feat(sync): wire state.json + divergence detection + dialogs
CI / test (3.10) (push) Successful in 3m16s
CI / test (3.11) (push) Failing after 3m25s
CI / test (3.12) (push) Failing after 3m25s
CI / build-pyz (push) Has been skipped
CI / release (push) Has been skipped

state.py: per-instance sync state. <pack>/.cloud-sync/state.json
(mode 600) records last_pulled_snapshot_id + last_pulled_at +
host_tag. Versioned schema. clear() on remote-empty.

sync.pull decision tree (replaces the unconditional restore):

  no token file
    → prompt_login_qt; on Skip return 0 (don't block launch)
  no state + remote empty
    → no-op
  no state + remote non-empty
    → restore (first-run on this machine)
  state.id == remote.id
    → skip restore (up to date)
  state.id != remote.id, no in-scope local edits since state.at
    → restore (fast-forward)
  state.id != remote.id, in-scope local edits since state.at
    → prompt_conflict_qt
      keep_local  → don't restore; push will overwrite cloud
      use_remote  → restore + update state
      cancel      → exit 1

sync.push: --json output parsed for snapshot_id; state.json updated
to that id after a successful backup. Skips silently if no token.

_find_modified_in_scope: walks include roots, filters via
_matches_any (restic-style globs: dir/, **/dir/, **/*.glob).
Stops at 50 hits; we only need 'any' + a sample for the dialog.

_format_dt: hand-rolled (no GNU-vs-Windows strftime quirks) →
'Thursday, October 21, 2021 at 7:12 PM'.

Restic JSON parsing helpers: _parse_snapshots, _parse_restic_time
(handles nanosecond precision), _parse_backup_summary.

tests/test_state.py: 19 new tests covering state read/write, scope-
aware mtime walk, exclude glob matching, restic output parsers.
Total: 52 green.
This commit is contained in:
2026-06-05 00:20:40 +02:00
parent fa04fade74
commit 7c9d33f952
3 changed files with 563 additions and 32 deletions
+84
View File
@@ -0,0 +1,84 @@
"""Per-instance sync state.
Tracks the snapshot id this pack was last synced to and when. Lives at
``<pack-folder>/.cloud-sync/state.json`` (mode 600).
Purpose: divergence detection. On ``pull``, if the remote latest id
differs from ``last_pulled_snapshot_id`` AND any in-scope local file
has mtime > ``last_pulled_at``, the local and remote diverged from a
common ancestor — surface the conflict dialog.
"""
from __future__ import annotations
import json
from dataclasses import dataclass
from datetime import datetime, timezone
from pathlib import Path
SCHEMA_VERSION = 1
@dataclass(frozen=True)
class State:
last_pulled_snapshot_id: str
last_pulled_at: datetime
host_tag: str = "cloud-sync"
def state_path(pack_folder: Path) -> Path:
return pack_folder / ".cloud-sync" / "state.json"
def read(pack_folder: Path) -> State | None:
"""Return parsed state or None if file missing / unreadable / wrong schema."""
p = state_path(pack_folder)
if not p.exists():
return None
try:
data = json.loads(p.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError):
return None
if data.get("schema") != SCHEMA_VERSION:
return None
try:
return State(
last_pulled_snapshot_id=data["last_pulled_snapshot_id"],
last_pulled_at=_parse_iso(data["last_pulled_at"]),
host_tag=data.get("host_tag", "cloud-sync"),
)
except (KeyError, ValueError):
return None
def write(pack_folder: Path, state: State) -> None:
"""Persist state. Creates parent dir + sets mode 600."""
p = state_path(pack_folder)
p.parent.mkdir(parents=True, exist_ok=True)
payload = {
"schema": SCHEMA_VERSION,
"last_pulled_snapshot_id": state.last_pulled_snapshot_id,
"last_pulled_at": state.last_pulled_at.astimezone(timezone.utc)
.isoformat()
.replace("+00:00", "Z"),
"host_tag": state.host_tag,
}
p.write_text(json.dumps(payload, indent=2) + "\n", encoding="utf-8")
p.chmod(0o600)
def clear(pack_folder: Path) -> None:
"""Remove state.json if present. Used when remote has zero snapshots."""
p = state_path(pack_folder)
p.unlink(missing_ok=True)
def _parse_iso(s: str) -> datetime:
"""Parse ISO-8601 with trailing Z or +HH:MM, return tz-aware UTC."""
if s.endswith("Z"):
s = s[:-1] + "+00:00"
dt = datetime.fromisoformat(s)
if dt.tzinfo is None:
dt = dt.replace(tzinfo=timezone.utc)
return dt.astimezone(timezone.utc)
+287 -32
View File
@@ -1,38 +1,52 @@
"""pull + push entry points.
"""pull + push entry points + divergence detection.
Both subprocess restic against ``rest:<scheme>://<id>:<password>@<host>/<id>/``
where the password is HTTP basic auth ONLY. Restic repos are initialised with
``--insecure-no-password`` so no encryption-at-rest password exists; protection
relies on:
Backend: restic-rest-server. URL form:
``rest:<scheme>://<id>:<password>@<host>/<id>/``
1. TLS in transit at the reverse proxy
2. ``--private-repos`` + htpasswd per user at restic-rest-server
3. ``--append-only`` to prevent client-side deletion
4. Disk-level encryption (LUKS) on the host
Password = HTTP basic auth ONLY. Restic repos are initialised with
``--insecure-no-password``; encryption-at-rest delegated to LUKS on the
host disk + TLS at the reverse proxy.
Defense-in-depth via repo encryption was dropped because the threat model
(homelab, operator-trusted) doesn't justify the password-coordination cost.
Divergence detection (pull only):
1. Fetch remote latest snapshot id + time via ``snapshots --json --latest 1``.
2. Read ``<pack>/.cloud-sync/state.json`` for the last-pulled snapshot id.
3. Branches:
- no state.json AND remote empty → no-op
- no state.json AND remote non-empty → first-run restore
- state.last_pulled == remote.id → up to date, skip
- state.last_pulled != remote.id, no local edits since
state.last_pulled_at → fast-forward restore
- state.last_pulled != remote.id, local edits since
state.last_pulled_at → CONFLICT dialog
"""
from __future__ import annotations
import fnmatch
import json
import sys
import urllib.parse
from datetime import datetime, timezone
from pathlib import Path
from . import restic, scope as scopemod
from . import restic, scope as scopemod, state as statemod
from .cli import Args
from .creds import read_credentials
from .ui import HeadlessProgress, Progress
def pull(args: Args, progress: Progress | None = None) -> int:
"""Restore latest snapshot's files into pack_folder.
If the repo has no snapshots yet, this is a no-op (first run on this
machine; nothing to restore).
"""
ui = progress or HeadlessProgress()
# First-run login. If the user declines, skip cloud sync without
# blocking the launch (return 0 — non-fatal for Prism PreLaunch).
if not args.token_file.exists():
if not _prompt_login_and_save(args, ui):
ui.set_status("Cloud sync skipped")
print("cloud-sync: no token; skipping pull")
return 0
ui.set_status("Reading credentials…")
discord_id, password = read_credentials(args.token_file)
@@ -56,18 +70,56 @@ def pull(args: Args, progress: Progress | None = None) -> int:
file=sys.stderr,
)
return 2
stripped = out.strip()
if stripped in ("", "null", "[]"):
snapshots = _parse_snapshots(out)
if not snapshots:
statemod.clear(args.pack_folder)
ui.set_status("No snapshots yet — nothing to pull")
print(
"cloud-sync: no snapshots yet for this user "
"(first run on this machine?); nothing to pull"
)
print("cloud-sync: no snapshots yet for this user; nothing to pull")
return 0
scope = scopemod.load(args.pack_folder)
_, exclude_from = scopemod.materialize_for_restic(args.pack_folder, scope)
remote = snapshots[0]
remote_id = remote["id"]
remote_time = _parse_restic_time(remote["time"])
local_state = statemod.read(args.pack_folder)
scope = scopemod.load(args.pack_folder)
decision: str
if local_state is None:
decision = "use_remote"
elif local_state.last_pulled_snapshot_id == remote_id:
ui.set_status("Cloud is up to date")
print("cloud-sync: already at latest snapshot")
return 0
else:
modified = _find_modified_in_scope(
args.pack_folder, scope, local_state.last_pulled_at
)
if not modified:
decision = "use_remote"
else:
decision = _ask_conflict(modified, remote_time)
if decision is None:
# UI unavailable in headless mode → conservative: cancel
ui.set_status("Conflict detected; no UI available")
print(
"cloud-sync: conflict detected (remote moved + local edits) "
"but headless mode can't prompt; aborting",
file=sys.stderr,
)
return 1
if decision == "cancel":
ui.set_status("Launch cancelled")
return 1
if decision == "keep_local":
ui.set_status("Keeping local; push will overwrite cloud on exit")
print("cloud-sync: keeping local copy")
return 0
# decision == "use_remote"
_, exclude_from = scopemod.materialize_for_restic(args.pack_folder, scope)
ui.set_status("Restoring files…")
code, _ = restic.run(
binary,
@@ -85,14 +137,27 @@ def pull(args: Args, progress: Progress | None = None) -> int:
if code != 0:
print(f"cloud-sync: restic restore failed (exit {code})", file=sys.stderr)
return 2
statemod.write(
args.pack_folder,
statemod.State(
last_pulled_snapshot_id=remote_id,
last_pulled_at=datetime.now(timezone.utc),
),
)
ui.set_status("Pull complete")
print("cloud-sync: pull ok")
return 0
def push(args: Args, progress: Progress | None = None) -> int:
"""Snapshot the in-scope files into the user's repo."""
ui = progress or HeadlessProgress()
if not args.token_file.exists():
ui.set_status("No cloud token; skipping push")
print("cloud-sync: no token; skipping push")
return 0
ui.set_status("Reading credentials…")
discord_id, password = read_credentials(args.token_file)
@@ -105,7 +170,7 @@ def push(args: Args, progress: Progress | None = None) -> int:
files_from, exclude_from = scopemod.materialize_for_restic(args.pack_folder, scope)
ui.set_status("Uploading snapshot…")
code, _ = restic.run(
code, out = restic.run(
binary,
[
"-r", repo, "--insecure-no-password",
@@ -114,6 +179,7 @@ def push(args: Args, progress: Progress | None = None) -> int:
"--exclude-file", str(exclude_from),
"--host", "cloud-sync",
"--tag", "auto",
"--json",
],
env=env,
cwd=args.pack_folder,
@@ -124,22 +190,211 @@ def push(args: Args, progress: Progress | None = None) -> int:
if code != 0:
print(f"cloud-sync: restic backup failed (exit {code})", file=sys.stderr)
return 2
new_id = _parse_backup_summary(out)
if new_id is not None:
statemod.write(
args.pack_folder,
statemod.State(
last_pulled_snapshot_id=new_id,
last_pulled_at=datetime.now(timezone.utc),
),
)
ui.set_status("Push complete")
print("cloud-sync: push ok")
return 0
# ---------------------------------------------------------------------------
# helpers
# divergence detection
# ---------------------------------------------------------------------------
def _restic_repo(base_url: str, discord_id: str, password: str) -> str:
"""Build rest:<scheme>://<id>:<pw>@<host>/<id>/
def _find_modified_in_scope(
pack_folder: Path, scope: scopemod.Scope, since: datetime
) -> list[tuple[Path, datetime]]:
"""Return (rel_path, mtime) for in-scope files newer than ``since``.
URL-embedded basic auth is universally supported by restic; alternative
env vars (RESTIC_REST_USERNAME, RESTIC_REST_PASSWORD) require 0.16+.
Walks each include root, skips paths matching any exclude glob. Stops
early at 50 hits (we only need to know "any" + sample some for the
conflict dialog). False positives are safer than false negatives —
a spurious conflict shows a dismissable dialog; a missed conflict
silently overwrites the user's edits.
"""
since_ts = since.timestamp()
hits: list[tuple[Path, datetime]] = []
for include in scope.include:
root = pack_folder / include.rstrip("/")
if not root.exists():
continue
candidates: list[Path] = [root] if root.is_file() else _walk_files(root)
for f in candidates:
try:
mtime = f.stat().st_mtime
except OSError:
continue
if mtime <= since_ts:
continue
rel = f.relative_to(pack_folder)
if _matches_any(rel, scope.exclude):
continue
hits.append((rel, datetime.fromtimestamp(mtime, tz=timezone.utc)))
if len(hits) >= 50:
return hits
return hits
def _walk_files(root: Path):
try:
for p in root.rglob("*"):
if p.is_file():
yield p
except OSError:
return
def _matches_any(rel: Path, patterns: list[str]) -> bool:
"""Restic-style glob match against the relative path.
Subset of restic's exclude semantics that covers our default scope:
``foo/`` — foo itself OR anything under it
``**/foo/`` — any ancestor directory named foo, recursively
``foo/bar*`` — fnmatch against the full relative path
``**/*.log`` — fnmatch the basename (and any tail subpath)
"""
rel_str = str(rel)
parts = rel.parts
for pat in patterns:
if pat.endswith("/"):
core = pat.rstrip("/")
if core.startswith("**/"):
tail = core[len("**/") :]
if tail in parts:
return True
continue
if rel_str == core or rel_str.startswith(core + "/"):
return True
continue
if fnmatch.fnmatch(rel_str, pat):
return True
if pat.startswith("**/"):
suffix = pat[len("**/") :]
if fnmatch.fnmatch(rel.name, suffix):
return True
for i in range(len(parts)):
if fnmatch.fnmatch("/".join(parts[i:]), suffix):
return True
return False
def _ask_conflict(
modified: list[tuple[Path, datetime]],
remote_time: datetime,
) -> str | None:
"""Show the conflict dialog. Returns choice or None if no UI available."""
try:
from .ui_qt import prompt_conflict_qt
except ImportError:
return None
newest = max(modified, key=lambda h: h[1])
return prompt_conflict_qt(
local_modified=_format_dt(newest[1]),
remote_modified=_format_dt(remote_time),
save_label="Minecraft save",
)
def _prompt_login_and_save(args: Args, ui: Progress) -> bool:
"""First-run login. Returns True if a token was saved, False if skipped."""
try:
from .ui_qt import prompt_login_qt
except ImportError:
ui.set_status("No token and no UI; can't prompt")
print(
"cloud-sync: no token at "
f"{args.token_file} and no Qt UI available",
file=sys.stderr,
)
return False
token = prompt_login_qt()
if token is None:
return False
args.token_file.parent.mkdir(parents=True, exist_ok=True)
args.token_file.write_text(token + "\n", encoding="utf-8")
args.token_file.chmod(0o600)
return True
# ---------------------------------------------------------------------------
# restic output parsing
# ---------------------------------------------------------------------------
def _parse_snapshots(out: str) -> list[dict]:
s = out.strip()
if not s or s == "null":
return []
try:
data = json.loads(s)
except json.JSONDecodeError:
return []
return data if isinstance(data, list) else []
def _parse_restic_time(s: str) -> datetime:
"""Restic emits e.g. ``2026-06-04T18:33:21.123456789Z``."""
if s.endswith("Z"):
s = s[:-1] + "+00:00"
# Python's fromisoformat doesn't accept nanosecond precision — trim to micro.
if "." in s:
head, _, tail = s.partition(".")
frac, _, tz = tail.partition("+")
if tz:
tz = "+" + tz
else:
frac, _, tz = tail.partition("-")
if tz:
tz = "-" + tz
s = f"{head}.{frac[:6]}{tz}"
return datetime.fromisoformat(s).astimezone(timezone.utc)
def _parse_backup_summary(out: str) -> str | None:
"""restic backup --json emits one JSON line per event; the final
``summary`` event carries ``snapshot_id``."""
for line in reversed(out.splitlines()):
line = line.strip()
if not line:
continue
try:
ev = json.loads(line)
except json.JSONDecodeError:
continue
if ev.get("message_type") == "summary" and "snapshot_id" in ev:
return str(ev["snapshot_id"])
return None
# ---------------------------------------------------------------------------
# misc helpers
# ---------------------------------------------------------------------------
def _format_dt(dt: datetime) -> str:
"""Format a tz-aware datetime as 'Thursday, October 21, 2021 at 7:12 PM'.
Hand-rolled instead of ``strftime("%-d")`` because GNU strftime's
leading-zero stripping syntax is platform-specific (``%-d`` vs ``%#d``).
"""
local = dt.astimezone()
weekday = local.strftime("%A")
month = local.strftime("%B")
hour = local.hour % 12 or 12
ampm = local.strftime("%p")
return f"{weekday}, {month} {local.day}, {local.year} at {hour}:{local.minute:02d} {ampm}"
def _restic_repo(base_url: str, discord_id: str, password: str) -> str:
raw = base_url.strip()
if raw.startswith("rest:"):
raw = raw[len("rest:"):]
+192
View File
@@ -0,0 +1,192 @@
"""state.json read/write + divergence helpers."""
from __future__ import annotations
import json
from datetime import datetime, timedelta, timezone
from pathlib import Path
import pytest
from cloud_sync import scope as scopemod
from cloud_sync import state as statemod
from cloud_sync.sync import (
_find_modified_in_scope,
_format_dt,
_matches_any,
_parse_backup_summary,
_parse_restic_time,
_parse_snapshots,
)
# ---- state.read/write ----
def test_read_missing_returns_none(tmp_path: Path):
assert statemod.read(tmp_path) is None
def test_write_then_read_roundtrip(tmp_path: Path):
dt = datetime(2026, 6, 5, 12, 34, 56, tzinfo=timezone.utc)
statemod.write(
tmp_path,
statemod.State(last_pulled_snapshot_id="abc123", last_pulled_at=dt),
)
got = statemod.read(tmp_path)
assert got is not None
assert got.last_pulled_snapshot_id == "abc123"
assert got.last_pulled_at == dt
def test_write_sets_mode_600(tmp_path: Path):
statemod.write(
tmp_path,
statemod.State(
last_pulled_snapshot_id="x",
last_pulled_at=datetime.now(timezone.utc),
),
)
mode = statemod.state_path(tmp_path).stat().st_mode & 0o777
assert mode == 0o600
def test_clear_idempotent(tmp_path: Path):
statemod.clear(tmp_path) # no-op when missing
statemod.write(
tmp_path,
statemod.State(
last_pulled_snapshot_id="x",
last_pulled_at=datetime.now(timezone.utc),
),
)
assert statemod.state_path(tmp_path).exists()
statemod.clear(tmp_path)
assert not statemod.state_path(tmp_path).exists()
def test_wrong_schema_returns_none(tmp_path: Path):
p = tmp_path / ".cloud-sync" / "state.json"
p.parent.mkdir(parents=True)
p.write_text(
json.dumps({"schema": 999, "last_pulled_snapshot_id": "x", "last_pulled_at": "2026-01-01T00:00:00Z"})
)
assert statemod.read(tmp_path) is None
def test_garbage_json_returns_none(tmp_path: Path):
p = tmp_path / ".cloud-sync" / "state.json"
p.parent.mkdir(parents=True)
p.write_text("{not json")
assert statemod.read(tmp_path) is None
# ---- _find_modified_in_scope ----
def _touch(p: Path, content: str = "x") -> None:
p.parent.mkdir(parents=True, exist_ok=True)
p.write_text(content)
def test_no_in_scope_changes_returns_empty(tmp_path: Path):
_touch(tmp_path / "options.txt")
# mtime is "now", `since` is the future
future = datetime.now(timezone.utc) + timedelta(hours=1)
hits = _find_modified_in_scope(tmp_path, scopemod.Scope(), future)
assert hits == []
def test_in_scope_file_modified_after_since_detected(tmp_path: Path):
_touch(tmp_path / "options.txt")
past = datetime.now(timezone.utc) - timedelta(hours=1)
hits = _find_modified_in_scope(tmp_path, scopemod.Scope(), past)
assert any(rel.name == "options.txt" for rel, _ in hits)
def test_excluded_path_skipped(tmp_path: Path):
# config/packwiz-installer.log matches exclude "**/*.log" AND "config/packwiz*"
_touch(tmp_path / "config" / "packwiz-installer.log")
past = datetime.now(timezone.utc) - timedelta(hours=1)
hits = _find_modified_in_scope(tmp_path, scopemod.Scope(), past)
assert hits == []
def test_only_excluded_changes_not_a_conflict(tmp_path: Path):
_touch(tmp_path / ".cloud-sync" / "files-from.txt") # excluded
_touch(tmp_path / "config" / "packwiz-installer.log") # excluded
past = datetime.now(timezone.utc) - timedelta(hours=1)
hits = _find_modified_in_scope(tmp_path, scopemod.Scope(), past)
assert hits == []
def test_walks_subdirectories(tmp_path: Path):
_touch(tmp_path / "config" / "fabric" / "custom.json")
past = datetime.now(timezone.utc) - timedelta(hours=1)
hits = _find_modified_in_scope(tmp_path, scopemod.Scope(), past)
assert any("custom.json" in str(rel) for rel, _ in hits)
# ---- _matches_any ----
def test_dir_pattern_matches_anything_under(tmp_path: Path):
assert _matches_any(Path(".cloud-sync/token"), [".cloud-sync/"])
assert _matches_any(Path("a/b/cache/x"), ["**/cache/"])
assert not _matches_any(Path("config/options.txt"), [".cloud-sync/"])
def test_glob_pattern(tmp_path: Path):
assert _matches_any(Path("config/packwiz-installer.log"), ["config/packwiz*"])
assert _matches_any(Path("foo/bar.log"), ["**/*.log"])
# ---- restic output parsing ----
def test_parse_snapshots_empty():
assert _parse_snapshots("") == []
assert _parse_snapshots("null") == []
assert _parse_snapshots("[]") == []
def test_parse_snapshots_one():
out = json.dumps([{"id": "abc", "time": "2026-06-05T12:00:00Z"}])
parsed = _parse_snapshots(out)
assert len(parsed) == 1
assert parsed[0]["id"] == "abc"
def test_parse_restic_time_with_nanos():
dt = _parse_restic_time("2026-06-04T18:33:21.123456789Z")
assert dt.tzinfo is timezone.utc
assert dt.year == 2026 and dt.day == 4 and dt.hour == 18
def test_parse_backup_summary_finds_snapshot_id():
out = (
'{"message_type":"status","percent_done":0.5}\n'
'{"message_type":"status","percent_done":1.0}\n'
'{"message_type":"summary","snapshot_id":"deadbeef","files_new":3}\n'
)
assert _parse_backup_summary(out) == "deadbeef"
def test_parse_backup_summary_missing_returns_none():
assert _parse_backup_summary("") is None
assert _parse_backup_summary("not json\nstill not json") is None
# ---- _format_dt ----
def test_format_dt_strips_leading_zero_on_hour():
# 7:12 PM, not 07:12 PM. Day of month also no leading zero.
dt = datetime(2021, 10, 21, 19, 12, tzinfo=timezone.utc)
out = _format_dt(dt)
# Output is local-time-converted, so don't pin the weekday/AM-PM exactly
# in case CI runs UTC vs PT; just check the formatting shape.
assert "October" in out or "Oct" not in out # full month name
assert ", 2021 at" in out
assert " AM" in out or " PM" in out
assert " 0:" not in out # no leading-zero hour