fix(gix): pass filename byte patterns to status and directory walks

Status and untracked-file filtering passed surrogateescaped strings into
GixPython. Converting those strings to Rust text raised `UnicodeEncodeError`,
while equivalent Git CLI queries correctly matched the original byte names.

Encode patterns with the shared lossless helper at the two native boundaries.
`StatusPlatform.into_iter()` and `Repository.dirwalk_iter()` already accept
byte patterns, so no new API, fallback or path matching implementation is needed.

The scripted SHA-1/SHA-256 regressions fail before the fix and pass on both
backends. They check matching deleted paths, nonmatching paths and a byte-bearing
character class matching an untracked file. Gix checks prohibit CLI execution.
Mypy and Ruff pass. Windows operand tests remain excluded because equivalent
arbitrary-byte CLI arguments cannot be represented there.

Author: codex

doc/gix-backend.md

@@ -395,6 +395,10 @@ and predate the current fallback; they do not establish native coverage.
 Reference names and revisions are encoded with `surrogateescape` before calling
 GixPython's byte-capable `Target.Symbolic`, `Repository.try_find_reference`,
 `Repository.rev_parse_single`, `Repository.rev_parse` and `RefEdit` APIs.
+Status and untracked-file filters likewise pass byte patterns to
+`StatusPlatform.into_iter()` and `Repository.dirwalk_iter()`. Their regression
+checks matching and nonmatching patterns containing undecodable bytes, and
+prohibits CLI fallback in Gix mode.
 This preserves non-UTF-8 packed tag names without interpreting reference storage
 in Python. `test_non_utf8_tag_names_preserve_bytes` covers lightweight and annotated
 tag enumeration in SHA-1 and SHA-256 repositories on both backends and all platforms.

git/_backend.py

@@ -515,7 +515,7 @@ def is_dirty(command: Any, index: bool, working_tree: bool, untracked: bool, sub
             .index_worktree_submodules("configured" if submodules else "all", check_dirty=submodules)
         )
         dirty = False
-        with status.into_iter(patterns) as entries:
+        with status.into_iter([safe_encode(pattern) for pattern in patterns]) as entries:
             for item in entries:
                 if item.summary() is None:
                     continue
@@ -556,7 +556,7 @@ def untracked_files(command: Any, args: Tuple[Any, ...], options: Dict[str, Any]
             raise _Unsupported("invalid pathspec")
         walk_options = repo.dirwalk_options().emit_untracked("matching").emit_tracked(False)
         result = []
-        with repo.dirwalk_iter(index, patterns, walk_options) as entries:
+        with repo.dirwalk_iter(index, [safe_encode(pattern) for pattern in patterns], walk_options) as entries:
             for item in entries:
                 entry = item.entry
                 if entry.status == "Untracked":

test/test_byte_pathspecs.py

@@ -0,0 +1,28 @@
+"""Path filters preserve Git's filename bytes at the native boundary."""
+
+import contextlib
+import os
+from unittest.mock import patch
+
+import pytest
+
+from git import Git, _backend
+from git.compat import safe_decode
+from test.lib.fixtures import fixture_repo
+
+
+@pytest.mark.skipif(os.name == "nt", reason="Windows CLI arguments cannot express arbitrary filename bytes")
+@pytest.mark.parametrize("object_format", ["sha1", "sha256"])
+def test_byte_pathspecs_preserve_matches(object_format):
+    with fixture_repo("pathspec-bytes.sh", object_format) as repo:
+        guard = (
+            patch.object(Git, "execute", side_effect=AssertionError("unexpected pathspec CLI fallback"))
+            if _backend.name == "gix"
+            else contextlib.nullcontext()
+        )
+        with guard:
+            # The tracked path is absent in the worktree, so only its matching filter is dirty.
+            assert repo.is_dirty(path=safe_decode(b"tracked-\xff"))
+            assert not repo.is_dirty(path=safe_decode(b"absent-\xff"))
+            assert repo._get_untracked_files(safe_decode(b"[u\xff]*")) == ["untracked"]
+            assert repo._get_untracked_files(safe_decode(b"absent-\xff")) == []

tests/fixtures/pathspec-bytes.sh

@@ -0,0 +1,9 @@
+#!/usr/bin/env bash
+set -eu
+git init --initial-branch=main --object-format="$1" repo
+cd repo
+blob=$(printf 'contents\n' | git hash-object -w --stdin)
+printf '100644 %s\ttracked-\377\0' "$blob" | git update-index -z --index-info
+git commit -m initial
+printf 'untracked\n' >untracked
+printf 'other\n' >other