[Feature] Improve CLI Realtime Logs

[antonio-amjr]

Fix: project-chip/certification-tool#1026

Description

Extends the existing real-time log streaming feature with two new capabilities:

• A new-execution notification banner
• A navigatable test hierarchy tree panel.

New execution notification

• Polls /api/status (every 3 s) after the current SSE stream ends, detecting when a new test run has started by comparing run_title and start_time
• Displays a dismissable banner (⚠ New execution started) with a Reload button that refreshes the page and a Dismiss button that stops polling and keeps the current view
• The page never auto-reloads — the user is always in control
• A newExecutionPending flag gates all incoming SSE data handlers; while the flag is set, events are buffered rather than discarded — if the reconnection turns out to be the same execution (network hiccup), the buffer is replayed in order; if it is a new execution, the buffer is discarded and the banner is shown
• Handles the async race between the connected SSE event and the /api/status fetch by setting the flag preemptively before the fetch resolves

Execution Tree panel

• Sidebar showing the full test hierarchy: Run → Suite → Case → Step, with real-time state icons (pending / executing / passed / failed / error / skipped)
• Tree is built from TestRunExecutionWithChildren in LogStreamHandler.init_tree() and broadcast as a tree_init SSE event; reconnecting clients receive the current snapshot directly from the shared tree_state dict
• State updates streamed as tree_update SSE events, applied incrementally without re-rendering the whole tree
• Each node is collapsible/expandable; toggles start in the expanded (▼) position matching the initially visible children
• Clicking a step node scrolls the log panel to the corresponding log entry using a two-tier strategy:
  • Tier 1 (precise): pattern-matches "Executing Test Step: <title>" log entries and tags them as DOM anchors; clicking jumps directly to that entry
  • Tier 2 (fallback): records the log-entry index at the moment a tree_update(step, executing) event fires and scrolls to that index if no anchor is found
• The tree does not auto-scroll to follow the currently executing step, so users can freely inspect earlier steps during a run

Files Changed

File	Change
log_stream_handler.py	Replaced single shared log_queue with a per-client broadcast set (_clients + _clients_lock); added _broadcast() which snapshots the set and puts events into every active client queue independently; added init_tree(), update_tree_node(), _build_tree(); mutable tree_state dict shared with HTTP server
logs_http_server.py	stream_logs() now creates a per-client queue.Queue, registers it in the broadcast set on connect, and removes it in a finally block on disconnect — concurrent or reconnecting clients each receive their own complete event stream; added /api/status endpoint; SSE stream routes tree_init / tree_update event types; sends tree_state snapshot to reconnecting clients
websocket.py	Calls get_log_stream_handler() on each test state update to forward tree events
logging.py	Added get_log_stream_handler() accessor
log_viewer.html	Split layout (tree sidebar + log panel), new-execution banner, full tree rendering and update logic, two-tier scroll anchoring

Unit Tests

Unit tests were updated to reflect the changes. Here's a summary of what changed in each file:

tests/test_run/test_log_stream_handler.py:

• test_log_queue_is_queue → test_clients_is_empty_set — verifies _clients is an empty set
• test_puts_none_sentinel_into_queue → test_broadcasts_none_sentinel_on_stop — registers a real client queue in h._clients, asserts it receives None after stop()
• test_skips_sentinel_when_queue_full → test_stop_does_not_raise_when_client_queue_full — pre-fills a client queue to capacity, asserts no raise
• All TestAddLogEntry tests — replaced h.log_queue reads with either patch.object(h, "_broadcast") (for the no-op case) or a client queue registered in h._clients (for all data-content assertions)

tests/test_run/test_logs_http_server.py:

• _make_handler — now sets explicit defaults: active_clients=set(), clients_lock=threading.Lock(), tree_state={}, tree_lock=None so stream_logs() never tries to deepcopy a MagicMock
• Added _pre_filled_queue(*items) helper — returns a pre-populated queue for use with patch("...queue.Queue", return_value=...)
• All TestStreamLogs tests — removed server_attrs={"log_queue": q}, now patch queue.Queue to return a pre-filled queue instead; test_no_log_queue_on_server_returns_early renamed to test_no_active_clients_on_server_returns_early with matching
  server_attrs
• TestLogsHTTPServerStart — all three tests now call start(active_clients=..., clients_lock=..., tree_state={}, ...) instead of start(log_queue=...); test_sets_server_attributes asserts mock_ths.active_clients and mock_ths.clients_lock instead of mock_ths.log_queue

Screenshots

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request

[gemini-code-assist]

Code Review

This pull request introduces real-time log streaming and execution tree visualization to the test runner, updating the execution hierarchy dynamically and streaming updates to a web-based log viewer via Server-Sent Events (SSE). Key feedback highlights critical thread-safety issues where the shared tree_state dictionary is mutated and read across threads without synchronization, suggesting the introduction of a threading.Lock. Additionally, the reviewer noted that multiple concurrent clients will compete for the single shared log queue, recommending a broadcast pattern with client-specific queues. Other issues include potential event loss during reconnection status checks due to lack of buffering, and the use of the deprecated global window.event in the HTML log viewer.