kshitijk4poor
060c0cd0f8
The resume token only recorded `cold_start_if_installed: bool` and execution re-read the mutable one-shot start-attestation marker to decide whether a Desktop-owned install still owed a cold-start. A concurrent `hermes gateway status`/`start` (`check_start_attestation`) consumes that marker between plan and execution, so the spawn was skipped and the token cleared with no gateway running. The marker now carries a `generation` nonce. The plan records the generation whose unclean death authorized the cold-start on the token (`attested_generation`); execution authorizes the spawn from the token, still re-checks live gateway PIDs, and consumes the marker only while it is still that generation — a newer marker written by a concurrent start keeps its own report. `attested_gateway_died` becomes `attested_death_generation` (`None` = undecidable, fail closed). Follow-up to #110020 review thread (a).
69 lines
2.9 KiB
Python
69 lines
2.9 KiB
Python
"""#84185: a Windows gateway cold-started after update that dies immediately
|
|
(e.g. a job object denying breakaway) must not be reported as started.
|
|
|
|
``_cold_start_windows_gateway_after_update`` used to print the success line
|
|
straight off a successful ``Popen`` return, which only proves the process was
|
|
created, not that it survived. This asserts the observable output: the
|
|
success line is gated on the process actually being found alive afterwards,
|
|
same as every other ``_spawn_detached`` caller.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import pytest
|
|
|
|
from hermes_cli import gateway as hermes_gateway
|
|
from hermes_cli import gateway_windows
|
|
from hermes_cli import main as cli_main
|
|
import hermes_cli.main_install_repair as main_install_repair
|
|
from hermes_cli import update_cmd
|
|
|
|
|
|
def _run_cold_start(monkeypatch, capsys, *, surviving_pids):
|
|
monkeypatch.setattr(cli_main, "_is_windows", lambda: True)
|
|
monkeypatch.setattr(main_install_repair, "_is_windows", lambda: True)
|
|
|
|
# The pre-spawn re-check (``all_profiles=True``) must find nothing
|
|
# running so the cold-start path proceeds and actually spawns.
|
|
monkeypatch.setattr(
|
|
hermes_gateway,
|
|
"find_gateway_pids",
|
|
lambda all_profiles=False: [] if all_profiles else surviving_pids,
|
|
)
|
|
monkeypatch.setattr(gateway_windows, "_spawn_detached", lambda: 4242)
|
|
# Avoid the real 6s/0.4s poll loop in _report_gateway_start.
|
|
monkeypatch.setattr(
|
|
gateway_windows, "_wait_for_gateway_ready", lambda *a, **k: surviving_pids
|
|
)
|
|
|
|
update_cmd._cold_start_windows_gateway_after_update()
|
|
|
|
return capsys.readouterr().out
|
|
|
|
|
|
def test_cold_start_raises_when_process_does_not_survive(monkeypatch, capsys):
|
|
with pytest.raises(RuntimeError, match="did not become ready"):
|
|
_run_cold_start(monkeypatch, capsys, surviving_pids=[])
|
|
|
|
assert "✓ Starting Windows gateway after update" not in capsys.readouterr().out
|
|
|
|
|
|
def test_failed_readiness_keeps_the_dead_attestation_for_the_retry(monkeypatch, tmp_path, capsys):
|
|
"""#110020 review: the attestation is the retry's only authority to spawn under Desktop
|
|
ownership, so a spawn that returns a PID but never becomes ready must NOT consume it —
|
|
otherwise the registered retry sees ownership with no marker and "succeeds" with no gateway."""
|
|
monkeypatch.setattr("hermes_cli.config.get_hermes_home", lambda: str(tmp_path))
|
|
gateway_windows._write_start_attestation([555], "direct spawn (PID 555)")
|
|
marker = tmp_path / "state" / "gateway.start-attestation.json"
|
|
|
|
with pytest.raises(RuntimeError, match="did not become ready"):
|
|
_run_cold_start(monkeypatch, capsys, surviving_pids=[])
|
|
assert marker.exists()
|
|
assert gateway_windows.attested_death_generation(current_pids=[]) is not None
|
|
|
|
|
|
def test_cold_start_reports_success_when_process_survives(monkeypatch, capsys):
|
|
out = _run_cold_start(monkeypatch, capsys, surviving_pids=[4242])
|
|
|
|
assert "✓ Gateway started via cold-start after update" in out
|