Add BEADS_DIR as a replacement for BEADS_DB to point to the .beads
directory instead of the database file directly.
Rationale:
- With --no-db mode, there's no .db file to point to
- The .beads directory is the logical unit (contains config.yaml, db
files, jsonl files)
- More intuitive: point to the beads directory not the database file
Implementation:
- Add BEADS_DIR environment variable support to FindDatabasePath()
- Priority order: BEADS_DIR > BEADS_DB > auto-discovery
- Maintain backward compatibility with BEADS_DB (now deprecated)
- Update --no-db mode to respect BEADS_DIR
- Update MCP integration (config.py, bd_client.py)
- Update documentation to show BEADS_DIR as preferred method
Testing:
- Backward compatibility: BEADS_DB still works
- BEADS_DIR works with regular database mode
- BEADS_DIR works with --no-db mode
- Priority: BEADS_DIR takes precedence over BEADS_DB
Follow-up issues for refactoring:
- bd-efe8: Refactor path canonicalization into helper function
- bd-c362: Extract database search logic into helper function
Closes bd-e16b
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
CRITICAL BUG: The previous fix had a race condition where the
importInProgress flag could be released twice, allowing two goroutines
to think they both hold the lock.
Bug scenario:
1. Goroutine A: acquires lock (CAS true)
2. Goroutine A: manually releases at line 208 for git dirty skip
3. Goroutine B: CAS succeeds, acquires lock
4. Goroutine A: defer runs, releases flag AGAIN (clears B lock)
5. Goroutine C: CAS succeeds - now TWO goroutines have lock
Root cause: Using both manual Store(false) AND defer Store(false)
created a window where the flag could be cleared twice.
Fix: Use a shouldDeferRelease flag to disable the deferred release
when we manually release early. This ensures exactly one release
per acquisition.
Testing: All auto-import tests still passing
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
CRITICAL FIX: The daemon could enter a corrupt state when auto-import
was triggered while uncommitted changes existed in .beads/ files. This
caused mysterious EOF errors requiring daemon restarts.
Root Causes Fixed:
1. No git dirty check before auto-import
2. Missing timeout protection (could hang indefinitely)
3. Race condition in importInProgress flag management
4. Improper git status parsing (false positives)
5. Context leak in export goroutine (accessing closed DB)
6. Data loss in triggerExport (missing deps/labels/comments)
Changes:
- Add hasUncommittedBeadsFiles() to check git status before import
- Properly parses git porcelain format ("XY filename")
- Ignores untracked files, only checks tracked .jsonl changes
- 5-second timeout on git command to prevent hangs
- Add 30-second timeout to import operations
- Prevents daemon from hanging on stuck imports
- Returns clear error message on timeout
- Fix race condition in importInProgress flag
- Explicitly release before early returns
- Prevents other requests seeing incorrect "import in progress"
- Fix context leak in onChanged export goroutine
- Check daemon shutdown state before export
- Suppress expected "database closed" errors during shutdown
- Pass storage/path as parameters to avoid closure issues
- Fix data loss in triggerExport()
- Populate dependencies, labels, comments (mirrors handleExport)
- Use atomic file write (temp + rename)
- Sort issues for consistent output
Impact: Prevents daemon corruption in collaborative workflows where
users frequently pull updates while having local uncommitted changes.
Testing: All auto-import tests passing
- TestDaemonAutoImportAfterGitPull ✓
- TestDaemonAutoImportDataCorruption ✓
- internal/autoimport tests ✓
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implement a new `bd status` command that provides a quick snapshot of the
issue database state, similar to how `git status` shows working tree state.
Features:
- Summary counts by state (open, in-progress, blocked, closed)
- Ready to work count
- Recent activity stats (last 7 days): created, closed, updated issues
- Support for --assigned flag to filter by current user
- JSON output format with --json flag
- Comprehensive test coverage
Usage examples:
bd status # Show summary
bd status --json # JSON output
bd status --assigned # Filter to assigned issues
bd status --no-daemon # Direct mode with recent activity
Note: Recent activity currently only works in direct mode (--no-daemon).
Daemon mode support marked with TODO for future enhancement.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implements a shorter, more natural syntax for adding comments to issues.
Users can now use 'bd comment <issue-id> <text>' instead of the more
verbose 'bd comments add <issue-id> <text>'.
Changes:
- cmd/bd/comments.go: Add commentCmd as top-level alias
- cmd/bd/comments_test.go: Add TestCommentAlias for verification
- commands/comments.md: Document the new shortcut
The alias delegates to the existing commentsAddCmd implementation,
ensuring all functionality (--file, --author flags) works identically.
Backwards compatibility is maintained - 'bd comments add' continues to work.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
The --json flag was being ignored because doctor.go defined a local
flag that shadowed the persistent --json flag from main.go. The local
flag wasn't bound to the global jsonOutput variable, so it remained
false even when --json was passed.
Fixed by removing the duplicate local flag definition. The doctor
command now correctly uses the global persistent flag.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add comprehensive guidance for managing AI-generated planning documents
(Claude slop) to the bd onboard command. Addresses GitHub issue #196.
Changes:
- Add Managing AI-Generated Planning Documents section to onboard output
- Recommend using history/ directory for ephemeral planning files
- List common planning doc types (PLAN.md, IMPLEMENTATION.md, etc.)
- Provide .gitignore example and benefits explanation
- Add to Important Rules: Store AI planning docs in history/
- Update AGENTS.md with the same guidance to demonstrate the pattern
- Add comprehensive tests for onboard command
This uses inverted control: bd outputs guidance, AI agents intelligently
integrate it into their project's documentation (AGENTS.md, CLAUDE.md,
.cursorrules, etc.), respecting existing conventions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Adds checkGitHooks() function to verify recommended hooks are installed
- Checks for pre-commit, post-merge, and pre-push hooks
- Warns if hooks are missing with install instructions
- Shows up early in diagnostics (even if .beads/ missing)
- Includes comprehensive test coverage
- Filed bd-6049 for broken --json flag
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Adds restart subcommand to bd daemons that gracefully stops a daemon
and starts a new one in the same workspace.
Features:
- Accepts workspace path or PID as target
- Graceful shutdown via RPC with SIGTERM fallback
- Starts new daemon with exec.Cmd in correct workspace directory
- Prefers workspace-local bd binary if present
- Supports --search and --json flags
- Proper error handling and user feedback
Closes bd-3ee2c7e9 (bd daemons command epic)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add migration for UNIQUE index on external_ref column (bd-897a)
- Add validation for duplicate external_ref in batch imports (bd-7315)
- Add query planner test to verify index usage (bd-f9a1)
- Add concurrent import tests for external_ref (bd-3f6a)
The migration detects existing duplicates and fails gracefully.
Batch imports now reject duplicates with clear error messages.
Tests verify the index is actually used by SQLite query planner.
Amp-Thread-ID: https://ampcode.com/threads/T-45ca66ed-3912-46c4-963c-caa7724a9a2f
Co-authored-by: Amp <amp@ampcode.com>
- Gate slow git E2E tests with testing.Short() (saves ~5s)
- Use shallow/shared clones for test repos
- Disable git hooks in test setup (major speedup)
- Reduce sync rounds from 3→1 (2 for dedup test)
- Add git speed configs (gc.auto=0, fsync=false, gpgSign=false)
Results:
- cmd/bd tests: 41s → 33s with -short (~20% faster)
- Full suite: >300s timeout → ~40s (no timeout!)
- E2E tests: 2-3s each → skipped with -short
Run full E2E tests with: go test ./...
Run fast tests with: go test -short ./...
- Changed bd import to call flushToJSONL() instead of markDirtyAndScheduleFlush()
- Ensures daemon FileWatcher detects changes within ~500ms instead of up to 30s
- Fixes race condition where MCP daemon served stale data after CLI import
- Created internal/syncbranch package with validation and env var support
- Added --branch flag to bd init command
- Enhanced bd config get/set to validate sync.branch
- Added BEADS_SYNC_BRANCH environment variable support
- Comprehensive tests for branch name validation
- Supports precedence: env var > database config > empty (current branch)
