
m.filter_loomio_activated = fake_activated
m.httpx.AsyncClient = FakeClient
m.loomio_fix_usernames = noop_fix
# 1. Disabled -> no Loomio call at all.
m.LOOMIO_SYNC_ENABLED = False
sent.clear()
await m.sync_loomio_memberships()
check("disabled: NO Loomio call", sent == [])
check("default: sync is ENABLED when the flag is unset", m.LOOMIO_SYNC_ENABLED is True)
# 2. Enabled: only the never-seen address is sent. old@ is excluded by its
# invite record; member@ is excluded because Loomio ALREADY has them
# (even though their loomio_invited_at is NULL — the backfill gap).
m.mark_loomio_invited("old@x.org")
m.LOOMIO_SYNC_ENABLED = True
sent.clear()
# Loomio group currently holds: an active member (member@), and the operator
# allowlist account (keep@) which is NOT an active portal member.
ROSTER = {"member@x.org": 10, "keep@x.org": 2}
# --- 1. ADD path: only the never-seen address is invited. ---
m.mark_loomio_invited("old@x.org") # recorded -> excluded by layer (b)
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
body = sent[0]["json"] if sent else {}
check("enabled: exactly one Loomio call", len(sent) == 1)
check("enabled: only the never-seen address is sent",
body.get("emails") == ["new@x.org"])
check("enabled: a member already in Loomio is excluded (NULL record)",
body = posts[0]["json"] if posts else {}
check("add: exactly one invite call", len(posts) == 1)
check("add: only the never-seen address is invited", body.get("emails") == ["new@x.org"])
check("add: a member ALREADY in Loomio is skipped silently",
"member@x.org" not in (body.get("emails") or []))
check("enabled: remove_absent is NOT sent (no evict/re-add path)",
"remove_absent" not in body)
check("add: the record guard excludes a previously-invited address",
"old@x.org" not in (body.get("emails") or []))
check("add: remove_absent is NEVER sent (no re-add path)", "remove_absent" not in body)
check("add: the newly-invited address is recorded",
m.get_loomio_invited_at("new@x.org") is not None)
check("add: nothing removed when nobody has lapsed", deletes == [])
# 3. The newly-invited address is recorded; a second run sends nothing.
check("new@ recorded after invite", m.get_loomio_invited_at("new@x.org") is not None)
sent.clear()
# --- 2. Second run adds nobody again. ---
ROSTER = {"member@x.org": 10, "keep@x.org": 2, "new@x.org": 99} # Loomio now has new@
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
check("second run: NO Loomio call (nobody new)", sent == [])
check("double-run: no second invite", posts == [])
check("double-run: no removals", deletes == [])
# 4. Fail-closed: if Loomio's member list can't be read, nothing is sent.
m.get_loomio_invited_at("new@x.org") # no-op; keep the record
conn = m.get_db()
conn.execute("UPDATE members SET loomio_invited_at = NULL WHERE email = 'new@x.org'")
conn.commit()
conn.close()
# --- 3. REMOVE path: a lapsed member is removed exactly once. ---
ROSTER = {"member@x.org": 10, "keep@x.org": 2, "new@x.org": 99, "lapsed@x.org": 77}
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
check("remove: the lapsed member is removed exactly once",
deletes == [f"{m.LOOMIO_BASE}/api/b3/memberships/77"])
check("remove: the live active member is NOT removed",
f"{m.LOOMIO_BASE}/api/b3/memberships/10" not in deletes)
check("remove: the operator allowlist is NEVER removed (even though not active)",
f"{m.LOOMIO_BASE}/api/b3/memberships/2" not in deletes)
check("remove: no re-invite is sent while removing", posts == [])
check("remove: it does not send present members to any endpoint",
all("keep@x.org" not in (p["json"].get("emails") or []) for p in posts))
# --- 4. Second run after the removal changes nothing. ---
ROSTER = {"member@x.org": 10, "keep@x.org": 2, "new@x.org": 99} # lapsed gone
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
check("remove double-run: nothing removed again", deletes == [])
check("remove double-run: nothing invited", posts == [])
# --- 5. A re-run against a STALE roster (member already gone) must not error:
# Loomio answers 404 for an absent membership, which is success. ---
ROSTER = {"member@x.org": 10, "keep@x.org": 2, "lapsed@x.org": 77}
DELETE_STATUS = 404
posts.clear(); deletes.clear()
err = None
try:
await m.sync_loomio_memberships()
except Exception as e: # pragma: no cover
err = e
check("idempotent remove: a 404 (already absent) is treated as success", err is None)
check("idempotent remove: still no re-invite", posts == [])
DELETE_STATUS = 200
# --- 6. FAIL CLOSED: an unreadable roster stops BOTH phases. ---
ROSTER = {"lapsed@x.org": 77}
GET_FAILS = True
sent.clear()
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
check("fail-closed: no invite sent when Loomio members unreadable", sent == [])
check("fail-closed: no invite when the roster is unreadable", posts == [])
check("fail-closed: no removal when the roster is unreadable", deletes == [])
GET_FAILS = False
# 5. Recording is idempotent — an existing timestamp is never overwritten.
# --- 7. The kill switch really pauses everything. ---
m.LOOMIO_SYNC_ENABLED = False
posts.clear(); deletes.clear()
await m.sync_loomio_memberships()
check("kill switch: no calls at all when LOOMIO_SYNC_ENABLED=0",
posts == [] and deletes == [])
m.LOOMIO_SYNC_ENABLED = True
# --- 8. Recording is idempotent — an existing timestamp is never overwritten. ---
first = m.get_loomio_invited_at("old@x.org")
m.mark_loomio_invited("old@x.org")
check("mark_loomio_invited never overwrites an existing timestamp",
m.get_loomio_invited_at("old@x.org") == first)
# --- 9. remove_loomio_membership status handling (200/204/404 ok, 500 not). ---
async with FakeClient() as c:
DELETE_STATUS = 204
check("remove_loomio_membership: 204 is success", await m.remove_loomio_membership(c, 1))
DELETE_STATUS = 500
check("remove_loomio_membership: 500 is a failure", not await m.remove_loomio_membership(c, 1))
DELETE_STATUS = 200
asyncio.run(main())
fails = sum(not ok for _, ok in results)