Commit cb910c1f authored by Ramon Nou's avatar Ramon Nou
Browse files

fix(metadata): restore whole-path regex matching for filtered scans

parent 33e98104

src/daemon/backend/metadata/rocksdb_backend.cpp

+3 −1
Changes for src/daemon/backend/metadata/rocksdb_backend.cpp: 3 added lines, 1 removed line.
Original line number Diff line number Diff line
@@ -798,7 +798,9 @@ RocksDBBackend::get_dirents_filtered_impl( 
        scanned_count++;



        bool matched = !S_ISDIR(mode);

        if(use_regex && !std::regex_search(std::string(rel_name), name_regex)) {

        // Preserve 0.9.6 whole-path matching. Substring regex_search with

        // wildcard filters such as .*01.* retries nonmatching path suffixes.

        if(use_regex && !std::regex_match(std::string(rel_name), name_regex)) {

            matched = false;

        }

        if(matched && filter_size != -1 && size != (size_t) filter_size) {

tests/integration/directories/test_sfind_filtered.py

+14 −11
Changes for tests/integration/directories/test_sfind_filtered.py: 14 added lines, 11 removed lines.
Original line number Diff line number Diff line
@@ -68,12 +68,12 @@ def test_sfind_filtered(test_workspace, request, monkeypatch, buffer_size, 
        ret = io_client.create_n_files(str(nested_dir), 10)

        assert ret.retval == 0



        # Search for "file_5" in the scoped subdir (should be file_5)

        # Server filters match the entire relative path, including subdirs.

        log.info(

            f"Running recursive sfind command: {sfind_bin} {test_dir} "

            "-regex \"(^|/)file_5$\" ...")

            "-regex \"(^|.*/)file_5$\" ...")

        # Should find file_5 in top dir AND file_5 in subdir

        ret = shell.run(str(sfind_bin), str(test_dir), "-regex", "(^|/)file_5$",

        ret = shell.run(str(sfind_bin), str(test_dir), "-regex", "(^|.*/)file_5$",

                        "-M", str(mount_dir), "-S", "1", "--server-side")



        sfind_stdout = ret.stdout.decode(errors='replace') if ret.stdout else ""
@@ -86,14 +86,18 @@ def test_sfind_filtered(test_workspace, request, monkeypatch, buffer_size, 
        # Matches: 2

        assert "MATCHED 2/111" in sfind_stdout



        # Both implementations must scan descendants, exclude directories

        # from matches, and report exact counts regardless of buffer size.

        # Both implementations scan descendants and exclude directories.

        # Server-side filters retain 0.9.6 whole-path matching, whereas the

        # client-side implementation uses substring matching.

        for server_side in (False, True):

            for count_only in (False, True):

                for pattern, expected in (("*file_5", 12),

                                          ("file_5", 12),

                                          ("*", 110),

                                          ("missing", 0)):

                for pattern, server_expected, client_expected in (

                        ("*file_5", 2, 12),

                        ("file_5", 1, 12),

                        ("*file_5*", 12, 12),

                        ("*", 110, 110),

                        ("missing", 0, 0)):

                    expected = server_expected if server_side else client_expected

                    flags = []

                    if server_side:

                        flags.append("--server-side")
@@ -110,8 +114,7 @@ def test_sfind_filtered(test_workspace, request, monkeypatch, buffer_size, 
                    if server_side and not count_only and pattern == "*file_5":

                        paths = {line for line in stdout.splitlines()

                                 if line.startswith(str(test_dir) + "/")}

                        assert paths == {str(test_dir / ("file_%d" % i))

                                         for i in [5] + list(range(50, 60))} | {

                        assert paths == {str(test_dir / "file_5"),

                                         str(nested_dir / "file_5")}



        # Failed RPCs must not be presented as a successful empty scan.