Fix issue 24

[rtsdque]

Problem

There's no way to reliably wait for the request queue to finish dispatching all queued heartbeats. The current workaround in the example watcher is a hardcoded sleep(1), which is both fragile and wasteful.

Fix

Added RequestQueue.wait_for_queue_empty(timeout) which polls the queue until it's empty, with an optional timeout in seconds. Also exposed it as ActivityWatchClient.wait_for_queue_empty(timeout) for convenience.

Behavior:

• Returns True if the queue empties before the timeout
• Returns False if the timeout is reached first
• Returns True immediately if the queue thread isn't running

Testing

• test_wait_for_queue_empty_basic — queue drains normally while running
• test_wait_for_queue_empty_not_running — returns True immediately if thread not started
• test_wait_for_queue_empty_timeout — returns False when queue can't drain in time
• All 6 tests passing, mypy clean, no new ruff issues introduced

Fixes #24

[greptile-apps]

Greptile Summary

This PR adds RequestQueue.wait_for_queue_empty(timeout) (and a thin wrapper on ActivityWatchClient) so callers can reliably block until all queued heartbeats have been dispatched, replacing the fragile sleep(1) workaround. It also hardens add_request against disk-full OSErrors and adds four new tests.

• wait_for_queue_empty polls qsize() and _current every 100 ms and uses the existing _stop_event to wake early; the is_alive() guard at entry handles the "thread never started" case.
• add_request now catches OSError and logs a warning instead of propagating the exception to the caller when the underlying SQLite queue can't be written.
• The new timeout test mocks _post with a 10-second sleep, which causes rq.join() to block for the full sleep duration and noticeably slows the test suite.

Confidence Score: 3/5

The core polling loop needs a guard for unexpected thread death before it is safe to call without a timeout.

The wait_for_queue_empty loop checks is_alive() once at entry but never again during polling. If the RequestQueue thread terminates without setting its stop event, any no-timeout call will spin forever burning CPU with no way to break out. This is a present defect on the new code path that should be addressed before merging.

aw_client/client.py — specifically the wait_for_queue_empty polling loop in RequestQueue.

Important Files Changed

Filename	Overview
aw_client/client.py	Adds wait_for_queue_empty to both ActivityWatchClient and RequestQueue, and wraps add_request's persistqueue.put in an OSError handler. The polling loop in RequestQueue.wait_for_queue_empty only checks is_alive() at entry — if the thread dies mid-wait with no timeout set, the loop spins indefinitely.
tests/test_requestqueue.py	Adds four new tests covering disk-full handling and wait_for_queue_empty scenarios. The timeout test will block rq.join() for ~10 s due to an unconditional 10-second sleep in the mock.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant Caller
    participant ActivityWatchClient
    participant RequestQueue
    participant PersistQueue

    Caller->>ActivityWatchClient: wait_for_queue_empty(timeout)
    ActivityWatchClient->>RequestQueue: wait_for_queue_empty(timeout)

    RequestQueue->>RequestQueue: is_alive()?
    alt thread not running
        RequestQueue-->>Caller: True (immediately)
    else thread alive
        loop poll every 100 ms
            RequestQueue->>PersistQueue: qsize()
            RequestQueue->>RequestQueue: _current is not None?
            alt queue empty and no current item
                RequestQueue-->>Caller: True
            else stop() called (stop_event set)
                RequestQueue-->>Caller: False
            else timeout exceeded
                RequestQueue-->>Caller: False
            else thread dies unexpectedly (no stop_event)
                Note over RequestQueue: loop spins indefinitely if timeout=None
            end
        end
    end

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant Caller
    participant ActivityWatchClient
    participant RequestQueue
    participant PersistQueue

    Caller->>ActivityWatchClient: wait_for_queue_empty(timeout)
    ActivityWatchClient->>RequestQueue: wait_for_queue_empty(timeout)

    RequestQueue->>RequestQueue: is_alive()?
    alt thread not running
        RequestQueue-->>Caller: True (immediately)
    else thread alive
        loop poll every 100 ms
            RequestQueue->>PersistQueue: qsize()
            RequestQueue->>RequestQueue: _current is not None?
            alt queue empty and no current item
                RequestQueue-->>Caller: True
            else stop() called (stop_event set)
                RequestQueue-->>Caller: False
            else timeout exceeded
                RequestQueue-->>Caller: False
            else thread dies unexpectedly (no stop_event)
                Note over RequestQueue: loop spins indefinitely if timeout=None
            end
        end
    end

Loading

_{Reviews (1): Last reviewed commit: "Implement wait_for_queue_empty method" | Re-trigger Greptile}

[greptile-apps]

Thread death leaves wait_for_queue_empty spinning indefinitely

is_alive() is checked once at entry, but not inside the polling loop. If the RequestQueue thread dies unexpectedly (e.g., an unhandled exception propagates out of run()'s inner loops without setting _stop_event), self.wait(0.1) will keep returning False on timeout every 100 ms, qsize() and _current will never change, and any call with timeout=None will spin forever. Adding if not self.is_alive(): return False as the first check inside the loop would break out cleanly in that scenario.

[greptile-apps]

rq.join() blocks for ~10 seconds in the timeout test

slow_post sleeps for 10 seconds unconditionally. After wait_for_queue_empty(timeout=0.5) returns, rq.stop() is called but the queue thread is already inside slow_post with no way to interrupt it. rq.join() then has to wait the remaining ~9.5 seconds before the thread exits. The test effectively adds ~10 s to the suite on every run. Using a much shorter sleep (e.g. 0.5 s) or a threading.Event that can be signaled by stop() would keep the test fast while still exercising the timeout path reliably.

[greptile-apps]

Missing newline at end of file. Both changed files (client.py and test_requestqueue.py) are missing a trailing newline, which can cause noisy diffs and violates POSIX text file conventions.

Suggested change

	assert result is False
	assert result is False

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!