fix(file-ops): allow file search in hidden roots

This commit is contained in:
ClawdIA 2026-04-27 15:31:15 -03:00 committed by Teknium
parent 9e2628ee7c
commit 64ad7dec0d
2 changed files with 97 additions and 5 deletions

View file

@ -2,6 +2,7 @@
import os
import pytest
import subprocess
from pathlib import Path
from unittest.mock import MagicMock
@ -388,6 +389,66 @@ class TestSearchPathValidation:
assert "search failed" in result.error.lower() or "Search error" in result.error
class TestSearchFilesFallbackHiddenPaths:
def _make_env(self):
env = MagicMock()
env.cwd = "/"
def execute(command, **kwargs):
completed = subprocess.run(
command,
shell=True,
text=True,
capture_output=True,
)
return {
"output": completed.stdout,
"returncode": completed.returncode,
}
env.execute = execute
return env
def test_hidden_root_with_hidden_ancestor_includes_files(self, tmp_path, monkeypatch):
"""Fallback find should include visible files when path is inside hidden root."""
root = tmp_path / ".hermes" / "logs"
root.mkdir(parents=True)
visible_file = root / "agent.log"
hidden_dir_file = root / ".hidden" / "secret.log"
nested_hidden_file = root / "nested" / ".secret.log"
visible_nested_file = root / "nested" / "visible.log"
for p in [visible_file, nested_hidden_file, visible_nested_file, hidden_dir_file]:
p.parent.mkdir(parents=True, exist_ok=True)
p.write_text("x")
ops = ShellFileOperations(self._make_env())
monkeypatch.setattr(ops, "_has_command", lambda command: command == "find")
result = ops._search_files("*.log", str(root), limit=50, offset=0)
assert result.error is None
assert set(result.files) == {str(visible_file), str(visible_nested_file)}
def test_normal_root_still_excludes_hidden_descendants(self, tmp_path, monkeypatch):
"""Fallback find should still exclude hidden descendant paths for normal roots."""
root = tmp_path / "repo"
root.mkdir()
visible_file = root / "agent.log"
visible_nested_file = root / "nested" / "visible.log"
hidden_dir_file = root / ".hidden" / "secret.log"
for p in [visible_file, visible_nested_file, hidden_dir_file]:
p.parent.mkdir(parents=True, exist_ok=True)
p.write_text("x")
ops = ShellFileOperations(self._make_env())
monkeypatch.setattr(ops, "_has_command", lambda command: command == "find")
result = ops._search_files("*.log", str(root), limit=50, offset=0)
assert result.error is None
assert set(result.files) == {str(visible_file), str(visible_nested_file)}
class TestShellFileOpsWriteDenied:
def test_write_file_denied_path(self, file_ops):
result = file_ops.write_file("~/.ssh/authorized_keys", "evil key")

View file

@ -987,6 +987,12 @@ class ShellFileOperations(FileOperations):
else:
search_pattern = pattern.split('/')[-1]
search_root = Path(path)
has_hidden_path_ancestor = any(
part not in (".", "..") and part.startswith(".")
for part in search_root.parts
)
# Prefer ripgrep: respects .gitignore, excludes hidden dirs by
# default, and has parallel directory traversal (~200x faster than
# find on wide trees). Mirrors _search_content which already uses rg.
@ -1002,17 +1008,25 @@ class ShellFileOperations(FileOperations):
)
# Exclude hidden directories (matching ripgrep's default behavior).
hidden_exclude = "-not -path '*/.*'"
hidden_exclude = "-not -path '*/.*'" if not has_hidden_path_ancestor else ""
hidden_filter_expr = f" {hidden_exclude}" if hidden_exclude else ""
cmd = f"find {self._escape_shell_arg(path)} {hidden_exclude} -type f -name {self._escape_shell_arg(search_pattern)} " \
f"-printf '%T@ %p\\n' 2>/dev/null | sort -rn | tail -n +{offset + 1} | head -n {limit}"
# Use shell pagination for standard roots. For hidden roots, gather full
# output so we can re-apply hidden-descendant filtering while allowing
# explicit hidden-root searches.
pagination_expr = ""
if not has_hidden_path_ancestor:
pagination_expr = f" | tail -n +{offset + 1} | head -n {limit}"
cmd = f"find {self._escape_shell_arg(path)}{hidden_filter_expr} -type f -name {self._escape_shell_arg(search_pattern)} " \
f"-printf '%T@ %p\\n' 2>/dev/null | sort -rn{pagination_expr}"
result = self._exec(cmd, timeout=60)
if not result.stdout.strip():
# Try without -printf (BSD find compatibility -- macOS)
cmd_simple = f"find {self._escape_shell_arg(path)} {hidden_exclude} -type f -name {self._escape_shell_arg(search_pattern)} " \
f"2>/dev/null | head -n {limit + offset} | tail -n +{offset + 1}"
cmd_simple = f"find {self._escape_shell_arg(path)}{hidden_filter_expr} -type f -name {self._escape_shell_arg(search_pattern)} " \
f"2>/dev/null | sort -rn{pagination_expr}"
result = self._exec(cmd_simple, timeout=60)
files = []
@ -1025,6 +1039,23 @@ class ShellFileOperations(FileOperations):
else:
files.append(line)
# For explicit hidden roots, find's path-based filtering excludes every
# file under the hidden path. Apply descendant filtering after command
# execution so only the explicit root ancestry is bypassed.
if has_hidden_path_ancestor:
normalized_root = search_root.resolve()
filtered_files = []
for file_path in files:
try:
rel_parts = Path(file_path).resolve().relative_to(normalized_root).parts
except ValueError:
rel_parts = Path(file_path).parts
if any(part not in (".", "..") and part.startswith(".") for part in rel_parts):
continue
filtered_files.append(file_path)
files = filtered_files[offset:offset + limit]
# pagination for standard roots is already applied in shell
return SearchResult(
files=files,
total_count=len(files)