mirror of
https://github.com/bytedance/deer-flow.git
synced 2026-10-27 05:30:39 +00:00
346ff7b253
* fix(memory): rank capacity eviction per record, not per fact_id select_facts_for_capacity keyed every score by fact_id, so an imported batch that repeats an id collapsed into a single dict entry: the last record's score won, and both the ranking pass and the evicted audit read that one value for every position. A 0.9-confidence record could therefore be evicted while its own 0.1-confidence duplicate survived. The import path keeps caller-supplied ids without checking them for uniqueness, and any over-cap import lands here. Score by position instead: a parallel list next to the indexed facts, so ranking and the audit read each record's own score. The public ``scores`` dict stays keyed by fact_id with last-write-wins semantics, keeping the DeerMem benchmark contract and the callers that never read it unchanged. Red on main: test_confidence_policy_scores_duplicate_fact_ids_per_record fails with the highest-confidence record evicted; green on this branch. * fix(utils): keep share-less UNC roots out of the extended-length prefix extended_length_path prefixed every absolute UNC path with \\?\UNC\, but Win32 only accepts \\?\UNC\server\share\... — a share-less root (\\server, \\server\, or a bare \) is not a valid device path, so every Win32 call handed the prefixed spelling fails. Only prefix when both the server and the share are present; otherwise return the ordinary spelling unchanged. Red on main: the \\server, \\server\, //, and //server cases of test_windows_extended_spelling assert the forged prefix; green on this branch.
67 lines
3.0 KiB
Python
67 lines
3.0 KiB
Python
"""Portability checks for host-visible path segments."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import os
|
|
from pathlib import Path
|
|
|
|
|
|
def extended_length_path(path: Path) -> Path:
|
|
"""Use Windows' extended namespace for local filesystem access.
|
|
|
|
Prefix before appending children: even a short root can contain a long
|
|
document filename. Normalize ordinary paths before adding the prefix;
|
|
extended paths bypass Win32's ordinary path normalization. This is lexical
|
|
only: callers must still resolve symlinks and enforce their own confinement.
|
|
Existing extended/device namespaces (including named pipes) are returned
|
|
unchanged, without normalizing their segments.
|
|
POSIX paths are unchanged. Do not use this spelling for Docker mount sources
|
|
or persisted relative paths.
|
|
"""
|
|
if os.name != "nt":
|
|
return path
|
|
if str(path).startswith(("\\\\?\\", "\\\\.\\")):
|
|
return path
|
|
absolute = os.path.abspath(path)
|
|
if absolute.startswith("\\\\"):
|
|
# Win32 only accepts \\?\UNC\server\share\...; a share-less UNC root
|
|
# (\\server, \\server\, or a bare \) is not a valid device path, so
|
|
# keep its ordinary spelling instead of forging an invalid one.
|
|
server, separator, share = absolute[2:].partition("\\")
|
|
if server and separator and share:
|
|
return Path("\\\\?\\UNC\\" + absolute[2:])
|
|
return Path(absolute)
|
|
return Path("\\\\?\\" + absolute)
|
|
|
|
|
|
# Windows also treats the ISO-8859-1 superscript digits ¹, ² and ³ as device numbers.
|
|
# https://learn.microsoft.com/en-us/windows/win32/fileio/naming-a-file
|
|
# CONIN$ and CONOUT$ open the console input and output buffers, not ordinary files.
|
|
# https://learn.microsoft.com/en-us/windows/console/console-handles
|
|
_WINDOWS_RESERVED_NAMES = frozenset({"CON", "PRN", "AUX", "NUL", "CONIN$", "CONOUT$"} | {f"COM{i}" for i in "123456789¹²³"} | {f"LPT{i}" for i in "123456789¹²³"})
|
|
|
|
|
|
def windows_incompatible_segment(segment: str) -> str | None:
|
|
"""Return a reason when *segment* is unsafe to create as a host path component.
|
|
|
|
Reserved device names and trailing dots or spaces are rejected on every
|
|
platform when a name is created: uploads, new custom-skill support files,
|
|
and sandbox writes of a path that is not already stored. Reads, deletes of
|
|
an existing support file, and access to a path that already exists on the
|
|
host do not use this check. Those names alias devices or are silently
|
|
stripped on Windows.
|
|
|
|
``.`` and ``..`` are ignored here; callers already reject traversal.
|
|
Do not reject colons, quotes, or control characters here.
|
|
"""
|
|
if segment in {"", ".", ".."}:
|
|
return None
|
|
if segment.endswith((" ", ".")):
|
|
return "trailing dot or space"
|
|
# Windows ignores ASCII spaces before a device name's extension (NUL .txt).
|
|
# Trim only for comparison; ordinary filenames must retain their spaces.
|
|
stem = segment.split(".", 1)[0].rstrip(" ")
|
|
if stem.upper() in _WINDOWS_RESERVED_NAMES:
|
|
return "reserved Windows device name"
|
|
return None
|