johno/beads
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: support git worktrees in hooks installation

Browse Source 
Use `git rev-parse --git-dir` instead of hardcoded `.git` path to find
the actual git directory. In worktrees, `.git` is a file containing a
gitdir pointer, not a directory.

Changes:
- Add getGitDir() helper in hooks.go
- Update installHooks(), uninstallHooks(), CheckGitHooks() to use it
- Update hooksInstalled(), detectExistingHooks(), installGitHooks() in init.go
- Update checkHooksQuick() in doctor.go
- Update GitHooks() in doctor/fix/hooks.go
- Update tests to use real git repos via `git init`

Fixes bd-63l

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Steve Yegge
2025-11-29 23:19:57 -08:00
parent fa9285a663
commit 0b13a0df3c
7 changed files with 127 additions and 58 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
.beads/issues.jsonl
View File

@@ -15,9 +15,9 @@
{"id":"bd-4l5","title":"bd prime: Detect ephemeral branches and adjust workflow output","description":"When 'bd prime' runs on a branch with no upstream (ephemeral branch), it should output a different SESSION CLOSE PROTOCOL.\n\n**Current output (wrong for ephemeral branches):**\n```\n[ ] 1. git status\n[ ] 2. git add \u003cfiles\u003e\n[ ] 3. bd sync\n[ ] 4. git commit -m \"...\"\n[ ] 5. bd sync\n[ ] 6. git push\n```\n\n**Needed output for ephemeral branches:**\n```\n[ ] 1. git status\n[ ] 2. git add \u003cfiles\u003e\n[ ] 3. bd sync --from-main (pull updates from main)\n[ ] 4. git commit -m \"...\"\n[ ] 5. (no push - branch is ephemeral)\n```\n\n**Detection:** `git rev-parse --abbrev-ref --symbolic-full-name @{u}` returns error code 128 if no upstream.\n\nAlso update Sync \u0026 Collaboration section to mention `bd sync --from-main` for ephemeral branches.\n\n**Use case:** Gastown polecats work on ephemeral local branches that are never pushed. Their code gets merged to main via local merge, and beads changes stay local (communicated via gm mail to Overseer).","status":"closed","priority":1,"issue_type":"feature","created_at":"2025-11-25T16:55:24.984104-08:00","updated_at":"2025-11-25T17:12:46.604978-08:00","closed_at":"2025-11-25T17:12:46.604978-08:00"}
{"id":"bd-4pv","title":"bd export only outputs 1 issue after auto-import corrupts database","description":"When auto-import runs and purges issues (due to git history backfill bug), subsequent 'bd export' only exports 1 issue even though the database should have many.\n\nReproduction:\n1. Have issues.jsonl with 55 issues\n2. Auto-import triggers and purges all issues via git history backfill\n3. Run 'bd export' - only exports 1 issue (the last one created before corruption)\n\nThe database gets into an inconsistent state where most issues are purged but export doesn't realize this.\n\nWorkaround: Rebuild database from scratch with 'rm .beads/beads.db \u0026\u0026 bd init --prefix bd'","status":"closed","priority":2,"issue_type":"bug","created_at":"2025-11-26T22:28:40.828866-08:00","updated_at":"2025-11-28T17:28:55.545056-08:00","closed_at":"2025-11-27T22:50:35.036227-08:00"}
{"id":"bd-4t7","title":"Auto-import runs during --no-auto-import operations via stats/ready commands","description":"Even when using --no-auto-import flag, certain commands like 'bd stats' and 'bd ready' still trigger auto-import internally, which can cause the git history backfill bug to corrupt data.\n\nExample:\n bd stats --no-auto-import\n # Still prints 'Purged bd-xxx (recovered from git history...)'\n\nThe flag should completely disable auto-import for the command, but it appears some code paths still trigger it.\n\nWorkaround: Use --allow-stale instead, or --sandbox mode.","status":"closed","priority":2,"issue_type":"bug","created_at":"2025-11-26T22:28:59.305898-08:00","updated_at":"2025-11-27T00:54:20.335013-08:00","closed_at":"2025-11-27T00:54:12.561872-08:00"}
{"id":"bd-53c","title":"bd sync corrupts issues.jsonl in multi-clone environments","description":"During parallel swarm operations on 2025-11-29, bd sync corrupted the issues database. Commit 93718056 purged all 80 issues from issues.jsonl.\n\nReproduction:\n1. Multiple clones of same repo (polecat swarm)\n2. Each clone runs bd sync in parallel\n3. One sync overwrites others, resulting in data loss\n\nEven in single-clone scenarios (~/src/beads), running bd sync after git pull seems to nuke the JSONL.\n\nWorkaround: Use bd import -i .beads/issues.jsonl --no-git-history instead of bd sync.\n\nRoot cause investigation needed. Options:\n1. Pessimistic locking (flock)\n2. Single point of merge (polecats on ephemeral branches)\n3. Beads-level locking (bd sync --lock)","status":"open","priority":0,"issue_type":"bug","created_at":"2025-11-29T16:30:30.592507-08:00","updated_at":"2025-11-29T16:30:30.592507-08:00"}
{"id":"bd-53c","title":"bd sync corrupts issues.jsonl in multi-clone environments","description":"During parallel swarm operations on 2025-11-29, bd sync corrupted the issues database. Commit 93718056 purged all 80 issues from issues.jsonl.\n\nReproduction:\n1. Multiple clones of same repo (polecat swarm)\n2. Each clone runs bd sync in parallel\n3. One sync overwrites others, resulting in data loss\n\nEven in single-clone scenarios (~/src/beads), running bd sync after git pull seems to nuke the JSONL.\n\nWorkaround: Use bd import -i .beads/issues.jsonl --no-git-history instead of bd sync.\n\nRoot cause investigation needed. Options:\n1. Pessimistic locking (flock)\n2. Single point of merge (polecats on ephemeral branches)\n3. Beads-level locking (bd sync --lock)","status":"closed","priority":0,"issue_type":"bug","created_at":"2025-11-29T16:30:30.592507-08:00","updated_at":"2025-11-29T23:19:57.254217-08:00","closed_at":"2025-11-29T23:15:02.435084-08:00"}
{"id":"bd-5bj","title":"Registry has cross-process race condition","description":"The global daemon registry (~/.beads/registry.json) can be corrupted when multiple daemons from different workspaces write simultaneously.\n\n**Root cause:**\n- Registry uses an in-process mutex but no file-level locking\n- Register() and Unregister() release the mutex between read and write\n- Multiple daemon processes can interleave their read-modify-write cycles\n\n**Evidence:**\nFound registry.json with double closing bracket: `]]` instead of `]`\n\n**Fix options:**\n1. Use file locking (flock/fcntl) around the entire read-modify-write cycle\n2. Use atomic write pattern (write to temp file, rename)\n3. Both (belt and suspenders)\n\n**Files:**\n- internal/daemon/registry.go:46-64 (readEntries)\n- internal/daemon/registry.go:67-87 (writeEntries)\n- internal/daemon/registry.go:90-108 (Register - the race window)","status":"closed","priority":2,"issue_type":"bug","created_at":"2025-11-27T13:55:50.426188-08:00","updated_at":"2025-11-27T14:07:06.22622-08:00","closed_at":"2025-11-27T14:07:06.22622-08:00"}
{"id":"bd-63l","title":"bd hooks install fails in git worktrees","description":"When bd is used in a git worktree, bd hooks install fails with 'mkdir .git: not a directory' because .git is a file (gitdir pointer) not a directory. Beads should detect and follow the .git gitdir pointer to install hooks in the correct location. This blocks normal worktree workflows.\n\n## Symptoms of this bug:\n- Git hooks don't install automatically\n- Auto-sync doesn't run (5-second debounce)\n- Hash mismatch warnings in bd output\n- Daemon fails to start with 'auto_start_failed'\n\n## Workaround:\nUse `git rev-parse --git-dir` to find the actual hooks directory and copy hooks manually:\n```bash\nmkdir -p $(git rev-parse --git-dir)/hooks\ncp -r .beads-hooks/* $(git rev-parse --git-dir)/hooks/\n```","status":"open","priority":1,"issue_type":"bug","created_at":"2025-11-29T00:27:59.111163003-07:00","updated_at":"2025-11-29T00:38:30.640871318-07:00"}
{"id":"bd-63l","title":"bd hooks install fails in git worktrees","description":"When bd is used in a git worktree, bd hooks install fails with 'mkdir .git: not a directory' because .git is a file (gitdir pointer) not a directory. Beads should detect and follow the .git gitdir pointer to install hooks in the correct location. This blocks normal worktree workflows.\n\n## Symptoms of this bug:\n- Git hooks don't install automatically\n- Auto-sync doesn't run (5-second debounce)\n- Hash mismatch warnings in bd output\n- Daemon fails to start with 'auto_start_failed'\n\n## Workaround:\nUse `git rev-parse --git-dir` to find the actual hooks directory and copy hooks manually:\n```bash\nmkdir -p $(git rev-parse --git-dir)/hooks\ncp -r .beads-hooks/* $(git rev-parse --git-dir)/hooks/\n```","status":"in_progress","priority":1,"issue_type":"bug","created_at":"2025-11-29T00:27:59.111163003-07:00","updated_at":"2025-11-29T23:19:57.254738-08:00"}
{"id":"bd-736d","title":"Refactor path canonicalization into helper function","description":"The path canonicalization logic (filepath.Abs + EvalSymlinks) is duplicated in 3 places:\n- beads.go:131-137 (BEADS_DIR handling)\n- cmd/bd/main.go:446-451 (--no-db cleanup)\n- cmd/bd/nodb.go:26-31 (--no-db initialization)\n\nRefactoring suggestion:\nExtract to a helper function like:\n func canonicalizePath(path string) string\n\nThis would:\n- Reduce code duplication\n- Make the logic easier to maintain\n- Ensure consistent behavior across all path handling\n\nRelated to bd-e16b implementation.","status":"closed","priority":3,"issue_type":"chore","created_at":"2025-11-02T18:33:47.727443-08:00","updated_at":"2025-11-25T22:27:33.738672-08:00","closed_at":"2025-11-25T22:27:33.738672-08:00"}
{"id":"bd-81a","title":"Add programmatic tip injection API","description":"Allow tips to be programmatically injected at runtime based on detected conditions. This enables dynamic tips (not just pre-defined ones) to be shown with custom priority and frequency.","status":"closed","priority":2,"issue_type":"feature","created_at":"2025-11-11T23:29:46.645583-08:00","updated_at":"2025-11-25T17:52:35.096882-08:00","closed_at":"2025-11-25T17:52:35.096882-08:00","dependencies":[{"issue_id":"bd-81a","depends_on_id":"bd-d4i","type":"blocks","created_at":"2025-11-11T23:29:46.646327-08:00","created_by":"daemon"}]}
{"id":"bd-8a5","title":"Refactor: deduplicate FindJSONLInDir and FindJSONLPath","description":"## Background\n\nAfter fixing bd-tqo, we now have two nearly identical functions for finding the JSONL file:\n- `autoimport.FindJSONLInDir(dbDir string)` in internal/autoimport/autoimport.go\n- `beads.FindJSONLPath(dbPath string)` in internal/beads/beads.go\n\nBoth implement the same logic:\n1. Prefer issues.jsonl\n2. Fall back to beads.jsonl for legacy support\n3. Skip deletions.jsonl and merge artifacts\n4. Default to issues.jsonl if nothing found\n\n## Problem\n\nCode duplication means bug fixes need to be applied in multiple places (as we just experienced with bd-tqo).\n\n## Proposed Solution\n\nExtract shared logic to a utility package that both can import. Options:\n1. Create `internal/jsonlpath` package with the core logic\n2. Have `autoimport` import `beads` and call `FindJSONLPath` (but APIs differ slightly)\n3. Move to `internal/utils` if appropriate\n\nNeed to verify no import cycles would be created.\n\n## Affected Files\n- internal/autoimport/autoimport.go\n- internal/beads/beads.go","status":"closed","priority":4,"issue_type":"task","created_at":"2025-11-26T23:45:18.974339-08:00","updated_at":"2025-11-29T22:06:06.330185-08:00","closed_at":"2025-11-28T23:07:08.912247-08:00"}