Merge remote-tracking branch 'origin/dev' into integration

This commit is contained in:
J. Nick Koston
2026-07-18 10:21:31 -10:00
466 changed files with 14801 additions and 3978 deletions
@@ -0,0 +1,30 @@
esphome:
name: test-ha-action-no-subscriber
friendly_name: Home Assistant Action No Subscriber Test
on_boot:
# Fires before any client is connected - dropped with a warning.
- homeassistant.action:
action: test.boot_action
host:
api:
on_client_connected:
# Fires at authentication time, before the client has subscribed to
# Home Assistant actions - dropped with a warning.
- homeassistant.action:
action: test.connected_action
logger:
level: DEBUG
button:
- platform: template
name: Send Action Button
id: send_action_button
on_press:
# Pressed only after the client has subscribed - must be delivered.
- homeassistant.action:
action: test.button_action
data:
value: subscribed
@@ -0,0 +1,6 @@
esphome:
name: zero-psk-provision-test
host:
api:
encryption:
logger:
@@ -0,0 +1,6 @@
esphome:
name: zero-psk-plaintext-test
host:
api:
encryption:
logger:
@@ -18,9 +18,6 @@ globals:
- id: interval_counter
type: int
initial_value: '0'
- id: retry_counter
type: int
initial_value: '0'
- id: defer_counter
type: int
initial_value: '0'
@@ -118,29 +115,7 @@ script:
id(timeout_counter) += 1;
});
// Test 10: set_retry with numeric ID
App.scheduler.set_retry(component1, 6001U, 50, 3,
[](uint8_t retry_countdown) {
id(retry_counter)++;
ESP_LOGI("test", "Numeric retry 6001 attempt %d (countdown=%d)",
id(retry_counter), retry_countdown);
if (id(retry_counter) >= 2) {
ESP_LOGI("test", "Numeric retry 6001 done");
return RetryResult::DONE;
}
return RetryResult::RETRY;
});
// Test 11: cancel_retry with numeric ID
App.scheduler.set_retry(component1, 6002U, 100, 5,
[](uint8_t retry_countdown) {
ESP_LOGE("test", "ERROR: Numeric retry 6002 should have been cancelled");
return RetryResult::RETRY;
});
App.scheduler.cancel_retry(component1, 6002U);
ESP_LOGI("test", "Cancelled numeric retry 6002");
// Test 12: defer with numeric ID (Component method)
// Test 10: defer with numeric ID (Component method)
class TestDeferComponent : public Component {
public:
void test_defer_methods() {
@@ -161,7 +136,7 @@ script:
static TestDeferComponent test_defer_component;
test_defer_component.test_defer_methods();
// Test 13: cancel_defer with numeric ID (Component method)
// Test 11: cancel_defer with numeric ID (Component method)
class TestCancelDeferComponent : public Component {
public:
void test_cancel_defer() {
@@ -181,8 +156,8 @@ script:
- id: report_results
then:
- lambda: |-
ESP_LOGI("test", "Final results - Timeouts: %d, Intervals: %d, Retries: %d, Defers: %d",
id(timeout_counter), id(interval_counter), id(retry_counter), id(defer_counter));
ESP_LOGI("test", "Final results - Timeouts: %d, Intervals: %d, Defers: %d",
id(timeout_counter), id(interval_counter), id(defer_counter));
sensor:
- platform: template
@@ -1,287 +0,0 @@
esphome:
debug_scheduler: true # Enable scheduler leak detection
name: scheduler-retry-test
on_boot:
priority: -100
then:
- logger.log: "Starting scheduler retry tests"
# Run all tests sequentially with delays
- script.execute: run_all_tests
host:
api:
logger:
level: VERY_VERBOSE
globals:
- id: simple_retry_counter
type: int
initial_value: '0'
- id: backoff_retry_counter
type: int
initial_value: '0'
- id: backoff_last_attempt_time
type: uint32_t
initial_value: '0'
- id: immediate_done_counter
type: int
initial_value: '0'
- id: cancel_retry_counter
type: int
initial_value: '0'
- id: empty_name_retry_counter
type: int
initial_value: '0'
- id: script_retry_counter
type: int
initial_value: '0'
- id: multiple_same_name_counter
type: int
initial_value: '0'
- id: const_char_retry_counter
type: int
initial_value: '0'
- id: static_char_retry_counter
type: int
initial_value: '0'
# Using different component types for each test to ensure isolation
sensor:
- platform: template
name: Simple Retry Test Sensor
id: simple_retry_sensor
lambda: return 1.0;
update_interval: never
- platform: template
name: Backoff Retry Test Sensor
id: backoff_retry_sensor
lambda: return 2.0;
update_interval: never
- platform: template
name: Immediate Done Test Sensor
id: immediate_done_sensor
lambda: return 3.0;
update_interval: never
binary_sensor:
- platform: template
name: Cancel Retry Test Binary Sensor
id: cancel_retry_binary_sensor
lambda: return false;
- platform: template
name: Empty Name Test Binary Sensor
id: empty_name_binary_sensor
lambda: return true;
switch:
- platform: template
name: Script Retry Test Switch
id: script_retry_switch
optimistic: true
- platform: template
name: Multiple Same Name Test Switch
id: multiple_same_name_switch
optimistic: true
script:
- id: run_all_tests
then:
# Test 1: Simple retry
- logger.log: "=== Test 1: Simple retry ==="
- lambda: |-
auto *component = id(simple_retry_sensor);
App.scheduler.set_retry(component, "simple_retry", 50, 3,
[](uint8_t retry_countdown) {
id(simple_retry_counter)++;
ESP_LOGI("test", "Simple retry attempt %d (countdown=%d)",
id(simple_retry_counter), retry_countdown);
if (id(simple_retry_counter) >= 2) {
ESP_LOGI("test", "Simple retry succeeded on attempt %d", id(simple_retry_counter));
return RetryResult::DONE;
}
return RetryResult::RETRY;
});
# Test 2: Backoff retry
- logger.log: "=== Test 2: Retry with backoff ==="
- lambda: |-
auto *component = id(backoff_retry_sensor);
App.scheduler.set_retry(component, "backoff_retry", 50, 4,
[](uint8_t retry_countdown) {
id(backoff_retry_counter)++;
uint32_t now = millis();
uint32_t interval = 0;
// Only calculate interval after first attempt
if (id(backoff_retry_counter) > 1) {
interval = now - id(backoff_last_attempt_time);
}
id(backoff_last_attempt_time) = now;
ESP_LOGI("test", "Backoff retry attempt %d (countdown=%d, interval=%dms)",
id(backoff_retry_counter), retry_countdown, interval);
if (id(backoff_retry_counter) == 1) {
ESP_LOGI("test", "First call was immediate");
} else if (id(backoff_retry_counter) == 2) {
ESP_LOGI("test", "Second call interval: %dms (expected ~50ms)", interval);
} else if (id(backoff_retry_counter) == 3) {
ESP_LOGI("test", "Third call interval: %dms (expected ~100ms)", interval);
} else if (id(backoff_retry_counter) == 4) {
ESP_LOGI("test", "Fourth call interval: %dms (expected ~200ms)", interval);
ESP_LOGI("test", "Backoff retry completed");
return RetryResult::DONE;
}
return RetryResult::RETRY;
}, 2.0f);
# Test 3: Immediate done
- logger.log: "=== Test 3: Immediate done ==="
- lambda: |-
auto *component = id(immediate_done_sensor);
App.scheduler.set_retry(component, "immediate_done", 50, 5,
[](uint8_t retry_countdown) {
id(immediate_done_counter)++;
ESP_LOGI("test", "Immediate done retry called (countdown=%d)", retry_countdown);
return RetryResult::DONE;
});
# Test 4: Cancel retry
- logger.log: "=== Test 4: Cancel retry ==="
- lambda: |-
auto *component = id(cancel_retry_binary_sensor);
App.scheduler.set_retry(component, "cancel_test", 30, 10,
[](uint8_t retry_countdown) {
id(cancel_retry_counter)++;
ESP_LOGI("test", "Cancel test retry attempt %d", id(cancel_retry_counter));
return RetryResult::RETRY;
});
// Cancel it after 100ms
App.scheduler.set_timeout(component, "cancel_timer", 100, []() {
bool cancelled = App.scheduler.cancel_retry(id(cancel_retry_binary_sensor), "cancel_test");
ESP_LOGI("test", "Retry cancellation result: %s", cancelled ? "true" : "false");
ESP_LOGI("test", "Cancel retry ran %d times before cancellation", id(cancel_retry_counter));
});
# Test 5: Empty name retry
- logger.log: "=== Test 5: Empty name retry ==="
- lambda: |-
auto *component = id(empty_name_binary_sensor);
App.scheduler.set_retry(component, "", 100, 5,
[](uint8_t retry_countdown) {
id(empty_name_retry_counter)++;
ESP_LOGI("test", "Empty name retry attempt %d", id(empty_name_retry_counter));
return RetryResult::RETRY;
});
// Try to cancel after 150ms
App.scheduler.set_timeout(component, "empty_cancel_timer", 150, []() {
bool cancelled = App.scheduler.cancel_retry(id(empty_name_binary_sensor), "");
ESP_LOGI("test", "Empty name retry cancel result: %s",
cancelled ? "true" : "false");
ESP_LOGI("test", "Empty name retry ran %d times", id(empty_name_retry_counter));
});
# Test 6: Component method
- logger.log: "=== Test 6: Component::set_retry method ==="
- lambda: |-
class TestRetryComponent : public Component {
public:
void test_retry() {
this->set_retry(50, 3,
[](uint8_t retry_countdown) {
id(script_retry_counter)++;
ESP_LOGI("test", "Component retry attempt %d", id(script_retry_counter));
if (id(script_retry_counter) >= 2) {
return RetryResult::DONE;
}
return RetryResult::RETRY;
}, 1.5f);
}
};
static TestRetryComponent test_component;
test_component.test_retry();
# Test 7: Multiple same name
- logger.log: "=== Test 7: Multiple retries with same name ==="
- lambda: |-
auto *component = id(multiple_same_name_switch);
// Set first retry
App.scheduler.set_retry(component, "duplicate_retry", 100, 5,
[](uint8_t retry_countdown) {
id(multiple_same_name_counter) += 1;
ESP_LOGI("test", "First duplicate retry - should not run");
return RetryResult::RETRY;
});
// Set second retry with same name (should cancel first)
App.scheduler.set_retry(component, "duplicate_retry", 50, 3,
[](uint8_t retry_countdown) {
id(multiple_same_name_counter) += 10;
ESP_LOGI("test", "Second duplicate retry attempt (counter=%d)",
id(multiple_same_name_counter));
if (id(multiple_same_name_counter) >= 20) {
return RetryResult::DONE;
}
return RetryResult::RETRY;
});
# Test 8: Const char* overloads
- logger.log: "=== Test 8: Const char* overloads ==="
- lambda: |-
auto *component = id(simple_retry_sensor);
// Test 8a: Direct string literal
App.scheduler.set_retry(component, "const_char_test", 30, 2,
[](uint8_t retry_countdown) {
id(const_char_retry_counter)++;
ESP_LOGI("test", "Const char retry %d", id(const_char_retry_counter));
return RetryResult::DONE;
});
# Test 9: Static const char* variable
- logger.log: "=== Test 9: Static const char* ==="
- lambda: |-
auto *component = id(backoff_retry_sensor);
static const char* STATIC_NAME = "static_retry_test";
App.scheduler.set_retry(component, STATIC_NAME, 20, 1,
[](uint8_t retry_countdown) {
id(static_char_retry_counter)++;
ESP_LOGI("test", "Static const char retry %d", id(static_char_retry_counter));
return RetryResult::DONE;
});
// Cancel with same static const char*
App.scheduler.set_timeout(component, "static_cancel", 10, []() {
static const char* STATIC_NAME = "static_retry_test";
bool result = App.scheduler.cancel_retry(id(backoff_retry_sensor), STATIC_NAME);
ESP_LOGI("test", "Static cancel result: %s", result ? "true" : "false");
});
# Wait for all tests to complete before reporting
- delay: 500ms
# Final report
- logger.log: "=== Retry Test Results ==="
- lambda: |-
ESP_LOGI("test", "Simple retry counter: %d (expected 2)", id(simple_retry_counter));
ESP_LOGI("test", "Backoff retry counter: %d (expected 4)", id(backoff_retry_counter));
ESP_LOGI("test", "Immediate done counter: %d (expected 1)", id(immediate_done_counter));
ESP_LOGI("test", "Cancel retry counter: %d (expected 2-4)", id(cancel_retry_counter));
ESP_LOGI("test", "Empty name retry counter: %d (expected 1-2)", id(empty_name_retry_counter));
ESP_LOGI("test", "Component retry counter: %d (expected 2)", id(script_retry_counter));
ESP_LOGI("test", "Multiple same name counter: %d (expected 20+)", id(multiple_same_name_counter));
ESP_LOGI("test", "Const char retry counter: %d (expected 1)", id(const_char_retry_counter));
ESP_LOGI("test", "Static char retry counter: %d (expected 1)", id(static_char_retry_counter));
ESP_LOGI("test", "All retry tests completed");
@@ -0,0 +1,86 @@
esphome:
name: wait-until-reentrant-restart
host:
api:
actions:
- action: start_self_restart
then:
- script.execute: retry_script
- action: start_stop_during_wait
then:
- globals.set:
id: gate_open
value: 'false'
# num 0 is a blocker: its condition never becomes true, so it is still
# waiting (already checked and set aside) when num 1 stops the script -
# it must be cancelled, not restored, so its timeout must never fire
- script.execute:
id: waiter
num: 0
- script.execute:
id: waiter
num: 1
- script.execute:
id: waiter
num: 2
- script.execute:
id: waiter
num: 3
# Give all three instances time to queue in the same wait_until
- delay: 100ms
- globals.set:
id: gate_open
value: 'true'
- delay: 200ms
- logger.log: "stop test complete"
logger:
level: DEBUG
globals:
- id: attempt
type: int
initial_value: '0'
- id: gate_open
type: bool
initial_value: 'false'
script:
# Self-restart retry pattern: when the wait_until times out, the rest of the
# script runs synchronously from inside the wait queue processing and restarts
# this same script - re-entering the same WaitUntilAction while it is still
# processing its queue. This used to corrupt the queue and crash.
- id: retry_script
mode: restart
then:
- wait_until:
condition:
lambda: 'return false;'
timeout: 20ms
- lambda: |-
id(attempt) += 1;
ESP_LOGD("test", "attempt %d done", id(attempt));
- if:
condition:
lambda: 'return id(attempt) < 5;'
then:
- script.execute: retry_script
else:
- logger.log: "retry test complete"
# Parallel waiters all queued in the same wait_until; the first one to pass the
# gate stops the script from its continuation, cancelling the other waiters
# while the queue is still being processed.
- id: waiter
mode: parallel
parameters:
num: int
then:
- wait_until:
condition:
lambda: 'return num != 0 && id(gate_open);'
timeout: 1s
- lambda: 'ESP_LOGD("test", "gate passed %d", num);'
- script.stop: waiter
@@ -0,0 +1,79 @@
"""Integration test for Home Assistant actions fired without a subscriber.
Home Assistant subscribes to device actions shortly after authenticating, while
on_client_connected (and similar triggers) fire right at authentication. Actions
fired before any client has subscribed cannot be delivered - they must produce a
warning in the log instead of vanishing silently.
"""
from __future__ import annotations
import asyncio
from aioesphomeapi import ButtonInfo, HomeassistantServiceCall
import pytest
from .types import APIClientConnectedFactory, RunCompiledFunction
@pytest.mark.asyncio
async def test_api_homeassistant_action_no_subscriber(
yaml_config: str,
run_compiled: RunCompiledFunction,
api_client_connected: APIClientConnectedFactory,
) -> None:
"""Undeliverable actions warn in the log; actions after subscribing arrive."""
loop = asyncio.get_running_loop()
boot_warning_future = loop.create_future()
connected_warning_future = loop.create_future()
button_action_future = loop.create_future()
def check_output(line: str) -> None:
if (
not boot_warning_future.done()
and "Home Assistant action 'test.boot_action' dropped; no client connected"
in line
):
boot_warning_future.set_result(True)
if (
not connected_warning_future.done()
and "Home Assistant action 'test.connected_action' dropped; "
"client has not subscribed to actions (yet)"
in line
):
connected_warning_future.set_result(True)
service_calls: list[HomeassistantServiceCall] = []
def on_service_call(service_call: HomeassistantServiceCall) -> None:
service_calls.append(service_call)
if (
service_call.service == "test.button_action"
and not button_action_future.done()
):
button_action_future.set_result(service_call)
async with run_compiled(yaml_config, line_callback=check_output):
# The on_boot action fires with no client connected at all.
await asyncio.wait_for(boot_warning_future, timeout=10.0)
async with api_client_connected() as client:
device_info = await client.device_info()
assert device_info.name == "test-ha-action-no-subscriber"
# on_client_connected fired at authentication, before this client
# subscribed to Home Assistant actions.
await asyncio.wait_for(connected_warning_future, timeout=5.0)
# After subscribing, actions must be delivered normally (and the
# dropped ones must not suddenly show up).
client.subscribe_service_calls(on_service_call)
entities, _ = await client.list_entities_services()
button = next(e for e in entities if isinstance(e, ButtonInfo))
client.button_command(button.key)
button_call = await asyncio.wait_for(button_action_future, timeout=5.0)
assert button_call.data == {"value": "subscribed"}
assert [call.service for call in service_calls] == ["test.button_action"]
@@ -0,0 +1,127 @@
"""Integration tests for provisioning the encryption key over a zero-PSK connection.
A device with `api: encryption:` but no key accepts Noise handshakes using the
well-known all-zeros PSK. The ephemeral X25519 exchange protects the key from
passive sniffing while it is provisioned; plaintext provisioning still works
but is deprecated.
"""
from __future__ import annotations
import asyncio
import base64
from aioesphomeapi import InvalidEncryptionKeyAPIError, RequiresEncryptionAPIError
import pytest
from .types import APIClientConnectedFactory, RunCompiledFunction
# The well-known provisioning PSK: base64 of 32 zero bytes
ZERO_PSK = base64.b64encode(bytes(32)).decode()
# A real key to provision
NEW_KEY = base64.b64encode(b"n" * 32)
# Time for the device to activate a newly saved key (100ms timer plus margin)
KEY_ACTIVATION_DELAY = 0.5
@pytest.fixture(autouse=True)
def isolated_preferences(monkeypatch: pytest.MonkeyPatch, tmp_path) -> None:
"""Keep host preferences per-test so every run starts unprovisioned."""
monkeypatch.setenv("ESPHOME_PREFDIR", str(tmp_path / "prefs"))
@pytest.mark.asyncio
async def test_api_zero_psk_provisioning(
yaml_config: str,
run_compiled: RunCompiledFunction,
api_client_connected: APIClientConnectedFactory,
) -> None:
"""Exercise the reject paths, then provision a key over the zero-PSK channel."""
async with run_compiled(yaml_config):
# --- Pre-provisioning reject paths (device state is unchanged) ---
# A wrong (non-zero) PSK fails against the zero provisioning PSK
with pytest.raises(InvalidEncryptionKeyAPIError):
async with api_client_connected(
noise_psk=base64.b64encode(b"w" * 32).decode(), timeout=5
) as client:
await client.device_info()
# A plaintext client and a zero-PSK client can be connected at the
# same time while the device is unprovisioned
async with (
api_client_connected() as plaintext_client,
api_client_connected(noise_psk=ZERO_PSK) as noise_client,
):
plaintext_info = await plaintext_client.device_info()
noise_info = await noise_client.device_info()
# Both transports advertise provisioning support so old and new
# clients can decide how to provision
assert plaintext_info.api_encryption_provisionable is True
assert noise_info.api_encryption_provisionable is True
# The all-zeros key is reserved as the provisioning PSK and is
# rejected on both transports
zero_key = base64.b64encode(bytes(32))
assert await noise_client.noise_encryption_set_key(zero_key) is False
assert await plaintext_client.noise_encryption_set_key(zero_key) is False
# --- Provision over the zero-PSK channel ---
# The unprovisioned device accepts the all-zeros PSK; the handshake's
# ephemeral-ephemeral DH encrypts everything that follows
async with api_client_connected(noise_psk=ZERO_PSK) as client:
device_info = await client.device_info()
assert device_info.name == "zero-psk-provision-test"
assert device_info.api_encryption_supported is True
assert device_info.api_encryption_provisionable is True
assert await client.noise_encryption_set_key(NEW_KEY) is True
# The device activates the new key shortly after responding
await asyncio.sleep(KEY_ACTIVATION_DELAY)
# The new key now works, and the device is no longer provisionable
async with api_client_connected(noise_psk=NEW_KEY.decode()) as client:
device_info = await client.device_info()
assert device_info.name == "zero-psk-provision-test"
assert device_info.api_encryption_provisionable is False
# The zero PSK no longer works
with pytest.raises(InvalidEncryptionKeyAPIError):
async with api_client_connected(noise_psk=ZERO_PSK, timeout=5) as client:
await client.device_info()
# Plaintext no longer works
with pytest.raises(RequiresEncryptionAPIError):
async with api_client_connected(timeout=5) as client:
await client.device_info()
@pytest.mark.asyncio
async def test_api_zero_psk_provisioning_plaintext(
yaml_config: str,
run_compiled: RunCompiledFunction,
api_client_connected: APIClientConnectedFactory,
) -> None:
"""The legacy plaintext provisioning path still works and warns."""
log_lines: list[str] = []
async with run_compiled(yaml_config, line_callback=log_lines.append):
async with api_client_connected() as client:
device_info = await client.device_info()
assert device_info.name == "zero-psk-plaintext-test"
assert await client.noise_encryption_set_key(NEW_KEY) is True
await asyncio.sleep(KEY_ACTIVATION_DELAY)
# The deprecation warning was logged
assert any("deprecated" in line for line in log_lines)
# The new key works; the zero PSK does not
async with api_client_connected(noise_psk=NEW_KEY.decode()) as client:
assert (await client.device_info()).name == "zero-psk-plaintext-test"
with pytest.raises(InvalidEncryptionKeyAPIError):
async with api_client_connected(noise_psk=ZERO_PSK, timeout=5) as client:
await client.device_info()
+43
View File
@@ -322,6 +322,49 @@ async def test_light_calls(
assert state.state is True
assert state.brightness == pytest.approx(0.75)
# Test 31: Setting brightness to 0 without an explicit state implicitly turns
# the light off; turning it back on (without an explicit brightness) then
# restores full brightness so the light is visible again.
client.light_command(key=rgbcw_light.key, state=True, brightness=0.5)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is True
assert state.brightness == pytest.approx(0.5)
# Brightness 0 with no explicit state -> implicit turn-off
client.light_command(key=rgbcw_light.key, brightness=0.0)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is False
assert state.brightness == pytest.approx(0.0)
# Turning on without an explicit brightness restores it to full brightness
client.light_command(key=rgbcw_light.key, state=True)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is True
assert state.brightness == pytest.approx(1.0)
# Test 31b: An explicit turn-on with brightness 0 still resets to full
# brightness - a turn-on must never leave the light on-but-invisible. This
# is the same path the restore logic exercises (set_state(true) +
# set_brightness(0) from a persisted brightness=0 turn-off).
client.light_command(key=rgbcw_light.key, state=True, brightness=0.0)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is True
assert state.brightness == pytest.approx(1.0)
# Test 32: Turning a light on when it already has nonzero brightness leaves
# the brightness unchanged (the reset only happens when brightness is 0).
client.light_command(key=rgbcw_light.key, state=True, brightness=0.4)
state = await wait_for_state_change(rgbcw_light.key)
assert state.brightness == pytest.approx(0.4)
client.light_command(key=rgbcw_light.key, state=False)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is False
client.light_command(key=rgbcw_light.key, state=True)
state = await wait_for_state_change(rgbcw_light.key)
assert state.state is True
assert state.brightness == pytest.approx(0.4)
# Final cleanup - turn all lights off
for light in lights:
client.light_command(
@@ -18,7 +18,6 @@ async def test_scheduler_numeric_id_test(
# Track counts
timeout_count = 0
interval_count = 0
retry_count = 0
defer_count = 0
# Events for each test completion
@@ -32,8 +31,6 @@ async def test_scheduler_numeric_id_test(
component_interval_fired = asyncio.Event()
zero_id_timeout_fired = asyncio.Event()
max_id_timeout_fired = asyncio.Event()
numeric_retry_done = asyncio.Event()
numeric_retry_cancelled = asyncio.Event()
numeric_defer_7001_fired = asyncio.Event()
numeric_defer_7002_fired = asyncio.Event()
numeric_defer_cancelled = asyncio.Event()
@@ -41,11 +38,10 @@ async def test_scheduler_numeric_id_test(
# Track interval counts
numeric_interval_count = 0
numeric_retry_count = 0
def on_log_line(line: str) -> None:
nonlocal timeout_count, interval_count, retry_count, defer_count
nonlocal numeric_interval_count, numeric_retry_count
nonlocal timeout_count, interval_count, defer_count
nonlocal numeric_interval_count
# Strip ANSI color codes
clean_line = re.sub(r"\x1b\[[0-9;]*m", "", line)
@@ -97,18 +93,6 @@ async def test_scheduler_numeric_id_test(
max_id_timeout_fired.set()
timeout_count += 1
# Check for numeric retry tests
elif "Numeric retry 6001 attempt" in clean_line:
match = re.search(r"attempt (\d+)", clean_line)
if match:
numeric_retry_count = int(match.group(1))
elif "Numeric retry 6001 done" in clean_line:
numeric_retry_done.set()
elif "Cancelled numeric retry 6002" in clean_line:
numeric_retry_cancelled.set()
# Check for numeric defer tests
elif "Component numeric defer 7001 fired" in clean_line:
numeric_defer_7001_fired.set()
@@ -122,14 +106,13 @@ async def test_scheduler_numeric_id_test(
# Check for final results
elif "Final results" in clean_line:
match = re.search(
r"Timeouts: (\d+), Intervals: (\d+), Retries: (\d+), Defers: (\d+)",
r"Timeouts: (\d+), Intervals: (\d+), Defers: (\d+)",
clean_line,
)
if match:
timeout_count = int(match.group(1))
interval_count = int(match.group(2))
retry_count = int(match.group(3))
defer_count = int(match.group(4))
defer_count = int(match.group(3))
final_results_logged.set()
async with (
@@ -200,23 +183,6 @@ async def test_scheduler_numeric_id_test(
except TimeoutError:
pytest.fail("Max ID timeout did not fire within 0.5 seconds")
# Wait for numeric retry tests
try:
await asyncio.wait_for(numeric_retry_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Numeric retry 6001 did not complete. Count: {numeric_retry_count}"
)
assert numeric_retry_count >= 2, (
f"Expected at least 2 numeric retry attempts, got {numeric_retry_count}"
)
# Verify numeric retry was cancelled
assert numeric_retry_cancelled.is_set(), (
"Numeric retry 6002 should have been cancelled"
)
# Wait for numeric defer tests
try:
await asyncio.wait_for(numeric_defer_7001_fired.wait(), timeout=0.5)
@@ -245,7 +211,4 @@ async def test_scheduler_numeric_id_test(
assert interval_count >= 3, (
f"Expected at least 3 interval fires, got {interval_count}"
)
assert retry_count >= 2, (
f"Expected at least 2 retry attempts, got {retry_count}"
)
assert defer_count >= 2, f"Expected at least 2 defer fires, got {defer_count}"
@@ -1,279 +0,0 @@
"""Test scheduler retry functionality."""
import asyncio
import re
import pytest
from .types import APIClientConnectedFactory, RunCompiledFunction
@pytest.mark.asyncio
async def test_scheduler_retry_test(
yaml_config: str,
run_compiled: RunCompiledFunction,
api_client_connected: APIClientConnectedFactory,
) -> None:
"""Test that scheduler retry functionality works correctly."""
# Track test progress
simple_retry_done = asyncio.Event()
backoff_retry_done = asyncio.Event()
immediate_done_done = asyncio.Event()
cancel_retry_done = asyncio.Event()
empty_name_retry_done = asyncio.Event()
component_retry_done = asyncio.Event()
multiple_name_done = asyncio.Event()
const_char_done = asyncio.Event()
static_char_done = asyncio.Event()
test_complete = asyncio.Event()
# Track retry counts
simple_retry_count = 0
backoff_retry_count = 0
immediate_done_count = 0
cancel_retry_count = 0
empty_name_retry_count = 0
component_retry_count = 0
multiple_name_count = 0
const_char_retry_count = 0
static_char_retry_count = 0
# Track specific test results
cancel_result = None
empty_cancel_result = None
backoff_intervals = []
def on_log_line(line: str) -> None:
nonlocal simple_retry_count, backoff_retry_count, immediate_done_count
nonlocal cancel_retry_count, empty_name_retry_count, component_retry_count
nonlocal multiple_name_count, const_char_retry_count, static_char_retry_count
nonlocal cancel_result, empty_cancel_result
# Strip ANSI color codes
clean_line = re.sub(r"\x1b\[[0-9;]*m", "", line)
# Simple retry test
if "Simple retry attempt" in clean_line:
if match := re.search(r"Simple retry attempt (\d+)", clean_line):
simple_retry_count = int(match.group(1))
elif "Simple retry succeeded on attempt" in clean_line:
simple_retry_done.set()
# Backoff retry test
elif "Backoff retry attempt" in clean_line:
if match := re.search(
r"Backoff retry attempt (\d+).*interval=(\d+)ms", clean_line
):
backoff_retry_count = int(match.group(1))
interval = int(match.group(2))
if backoff_retry_count > 1: # Skip first (immediate) call
backoff_intervals.append(interval)
elif "Backoff retry completed" in clean_line:
backoff_retry_done.set()
# Immediate done test
elif "Immediate done retry called" in clean_line:
immediate_done_count += 1
immediate_done_done.set()
# Cancel retry test
elif "Cancel test retry attempt" in clean_line:
cancel_retry_count += 1
elif "Retry cancellation result:" in clean_line:
cancel_result = "true" in clean_line
cancel_retry_done.set()
# Empty name retry test
elif "Empty name retry attempt" in clean_line:
if match := re.search(r"Empty name retry attempt (\d+)", clean_line):
empty_name_retry_count = int(match.group(1))
elif "Empty name retry cancel result:" in clean_line:
empty_cancel_result = "true" in clean_line
elif "Empty name retry ran" in clean_line:
empty_name_retry_done.set()
# Component retry test
elif "Component retry attempt" in clean_line:
if match := re.search(r"Component retry attempt (\d+)", clean_line):
component_retry_count = int(match.group(1))
if component_retry_count >= 2:
component_retry_done.set()
# Multiple same name test
elif "Second duplicate retry attempt" in clean_line:
if match := re.search(r"counter=(\d+)", clean_line):
multiple_name_count = int(match.group(1))
if multiple_name_count >= 20:
multiple_name_done.set()
# Const char retry test
elif "Const char retry" in clean_line:
if match := re.search(r"Const char retry (\d+)", clean_line):
const_char_retry_count = int(match.group(1))
const_char_done.set()
# Static const char retry test
elif "Static const char retry" in clean_line:
if match := re.search(r"Static const char retry (\d+)", clean_line):
static_char_retry_count = int(match.group(1))
static_char_done.set()
elif "Static cancel result:" in clean_line:
# This is part of test 9, but we don't track it separately
pass
# Test completion
elif "All retry tests completed" in clean_line:
test_complete.set()
async with (
run_compiled(yaml_config, line_callback=on_log_line),
api_client_connected() as client,
):
# Verify we can connect
device_info = await client.device_info()
assert device_info is not None
assert device_info.name == "scheduler-retry-test"
# Wait for simple retry test
try:
await asyncio.wait_for(simple_retry_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Simple retry test did not complete. Count: {simple_retry_count}"
)
assert simple_retry_count == 2, (
f"Expected 2 simple retry attempts, got {simple_retry_count}"
)
# Wait for backoff retry test
try:
await asyncio.wait_for(backoff_retry_done.wait(), timeout=3.0)
except TimeoutError:
pytest.fail(
f"Backoff retry test did not complete. Count: {backoff_retry_count}"
)
assert backoff_retry_count == 4, (
f"Expected 4 backoff retry attempts, got {backoff_retry_count}"
)
# Verify backoff intervals (allowing for timing variations)
assert len(backoff_intervals) >= 2, (
f"Expected at least 2 intervals, got {len(backoff_intervals)}"
)
if len(backoff_intervals) >= 3:
# First interval should be ~50ms (very wide tolerance for heavy system load)
assert 20 <= backoff_intervals[0] <= 150, (
f"First interval {backoff_intervals[0]}ms not ~50ms"
)
# Second interval should be ~100ms (50ms * 2.0)
assert 50 <= backoff_intervals[1] <= 250, (
f"Second interval {backoff_intervals[1]}ms not ~100ms"
)
# Third interval should be ~200ms (100ms * 2.0)
assert 100 <= backoff_intervals[2] <= 500, (
f"Third interval {backoff_intervals[2]}ms not ~200ms"
)
# Wait for immediate done test
try:
await asyncio.wait_for(immediate_done_done.wait(), timeout=3.0)
except TimeoutError:
pytest.fail(
f"Immediate done test did not complete. Count: {immediate_done_count}"
)
assert immediate_done_count == 1, (
f"Expected 1 immediate done call, got {immediate_done_count}"
)
# Wait for cancel retry test
try:
await asyncio.wait_for(cancel_retry_done.wait(), timeout=3.0)
except TimeoutError:
pytest.fail(
f"Cancel retry test did not complete. Count: {cancel_retry_count}"
)
assert cancel_result is True, "Retry cancellation should have succeeded"
assert 2 <= cancel_retry_count <= 5, (
f"Expected 2-5 cancel retry attempts before cancellation, got {cancel_retry_count}"
)
# Wait for empty name retry test
try:
await asyncio.wait_for(empty_name_retry_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Empty name retry test did not complete. Count: {empty_name_retry_count}"
)
# Empty name retry should run at least once before being cancelled
assert 1 <= empty_name_retry_count <= 3, (
f"Expected 1-3 empty name retry attempts, got {empty_name_retry_count}"
)
assert empty_cancel_result is True, (
"Empty name retry cancel should have succeeded"
)
# Wait for component retry test
try:
await asyncio.wait_for(component_retry_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Component retry test did not complete. Count: {component_retry_count}"
)
assert component_retry_count >= 2, (
f"Expected at least 2 component retry attempts, got {component_retry_count}"
)
# Wait for multiple same name test
try:
await asyncio.wait_for(multiple_name_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Multiple same name test did not complete. Count: {multiple_name_count}"
)
# Should be 20+ (only second retry should run)
assert multiple_name_count >= 20, (
f"Expected multiple name count >= 20 (second retry only), got {multiple_name_count}"
)
# Wait for const char retry test
try:
await asyncio.wait_for(const_char_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Const char retry test did not complete. Count: {const_char_retry_count}"
)
assert const_char_retry_count == 1, (
f"Expected 1 const char retry call, got {const_char_retry_count}"
)
# Wait for static char retry test
try:
await asyncio.wait_for(static_char_done.wait(), timeout=1.0)
except TimeoutError:
pytest.fail(
f"Static char retry test did not complete. Count: {static_char_retry_count}"
)
assert static_char_retry_count == 1, (
f"Expected 1 static char retry call, got {static_char_retry_count}"
)
# Wait for test completion
try:
await asyncio.wait_for(test_complete.wait(), timeout=1.0)
except TimeoutError:
pytest.fail("Test did not complete within timeout")
@@ -0,0 +1,89 @@
"""Integration test for wait_until queue reentrancy.
When a wait_until completes, the rest of the action chain runs synchronously
from inside the wait queue processing. That chain can re-enter the very same
WaitUntilAction - for example a script with mode: restart that executes itself
as a retry pattern, or a waiter that stops its own script. Both used to mutate
the std::list while it was being iterated, corrupting it and crashing the
device (Guru Meditation StoreProhibited in _M_transfer).
"""
from __future__ import annotations
import asyncio
import re
import pytest
from .types import APIClientConnectedFactory, RunCompiledFunction
@pytest.mark.asyncio
async def test_wait_until_reentrant_restart(
yaml_config: str,
run_compiled: RunCompiledFunction,
api_client_connected: APIClientConnectedFactory,
) -> None:
"""Test that re-entering a wait_until from its own continuation is safe."""
retry_complete = asyncio.Event()
stop_complete = asyncio.Event()
attempt_pattern = re.compile(r"attempt (\d+) done")
gate_pattern = re.compile(r"gate passed (\d+)")
attempts: list[int] = []
gate_passed: list[int] = []
def check_output(line: str) -> None:
"""Check log output for expected messages."""
if mo := attempt_pattern.search(line):
attempts.append(int(mo.group(1)))
elif mo := gate_pattern.search(line):
gate_passed.append(int(mo.group(1)))
elif "retry test complete" in line:
retry_complete.set()
elif "stop test complete" in line:
stop_complete.set()
async with (
run_compiled(yaml_config, line_callback=check_output),
api_client_connected() as client,
):
device_info = await client.device_info()
assert device_info is not None
assert device_info.name == "wait-until-reentrant-restart"
_, services = await client.list_entities_services()
self_restart_service = next(
(s for s in services if s.name == "start_self_restart"), None
)
assert self_restart_service is not None, "start_self_restart not found"
stop_service = next(
(s for s in services if s.name == "start_stop_during_wait"), None
)
assert stop_service is not None, "start_stop_during_wait not found"
# Scenario 1: the wait_until timeout continuation restarts its own
# script five times, re-entering the same wait_until each time.
await client.execute_service(self_restart_service, {})
try:
await asyncio.wait_for(retry_complete.wait(), timeout=10.0)
except TimeoutError:
pytest.fail(f"Self-restart retry did not finish. Attempts: {attempts}")
assert attempts == [1, 2, 3, 4, 5], attempts
# Scenario 2: the first waiter through the gate stops the script while
# the other waiters are still queued in the same wait_until; both the
# not-yet-checked waiters (2, 3) and the already-checked still-waiting
# blocker (0) must be cancelled, not fired.
await client.execute_service(stop_service, {})
try:
await asyncio.wait_for(stop_complete.wait(), timeout=10.0)
except TimeoutError:
pytest.fail(f"Stop-during-wait did not finish. Gate passed: {gate_passed}")
assert gate_passed == [1], gate_passed
# If the cancelled blocker had been kept, its 1s wait_until timeout
# would still fire - give it the chance and check it stays silent.
await asyncio.sleep(1.5)
assert gate_passed == [1], gate_passed