fix: size the write-settle window to outlast its own confirm poll (issues #17, #53)

mark_write_pending's window was a fixed few seconds, started before the
PUT and the confirming /device/0 refresh that follows it -- but that
refresh is a full summary poll, which can legitimately take far longer
than that on these AC devices. The window routinely expired while the
confirm poll was still in flight, so a stale read (the device's own
resource tree hadn't caught up to the instant physical change yet)
landed unprotected and reverted the optimistic write, with nothing to
correct it again until the next scheduled summary poll. That's the
20-60s lag both issues report even after the earlier optimistic-apply
fix.

Size the window to cover the PUT and confirm-poll timeouts combined,
and release it as soon as that round trip actually completes (success
or failure) instead of leaving it open for the rest of a now much
longer window.
This commit is contained in:
Marc Billow committed 2026-07-23 22:11:26 +00:00
1 parent 30e814f242
commit 615c5d5aa9
3 files changed
+95 -16

No files matched your search

+44 -13
View File
@@ -643,7 +643,11 @@ async def test_write_marks_href_pending_before_post(
hass: HomeAssistant, mock_entry, mock_coordinator_observe_session
) -> None:
"""async_send_command must call mark_write_pending before POSTing so a
slow-to-settle device can't immediately revert the optimistic write."""
slow-to-settle device can't immediately revert the optimistic write --
and the window must still be open through the POST itself (checked from
inside the fake POST, since the guard is released once the write's own
confirming refresh completes -- see
test_send_command_applies_write_optimistically_before_settling)."""
from custom_components.localthings.registry.discovery import BoundEntity
from custom_components.localthings.registry.entities import NumberDesc
@@ -658,21 +662,39 @@ async def test_write_marks_href_pending_before_post(
desc = NumberDesc(key='test', field='value', write_fn=_write_fn)
bound = BoundEntity(href='/test/vs/0', capability=coordinator.bound[0].capability, desc=desc)
settling_during_post = None
def _post(*a, **k):
nonlocal settling_during_post
settling_during_post = coordinator._observe._settle_until.get('/test/vs/0') is not None
return (0x44, b'')
with patch.object(fake, 'subscribe'):
fake.post = lambda *a, **k: (0x44, b'')
fake.post = _post
await coordinator.async_send_command(bound, 5)
assert coordinator._observe._settle_until.get('/test/vs/0') is not None
assert settling_during_post is True
# Released once the write's own confirming refresh has completed, not
# left open for the rest of what is now a much longer settle window.
assert coordinator._observe._settle_until.get('/test/vs/0') is None
async def test_send_command_applies_write_optimistically_before_settling(
hass: HomeAssistant, mock_entry, mock_coordinator_observe_session
hass: HomeAssistant, mock_entry, mock_coordinator_observe_session, fridge_resources,
) -> None:
"""The cache must reflect a write immediately, and stay put against a
stale echo for the rest of the settle window -- otherwise
mark_write_pending has nothing to protect and just delays the real
device confirmation instead, which reads to a user as the write being
silently reverted (issue #27)."""
silently reverted (issue #27).
The settle window now spans the whole POST + confirming-refresh round
trip async_send_command awaits (see coordinator._POST_TIMEOUT_S /
_POLL_TIMEOUT_S), not a fixed few seconds -- so this drives the stale
echo through _poll_once itself, racing in while that confirming refresh
is still in flight (issues #17/#53: a full summary poll on these
devices can legitimately take far longer than a short fixed window,
which let a stale read land unprotected and revert the write)."""
from custom_components.localthings.registry.discovery import BoundEntity
from custom_components.localthings.registry.entities import NumberDesc
@@ -687,20 +709,29 @@ async def test_send_command_applies_write_optimistically_before_settling(
desc = NumberDesc(key='test', field='value', write_fn=_write_fn)
bound = BoundEntity(href='/test/vs/0', capability=coordinator.bound[0].capability, desc=desc)
with patch.object(fake, 'subscribe'):
def _stale_confirm_poll():
# Stand-in for the write's own confirming /device/0 poll racing
# against an independent stale update for the same href (e.g. the
# device's resource tree hadn't caught up internally yet, even
# though the physical change was instant).
coordinator._observe.apply('/test/vs/0', {'value': 0}, source='poll')
return fridge_resources
with (
patch.object(fake, 'subscribe'),
patch.object(LocalThingsCoordinator, '_poll_once', side_effect=_stale_confirm_poll),
):
fake.post = lambda *a, **k: (0x44, b'')
await coordinator.async_send_command(bound, 5)
# The optimistic value is visible right away, without waiting for a
# device response.
# device response, and survived the stale confirm-poll race above.
assert coordinator._cache.get('/test/vs/0') == {'value': 5}
# A stale update racing in behind the write (e.g. a poll/notify that
# was already in flight before the PUT) must not clobber it while the
# settle window is open.
# The write's own round trip is done, so the guard has been released --
# a later, genuinely new update is no longer held back.
applied = coordinator._observe.apply('/test/vs/0', {'value': 0}, source='observe')
assert applied is False
assert coordinator._cache.get('/test/vs/0') == {'value': 5}
assert applied is True
class TestRemoteControlEnabled:
@@ -803,7 +834,7 @@ async def test_send_command_allowed_when_remote_control_enabled(
fake.post = lambda *a, **k: (0x44, b'')
await coordinator.async_send_command(bound, 5)
assert coordinator._observe._settle_until.get('/test/vs/0') is not None
assert coordinator._cache.get('/test/vs/0') == {'value': 5}
async def test_send_command_remote_control_check_precedes_validate_fn(