🐛 fix(cache): hash filesystem path bytes

docs/changelog/+non-utf8-cache-path.bugfix.rst

@@ -0,0 +1 @@
+Handle interpreter paths containing non-UTF-8 filesystem bytes when using DiskCache.

src/python_discovery/_cache.py

@@ -8,6 +8,7 @@
 import stat
 from contextlib import contextmanager, suppress
 from hashlib import sha256
+from os import fsencode
 from pathlib import Path
 from tempfile import NamedTemporaryFile
 from typing import TYPE_CHECKING, Final, Protocol, runtime_checkable
@@ -173,7 +174,7 @@ def py_info(self, path: Path) -> ContentStore:
         """
         if not self._trusted_folder():
             return NoOpContentStore()
-        key = sha256(str(path).encode("utf-8")).hexdigest()
+        key = sha256(fsencode(path)).hexdigest()
         return DiskContentStore(self._py_info_dir, key)
 
     def _trusted_folder(self) -> bool:

tests/test_cache.py

@@ -3,13 +3,15 @@
 import os
 import stat
 import sys
+from os import fsdecode
 from pathlib import Path
 from threading import Event, Thread
 from typing import TYPE_CHECKING
 
 import pytest
 from filelock import FileLock, Timeout
 
+from python_discovery import get_interpreter
 from python_discovery._cache import DiskCache, DiskContentStore, NoOpCache, NoOpContentStore
 
 if TYPE_CHECKING:
@@ -118,6 +120,19 @@ def test_disk_cache_py_info(tmp_path: Path) -> None:
     assert isinstance(store, DiskContentStore)
 
 
+@pytest.mark.skipif(os.name == "nt", reason="Windows paths do not use surrogateescape")
+def test_disk_cache_non_utf8_interpreter_path(tmp_path: Path) -> None:
+    directory = tmp_path / fsdecode(b"interpreter-\xff")
+    directory.mkdir()
+    executable = directory / "python3"
+    executable.symlink_to(sys.executable)
+
+    info = get_interpreter(str(executable), cache=DiskCache(tmp_path / "cache"))
+
+    assert info is not None
+    assert info.executable == str(executable)
+
+
 def test_disk_cache_py_info_clear(tmp_path: Path) -> None:
     cache = DiskCache(tmp_path)
     store = cache.py_info(Path("/some/path"))