fix(codex): resolve peer-review findings R1-R6 and T1-T2
- R1: use herdr send-keys C-c for graceful stop - R2: preserve JSONL fallback discovery when database is empty - R3: apply workspace-scoped SQLite query before bounding results - R4: tighten artifact validation to fail closed on missing/invalid metadata - R5: verify actual rollout presence before reporting resumable in status - R6: remove unconditional drift-B backend command fallback - T1: parameterize artifact metadata regression with verified finder fixture - T2: seed unmanaged and managed sessions in mock herdr for drift-B testing - Add final loop review reports (100% PASS)
This commit is contained in:
Godopu committed
1 parent
76f10b2096
commit
47640e2fd7
10 files changed
+504
-34
No files matched your search
@@ -1092,8 +1092,156 @@ def test_orc_onboard_codex_flag_tolerant_regex():
|
||||
assert res.stdout.strip() == "12345678-1234-1234-1234-123456789abc"
|
||||
|
||||
|
||||
def test_stop_session_codex_sends_c_c_chord_not_text_prompt(mam_sandbox, mock_herdr, mock_agents):
|
||||
"""R1: Verify graceful stop sends C-c key chord rather than text prompt for codex."""
|
||||
import subprocess, json
|
||||
# 1. Create a codex session
|
||||
create_script = mam_sandbox / ".agents" / "skills" / "multi-agent-mux-create" / "scripts" / "create_session.sh"
|
||||
res = subprocess.run([
|
||||
"bash", str(create_script),
|
||||
"--workspace", str(mam_sandbox),
|
||||
"--agent", "codex",
|
||||
"--role", "creator"
|
||||
], capture_output=True, text=True, cwd=str(mam_sandbox))
|
||||
assert res.returncode == 0, res.stderr
|
||||
|
||||
yaml_path = mam_sandbox / ".mam" / "agent-sessions.yaml"
|
||||
import yaml
|
||||
with open(yaml_path) as f:
|
||||
data = yaml.safe_load(f)
|
||||
session_name = data["herdr_sessions"][0]["name"]
|
||||
|
||||
# 2. Stop the session
|
||||
stop_script = mam_sandbox / ".agents" / "skills" / "multi-agent-mux-stop" / "scripts" / "stop_session.sh"
|
||||
res_stop = subprocess.run([
|
||||
"bash", str(stop_script),
|
||||
"--session", session_name,
|
||||
"--agent", "codex"
|
||||
], capture_output=True, text=True, cwd=str(mam_sandbox))
|
||||
assert res_stop.returncode == 0, res_stop.stderr
|
||||
|
||||
# 3. Verify mock_herdr recorded C-c send-keys call and did NOT record Ctrl+C prompt
|
||||
with open(mock_herdr) as f:
|
||||
state = json.load(f)
|
||||
calls = state.get("calls", [])
|
||||
pane_send_keys = [c for c in calls if "send-keys" in c and "C-c" in c]
|
||||
agent_prompts = [c for c in calls if "prompt" in c and any("Ctrl+C" in str(arg) for arg in c)]
|
||||
assert len(pane_send_keys) >= 1
|
||||
assert len(agent_prompts) == 0
|
||||
|
||||
|
||||
def test_status_resume_on_disk_codex_stale_and_valid(mam_sandbox):
|
||||
"""R5: Verify status.sh resume_on_disk reports MISSING for stale DB rows and yes for existing files."""
|
||||
import sqlite3, json, subprocess
|
||||
codex_home = mam_sandbox / ".codex"
|
||||
codex_home.mkdir(parents=True, exist_ok=True)
|
||||
db_path = codex_home / "state_5.sqlite"
|
||||
conn = sqlite3.connect(str(db_path))
|
||||
conn.execute("CREATE TABLE threads (id TEXT PRIMARY KEY, cwd TEXT, rollout_path TEXT)")
|
||||
|
||||
# 1. Stale row: rollout_path does not exist
|
||||
stale_uuid = "stale-1234-1234-1234-123456789abc"
|
||||
conn.execute("INSERT INTO threads (id, cwd, rollout_path) VALUES (?, ?, ?)",
|
||||
(stale_uuid, str(mam_sandbox), str(codex_home / "nonexistent.jsonl")))
|
||||
|
||||
# 2. Valid row: rollout_path exists
|
||||
valid_uuid = "valid-1234-1234-1234-123456789abc"
|
||||
valid_file = codex_home / f"rollout-{valid_uuid}.jsonl"
|
||||
valid_file.write_text(json.dumps({"payload": {"id": valid_uuid, "cwd": str(mam_sandbox)}}) + "\n")
|
||||
conn.execute("INSERT INTO threads (id, cwd, rollout_path) VALUES (?, ?, ?)",
|
||||
(valid_uuid, str(mam_sandbox), str(valid_file)))
|
||||
conn.commit()
|
||||
conn.close()
|
||||
|
||||
yaml_path = mam_sandbox / ".mam" / "agent-sessions.yaml"
|
||||
yaml_path.write_text(f"""herdr_sessions:
|
||||
- name: test-stale-codex
|
||||
status: running
|
||||
codex_session_id_own: {stale_uuid}
|
||||
pane:
|
||||
cwd: {mam_sandbox}
|
||||
- name: test-valid-codex
|
||||
status: running
|
||||
codex_session_id_own: {valid_uuid}
|
||||
pane:
|
||||
cwd: {mam_sandbox}
|
||||
""")
|
||||
script_path = mam_sandbox / "skills" / "multi-agent-mux-status" / "scripts" / "status.sh"
|
||||
env = {**os.environ, "CODEX_HOME": str(codex_home)}
|
||||
res = subprocess.run(["bash", str(script_path), "--json"], capture_output=True, text=True, env=env)
|
||||
assert res.returncode == 0
|
||||
data = json.loads(res.stdout)
|
||||
details = {d["name"]: d["resume_state"] for d in data["sessions_detail"]}
|
||||
assert details["test-stale-codex"] == "MISSING"
|
||||
assert details["test-valid-codex"] == "yes"
|
||||
|
||||
|
||||
def test_reconcile_drift_b_unmanaged_session_ignored(mam_sandbox, mock_herdr):
|
||||
"""R6 / T2: Verify drift-B ignores unmanaged in-workspace sessions while registering managed positive control."""
|
||||
import subprocess, json, yaml
|
||||
reconcile_script = mam_sandbox / ".agents" / "skills" / "multi-agent-mux-monitor" / "scripts" / "reconcile.sh"
|
||||
assert reconcile_script.exists()
|
||||
|
||||
# Create dummy yaml with no sessions
|
||||
yaml_path = mam_sandbox / ".mam" / "agent-sessions.yaml"
|
||||
yaml_path.parent.mkdir(parents=True, exist_ok=True)
|
||||
yaml_path.write_text("herdr_sessions: []\n")
|
||||
|
||||
unmanaged_name = "personal-session"
|
||||
managed_name = "test-proj-creator-codex"
|
||||
|
||||
# Seed mock_herdr with an unmanaged session and a managed control session running codex in workspace
|
||||
state = {
|
||||
"workspaces": [{"workspace_id": "w1", "label": "default", "cwd": str(mam_sandbox)}],
|
||||
"agents": {
|
||||
unmanaged_name: {
|
||||
"name": unmanaged_name,
|
||||
"status": "running",
|
||||
"cwd": str(mam_sandbox),
|
||||
"command": "codex",
|
||||
"pid": 12345
|
||||
},
|
||||
managed_name: {
|
||||
"name": managed_name,
|
||||
"status": "running",
|
||||
"cwd": str(mam_sandbox),
|
||||
"command": "codex",
|
||||
"pid": 54321
|
||||
}
|
||||
},
|
||||
"calls": []
|
||||
}
|
||||
with open(mock_herdr, "w") as f:
|
||||
json.dump(state, f)
|
||||
|
||||
# Assert candidate is visible to listing and pane metadata lookup before running reconcile
|
||||
r_ls = subprocess.run(["herdr", "ls", "-F", "#{session_name}|#{session_created}"], capture_output=True, text=True)
|
||||
assert r_ls.returncode == 0
|
||||
assert unmanaged_name in r_ls.stdout
|
||||
assert managed_name in r_ls.stdout
|
||||
|
||||
r_pm = subprocess.run(["herdr", "list-panes", "-t", unmanaged_name, "-F", "#{pane_pid}|#{pane_current_path}|#{pane_current_command}"], capture_output=True, text=True)
|
||||
assert r_pm.returncode == 0
|
||||
assert f"12345|{str(mam_sandbox)}|codex" in r_pm.stdout
|
||||
|
||||
res = subprocess.run(["bash", str(reconcile_script), "--once", "--emit-diff"],
|
||||
capture_output=True, text=True, cwd=str(mam_sandbox),
|
||||
env={**os.environ, "WORKSPACE_ROOT": str(mam_sandbox)})
|
||||
assert res.returncode == 0, res.stderr
|
||||
|
||||
out = json.loads(res.stdout)
|
||||
drifts = out.get("drifts", [])
|
||||
|
||||
# Personal session must NOT be registered as drift B
|
||||
assert not any(d.get("name") == unmanaged_name for d in drifts)
|
||||
|
||||
# Managed positive control MUST be registered as drift B
|
||||
b_registered = [d.get("name") for d in drifts if d.get("class") == "B"]
|
||||
assert managed_name in b_registered
|
||||
|
||||
# Verify YAML updated with managed session but without unmanaged session
|
||||
with open(yaml_path) as f:
|
||||
data = yaml.safe_load(f)
|
||||
registered_names = [s.get("name") for s in data.get("herdr_sessions", [])]
|
||||
assert unmanaged_name not in registered_names
|
||||
assert managed_name in registered_names
|
||||
Reference in new issue
Block a user