Fix dashboard registration to use the real Lovelace API (P1-P3)

P1: The registration was a silent no-op in production — it read
hass.data["lovelace"].dashboards as the DashboardsCollection and
awaited a sync async_items(), and passed "strategy"/"mode" keys the
real create schema rejects. Per the real lovelace source
(components/lovelace/__init__.py, _create_map_dashboard): the metadata
collection lives at hass.data["lovelace"]["dashboards_collection"],
async_items() is a SYNC @callback, the create payload carries metadata
only (url_path/title/icon/show_in_sidebar/allow_single_word), and the
dashboard CONFIG (the strategy reference) is saved as a SECOND step
through the content manager registered under
hass.data["lovelace"]["dashboards"][url_path]. The strategy is saved
only for a dashboard created in this run (or whose config already
carries ours); an existing dashboard skips both steps, so a user's
dashboard config is never overwritten. Collection unavailable / save
raising still warns and continues, never crashing setup.

P2: Declare "lovelace" in manifest.json dependencies (alongside
frontend/http) so hass.data["lovelace"] is populated before setup
instead of relying on boot order.

P3: Guard the check-then-create with a per-hass asyncio.Lock so two
concurrent entry setups cannot both pass the exists-check and race on
the url_path create. Per-hass (attribute on the hass object) rather
than a bare module-level lock, which binds to the first event loop
that awaits it.

The lifecycle tests' fake is rebuilt to mirror the real API shape
(sync async_items, async async_create_item, a dashboards content-
manager dict whose entries expose async async_save), with assertions
that the strategy config was saved with the custom:nestquest-party
type, that a user's existing config is not overwritten, that a failing
save still leaves setup green, and that concurrent registrations
create the dashboard exactly once (verified to fail without the
guard).
This commit is contained in:
2026-09-23 10:47:54 -04:00
parent f668cf95b0
commit 710ef0dcf2
3 changed files with 290 additions and 51 deletions
+155 -19
View File
@@ -2,7 +2,6 @@
from __future__ import annotations
import asyncio
from types import SimpleNamespace
import pytest
@@ -16,6 +15,7 @@ from custom_components.nestquest.const import (
from custom_components.nestquest.dashboard import (
DASHBOARD_STRATEGY_TYPE,
DASHBOARD_URL_PATH,
async_register_dashboard,
)
@@ -23,29 +23,86 @@ def _wire(entry, registry):
return wire_entry_to_registry(entry, registry)
class FakeDashboardsCollection:
"""Stand-in for the Lovelace dashboards collection.
class FakeDashboardStore:
"""Stand-in for a per-dashboard CONTENT manager (LovelaceStorage).
Records ``async_create_item`` calls the way the real collection
would persist them, so tests can assert exactly what the setup path
registered — and that a reload registers nothing more.
In the real API the dashboard CONFIG (the strategy reference) is
saved through THIS manager's ``async_save``, NOT through the
metadata collection — so every strategy-save assertion reads here.
"""
def __init__(self, items=None):
def __init__(self) -> None:
self.saved: list[dict] = []
self.config: dict | None = None
async def async_save(self, config: dict) -> None:
self.saved.append(dict(config))
self.config = dict(config)
class FakeDashboardsCollection:
"""Stand-in mirroring the REAL Lovelace dashboards API shapes.
Faithful to ``homeassistant/components/lovelace`` +
``helpers/collection.py``: ``async_items()`` is a SYNC ``@callback``
(ObservableCollection.async_items returns list(data.values())),
``async_create_item`` IS async, and — mirroring lovelace's
``storage_dashboard_changed`` listener — a successful create
registers a fresh CONTENT manager under
``hass.data["lovelace"]["dashboards"][url_path]``, which is where
the dashboard config is then saved. Records ``async_create_item``
calls so tests can assert exactly what the setup path registered —
and that a reload registers nothing more.
"""
def __init__(self, hass, items=None) -> None:
self.hass = hass
self.items = list(items or [])
self.created: list[dict] = []
#: url_path -> content manager, mirroring the real listener's
# registrations (pre-existing items included).
self.stores: dict[str, FakeDashboardStore] = {}
async def async_items(self):
def async_items(self):
"""SYNC, like the real ``ObservableCollection.async_items``."""
return list(self.items)
async def async_create_item(self, item_config: dict) -> None:
# The real collection awaits validation and change notification
# around the item insertion; yield first so two concurrent
# callers without a single-flight guard would BOTH get here.
await asyncio.sleep(0)
self.created.append(dict(item_config))
self.items.append(dict(item_config))
self.stores[item_config["url_path"]] = FakeDashboardStore()
self.hass.data["lovelace"]["dashboards"][item_config["url_path"]] = (
self.stores[item_config["url_path"]]
)
def _install_dashboard_collection(hass, items=None) -> FakeDashboardsCollection:
collection = FakeDashboardsCollection(items)
hass.data["lovelace"] = SimpleNamespace(dashboards=collection)
def _install_dashboard_collection(
hass, items=None
) -> FakeDashboardsCollection:
"""Install the REAL-API-shaped lovelace surfaces on the hass stub.
Mirrors what lovelace's ``async_setup`` leaves behind in storage
mode: ``hass.data["lovelace"]`` is a DICT with the metadata
collection under ``dashboards_collection`` and the per-url_path
CONTENT managers under ``dashboards``. Any pre-existing item gets
a content manager, exactly as lovelace's change listener would
have registered one when the dashboard was created.
"""
collection = FakeDashboardsCollection(hass, items)
hass.data["lovelace"] = {
"mode": "storage",
"dashboards": {},
"dashboards_collection": collection,
}
for item in collection.items:
url_path = item.get("url_path")
store = FakeDashboardStore()
collection.stores[url_path] = store
hass.data["lovelace"]["dashboards"][url_path] = store
return collection
@@ -305,52 +362,76 @@ async def test_unload_shuts_coordinator_even_when_listener_removal_raises() -> N
async def test_setup_creates_the_panel_dashboard(hass, make_entry) -> None:
"""A fresh install gets the storage-mode strategy dashboard, with the
stable url_path, a title/icon, sidebar visibility and no YAML."""
"""A fresh install gets the storage-mode strategy dashboard via the
REAL two-step Lovelace sequence: the METADATA (url_path, title,
icon, sidebar, single-word allowance — nothing else: the real
create schema rejects unknown keys like ``strategy``) through the
collection, then the dashboard CONFIG (the strategy reference)
through the fresh dashboard's content manager."""
collection = _install_dashboard_collection(hass)
entry = _wire(make_entry(), hass.registry)
assert await async_setup_entry(hass, entry) is True
# Step 1: the metadata create through the collection.
assert len(collection.created) == 1
created = collection.created[0]
assert created["url_path"] == DASHBOARD_URL_PATH
assert created["mode"] == "storage"
assert created["strategy"] == {"type": DASHBOARD_STRATEGY_TYPE}
assert created["title"]
assert created["show_in_sidebar"] is True
assert created["allow_single_word"] is True
assert "icon" in created
# The create payload carries METADATA only: a ``strategy`` key is
# rejected by the real collection's create schema (the silent
# no-op the first implementation shipped).
assert "strategy" not in created
assert "mode" not in created
# Step 2: the strategy CONFIG saved through the content manager
# the create registered, with the custom strategy type.
store = hass.data["lovelace"]["dashboards"][DASHBOARD_URL_PATH]
assert store is collection.stores[DASHBOARD_URL_PATH]
assert store.saved == [{"strategy": {"type": DASHBOARD_STRATEGY_TYPE}}]
assert store.config == {"strategy": {"type": DASHBOARD_STRATEGY_TYPE}}
async def test_setup_reload_does_not_duplicate_the_dashboard(
hass, make_entry
) -> None:
"""A reload (or a second entry) must not create a second dashboard:
the one under the stable url_path is reused as-is."""
the one under the stable url_path is reused as-is — BOTH steps are
skipped, so the strategy config is not re-saved either."""
collection = _install_dashboard_collection(hass)
entry = _wire(make_entry(), hass.registry)
assert await async_setup_entry(hass, entry) is True
assert len(collection.created) == 1
store = hass.data["lovelace"]["dashboards"][DASHBOARD_URL_PATH]
assert len(store.saved) == 1
# A reload: setup again on the same hass (the reload cycle) —
# the existing dashboard must be left alone.
await async_unload_entry(hass, entry)
assert await async_setup_entry(hass, entry) is True
assert len(collection.created) == 1
assert len(store.saved) == 1
# And a second entry on the same install.
other = _wire(make_entry(entry_id="other"), hass.registry)
assert await async_setup_entry(hass, other) is True
assert len(collection.created) == 1
assert len(store.saved) == 1
async def test_setup_leaves_an_existing_user_dashboard_alone(
hass, make_entry
) -> None:
"""A dashboard already occupying the stable url_path — the user's
own — is never overwritten, deleted or duplicated."""
own — is never overwritten, deleted or duplicated: no metadata
create AND no strategy save against its content manager."""
users_dashboard = {"url_path": DASHBOARD_URL_PATH, "title": "My Board"}
collection = _install_dashboard_collection(hass, items=[users_dashboard])
entry = _wire(make_entry(), hass.registry)
assert await async_setup_entry(hass, entry) is True
assert collection.created == []
assert collection.items[0] is users_dashboard
users_store = hass.data["lovelace"]["dashboards"][DASHBOARD_URL_PATH]
assert users_store.saved == []
assert users_store.config is None
async def test_setup_succeeds_without_the_dashboards_collection(
@@ -376,8 +457,63 @@ async def test_setup_survives_a_failing_dashboards_collection(
async def async_create_item(self, item_config):
raise PermissionError("not an admin")
collection = _RefusingCollection()
hass.data["lovelace"] = SimpleNamespace(dashboards=collection)
collection = _RefusingCollection(hass)
hass.data["lovelace"] = {
"mode": "storage",
"dashboards": {},
"dashboards_collection": collection,
}
entry = _wire(make_entry(), hass.registry)
assert await async_setup_entry(hass, entry) is True
assert collection.created == []
async def test_setup_survives_a_failing_content_manager_save(
hass, make_entry
) -> None:
"""A content manager that refuses the config save after the
metadata create succeeded (storage write failure) logs a warning
and setup still succeeds — the dashboard exists, the strategy
simply was not persisted through it."""
class _RefusingStore(FakeDashboardStore):
async def async_save(self, config):
raise OSError("disk full")
class _StorelessCreateCollection(FakeDashboardsCollection):
async def async_create_item(self, item_config):
await asyncio.sleep(0)
self.created.append(dict(item_config))
self.items.append(dict(item_config))
self.stores[item_config["url_path"]] = _RefusingStore()
self.hass.data["lovelace"]["dashboards"][
item_config["url_path"]
] = self.stores[item_config["url_path"]]
collection = _StorelessCreateCollection(hass)
hass.data["lovelace"] = {
"mode": "storage",
"dashboards": {},
"dashboards_collection": collection,
}
entry = _wire(make_entry(), hass.registry)
assert await async_setup_entry(hass, entry) is True
assert len(collection.created) == 1
store = hass.data["lovelace"]["dashboards"][DASHBOARD_URL_PATH]
assert store.saved == []
async def test_concurrent_registrations_create_the_dashboard_once(hass) -> None:
"""Two CONCURRENT registrations (the race two entry setups run)
must produce exactly ONE dashboard: the single-flight guard makes
the second caller's exists-check run after the first caller's
create completed, so it skips both steps. The fake's create
yields before recording, so without the guard both callers would
pass the exists-check and the dashboard would be created twice."""
collection = _install_dashboard_collection(hass)
results = await asyncio.gather(
async_register_dashboard(hass),
async_register_dashboard(hass),
)
assert results == [True, True]
assert len(collection.created) == 1
assert len(collection.stores[DASHBOARD_URL_PATH].saved) == 1