Implement the RSYNC_COMPAT Phase-2 filter/parser feature group:
- --files-from=FILE (repeatable) plus -0/--from0 NUL delimiters: parse the
source file list relative to the source root into a shared read-only
allow-set; the scanner transfers listed files and the whole subtree of
listed directories and prunes everything else in single- and
multithreaded mode. Absolute/'..' entries and missing files are hard
CLI errors.
- --filter=RULE: rsync-style +/- rules (anchored '/', dir-only trailing '/',
word include/exclude forms) evaluated first-match-wins with a default of
include, as an independent layer from legacy --exclude/--include.
Unsupported directives (merge/hide/... ) are rejected explicitly. -f stays
sendfile.
- -C/--cvs-exclude: well-known rsync CVS default exclude set.
- -F: per-directory .rsync-filter files read during traversal and applied to
the owning directory's subtree (single + parallel), never transferred.
- Delete manifest still derives from what was actually sent.
Client-only config fields; no wire/protocol change. Adds unit coverage
(CLI parse, allow-set and filter scanning single+parallel) and integration
tests (TestFilesFrom, TestFilters). RSYNC_COMPAT matrix rows updated:
5 rows move to Implemented (Summary 62/3/5/1/76 = 147).
Capture the transfer root's device (st_dev) at scanner creation and skip
descending into any subdirectory on a different device (a mount point).
Implemented sender/client-side only: sequential BFS and parallel (-m) root
scan apply the same scanner_same_filesystem decision; no wire/protocol change
and default behavior is unchanged. Unit tests cover the pure decision, same
device scanning in both modes, and CLI parsing; integration tests prove -x
leaves a single-filesystem tree byte-identical and, when root can mount a
tmpfs, skips a genuine cross-device subtree.
- Add log_perror() helper (context + strerror(errno)) to the log module
- Replace all bare perror() calls with log_perror() so errors are routed
through the unified logger (stderr sink + optional --log-file sink)
- Convert fprintf(stderr, "Error:/Warning: ...") in client code to
log_message(); raw fprintf kept only for progress/stats output
- parallel_scanner_init(): result queue + sync primitive setup with unwinding
- batch_files(): root-file chunk batching, reusable by other scan paths
- scan_root_directory()/scan_root_entry(): root-dir scanning
- spawn_parallel_workers(): worker thread creation with per-thread arg setup
Main function reduced from ~230 to ~40 lines
- Restore PROTOCOL_VERSION to a forward-compatible 2.1.0 and document wire format
- Add NULL guard to config_delete
- Add pipeline cancellation flag and cancellation-aware queue helper
- Join running threads before destroying pipeline contexts on creation failure
- Fix NULL dereference and memory leaks in manifest/chunk handling
- Fix TLS/TCP socket fd leak on connect error paths
- Add compression-level range validation (1-22)
- Close previous log file before opening a new one
- Use getline for unbounded pattern-file lines
- Fix thread-unsafe localtime() and add log level bounds check
- Fix file_load_data to clean up data on read size mismatch
- Add hard ceiling to decompression buffer growth
- Fix mkdir_r bounds check and restore glob comments
- Add send_str NULL guard and mutex-protect bandwidth limiter
- Update AGENTS.md for per-thread io_ssl contract
- send_n_data: void* data → const void* data
- send_str: char* data → const char* data
- send_data: Data* data → const Data* data
- Remove duplicate typedef struct ssl_st SSL in protocol.h
- Add NULL check after malloc in directory_scanner_create
- Fix ArrayList leak in directory_scanner_next when no files found
- Remove TSan from CI matrix (Docker 'unexpected memory mapping')
- Keep ASan which now passes clean
- Reformat all C/H files to match .clang-format (LLVM style)
- Fix 26 cppcheck const-correctness warnings (constParameterPointer,
constVariablePointer, constVariable)
- Update function declarations in headers to match const parameters
Feature changes across 15 files:
- -a/--archive: enables -c -m -M (no -s, which slows transfers)
- -n/--dry-run: scan + print without connecting or transferring
- -p <port>: custom SSH port (passed as -p to ssh via execvp)
- --exclude <pattern>: glob-based filename filtering in scanner
- --delete: sender collects file manifest; receiver deletes unlisted files
Protocol: added STATUS_MANIFEST, use_delete field in Config wire format.
New utilities: glob_match(), delete_extras() with recursive directory walk.
Scanner: accepts exclude patterns; skips matching entries.
SSH transport: switched from execlp to execvp for dynamic port arg.
Server + multiprocessing: handle STATUS_MANIFEST in both single and
multithreaded receive paths.
Builds clean, all 7 tests pass.
- Reduced test data from 50MB to 25MB (test.py)
- Increased SSH socketpair buffer to 1MB via setsockopt (transport_ssh.c)
- Made chunk size configurable: new Config field, --chunk-size CLI flag,
sent over wire, and used in scanner instead of DESIRED_CHUNK_SIZE
- Fixed DESIRED_CHUNK_SIZE macro parentheses (chunk.h)
- Remove FileReceive struct; use File everywhere with nullable FileMetadata*
- Remove struct stat from File; file size lives in Data->size (data_create_reserve)
- Add FileMetadata struct (mode, uid, gid, mtime) sent conditionally over wire
- Add -M / --preserve flag to client
- Restore permissions, ownership, timestamps on disk write
- Wire format uses per-file present flag for metadata (zero overhead when off)
- Fix scanner: when chunk fills up mid-directory, save DIR handle so
remaining files in that directory are not skipped (pre-existing bug)
- Fix -c -s: add STATUS_CHUNK to protocol, send it before chunk data,
handle it on the server side for both single and multithreaded paths
- Extract chunk_serialize/chunk_deserialize from chunk_compress/decompress
- Remove dead declarations (file_receive_from_buffer, etc.)
- Fix memory leak in receive_file_receive (free -> data_destroy)
- test.py generates ~50 MB of test data across bulk files
- All 8 integration tests + 7 unit tests pass
TapTap
Theo Tappe