ref(replay): filter out events before replay start for log messages

[michellewzhang]

related to: #102897

backend PR for filtering out events that occur before the replay start time, so we don't have to generate log messages for those. we filter out these events on the frontend as well -- they aren't shown as breadcrumbs.

also fix some timestamp inconsistencies in tests

[cursor]

Bug: Prematurely Emitting Pre-Start Error Events

Remaining error events are yielded without checking if they occurred before replay_start_ms. After processing all segment events, the code yields all remaining errors regardless of their timestamps, allowing error messages from before the replay started to appear in the logs.

src/sentry/replays/usecases/summarize.py#L310-L320

sentry/src/sentry/replays/usecases/summarize.py

Lines 310 to 320 in 9a5ea32

	# Yield any remaining error messages
	while error_idx < len(error_events):
	error = error_events[error_idx]
	if error["category"] == "error":
	yield generate_error_log_message(error)
	elif error["category"] == "feedback":
	yield generate_feedback_log_message(error)
	error_idx += 1

 

---

[cursor]

Bug: Feedback breadcrumb timestamp in seconds causes filtering error

The feedback breadcrumb timestamp in the test is in seconds instead of milliseconds. The timestamp field should be float((now - timedelta(minutes=3)).timestamp() * 1000) to match the expected millisecond format for segment events, causing this feedback to be incorrectly filtered out as occurring before the replay start.

tests/sentry/replays/usecases/test_summarize.py#L1404-L1405

sentry/tests/sentry/replays/usecases/test_summarize.py

Lines 1404 to 1405 in 3488fa5

	self.store_replay(dt=now - timedelta(minutes=10), segment_id=0, trace_ids=[trace_id])

 

---

[aliu39]

Suggested change

	segment_data, error_events, project_id, is_mobile_replay, replay_start
	segment_data, error_events, project_id, is_mobile_replay=is_mobile_replay, replay_start=replay_start

and same for L598 - best to be explicit w kwargs

[aliu39]

Suggested change

	replay_start_ms = _parse_iso_timestamp_to_ms(replay_start) if replay_start else 0.0
	replay_start_ms = _parse_iso_timestamp_to_ms(replay_start) if replay_start else 0.0
	while error_events[error_idx]["timestamp"] < replay_start_ms:
	error_idx += 1

lets you avoid the ifs below when yielding errors

[aliu39]

would make these errors direct connected rather than trace connected, cuz the trace query already uses the replay range filter

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
29528	1	29527	239

View the top 1 failed test(s) by shortest run time

tests.sentry.replays.usecases.test_summarize.RpcGetReplaySummaryLogsTestCase::test_rpc_filters_out_events_before_replay_start

Stack Traces | 15.6s run time

#x1B[1m#x1B[.../replays/usecases/test_summarize.py#x1B[0m:1620: in test_rpc_filters_out_events_before_replay_start
    assert len(logs) == 2
#x1B[1m#x1B[31mE   assert 1 == 2#x1B[0m
#x1B[1m#x1B[31mE    +  where 1 = len(["Logged: 'world' at 1762472342648.16"])#x1B[0m

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}