Background threads (spawn_indexer, spawn_visual_height_rebuild) previously
held mmap during entire file scan, risking SIGBUS if file was truncated
externally. Now uses BufReader streaming scan with mmap created only
after scan completes, plus stat validation.
Changes:
- spawn_indexer: replace mmap scan with BufReader fill_buf/consume loop,
create mmap post-scan with fd stat validation
- spawn_visual_height_rebuild: replace mmap/FileReader with sequential
BufReader scan, discard results on line count mismatch
- FileReader::open/reload/update_for_append: add stat-after-mmap check
- LineIndex: make fields pub(crate) for direct construction from scan loop
- Add 3 regression tests for truncation scenarios
spawn_indexer builds LineIndex from mmap snapshot but IndexCache::save()
re-opened the file to compute the hash. If the file changed between those
two steps, the cached index would be stored under the wrong hash.
- Add IndexCache::save_with_hash() that computes hash from in-memory data
- Add compute_data_hash() public function (same algorithm as compute_file_hash)
- Update spawn_indexer, FileReader::save_cache, and test callers
Replace direct indexing with .get()? to return None instead of panicking
when sampled_offsets is shorter than expected (e.g. corrupt bincode cache).
Closes#3
Replace LineIndex::from_reader(BufReader) with LineIndex::from_bytes(&mmap)
in both open() and reload(), ensuring the line index is always built from
the same mmap snapshot rather than a separate read through the file descriptor.
This closes the race window where an external file modification between
mmap() and from_reader() could cause line_index offsets to disagree with
the mmap data, leading to get_line() returning wrong content or panicking.
Closes#2
File truncation/rotation during mmap lifetime caused SIGBUS crash
because update_for_append() ignored new_size < old_size, leaving
a stale mapping that would fault on access.
Introduce AppendStatus enum (Unchanged/Appended/Reloaded) so the
caller can distinguish shrink events from normal appends. On shrink,
reload() rebuilds the mmap and line index. The TUI layer clamps
cursor and invalidates viewport cache on Reloaded, matching the
existing handle_file_truncated() behavior.
Fixes#1
Add SCAN_AHEAD_LIMIT (10000 lines) to get_line() in Sampling state. Without this, jumping to end-of-file (G) during progressive loading would scan the entire file byte-by-byte on the main thread, blocking the UI and consuming excessive memory. Lines beyond the scanned region + limit now return None, which the TUI renders as empty.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Complete FileWatcher implementation using notify 8.x crate with get_inode() for cross-platform file identity. Support append detection for incremental index updates and truncate detection for full reloads.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Add ProgressiveFileReader state machine (Sampling/Ready/Error) with scanned_newlines cache for O(1) line lookups. Implement spawn_indexer() for background mmap-based indexing with cancellation and progress reporting. Add VisualHeightIndex with prefix-sum + binary search for O(log N) visual height queries. Register all new io submodules and extend FileReader with reload(), save_cache(), and accessor methods.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Move wrap_line_chars and format_json_line from app.rs to core/io/wrap.rs with MAX_WRAP_INPUT_LEN guard. Add serde derives, pub getters, and extend_from_bytes() to LineIndex for incremental index building.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Implement head/tail 64KB sampling to estimate total line count without scanning the entire file. Also provide read_first_lines() for reading the first N lines for immediate display during progressive loading.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Add xxhash-rust and bincode workspace dependencies for fast hashing and serialization. Implement cache_util for cache directory/path resolution with versioning, and IndexCache for saving/loading line indices to disk with file-hash validation.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Replace Vec<u8> with memmap2::Mmap for file content. Build line index via BufReader streaming (from_reader) instead of scanning mmap'd data, keeping RSS at ~3MB for 5GB files instead of ~5GB.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Rewrite LineIndex to use sparse sampling (every 256 lines) instead of per-line offsets. Add from_reader() for low-RSS streaming index construction via BufReader fill_buf()/consume(), reducing 5GB file RSS from 5122MB to 3.4MB.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Add level detection that tries JSON parsing first (trusts level field), then falls back to keyword scanning with word-boundary checks. Supports SEVERE, FATAL, ERROR, WARN, INFO, DEBUG, TRACE and their abbreviations.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Replace empty AppConfig stub with ColorConfig struct that stores level-to-color mappings as strings. Supports loading/saving from XDG config directory via TOML, with graceful degradation on missing/invalid files.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
