fix(explorer): better profile thread selection logic

[roaga]

Instead of just selecting the thread from the first sample, we now prioritize the thread with the most in-app frames. And if no in-app frames are present, we show system frames.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
29877	1	29876	244

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

tests.sentry.seer.explorer.test_explorer_utils.TestConvertProfileToExecutionTree::test_convert_profile_selects_thread_with_most_in_app_frames

Stack Traces | 0.043s run time

#x1B[1m#x1B[.../seer/explorer/test_explorer_utils.py#x1B[0m:618: in test_convert_profile_selects_thread_with_most_in_app_frames
    assert len(root.children) == 2
#x1B[1m#x1B[31mE   AssertionError: assert 1 == 2#x1B[0m
#x1B[1m#x1B[31mE    +  where 1 = len([ExecutionTreeNode(function='worker_function_2', module='app', filename='worker.py', lineno=30, in_app=True, children=...ker.py:40/worker_function_2:worker.py:30', sample_count=1, first_seen_ns=1000000, last_seen_ns=1000000, duration_ns=0)])#x1B[0m
#x1B[1m#x1B[31mE    +    where [ExecutionTreeNode(function='worker_function_2', module='app', filename='worker.py', lineno=30, in_app=True, children=...ker.py:40/worker_function_2:worker.py:30', sample_count=1, first_seen_ns=1000000, last_seen_ns=1000000, duration_ns=0)] = ExecutionTreeNode(function='worker_function_3', module='app', filename='worker.py', lineno=40, in_app=True, children=[... node_id='/worker_function_3:worker.py:40', sample_count=1, first_seen_ns=1000000, last_seen_ns=1000000, duration_ns=0).children#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.}

[cursor]

Bug: MainThread selection ignores whether thread has samples

When no threads have in_app frames, the code attempts to select MainThread from metadata without verifying it actually contains any samples. If MainThread exists in metadata but never appears in the samples data, selected_thread_id will be set to a thread that has no profiling data, causing all samples to be skipped and an empty execution tree to be returned instead of falling back to the first sample's thread.

 

[JoshFerge]

curious what the logic is on issue details for this, but makes sense to me

[roaga]

curious what the logic is on issue details for this, but makes sense to me

From dogfooding in the UI, it doesn't seem like there is any consistent logic besides selecting the first available thread, which isn't a great UX...