Windows file in use: every state-file swap goes through atomic.replace (retry); a sharing violation no longer says run as yourself

serve/runconfig.py

@@ -15,6 +15,7 @@ import shutil
 from pathlib import Path
 
 from strata_control import user_args
+from strata_control import atomic as _atomic_mod
 
 # (key, kind, help).  kind: "bool", "int>=0", "num>=0", ("enum", values), "names", or ("sampling", check) for a key
 # of the "sampling" block, ("arg", flag) for an engine option kept in "args".
@@ -179,7 +180,7 @@ def save(path: str | Path, cfg: dict) -> Path:
     shutil.copyfile(path, bak)
     tmp = path.with_name(path.name + ".tmp")
     tmp.write_text(json.dumps(cfg, indent=1), encoding="utf-8")
-    os.replace(tmp, path)
+    _atomic_mod.replace(tmp, path)
     return bak
 
 

strata_cli/bench.py

@@ -33,6 +33,7 @@ from pathlib import Path
 
 from strata_cli.client import ClientError, Endpoint, _NoRedirect
 from strata_control.profiles import ProfileError
+from strata_control import atomic as _atomic_mod
 
 
 SCHEMA = 1
@@ -685,7 +686,7 @@ def _progress_write(path: Path, record: dict, *, first: bool = False) -> None:
     temporary = path.with_name(".progress-" + uuid.uuid4().hex + ".tmp")
     try:
         _write_new(temporary, encoded)
-        os.replace(temporary, path)
+        _atomic_mod.replace(temporary, path)
     finally:
         temporary.unlink(missing_ok=True)
 

strata_cli/cli.py

@@ -1982,10 +1982,15 @@ def _access_retry(a: argparse.Namespace, error: BaseException) -> int | None:
     if not (isinstance(error, PermissionError) or "PermissionError" in str(error)):
         return None
     as_json = bool(getattr(a, "json", False) or getattr(a, "output", None) in ("json", "jsonl"))
+    in_use = private_files.is_sharing_violation(error)
     if getattr(a, "_access_retried", False):
-        return _error(f"{private_files.ACCESS_MESSAGE} (details: {error})", 5, as_json)
+        message = private_files.IN_USE_MESSAGE if in_use else private_files.ACCESS_MESSAGE
+        return _error(f"{message} (details: {error})", 5, as_json)
     a._access_retried = True
-    private_files.repair_known_roots()
+    if in_use:
+        time.sleep(1.5)                                  # a scanner or a sync tool lets go of the file in a moment; repairing would not help
+    else:
+        private_files.repair_known_roots()
     return _run(a)
 
 

strata_cli/connections.py

@@ -6,12 +6,18 @@ CLI resolve the same saved connection without importing server or installer code
 
 from __future__ import annotations
 
+import os
 try:
     from strata_control.private_files import private_mkdir as _private_mkdir
 except ImportError:                                     # a thin bundle (monitor only) ships without strata_control
     def _private_mkdir(path, *, parents=False, exist_ok=True):
         path.mkdir(mode=0o700, parents=parents, exist_ok=exist_ok)
         return path
+try:
+    from strata_control import atomic as _atomic_mod
+except ImportError:                                     # the same thin bundle: a plain replace
+    class _atomic_mod:                                  # noqa: N801
+        replace = staticmethod(os.replace)
 import hashlib
 import json
 import os
@@ -158,7 +164,7 @@ def _replace(path: Path, data: bytes) -> None:
             stream.write(data)
             stream.flush()
             os.fsync(stream.fileno())
-        os.replace(name, path)
+        _atomic_mod.replace(name, path)
         if os.name == "posix":
             try:
                 parent = os.open(path.parent, os.O_RDONLY)

strata_cli/launch/runner.py

@@ -21,6 +21,7 @@ import sys
 import tempfile
 import threading
 import time
+from strata_control import atomic as _atomic_mod
 from dataclasses import dataclass, field
 from pathlib import Path
 from typing import Any, Callable
@@ -277,7 +278,7 @@ def write_private(path: Path, data: dict) -> None:
             os.chmod(tmp, 0o600)
         except OSError:
             pass
-        os.replace(tmp, path)
+        _atomic_mod.replace(tmp, path)
     finally:
         if os.path.exists(tmp):
             os.unlink(tmp)

strata_cli/launch/store.py

@@ -13,6 +13,7 @@ import json
 import os
 import tempfile
 import time
+from strata_control import atomic as _atomic_mod
 from pathlib import Path
 from typing import Any
 
@@ -47,7 +48,7 @@ def _private_write(path: Path, payload: bytes) -> None:
             os.chmod(tmp, 0o600)
         except OSError:
             pass
-        os.replace(tmp, path)
+        _atomic_mod.replace(tmp, path)
     finally:
         if os.path.exists(tmp):
             os.unlink(tmp)

strata_cli/tune_commands.py

@@ -10,6 +10,7 @@ import sys
 import tempfile
 import threading
 import time
+from strata_control import atomic as _atomic_mod
 from pathlib import Path
 from typing import Any
 
@@ -151,7 +152,7 @@ def _write_status(path: str | None, payload: dict[str, Any]) -> None:
     try:
         tmp = path + ".tmp"
         Path(tmp).write_text(json.dumps(payload, ensure_ascii=False, default=str), encoding="utf-8")
-        os.replace(tmp, path)
+        _atomic_mod.replace(tmp, path)
     except OSError:
         pass
 

strata_control/calibration.py

@@ -26,6 +26,7 @@ from pathlib import Path
 from strata_control import profiles
 from strata_control.doctor import inspect as inspect_components
 from strata_control.profiles import ProfileError
+from strata_control import atomic as _atomic_mod
 
 
 SUITE = "legacy-decode-v1"
@@ -412,7 +413,7 @@ def _replace_bytes(path: Path, payload: bytes) -> None:
             stream.write(payload)
             stream.flush()
             os.fsync(stream.fileno())
-        os.replace(temp, path)
+        _atomic_mod.replace(temp, path)
     finally:
         if os.path.exists(temp):
             os.unlink(temp)

strata_control/components.py

@@ -52,6 +52,7 @@ from typing import Any, Iterable, Sequence
 SCHEMA_VERSION = 1
 
 from strata_control.targets import BACKEND_MATRIX, ComponentError, backend_supported, host_platform  # noqa: E402,F401
+from strata_control import atomic as _atomic_mod
 #: Component kinds, so a manifest cannot claim a role the architecture does not have.
 KINDS = ("application", "runtime", "agent-runtime", "tuning", "engine", "prepared-data")
 
@@ -426,7 +427,7 @@ class ComponentStore:
     def _save_state(self, data: dict[str, Any]) -> None:
         temp = self.state_path.parent / (self.state_path.name + ".tmp")
         temp.write_text(json.dumps(data, indent=1, sort_keys=True) + "\n", encoding="utf-8")
-        os.replace(temp, self.state_path)
+        _atomic_mod.replace(temp, self.state_path)
 
     # ------------------------------------------------------------------ cache and installation
 
@@ -487,7 +488,7 @@ class ComponentStore:
                            indent=1) + "\n", encoding="utf-8")
             if destination.exists():
                 shutil.rmtree(destination, ignore_errors=True)
-            os.replace(staging, destination)
+            _atomic_mod.replace(staging, destination)
         except BaseException:
             # Leave nothing that looks complete.
             shutil.rmtree(staging, ignore_errors=True)
@@ -590,7 +591,7 @@ class ComponentStore:
         if active.exists():
             if previous.exists():
                 shutil.rmtree(previous, ignore_errors=True)
-            os.replace(active, previous)
+            _atomic_mod.replace(active, previous)
             moved_aside = str(previous)
 
         # Record what is being displaced BEFORE the activation, so a crash between the two leaves a
@@ -611,7 +612,7 @@ class ComponentStore:
         if incoming.exists():
             shutil.rmtree(incoming, ignore_errors=True)
         shutil.copytree(source, incoming, symlinks=False)
-        os.replace(incoming, active)
+        _atomic_mod.replace(incoming, active)
 
         state = self.state()
         entry = state["components"].setdefault(component.id, {})
@@ -636,8 +637,8 @@ class ComponentStore:
             abandoned = self.root / "previous" / f"{component_id}.replaced"
             if abandoned.exists():
                 shutil.rmtree(abandoned, ignore_errors=True)
-            os.replace(active, abandoned)
-        os.replace(previous, active)
+            _atomic_mod.replace(active, abandoned)
+        _atomic_mod.replace(previous, active)
         state = self.state()
         entry = state["components"].get(component_id, {})
         restored = entry.get("rollback", {})

strata_control/preset_marks.py

@@ -26,6 +26,7 @@ import uuid
 from pathlib import Path
 
 from strata_control import paths
+from strata_control import atomic as _atomic_mod
 
 SCHEMA_VERSION = 1
 IDENTITY_KEYS = ("version", "build_id", "manifest_sha256", "backend", "platform", "variant")
@@ -58,7 +59,7 @@ def _write(path: Path, data: dict) -> None:
         with os.fdopen(descriptor, "w", encoding="utf-8") as stream:
             json.dump(data, stream, indent=1, sort_keys=True)
             stream.write("\n")
-        os.replace(temp, path)
+        _atomic_mod.replace(temp, path)
     finally:
         if os.path.exists(temp):
             os.unlink(temp)

strata_control/presets.py

@@ -29,6 +29,7 @@ from pathlib import Path
 from typing import Any
 
 from strata_control import tune_knobs, tune_machine
+from strata_control import atomic as _atomic_mod
 
 SCHEMA = "strata.preset/1"
 RECHECK_SET = "recheck/1"
@@ -57,7 +58,7 @@ def _atomic(path: Path, payload: bytes) -> None:
             fh.write(payload)
             fh.flush()
             os.fsync(fh.fileno())
-        os.replace(tmp, path)
+        _atomic_mod.replace(tmp, path)
     finally:
         if os.path.exists(tmp):
             os.unlink(tmp)

strata_control/private_files.py

@@ -383,7 +383,19 @@ def read_bytes_repairing(path: str | os.PathLike) -> bytes:
         return target.read_bytes()
 
 
-ACCESS_MESSAGE = "Strata can't open its settings file. Run Strata once as yourself (not as administrator) or press Repair."
+ACCESS_MESSAGE = ("Strata is not allowed to open one of its own files. Run `strata repair-access` (or press Repair in the app); "
+                "these files were probably made while Strata ran as administrator, so run it as your normal user from now on.")
+IN_USE_MESSAGE = ("A file Strata needs is in use by another program (usually antivirus, a backup or a sync tool). "
+                  "Wait a moment and run the same command again.")
+
+
+def is_sharing_violation(error: BaseException) -> bool:
+    """Windows "file in use" (WinError 32 or 33), as opposed to a real refusal (WinError 5): waiting helps, repairing does not."""
+    code = getattr(error, "winerror", None)
+    if code in (32, 33):
+        return True
+    text = str(error)
+    return "WinError 32" in text or "WinError 33" in text or "being used by another process" in text
 
 
 _SKIP_DIRS = {"models", "node_modules", "harness", "python", "host", "node", "control", "__pycache__", ".git"}

strata_control/procs.py

@@ -30,6 +30,7 @@ from dataclasses import dataclass
 from pathlib import Path
 
 from strata_control import paths
+from strata_control import atomic as _atomic_mod
 
 WINDOWS = os.name == "nt"
 SCHEMA = 1
@@ -192,7 +193,7 @@ def write_record(pid: int, role: str, *, profile: str | None = None, cmd: list[s
         path = procs_dir() / f"{pid}.json"
         tmp = path.with_suffix(f".{os.getpid()}.tmp")
         tmp.write_text(json.dumps(record), encoding="utf-8")
-        os.replace(tmp, path)
+        _atomic_mod.replace(tmp, path)
         return token
     except (OSError, ValueError):
         return None

strata_control/profiles.py

@@ -17,6 +17,7 @@ import shlex
 import stat
 import subprocess
 import tempfile
+from strata_control import atomic as _atomic_mod
 from contextlib import contextmanager
 from pathlib import Path
 
@@ -216,7 +217,7 @@ def _atomic(path: Path, data: bytes) -> None:
             stream.write(data)
             stream.flush()
             os.fsync(stream.fileno())
-        os.replace(temp_name, path)
+        _atomic_mod.replace(temp_name, path)
         if os.name != "nt":
             try:
                 parent_fd = os.open(path.parent, os.O_RDONLY)

strata_control/tune_apply.py

@@ -42,6 +42,7 @@ from pathlib import Path
 from typing import Any, Callable, Sequence
 
 from strata_control import user_args
+from strata_control import atomic as _atomic_mod
 
 SCHEMA_VERSION = 1
 
@@ -66,7 +67,7 @@ def _atomic_write(path: Path, payload: bytes) -> None:
             stream.write(payload)
             stream.flush()
             os.fsync(stream.fileno())
-        os.replace(temp, path)
+        _atomic_mod.replace(temp, path)
     finally:
         if os.path.exists(temp):
             os.unlink(temp)

strata_control/updater.py

@@ -44,6 +44,7 @@ from typing import Callable
 
 from strata_control import acquisition, components, engine_component as ec, feed as feedmod, longpath, preset_marks, signing, transfer
 from strata_control.targets import host_platform
+from strata_control import atomic as _atomic_mod
 
 STATE_DIR = ".updates"
 STAGE_DIR = ".s"                 # staging is short on purpose: <components>\.s\<4 hex>\engine\<folder> (see longpath.py)
@@ -152,7 +153,7 @@ class State:
             with os.fdopen(descriptor, "w", encoding="utf-8") as stream:
                 json.dump(self.data, stream, indent=1, sort_keys=True)
                 stream.write("\n")
-            os.replace(temp, self.path)
+            _atomic_mod.replace(temp, self.path)
         finally:
             if os.path.exists(temp):
                 os.unlink(temp)
@@ -948,7 +949,7 @@ def update_control(ctx: Context, feed: feedmod.Feed, state: State) -> dict:
         pointer = control_dir(root) / "active.json"
         temp = pointer.with_suffix(".tmp")
         temp.write_text(json.dumps({"version": pkg.version, "folder": pkg.version, "sha256": pkg.sha256}), encoding="utf-8")
-        os.replace(temp, pointer)
+        _atomic_mod.replace(temp, pointer)
         keep = {pkg.version} | ({previous["folder"]} if previous else set())
         for entry_dir in control_dir(root).iterdir():
             if entry_dir.is_dir() and entry_dir.name not in keep:
@@ -1022,7 +1023,7 @@ def update_app(ctx: Context, feed: feedmod.Feed) -> dict:
     target_dir.mkdir(parents=True, exist_ok=True)
     name = Path(transfer.resolve_ref(transfer.base_of(feed.source), entry["url"]).replace("\\", "/")).name or f"strata-app-{feed.app.version}"
     target = target_dir / name
-    os.replace(archive, target)
+    _atomic_mod.replace(archive, target)
     result.update({"changed": True, "status": "downloaded", "installed_after": None, "version": feed.app.version, "path": str(target),
                    "note": "Verified. Run the installer to finish; models and settings are kept."})
     return result

tests/terminal/test_file_in_use.py

@@ -0,0 +1,54 @@
+"""Windows 'file in use': every swap of a state file waits and retries (atomic.replace), and the message does not blame permissions."""
+
+import re
+import unittest
+from pathlib import Path
+from unittest import mock
+
+from strata_control import atomic, private_files
+
+ROOT = Path(__file__).resolve().parents[2]
+SITES = ["strata_control/calibration.py", "strata_control/components.py", "strata_control/presets.py",
+         "strata_control/preset_marks.py", "strata_control/procs.py", "strata_control/profiles.py",
+         "strata_control/tune_apply.py", "strata_control/updater.py", "strata_cli/bench.py", "strata_cli/connections.py",
+         "strata_cli/launch/runner.py", "strata_cli/launch/store.py", "strata_cli/tune_commands.py", "serve/runconfig.py"]
+
+
+class Sites(unittest.TestCase):
+    def test_no_bare_os_replace_is_left_in_the_state_file_writers(self):
+        for rel in SITES:
+            text = (ROOT / rel).read_text(encoding="utf-8")
+            bare = [l for l in text.splitlines() if re.search(r"(?<![_\w.])os\.replace\(", l) or re.search(r"[^_]os\.replace\(", l)]
+            # updater._move_dir keeps its own longer retry loop
+            bare = [l for l in bare if not (rel.endswith("updater.py") and "source, destination" in l)]
+            self.assertEqual(bare, [], rel)
+            self.assertIn("_atomic_mod.replace(", text, rel)
+
+
+class Retry(unittest.TestCase):
+    def test_a_held_file_is_waited_for_on_windows(self):
+        calls = []
+
+        def flaky(src, dst):
+            calls.append(1)
+            if len(calls) < 3:
+                raise PermissionError(13, "in use")
+        with mock.patch.object(atomic.os, "replace", flaky), mock.patch.object(atomic.os, "name", "nt"):
+            atomic.replace("a", "b", sleep=lambda s: None)
+        self.assertEqual(len(calls), 3)
+
+
+class Message(unittest.TestCase):
+    def test_sharing_violation_is_told_from_a_refusal(self):
+        in_use = PermissionError(13, "The process cannot access the file because it is being used by another process")
+        in_use.winerror = 32
+        denied = PermissionError(13, "Access is denied")
+        denied.winerror = 5
+        self.assertTrue(private_files.is_sharing_violation(in_use))
+        self.assertFalse(private_files.is_sharing_violation(denied))
+        self.assertNotIn("as yourself", private_files.ACCESS_MESSAGE)
+        self.assertIn("another program", private_files.IN_USE_MESSAGE)
+
+
+if __name__ == "__main__":
+    unittest.main()