agent-proxy: wire pytest-asyncio so the 5 skipped async tests actually run #25
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Context
Follow-up flagged in the agent-proxy#22 retrospective. After that fix the suite is 22 passed, 5 skipped, 0 failed. The 5 skips are async test cases that need
pytest-asynciowired in - a pre-existing harness gap, not caused by the crash fix. Skipped async tests read as green while actually asserting nothing on the async paths (the queue worker, streaming dispatch, the resilience retry/fallback loop).Ask
pytest-asyncioto[project.optional-dependencies].devinpyproject.tomland regenerateuv.lock.asyncio_mode = "auto"under[tool.pytest.ini_options], or per-test markers - whichever matches the existing test style) so the 5 skipped async cases actually run.Acceptance
pytest-asynciois a declared dev dep and captured inuv.lock.Refs
🔒 Reserved by
ward agent --driver claude— containerengineer-claude-agent-proxy-25on hostkais-macbook-pro-2.localis carrying this issue (reserved 2026-07-02T08:24:12Z). Concurrentward agentruns are blocked until it finishes or the reservation goes stale (2h0m0s TTL);--forceoverrides.— Claude (she/her), via
ward agentWARD-OUTCOME: done - pytest-asyncio wired via asyncio_mode="auto"; suite now 27 passed / 0 skipped on main.
Honestly the smoothest kind of task. The five skips were pure harness gap, not broken code: adding pytest-asyncio to the dev extra, an asyncio_mode = "auto" block matching the marker-free test style, and a uv.lock regen was the whole fix. All five async tests passed unchanged the moment the plugin loaded, so there was nothing to fix in the queue/upstream paths - they were correct all along, just never exercised.
One thing that fought back a little: uv lock pulled pytest 7.4.4 -> 9.1.1 as a transitive consequence, a bigger jump than I expected. Everything still passes on 9.1.1, so I left it, but it is worth a glance if any pytest 8/9 deprecations bite later. Confidence is high - green on main via the canonical ward test. No follow-ups worth filing; the resilience retry/fallback loop the issue mentions is covered by the now-running tests.