hermes-agent/tests/test_run_tests_parallel.py
Teknium 597615ade4
fix(ci): make tests, workflows, and attribution reliable under load (#66373)
* feat(attribution): conflict-free contributor mappings via contributors/emails/ directory

The AUTHOR_MAP dict in scripts/release.py was a merge-conflict magnet:
every concurrent salvage PR appended entries to the same lines of the
same file, so parallel PRs re-conflicted on every merge to main.

New system: one file per email under contributors/emails/ — filename is
the commit-author email, first non-comment line is the GitHub login.
File additions never conflict, so any number of PRs can add mappings
concurrently.

- scripts/release.py: AUTHOR_MAP is now LEGACY_AUTHOR_MAP (frozen)
  merged with the directory at import time (directory wins). All
  existing consumers (resolve_author, contributor_audit.py) unchanged.
- scripts/add_contributor.py: idempotent CLI to add a mapping; refuses
  conflicting reassignments (incl. against the legacy map), validates
  email/login shapes.
- contributor-check.yml: attribution gate now accepts a mapping file OR
  a legacy entry; failure message prints the exact add_contributor
  command. Also auto-resolves bare <login>@users.noreply.github.com
  emails is intentionally NOT added (kept id+login form only, matching
  previous behavior).
- contributor_audit.py: guidance now points at add_contributor.py.
- tests/scripts/test_contributor_map.py: 12 tests covering loader,
  merge precedence, CLI idempotency/conflict/validation, subprocess E2E.

* feat(ci): one-shot per-file flake retry in the parallel test runner

A failing test FILE is re-run once in a fresh subprocess. Pass-on-retry
counts as green but is loudly reported in a '⚠ FLAKY' summary section
(with both attempts' output preserved) so the flake gets fixed instead
of eating a full-run rerun. Deterministic failures fail both attempts —
regressions cannot be laundered green.

- --file-retries N / HERMES_TEST_FILE_RETRIES (default 1, 0 disables)
- E2E verified: simulated first-run-fail flake goes green with banner;
  deterministic failure still exits 1; retries=0 restores old behavior.

This converts the dominant CI failure mode (one timing-sensitive test
flaking a 4600-test shard, requiring a manual 10-minute rerun and an
agent triage loop) into a self-healing retry that costs one file's
runtime.

* test(approval): loosen wall-clock perf bounds 0.15s -> 2.0s

These guard against catastrophic regex backtracking (seconds-to-minutes
class), but 0.15s is within scheduler-stall noise on loaded shared CI
runners — test_max_accepted_separator_free_input_is_fast failed a CI
shard this week on runner load alone. 2.0s still catches the regression
class with zero flake surface.

* fix(ci): job timeouts everywhere + retries on all network installs

Reliability pass over every workflow:
- timeout-minutes on all 21 jobs that lacked one (a hung job previously
  burned the 6-hour default runner budget)
- ./.github/actions/retry wrapped around every network-fetching install
  that lacked it: pip installs (deploy-site, skills-index), npm ci
  (deploy-site website, upload_to_pypi web + ui-tui), uv sync (docker
  test deps). Deterministic build steps (npm run build) deliberately
  NOT retried — split into separate steps so a real build failure fails
  fast instead of retrying 3x.

* docs(agents): document the file-retry flake policy

* fix(ci): curl retries on deploy hook + skills-index probe

* fix(ci): kill the remaining transient-failure classes in workflows + Dockerfile

From the workflow reliability audit:
- tests.yml: duration-cache restore had NO restore-keys while saves use
  run_id-suffixed keys — the cache never matched once, so LPT slicing
  always ran blind and unbalanced slices pushed heavy files toward the
  per-file timeout. One-line restore-keys fixes slice balancing.
- Label gates (lint ci-reviewed, supply-chain mcp-catalog-reviewed):
  'gh pr view || true' turned an API blip into 'label absent' → false
  BLOCKING failure. Now 3x retry, and API failure is reported as an API
  failure instead of a missing label.
- detect-changes action: compare API retried before failing open (was
  silently running all lanes on any blip).
- uv-lockfile-check: 'uv lock --check' resolves against PyPI — retried
  so registry blips don't read as 'lockfile stale'.
- docker.yml merge job: imagetools create retried (Docker Hub eventual
  consistency on just-pushed digests).
- Dockerfile: apt-get Acquire::Retries=3; s6-overlay ADDs converted to
  curl --retry 3 (ADD cannot retry; checksums still enforced); npm
  --fetch-retries=5; playwright chromium fetch retried 3x.
- Advisory artifact uploads (per-slice durations, ci-timings report)
  get continue-on-error so an artifact-service blip can't fail a green
  test slice.

* fix(tests): kill the two root-cause flakes — leaking pre-warm timer + env-dependent provider list

- test_tui_gateway_server.py: session.create / non-eager session.resume
  arm a 50ms threading.Timer (_schedule_agent_build) that outlives its
  test and fires into the NEXT test's _make_agent mock, racily
  corrupting captured state (the recurring session_resume shard
  failures). Replaced the per-test whack-a-mole stub with a module-wide
  autouse fixture; the 3 worker-lifecycle tests that genuinely need the
  deferred build opt back in via @pytest.mark.real_agent_prewarm (new
  marker in pyproject).
- test_api_key_providers.py: PROVIDER_ENV_VARS is now derived from the
  live PROVIDER_REGISTRY instead of a hand-list that had drifted
  (missing HF_TOKEN / DEEPINFRA_API_KEY) — resolve_provider('auto')
  tests failed on any machine with HF_TOKEN exported. E2E-verified with
  HF_TOKEN/DEEPINFRA_API_KEY set: 42/42 pass.

* test: de-flake 30 timing-sensitive test files for loaded CI runners

Root-cause fixes from the flake audit (session-DB mining + repo sweep):

Event-based sync instead of sleep-sync:
- title_generator: mock sets threading.Event, wait(10) replaces
  sleep(0.3) hoping the daemon thread got scheduled
- docker zombie_reaping / profile_gateway: poll-for-state helpers
  replace fixed 1-3s sleeps (s6 transitions + SIGCHLD reaping are async)
- process_registry tree test: select()-bounded readline replaces an
  unbounded blocking read (parent wedge now fails THIS test with a clear
  message instead of an opaque rc=124 file kill); SIGTERM grace 1s->2s
  (the 1s partition window mid-interpreter-startup is how a child PID
  escaped the live-system guard in CI)

Timeout raises (loaded 8-way-sliced runners see ~5s scheduling floors;
all of these complete in ms-to-1s when healthy so the raises cost
nothing on green runs):
- subprocess/thread waits <= 2s raised to 10-15s across mcp_tool,
  mcp_circuit_breaker, mcp_reconnect_retry_reset, mcp_parked_self_probe,
  mcp_cancelled_error_propagation, registry, clarify_gateway, interrupt,
  voice_cli_integration, docker_environment, session_store_lock_io,
  planned_stop_watcher, cli_interrupt_subagent, thread_scoped_output
  (joins now also assert not is_alive() so stragglers fail loudly)
- wall-clock discrimination ceilings loosened where the guarded hang is
  10x larger: local_background_child_hang 4s->10s, interrupt_cleanup
  setup 5s->20s + pgid-exit 30s->60s, mcp_stability grandchild spinup
  5s->15s, protocol/gil-starvation fast-handler 0.5s->2s,
  iso_certify_seam 1.5s->5s, wait_for_mcp_discovery 0.1s->1s
- narrow assertion windows widened: honcho first-turn wait 0.4..0.65 ->
  0.25..2.0 (property is bounded-not-hung, not an exact wall-clock);
  compression fork-lock TTL 1s->3s (12 refresh chances per lease);
  compression-lock expiry margins symmetric (ttl 0.05->0.5, sleep 1.0)
- telegram hung-DNS bound 1.0->1.4 (fake hang is 1.5s — must stay under)

* fix(tests): repair indentation from de-flake batch edit

* fix(tests): harden env isolation and replace remaining sleep-sync races

The full 42k-test run and complete npm check surfaced three more classes:

- Environment isolation: local ~/.honcho defaultHost and SSH_* variables
  leaked into Python/TUI tests. Pin the default Honcho host in the
  hermetic fixture, isolate the one fallback test from ~/.honcho, and
  blank SSH_* around terminalSetup tests. This flipped 20 false failures
  back to deterministic behavior on developer machines.
- Background-thread sleep-sync: Honcho async writer tests patched
  time.sleep globally, then busy-polled with that same mocked sleep. Under
  full-suite load the poller could starve the writer. Each test now waits
  on an Event emitted by the exact flush/retry transition; 30/30 passed
  under 15-way contention.
- Desktop streaming: the test slept 80ms and assumed a 500ms timer could
  not fire before its assertion. A loaded runner descheduled the test for
  >500ms and both chunks arrived. Producer controls now gate second-chunk
  and completion transitions explicitly.

Also make file-retry observability complete: a self-healed flaky file now
prints BOTH attempts' full output in the FLAKY summary. Two behavioral
runner tests prove pass-on-retry is green+loud+traceback-preserving, while
a deterministic failure remains red.

* refactor(ci): use gh bot pat, better retries

refactor(ci): use retry action for PR label fetch
the retry action now captures stdout as a step output, so it can serve
double duty: retry + output capture for commands like 'gh pr view' whose
result must be consumed by later steps.

Retry action gains:
- 'stdout' output (heredoc-delimited to preserve newlines)
- tee to temp file so stdout still streams to the job log
- step id 'retry' for output reference

Both lint.yml and supply-chain-audit.yml now use the retry action
directly with 'command: gh pr view ...' and read
steps.<id>.outputs.stdout.

ci: use AUTOFIX_BOT_PAT for all gh CLI / GitHub API auth

Replace secrets.GITHUB_TOKEN and github.token with
secrets.AUTOFIX_BOT_PAT across all workflows and composite actions
that use the gh CLI or GitHub API. The PAT has consistent permissions
across fork PRs (where GITHUB_TOKEN is read-only), avoids API rate
limit sharing with the default token, and is already used by
js-autofix.yml for the same reasons.

19 sites swapped across 9 files:
- lint.yml (3): label fetch, comment post/edit, comment update
- supply-chain-audit.yml (5): scan, critical comment, unbounded dep
  comment, label fetch, mcp-catalog comment
- lockfile-diff.yml (1): PR comment post/update
- skills-index-freshness.yml (1): issue creation on degraded probe
- skills-index.yml (2): index build, trigger deploy workflow
- upload_to_pypi.yml (2): release view poll, release upload
- ci.yml (1): timings report
- deploy-site.yml (2): skills index crawl
- detect-changes/action.yml (1): compare API call

---------

Co-authored-by: ethernet <arilotter@gmail.com>
2026-07-17 20:55:24 +00:00

361 lines
13 KiB
Python

"""Verify scripts/run_tests_parallel.py kills test-spawned grandchildren.
Setup
-----
A test in this file spawns a long-lived Python grandchild that writes
its PID + a nonce to a tempfile, then exits without cleaning up.
With the old ``subprocess.run`` runner, that grandchild would orphan
and outlive the test (and the whole runner). With the current Popen +
``start_new_session`` + ``_kill_tree`` runner, the grandchild gets
SIGKILL'd via process-group kill when its file's pytest exits.
The leaker test always passes — its only job is to spawn a grandchild
and walk away. The verifier runs the runner over the leaker file in a
subprocess, then waits for the grandchild PID to disappear from the
kernel's process table.
POSIX-only: Windows has its own grandchild lifecycle (no shared session,
``taskkill /F /T`` semantics). Marked accordingly.
"""
from __future__ import annotations
import json
import os
import subprocess
import sys
import textwrap
import time
from pathlib import Path
import pytest
# Both tests share the same handoff file: the leaker writes here, the
# verifier reads here. We park it in $TMPDIR with a unique-per-run name
# so concurrent invocations of the suite don't clobber each other.
_HANDOFF_DIR = Path(os.environ.get("TMPDIR", "/tmp")) / "hermes-isolation-probe"
_HANDOFF_DIR.mkdir(exist_ok=True)
def _handoff_path_for(nonce: str) -> Path:
return _HANDOFF_DIR / f"grandchild-{nonce}.json"
def _pid_alive(pid: int) -> bool:
"""POSIX: send signal 0 to probe whether ``pid`` is still alive.
``os.kill(pid, 0)`` raises ``ProcessLookupError`` if the process is
gone, ``PermissionError`` if it exists but we can't signal it
(someone else's pid). We treat PermissionError as "alive" because
the process exists and that's all we need to know.
"""
if sys.platform == "win32": # pragma: no cover — POSIX-only test
# On Windows we'd use OpenProcess + GetExitCodeProcess; this
# test is skipped on Windows so the path is unreachable.
raise RuntimeError("_pid_alive POSIX-only")
try:
os.kill(pid, 0)
except ProcessLookupError:
return False
except PermissionError:
return True
return True
@pytest.mark.skipif(sys.platform == "win32", reason="POSIX-only probe")
@pytest.mark.live_system_guard_bypass
def test_grandchild_leak_is_killed_by_runner(tmp_path: Path) -> None:
"""Run the parallel runner over a probe file and verify cleanup.
1. Materialize a probe file that spawns a long-lived grandchild and
writes its PID to disk before exiting.
2. Invoke ``scripts/run_tests_parallel.py`` against the probe file.
3. Wait for the grandchild PID to vanish (poll for ~5s).
4. Assert the runner exited cleanly AND the grandchild is dead.
"""
repo_root = Path(__file__).resolve().parent.parent
runner = repo_root / "scripts" / "run_tests_parallel.py"
assert runner.exists(), f"runner missing at {runner}"
# Probe lives in a temp dir, NOT under tests/, so the regular suite
# never picks it up — only our explicit invocation does.
probe_dir = tmp_path / "probe"
probe_dir.mkdir()
probe = probe_dir / "test_probe_leaker.py"
nonce = f"{os.getpid()}-{int(time.time() * 1000)}"
handoff = _handoff_path_for(nonce)
if handoff.exists():
handoff.unlink()
probe_src = textwrap.dedent(f"""
import json, os, subprocess, sys, time
from pathlib import Path
HANDOFF = Path({str(handoff)!r})
def test_spawns_grandchild_and_walks_away():
# Long-lived grandchild: detached, ignores SIGTERM (we want
# SIGKILL or process-group kill to be the only thing that
# works, simulating a misbehaving server).
child = subprocess.Popen(
[
sys.executable, "-c",
"import os, signal, sys, time; "
"signal.signal(signal.SIGTERM, signal.SIG_IGN); "
"sys.stdout.write(f'gc-pgid={{os.getpgid(0)}} gc-pid={{os.getpid()}}\\\\n'); "
"sys.stdout.flush(); "
"time.sleep(600)",
],
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
# IMPORTANT: do NOT pass start_new_session here. We want
# the grandchild to inherit the pytest subprocess's
# process group, so when the runner kills the group the
# grandchild dies too.
)
# Read the first line so we can record gc's pgid in the
# handoff, then walk away — don't close the pipe (would
# signal EOF and let the child see SIGPIPE on next write).
first_line = child.stdout.readline().decode().strip()
HANDOFF.write_text(json.dumps({{
"pid": child.pid,
"diag": first_line,
"test_pid": os.getpid(),
"test_pgid": os.getpgid(0),
}}))
assert child.pid > 0
""").strip()
probe.write_text(probe_src + "\n")
# Run the parallel runner against just the probe file. The runner
# discovers under ``tests/`` by default, so we override via --paths.
proc = subprocess.run(
[
sys.executable,
str(runner),
"--paths",
str(probe_dir),
"-j",
"1",
# Tight per-file timeout: the probe finishes in <1s, no
# need for 10min.
"--file-timeout",
"30",
],
cwd=repo_root,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
text=True,
timeout=60,
)
assert handoff.exists(), (
f"probe never wrote handoff file; runner output:\n{proc.stdout}"
)
handoff_data = json.loads(handoff.read_text())
grandchild_pid = handoff_data["pid"]
diag = handoff_data.get("diag", "(no diag)")
test_pid = handoff_data.get("test_pid")
test_pgid = handoff_data.get("test_pgid")
handoff.unlink()
# The runner must have exited cleanly (probe test passes).
assert proc.returncode == 0, (
f"runner exited {proc.returncode}; output:\n{proc.stdout}"
)
# The grandchild must be gone. Poll for a bit because process-group
# SIGKILL + reaping isn't synchronous; on a loaded box it can take
# a beat.
deadline = time.monotonic() + 5.0
while time.monotonic() < deadline:
if not _pid_alive(grandchild_pid):
break
time.sleep(0.05)
else:
# Test cleanup: kill the leaked grandchild ourselves so a
# FAILED assertion doesn't leave a sleep(600) running.
try:
os.kill(grandchild_pid, 9)
except ProcessLookupError:
pass
pytest.fail(
f"grandchild PID {grandchild_pid} survived runner exit; "
f"diag={diag!r} test_pid={test_pid} test_pgid={test_pgid}; "
f"runner output:\n{proc.stdout}"
)
# ── Bare pytest-flag passthrough ─────────────────────────────────────────────
#
# The runner routes any token starting with ``-`` that isn't one of its own
# options (``-j``/``--jobs``, ``--paths``, ``--slice``, ``--file-timeout``,
# ``--generate-slices``, ``--files``, ``--include-integration``) straight
# through to each per-file pytest invocation — no ``--`` separator required.
# Before this, a bare ``-q`` errored out with "unrecognized arguments",
# forcing a retry on every run. These tests are behavior contracts, not
# snapshots: they assert that bare flags reach pytest and that value-taking
# flags (``-k expr``) keep their value instead of having it stolen by the
# positional-path discovery.
def _make_probe_dir(tmp_path: Path) -> Path:
"""Two trivial passing tests, one named test_alpha, one test_beta."""
probe_dir = tmp_path / "probe"
probe_dir.mkdir()
(probe_dir / "test_flagprobe.py").write_text(
"def test_alpha():\n assert True\n\n"
"def test_beta():\n assert True\n"
)
return probe_dir
def _run_runner(probe_dir: Path, *extra: str) -> subprocess.CompletedProcess:
repo_root = Path(__file__).resolve().parent.parent
runner = repo_root / "scripts" / "run_tests_parallel.py"
return subprocess.run(
[sys.executable, str(runner), "--paths", str(probe_dir),
"-j", "1", "--file-timeout", "30", *extra],
cwd=repo_root,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
text=True,
timeout=60,
)
def test_bare_q_flag_passes_through(tmp_path: Path) -> None:
"""A bare ``-q`` (no ``--``) runs clean instead of erroring out."""
probe_dir = _make_probe_dir(tmp_path)
proc = _run_runner(probe_dir, "-q")
assert proc.returncode == 0, proc.stdout
assert "unrecognized arguments" not in proc.stdout
def test_bare_value_flag_keeps_its_value(tmp_path: Path) -> None:
"""``-k test_alpha`` reaches pytest as a selector, not as a path.
The value token (``test_alpha``) must NOT be swallowed by the runner's
positional-path discovery — if it were, discovery would look for a path
named ``test_alpha``, find nothing, and the run would degrade. We assert
the run succeeds AND only one of the two tests was selected (proving the
``-k`` filter actually applied inside pytest).
"""
probe_dir = _make_probe_dir(tmp_path)
proc = _run_runner(probe_dir, "-k", "test_alpha")
assert proc.returncode == 0, proc.stdout
# Exactly one test selected: the per-file summary shows "1✓" (1 passed).
# test_beta is deselected by the -k filter.
assert "1✓" in proc.stdout or "1 passed" in proc.stdout, proc.stdout
assert "2✓" not in proc.stdout, (
f"both tests ran — -k filter did not apply:\n{proc.stdout}"
)
def test_explicit_double_dash_still_works(tmp_path: Path) -> None:
"""The legacy ``--`` separator keeps working alongside bare flags."""
probe_dir = _make_probe_dir(tmp_path)
proc = _run_runner(probe_dir, "-q", "--", "--tb=short")
assert proc.returncode == 0, proc.stdout
assert "unrecognized arguments" not in proc.stdout
def test_positional_path_not_treated_as_flag(tmp_path: Path) -> None:
"""A positional path arg still overrides discovery (not routed to pytest)."""
probe_dir = _make_probe_dir(tmp_path)
repo_root = Path(__file__).resolve().parent.parent
runner = repo_root / "scripts" / "run_tests_parallel.py"
# Pass the probe dir positionally (no --paths), plus a bare -q.
proc = subprocess.run(
[sys.executable, str(runner), str(probe_dir), "-j", "1",
"--file-timeout", "30", "-q"],
cwd=repo_root, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
text=True, timeout=60,
)
assert proc.returncode == 0, proc.stdout
# Discovery found the probe file (2 tests), proving the positional path
# was consumed as a root, not forwarded to pytest as a bad flag.
assert "test_flagprobe.py" in proc.stdout, proc.stdout
def test_file_retry_self_heals_and_prints_both_attempts(tmp_path: Path) -> None:
"""A pass-on-retry is green, loud, and retains the failing traceback."""
repo_root = Path(__file__).resolve().parent.parent
runner = repo_root / "scripts" / "run_tests_parallel.py"
marker = tmp_path / "ran-once"
probe = tmp_path / "test_flaky_probe.py"
probe.write_text(
textwrap.dedent(
f"""
from pathlib import Path
def test_flaky_once():
marker = Path({str(marker)!r})
if not marker.exists():
marker.write_text("failed once")
assert False, "simulated first-attempt flake"
assert True
"""
),
encoding="utf-8",
)
proc = subprocess.run(
[
sys.executable,
str(runner),
"--files",
str(probe),
"--file-retries",
"1",
"-j",
"1",
"-q",
],
cwd=repo_root,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
text=True,
timeout=60,
)
assert proc.returncode == 0, proc.stdout
assert "FLAKY file" in proc.stdout
assert "simulated first-attempt flake" in proc.stdout
assert "first-attempt output" in proc.stdout
assert "retry output" in proc.stdout
def test_file_retry_does_not_launder_deterministic_failure(tmp_path: Path) -> None:
"""A real regression fails both attempts and the runner remains red."""
repo_root = Path(__file__).resolve().parent.parent
runner = repo_root / "scripts" / "run_tests_parallel.py"
probe = tmp_path / "test_red_probe.py"
probe.write_text(
"def test_always_red():\n assert False, 'deterministic regression'\n",
encoding="utf-8",
)
proc = subprocess.run(
[
sys.executable,
str(runner),
"--files",
str(probe),
"--file-retries",
"1",
"-j",
"1",
"-q",
],
cwd=repo_root,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
text=True,
timeout=60,
)
assert proc.returncode == 1, proc.stdout
assert "deterministic regression" in proc.stdout
assert "FLAKY file" not in proc.stdout