mirror of
https://github.com/esphome/esphome.git
synced 2026-06-24 12:33:10 +00:00
3a4f67def8
The is_failed() skip exists in two execution paths: the heap loop in call() and should_skip_item_() (defer queue / delay:0). The previous commit only exempted SELF_POINTER items from the heap-path check, so on multi-threaded platforms a delay:0 continuation whose host component had failed would still be silently dropped. Extract a single is_item_failed_() helper (with the SELF_POINTER exemption) and use it from both paths so they cannot drift again. Add an integration test that schedules a delay from a component that marks itself failed and asserts the continuation still fires (verified to fail without the exemption).
33 lines
860 B
YAML
33 lines
860 B
YAML
esphome:
|
|
name: scheduler-delay-failed
|
|
|
|
host:
|
|
api:
|
|
logger:
|
|
level: DEBUG
|
|
|
|
globals:
|
|
- id: started
|
|
type: bool
|
|
restore_value: false
|
|
initial_value: "false"
|
|
|
|
# The interval fires with itself as the current component, schedules a delay (which
|
|
# captures that component for log attribution), then marks itself failed. The delay
|
|
# continuation must still fire: a failed component must not drop an already-scheduled
|
|
# delay, because the SELF_POINTER scheduler item stores the component for attribution
|
|
# only.
|
|
interval:
|
|
- interval: 100ms
|
|
id: host_interval
|
|
then:
|
|
- if:
|
|
condition:
|
|
lambda: "return !id(started);"
|
|
then:
|
|
- lambda: |-
|
|
id(started) = true;
|
|
id(host_interval)->mark_failed();
|
|
- delay: 200ms
|
|
- logger.log: "DELAY_FIRED_AFTER_FAIL"
|