nisse-dev/tests/test_git.py::test_a_failed_push_still_kills_its_agent failed on main's push run for 9150c4a (the #170 merge):
AssertionError: assert {'2304'} == set()
_agents() lists every live ssh-agent on the machine (pgrep -x ssh-agent). The unit tier runs under pytest-xdist (-n auto), so another worker's push can start its own agent between before and the assertion. The runs on main before and after this one passed.
Possible fix: count only the agents this test started. One way is to match the agent's socket path, if GitRepo.push puts it somewhere the test can name, such as under tmp_path. Another is to put the ssh-agent tests in one xdist group (--dist loadgroup). Either way, add a test that fails while a second agent runs alongside.
Also seen in the same session: tests/integration/test_delegation_live.py::test_delegate_completes_a_scoped_subtask hit its 12-turn cap on claude-haiku-5-5 on #171's first CI run. That's a model-variance flake on a live test. It's noted here only to track how often it happens.
`nisse-dev/tests/test_git.py::test_a_failed_push_still_kills_its_agent` failed on main's push run for 9150c4a (the #170 merge):
```
AssertionError: assert {'2304'} == set()
```
`_agents()` lists every live `ssh-agent` on the machine (`pgrep -x ssh-agent`). The unit tier runs under pytest-xdist (`-n auto`), so another worker's push can start its own agent between `before` and the assertion. The runs on main before and after this one passed.
**Possible fix:** count only the agents this test started. One way is to match the agent's socket path, if `GitRepo.push` puts it somewhere the test can name, such as under `tmp_path`. Another is to put the ssh-agent tests in one xdist group (`--dist loadgroup`). Either way, add a test that fails while a second agent runs alongside.
Also seen in the same session: `tests/integration/test_delegation_live.py::test_delegate_completes_a_scoped_subtask` hit its 12-turn cap on claude-haiku-5-5 on #171's first CI run. That's a model-variance flake on a live test. It's noted here only to track how often it happens.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
nisse-dev/tests/test_git.py::test_a_failed_push_still_kills_its_agentfailed on main's push run for9150c4a(the #170 merge):_agents()lists every livessh-agenton the machine (pgrep -x ssh-agent). The unit tier runs under pytest-xdist (-n auto), so another worker's push can start its own agent betweenbeforeand the assertion. The runs on main before and after this one passed.Possible fix: count only the agents this test started. One way is to match the agent's socket path, if
GitRepo.pushputs it somewhere the test can name, such as undertmp_path. Another is to put the ssh-agent tests in one xdist group (--dist loadgroup). Either way, add a test that fails while a second agent runs alongside.Also seen in the same session:
tests/integration/test_delegation_live.py::test_delegate_completes_a_scoped_subtaskhit its 12-turn cap on claude-haiku-5-5 on #171's first CI run. That's a model-variance flake on a live test. It's noted here only to track how often it happens.🤖 Generated with Claude Code
https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk