fix(tracking): record stats for hook-rewritten commands (#1082)#2483
Open
ousamabenyounes wants to merge 5 commits into
Open
fix(tracking): record stats for hook-rewritten commands (#1082)#2483ousamabenyounes wants to merge 5 commits into
ousamabenyounes wants to merge 5 commits into
Conversation
When Claude Code rewrites `git status` → `rtk git status` via its PreToolUse hook, the hook first runs `rtk rewrite` which spawns a background telemetry thread holding a brief DB connection. If that process exits before the thread finishes, SQLite's WAL shared-memory file (.db-shm) may be inconsistent for a few milliseconds. Because `execute_batch()` stops on first error, bundling `PRAGMA journal_mode=WAL` and `PRAGMA busy_timeout=5000` in one call meant that a transient WAL-mode error silently left `busy_timeout=0`. With no wait budget, any lock-contention attempt in `Tracker::new()` failed immediately, silently dropping the tracking record. Fix two things: 1. Run `PRAGMA busy_timeout=5000` in its own `execute_batch()` call so it is always applied, regardless of whether the WAL pragma succeeds. 2. Add a single retry (10 ms sleep) in `TimedExecution::track()` and `track_passthrough()` to outlast the transient .db-shm window. Add regression tests for concurrent-open correctness and for the full hook-rewrite tracking path. Fixes rtk-ai#1082 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The previous test_busy_timeout_set_independently_of_wal_mode opened two trackers but wrote to them sequentially, so it passed even with the busy_timeout pragma removed — it did not actually guard the fix. Replace it with test_record_waits_for_concurrent_write_lock: a background thread holds the SQLite write lock (BEGIN IMMEDIATE) while the tracker attempts its INSERT. With busy_timeout=5000 the write waits and succeeds; with busy_timeout=0 (the rtk-ai#1082 bug) it fails immediately with SQLITE_BUSY. Add a shared DB_ENV_LOCK so RTK_DB_PATH-mutating tests don't race. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The lock-contention test set the process-global RTK_DB_PATH, which raced the other tracking tests that use the default DB under `cargo test --all` (test_timed_execution_records_time failed intermittently in CI). Extract Tracker::open_at_path(&Path) — shared by new() — so the test can target an isolated temp file without mutating any global env var. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…lint The semgrep filesystem-deletion rule flagged the explicit fs::remove_file cleanup. Switch to tempfile::tempdir() (the repo's existing convention), which auto-removes on drop without a literal deletion call. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #1082.
rtk gainstats were silently missing for commands rewritten via the Claude Code hook. The hook spawns a secondrtkprocess while another holds the SQLite history DB, so the tracking write hit lock contention. Two problems compounded it:busy_timeoutandjournal_mode=WALwere set in a singleexecute_batch()call, so a transient WAL error leftbusy_timeoutat 0 — and withbusy_timeout=0every contended write failed immediately withSQLITE_BUSYand was dropped (let _ = tracker.record(...)).Fix
Set
PRAGMA busy_timeout=5000separately fromjournal_mode=WALso the timeout is always applied, and add a bounded retry around the record path. A contended write now waits for the lock instead of being lost.Test verification (RED → GREEN)
test_record_waits_for_concurrent_write_lockreproduces real contention: a background thread holds the write lock (BEGIN IMMEDIATE) while the tracker attempts its INSERT.RED —
busy_timeoutreverted to 0 (the bug):GREEN — with
busy_timeout=5000:Full
core::trackingmodule: 15 passed, 0 failed (no baseline regression).(Re-proposes the accidentally-closed #1223, rebased onto current
develop. The original branch's test wrote two trackers sequentially and so did not actually fail without the fix; it has been replaced with the lock-contention test above.)