fix(dashboard): pin machine-dashboard reroute to the machine root, not $HOME/.hermes (#46487)

The unified machine-dashboard reroute (cmd_dashboard) re-execs a named-profile
dashboard launch as the machine dashboard and dropped HERMES_HOME from the
child env with the comment "so the child binds the machine root". That holds
for a standard install (root == ~/.hermes) but breaks the Docker layout: the
published image sets `ENV HERMES_HOME=/opt/data`, so once HERMES_HOME is unset
the child falls back to $HOME/.hermes = /opt/data/.hermes — an empty,
auto-seeded home.

Two user-visible symptoms, one root cause (reported via support):

1. Dashboard Profiles page shows only an empty `default` — the real
   default/oracle/saga profiles live under /opt/data/profiles, but the
   rerouted child resolves _get_profiles_root() to /opt/data/.hermes/profiles.

2. The "Update Hermes" button runs `hermes update` inside the container
   repeatedly instead of bailing with the docker-update guidance. The Docker
   guard keys off detect_install_method(), which reads
   $HERMES_HOME/.install_method; the image stamps that at /opt/data, but the
   misresolved home has no stamp, no HERMES_MANAGED, and no .git → falls
   through to "pip", so the guard never fires.

The reporter's workaround was to bind-mount the host dir at both /opt/data and
/opt/data/.hermes so the two paths converge (at the cost of a self-referential
recursion).

Fix: resolve the machine root explicitly with get_default_hermes_root() and set
it on the child env instead of popping HERMES_HOME. That helper returns the
root for both layouts — ~/.hermes for a standard install, and /opt/data for
Docker (it strips a trailing profiles/<name>). Falls back to the old pop
behaviour only if root resolution raises, so the reroute is never blocked.

Regression tests in test_dashboard_unified_launch.py: the existing standard-
install test now asserts the child carries HERMES_HOME == get_default_hermes_root()
(not absent), and a new test_reexec_pins_docker_machine_root covers the Docker
layout (HERMES_HOME=/opt/data/profiles/oracle → child gets /opt/data). Both
fail against the pre-fix pop behaviour (mutation-verified).
This commit is contained in:
Ben Barclay 2026-06-15 15:33:15 +10:00 committed by GitHub
parent c2b7669ad3
commit 80f8ffc74c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 59 additions and 4 deletions

View File

@ -10753,8 +10753,24 @@ def cmd_dashboard(args):
if getattr(args, "skip_build", False):
reexec_argv.append("--skip-build")
env = os.environ.copy()
# Drop the profile HERMES_HOME so the child binds the machine root.
env.pop("HERMES_HOME", None)
# Pin the child to the machine ROOT, not the launching profile's
# HERMES_HOME. We must resolve the root explicitly instead of just
# dropping HERMES_HOME: in the Docker layout the machine root is
# /opt/data (set via `ENV HERMES_HOME=/opt/data`), so an unset
# HERMES_HOME falls back to $HOME/.hermes = /opt/data/.hermes — an
# empty, auto-seeded home where the dashboard sees only the default
# profile and the install-method stamp is missing (so the Docker
# update-button guard also misfires). get_default_hermes_root()
# returns the root for both layouts: ~/.hermes for a standard install
# and /opt/data for Docker (it strips a trailing profiles/<name>).
# See the support report for the double-mount workaround this avoids.
try:
from hermes_constants import get_default_hermes_root
env["HERMES_HOME"] = str(get_default_hermes_root())
except Exception:
# Best-effort: if root resolution fails, fall back to the prior
# behaviour (drop HERMES_HOME) rather than block the reroute.
env.pop("HERMES_HOME", None)
# On Windows, os.execvpe() does not truly replace the process — it
# spawns via CreateProcess then the parent exits. Under Python 3.14+
# this can crash with STATUS_ACCESS_VIOLATION (0xC0000005) when

View File

@ -57,6 +57,7 @@ class TestUnifiedDashboardRouting:
assert opened == ["http://127.0.0.1:9119/?profile=worker_x"]
def test_profile_launch_reexecs_machine_dashboard(self, main_mod, monkeypatch):
monkeypatch.delenv("HERMES_HOME", raising=False)
monkeypatch.setattr(
"hermes_cli.profiles.get_active_profile_name", lambda: "worker_x"
)
@ -79,8 +80,46 @@ class TestUnifiedDashboardRouting:
assert "-p" in argv and argv[argv.index("-p") + 1] == "default"
assert "--open-profile" in argv
assert argv[argv.index("--open-profile") + 1] == "worker_x"
# Profile HERMES_HOME dropped so the child binds the machine root.
assert "HERMES_HOME" not in env
# The child is pinned to the machine ROOT, not the launching profile's
# HERMES_HOME. For a standard install (HERMES_HOME unset) that root is
# the platform-native default (~/.hermes), NOT dropped — see the Docker
# test below for why we resolve explicitly instead of popping.
from hermes_constants import get_default_hermes_root
assert env.get("HERMES_HOME") == str(get_default_hermes_root())
def test_reexec_pins_docker_machine_root(self, main_mod, monkeypatch):
"""In the Docker layout (HERMES_HOME=/opt/data, profiles under
/opt/data/profiles/<name>) the reroute must pin the child to the
machine root /opt/data NOT drop HERMES_HOME.
Dropping it makes the child fall back to $HOME/.hermes
(= /opt/data/.hermes), an empty auto-seeded home, so the dashboard
shows only the default profile and the .install_method stamp is
missing (which also misfires the Docker update-button guard).
Regression test for the support report.
"""
monkeypatch.setenv("HERMES_HOME", "/opt/data/profiles/oracle")
monkeypatch.setattr(
"hermes_cli.profiles.get_active_profile_name", lambda: "oracle"
)
monkeypatch.setattr(main_mod, "_dashboard_listening", lambda host, port: False)
execs = []
def fake_exec(exe, argv, env):
execs.append((exe, argv, env))
raise SystemExit(0)
monkeypatch.setattr(main_mod.os, "execvpe", fake_exec)
with pytest.raises(SystemExit):
main_mod.cmd_dashboard(_args())
assert len(execs) == 1
_exe, _argv, env = execs[0]
# get_default_hermes_root() strips the trailing profiles/<name>, so the
# child binds /opt/data — where the real default/oracle/saga profiles
# and the .install_method stamp actually live.
assert env.get("HERMES_HOME") == "/opt/data"
def test_desktop_profile_backend_skips_machine_dashboard_reroute(self, main_mod, monkeypatch):
"""A desktop-spawned named-profile backend (HERMES_DESKTOP=1) must NOT