When multiple bd commands are run in parallel, they can race during database
migrations, causing "duplicate column name" errors. This happens because:
1. Process A checks if column exists → false
2. Process B checks if column exists → false
3. Process A adds column → succeeds
4. Process B adds column → FAILS (duplicate column)
Changes:
- Wrap RunMigrations in BEGIN EXCLUSIVE transaction to serialize migrations
- Disable foreign keys BEFORE the transaction (PRAGMA must be called outside tx)
- Convert nested BEGIN/COMMIT in migrations 010, 022, 025 to use SAVEPOINTs
(SQLite does not support nested transactions)
- Remove redundant PRAGMA foreign_keys calls from individual migrations
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The test was creating an empty .beads/ directory after simulating rm -rf,
but FindBeadsDir() requires some project files (config.yaml, metadata.json,
or *.db/*.jsonl) to recognize a directory as a valid beads directory.
Updated the test to create a minimal config.yaml like bd init actually does
before auto-import runs. This makes the test more realistic and fixes the
flakiness.
FindJSONLInDir() was returning interactions.jsonl when issues.jsonl
didn't exist. This caused bd sync to write issue data to the wrong
file after fresh init.
Add interactions.jsonl to the skip list alongside deletions.jsonl
and merge artifacts, so the function correctly defaults to issues.jsonl.
When using bd create -f with a running daemon, the command would fail
with 'database not initialized' because it bypassed the daemon RPC path.
The fix adds createIssuesFromMarkdownViaDaemon which:
- Parses the markdown file (no store access needed)
- Converts issue templates to CreateArgs
- Uses daemon Batch RPC to create all issues efficiently
- Preserves all features: labels, dependencies, hooks, JSON output
Tested with daemon and non-daemon modes, verifying priority 0 is
preserved correctly.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Migration 026_additional_indexes adds 5 indexes identified during schema review (bd-h0we):
- idx_issues_updated_at: For GetStaleIssues date filtering
- idx_issues_status_priority: For common list query patterns
- idx_labels_label_issue: Covering index for label lookups
- idx_dependencies_issue_type: For blocked issues queries
- idx_events_issue_type: For close reason queries
These indexes improve query performance for common operations,
particularly at scale (10K+ issues).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
External dependencies (external:project:capability) are now visible in
the dependency tree output. Previously they were invisible because the
recursive CTE only JOINed against the issues table.
Changes:
- GetDependencyTree now fetches external deps and adds them as synthetic
leaf nodes with resolution status (satisfied/blocked)
- formatTreeNode displays external deps with special formatting
- Added helper parseExternalRefParts for parsing external refs
Test coverage added for:
- External deps appearing in dependency tree
- Cycle detection ignoring external refs
- CheckExternalDep when target has no .beads directory
- Various invalid external ref format variations
Closes: bd-vks2, bd-mv6h, bd-d9mu
GetBlockedIssues was showing external deps as blocking even when they
were satisfied (had a closed issue with provides:capability label).
Added filterBlockedByExternalDeps() which:
- Collects all external refs from blocked issues
- Checks each with CheckExternalDeps() in batch
- Filters satisfied refs from BlockedBy lists
- Updates BlockedByCount accordingly
- Removes issues with no remaining blockers (unless status=blocked/deferred)
This matches the behavior of GetReadyWork which already filters by
external deps correctly.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove TestFindJSONLPathDefault from .test-skip (now passes)
- Add explanatory comments to 24 ignored error locations in cmd/bd:
- Cobra flag methods (MarkHidden, MarkRequired, MarkDeprecated)
- Best-effort cleanup/close operations
- Process signaling operations
Part of code health review epic bd-tggf.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add git.GetRepoRoot() with Windows path normalization
- Update beads.findGitRoot() to delegate to git.GetRepoRoot()
- Replace findBeadsDir() with beads.FindBeadsDir() across 8 files
- Remove duplicate findBeadsDir() and findGitRoot() function definitions
- Remove dead test code (TestInfoCommand, TestInfoWithNoDaemon)
- Update tests to work with consolidated APIs
Part of Code Health Review Dec 2025 epic (bd-tggf).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add EntityRef type for structured entity references with URI support
- Add Creator field to Issue for tracking who created work
- Add Validation type and Validations field for proof-of-stake approvals
- Fix RemoveDependency FK violation on external deps (bd-a3sj)
- Include all new fields in content hash computation
- Full test coverage for all new types
🤝 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add bd mol current: Shows current position in molecule workflow
- Displays all steps with status indicators (done/current/ready/blocked/pending)
- Infers molecule from in_progress issues when no ID given
- Supports --for flag to check another agent's molecules
Add bd close --continue: Auto-advances to next molecule step
- After closing, finds parent molecule and next ready step
- Auto-claims next step by default (--no-auto to skip)
- Shows molecule completion message when all steps closed
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The daemon.log was showing duplicate log messages:
- 'File change detected' appeared 4x for a single file change
- 'Git ref change detected' appeared 2x for a single ref update
Root cause:
- fsnotify generates multiple events (Write, Create, Chmod) for single file ops
- Both parent directory and file watchers can trigger for the same change
- Multiple git ref files may be updated simultaneously
Fix:
- Add log deduplication with 500ms window (matching debouncer window)
- Track last log time for file changes and git ref changes separately
- Only log if enough time has passed since last log of same type
- Still trigger debouncer for every event (functionality unchanged)
This reduces log noise while maintaining full functionality.
Improves the version bump workflow with missing local installation steps:
- --install: Now installs bd to BOTH ~/go/bin AND ~/.local/bin
- --mcp-local: Install beads-mcp from local source via uv/pip
- --restart-daemons: Restart all bd daemons to pick up new version
- --all: Shorthand for --install --mcp-local --restart-daemons
Also updated RELEASING.md with flag documentation and recommended workflow.
Closes bd-of2p
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The error message 'path exists but is not a valid git worktree' was appearing
in daemon.log when the daemon attempted to use an existing worktree that was
in the git worktree list but had other issues (broken sparse checkout, etc.).
Root cause:
- CreateBeadsWorktree only checked isValidWorktree (is it in git worktree list)
- CheckWorktreeHealth was called separately and checked additional things
- If the worktree passed isValidWorktree but failed health check, an error
was logged and repair was attempted
Fix:
- CreateBeadsWorktree now performs a full health check when it finds an
existing worktree that's in the git worktree list
- If the health check fails, it automatically removes and recreates the
worktree
- Removed redundant CheckWorktreeHealth calls in daemon_sync_branch.go and
syncbranch/worktree.go since CreateBeadsWorktree now handles this internally
This eliminates the confusing error message and ensures worktrees are always
in a healthy state after CreateBeadsWorktree returns successfully.
- Update Event-Driven Daemon Mode section from 'Experimental' to 'Default'
- Document that event-driven mode is default since v0.21.0
- Add detailed architecture diagram showing export and import flows
- Add remote-sync-interval configuration section with examples
- Document BEADS_REMOTE_SYNC_INTERVAL environment variable
- Add config.yaml example for remote-sync-interval
- Update 'Enabling Event-Driven Mode' to reflect default behavior
- Add 'Switch to Polling Mode' section for edge cases
The documentation now accurately reflects the current implementation where
event-driven mode is the default and includes periodic remote sync for
multi-clone workflows.
Add --auto-pull flag to control whether the daemon periodically pulls from
remote to check for updates from other clones.
Configuration precedence:
1. --auto-pull CLI flag (highest)
2. BEADS_AUTO_PULL environment variable
3. daemon.auto_pull in database config
4. Default: true when sync.branch is configured
When auto_pull is enabled, the daemon creates a remoteSyncTicker that
periodically calls doAutoImport() to pull remote changes. When disabled,
users must manually run 'git pull' to sync remote changes.
Changes:
- cmd/bd/daemon.go: Add --auto-pull flag and config reading logic
- cmd/bd/daemon_event_loop.go: Gate remoteSyncTicker on autoPull parameter
- cmd/bd/daemon_lifecycle.go: Add auto-pull to status output and spawn args
- internal/rpc/protocol.go: Add AutoPull field to StatusResponse
- internal/rpc/server_core.go: Add autoPull to Server struct and SetConfig
- internal/rpc/server_routing_validation_diagnostics.go: Include in status
- Tests updated to pass autoPull parameter
Closes #TBD
Modernize sorting code to use Go 1.21+ slices package:
- Replace sort.Slice with slices.SortFunc across 16 files
- Use cmp.Compare for orderable types (strings, ints)
- Use time.Time.Compare for time comparisons
- Use cmp.Or for multi-field sorting
- Use slices.SortStableFunc where stability matters
Benefits: cleaner 3-way comparison, slightly better performance,
modern idiomatic Go.
Part of GH#692 refactoring epic.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(daemon): include tombstones in exportToJSONLWithStore for sync propagation
The daemon's exportToJSONLWithStore() function was using an empty
IssueFilter which defaults to IncludeTombstones: false. This caused
deleted issues (tombstones) to be excluded from JSONL exports during
daemon sync cycles.
Bug scenario:
1. User runs `bd delete <issue>` with daemon active
2. Database correctly marks issue as tombstone
3. Main .beads/issues.jsonl correctly shows status:"tombstone"
4. But sync branch worktree JSONL still showed status:"open"
5. Other clones would not see the deletion
The fix adds IncludeTombstones: true to match the behavior of
exportToJSONL() in sync.go, ensuring tombstones propagate to other
clones and prevent resurrection of deleted issues.
Adds regression test TestExportToJSONLWithStore_IncludesTombstones
that verifies tombstones are included in daemon JSONL exports.
* fix: resolve all golangci-lint errors (cherry-pick from fix/linting-errors)
Cherry-picked linting fixes to ensure CI passes.
---------
Co-authored-by: Charles P. Cross <cpdata@users.noreply.github.com>
* fix(daemon): handle diverged sync branch with fetch-rebase-retry on push
When pushing to the sync branch, if the remote has newer commits that
the local worktree doesn't have, the push would fail with "fetch first"
error and the daemon would log the failure without recovery.
Bug scenario:
1. Clone A pushes commit X to sync branch
2. Clone B has local commit Y (not based on X)
3. Clone B's push fails with "fetch first" error
4. Without this fix: daemon logs failure and stops
5. With this fix: daemon fetches, rebases Y on X, and retries push
This fix adds fetch-rebase-retry logic to gitPushFromWorktree():
1. Detect push rejection due to remote having newer commits
2. Fetch the latest remote sync branch
3. Rebase local commits on top of remote
4. Retry the push
If rebase fails (e.g., due to conflicts), the rebase is aborted and
an error is returned with helpful context.
This allows multiple clones to push to the same sync branch without
manual intervention, as long as the changes don't conflict.
Adds integration test TestGitPushFromWorktree_FetchRebaseRetry that
verifies the fetch-rebase-retry behavior with diverged branches.
* fix: resolve all golangci-lint errors (cherry-pick from fix/linting-errors)
Cherry-picked linting fixes to ensure CI passes.
---------
Co-authored-by: Charles P. Cross <cpdata@users.noreply.github.com>
* fix(daemon): add periodic remote sync to event-driven mode
The event-driven daemon mode only triggered imports when the local JSONL
file changed (via file watcher) or when the fallback ticker fired (only
if watcher failed). This meant the daemon wouldn't see updates pushed
by other clones until something triggered a local file change.
Bug scenario:
1. Clone A creates an issue and daemon pushes to sync branch
2. Clone B's daemon only watched local file changes
3. Clone B would not see the new issue until something triggered local change
4. With this fix: Clone B's daemon periodically calls doAutoImport
This fix adds a 30-second periodic remote sync ticker that calls
doAutoImport(), which includes syncBranchPull() to fetch and import
updates from the remote sync branch.
This is essential for multi-clone workflows where:
- Clone A creates an issue and daemon pushes to sync branch
- Clone B's daemon needs to periodically pull to see the new issue
- Without periodic sync, Clone B would only see updates if its local
JSONL file happened to change
The 30-second interval balances responsiveness with network overhead.
Adds integration test TestEventDrivenLoop_PeriodicRemoteSync that
verifies the event-driven loop starts with periodic sync support.
* feat(daemon): add configurable interval for periodic remote sync
- Add BEADS_REMOTE_SYNC_INTERVAL environment variable to configure
the interval for periodic remote sync (default: 30s)
- Add getRemoteSyncInterval() function to parse the env var
- Minimum interval is 5s to prevent excessive load
- Setting to 0 disables periodic sync (not recommended)
- Add comprehensive integration tests for the configuration
Valid duration formats:
- "30s" (30 seconds)
- "1m" (1 minute)
- "5m" (5 minutes)
Tests added:
- TestEventDrivenLoop_HasRemoteSyncTicker
- TestGetRemoteSyncInterval_Default
- TestGetRemoteSyncInterval_CustomValue
- TestGetRemoteSyncInterval_MinimumEnforced
- TestGetRemoteSyncInterval_InvalidValue
- TestGetRemoteSyncInterval_Zero
- TestSyncBranchPull_FetchesRemoteUpdates
* fix: resolve all golangci-lint errors (cherry-pick from fix/linting-errors)
Cherry-picked linting fixes to ensure CI passes.
* feat(daemon): add config.yaml support for remote-sync-interval
- Add remote-sync-interval to .beads/config.yaml as alternative to
BEADS_REMOTE_SYNC_INTERVAL environment variable
- Environment variable takes precedence over config.yaml (follows
existing pattern for flush-debounce)
- Add config binding in internal/config/config.go
- Update getRemoteSyncInterval() to use config.GetDuration()
- Add doctor validation for remote-sync-interval in config.yaml
Configuration sources (in order of precedence):
1. BEADS_REMOTE_SYNC_INTERVAL environment variable
2. remote-sync-interval in .beads/config.yaml
3. DefaultRemoteSyncInterval (30s)
Example config.yaml:
remote-sync-interval: "1m"
---------
Co-authored-by: Charles P. Cross <cpdata@users.noreply.github.com>
The interactions.jsonl file is an append-only audit log that should be
tracked in git (synced with bd sync), but it was missing from the
negation patterns in .beads/.gitignore.
This adds !interactions.jsonl alongside !issues.jsonl and !metadata.json
for consistency and clarity that this file should be committed.
Co-authored-by: Christian Catalan <crcatala@gmail.com>
