Files
Codewalkers/docs/git-process-logging.md
Lukas May 9894cdd06f feat: add diffBranchesStat and diffFileSingle to BranchManager
Adds FileStatEntry type and two new primitives to the BranchManager
port and SimpleGitBranchManager adapter, enabling split diff
procedures in the tRPC layer without returning raw multi-megabyte diffs.

- FileStatEntry captures path, status, additions/deletions, oldPath
  (renames), and optional projectId for multi-project routing
- diffBranchesStat uses --name-status + --numstat, detects binary
  files (shown as - / - in numstat), handles spaces in filenames
- diffFileSingle returns raw unified diff for a single file path
2026-03-06 19:33:47 +01:00

163 lines
6.6 KiB
Markdown

# Git, Process, and Logging Modules
Three infrastructure modules supporting agent execution.
## Git Module (`apps/server/git/`)
Manages git worktrees for isolated agent workspaces.
### Architecture
- **Port**: `WorktreeManager` interface
- **Adapter**: `SimpleGitWorktreeManager` using simple-git library
### WorktreeManager Methods
| Method | Purpose |
|--------|---------|
| `create(id, branch, baseBranch?)` | Create worktree with new branch (default base: 'main') |
| `remove(id)` | Clean up worktree directory |
| `list()` | All worktrees including main |
| `get(id)` | Specific worktree by ID |
| `diff(id)` | Changed files vs HEAD |
| `merge(id, targetBranch)` | Merge worktree branch into target |
### Worktree Storage
Worktrees stored in `.cw-worktrees/` subdirectory of the repo. Each agent gets a worktree at `.cw-worktrees/agent/<alias>/`.
### Merge Flow
1. Check out target branch
2. `git merge <source> --no-edit`
3. On success: emit `worktree:merged`
4. On conflict: `git merge --abort`, emit `worktree:conflict` with conflicting file list
5. Restore original branch
### BranchManager (`apps/server/git/branch-manager.ts`)
- **Port**: `BranchManager` interface
- **Adapter**: `SimpleGitBranchManager` using simple-git
| Method | Purpose |
|--------|---------|
| `ensureBranch(repoPath, branch, baseBranch)` | Create branch from base if it doesn't exist (idempotent) |
| `mergeBranch(repoPath, source, target)` | Merge via ephemeral worktree, returns conflict info |
| `diffBranches(repoPath, base, head)` | Three-dot diff between branches |
| `diffBranchesStat(repoPath, base, head)` | Per-file metadata (path, status, additions, deletions) — no hunk content. Binary files included with `status: 'binary'` and counts of 0. Returns `FileStatEntry[]`. |
| `diffFileSingle(repoPath, base, head, filePath)` | Raw unified diff for a single file (three-dot diff). `filePath` must be URL-decoded. Returns empty string for binary files. |
| `deleteBranch(repoPath, branch)` | Delete local branch (no-op if missing) |
| `branchExists(repoPath, branch)` | Check local branches |
| `remoteBranchExists(repoPath, branch)` | Check remote tracking branches (`origin/<branch>`) |
| `listCommits(repoPath, base, head)` | List commits head has that base doesn't (with stats) |
| `diffCommit(repoPath, commitHash)` | Get unified diff for a single commit |
| `getMergeBase(repoPath, branch1, branch2)` | Get common ancestor commit hash |
| `pushBranch(repoPath, branch, remote?)` | Push branch to remote (default: 'origin') |
| `checkMergeability(repoPath, source, target)` | Dry-run merge check via `git merge-tree --write-tree` (git 2.38+). Returns `{ mergeable, conflicts? }` with no side effects |
`remoteBranchExists` is used by `registerProject` and `updateProject` to validate that a project's default branch actually exists in the cloned repository before saving.
### Project Clones
- `cloneProject(url, destPath)` — Simple git clone wrapper
- `ensureProjectClone(project, workspaceRoot)` — Idempotent: checks if clone exists, clones if not
- `getProjectCloneDir(name, id)` — Canonical path: `repos/<sanitized-name>-<id>/`
### ProjectSyncManager (`apps/server/git/remote-sync.ts`)
Fetches remote changes for project clones and fast-forwards the local `defaultBranch`. Safe to run with active worktrees — only updates remote-tracking refs and the base branch (never checked out directly by any worktree).
**Constructor**: `ProjectSyncManager(projectRepository, workspaceRoot, eventBus?)`
| Method | Purpose |
|--------|---------|
| `syncProject(projectId)` | `git fetch origin` + `git merge --ff-only origin/<defaultBranch>`, updates `lastFetchedAt` |
| `syncAllProjects()` | Sync all registered projects sequentially |
| `getSyncStatus(projectId)` | Returns `{ ahead, behind, lastFetchedAt }` via `rev-list --left-right --count` |
| `getInitiativeDivergence(projectId, branch)` | Ahead/behind between `defaultBranch` and an initiative branch |
**Sync flow during phase dispatch**: `dispatchNextPhase()` syncs all linked project clones *before* creating initiative/phase branches, so branches fork from up-to-date remote state. Sync is best-effort — failures are logged but don't block dispatch.
**CLI**: `cw project sync [name] --all`, `cw project status [name]`
**tRPC**: `syncProject`, `syncAllProjects`, `getProjectSyncStatus`
### Events Emitted
`worktree:created`, `worktree:removed`, `worktree:merged`, `worktree:conflict`, `project:synced`, `project:sync_failed`
---
## Process Module (`apps/server/process/`)
Spawns, tracks, and controls child processes.
### Classes
**ProcessRegistry** — In-memory metadata store:
- `register(info)`, `unregister(id)`, `get(id)`, `getAll()`, `getByPid(pid)`, `updateStatus(id, status)`
**ProcessManager** — Lifecycle management:
| Method | Purpose |
|--------|---------|
| `spawn(options)` | Spawn detached process (survives parent exit) |
| `stop(id)` | SIGTERM → wait 5s → SIGKILL |
| `stopAll()` | Stop all running processes in parallel |
| `restart(id)` | Stop + re-spawn with same options |
| `isRunning(id)` | Check with `process.kill(pid, 0)` |
### Spawn Details
- Uses `execa` with `detached: true`, `stdio: 'ignore'`
- Calls `subprocess.unref()` so parent can exit
- Exit handler updates registry and emits events
### Events Emitted
`process:spawned`, `process:stopped`, `process:crashed`
---
## Logger Module (`apps/server/logger/`)
Structured logging via **pino**.
### Usage
```typescript
import { createModuleLogger } from '../logger/index.js';
const log = createModuleLogger('my-module');
log.info({ key: 'value' }, 'message');
```
### Configuration
| Env Var | Effect |
|---------|--------|
| `CW_LOG_LEVEL` | Override log level |
| `CW_LOG_PRETTY` | Set to `'1'` for human-readable output |
| `NODE_ENV=development` | Default to 'debug' level |
### Output
- Default: JSON to stderr (fd 2)
- Pretty mode: pino-pretty to stdout with colors and timestamps
---
## Logging Module (`apps/server/logging/`)
File-based per-process output capture (separate from pino).
### Classes
**LogManager** — Directory management:
- Base dir: `~/.cw/logs/`
- Structure: `{processId}/stdout.log`, `{processId}/stderr.log`
- `cleanOldLogs(retainDays)` — removes old directories by mtime
**ProcessLogWriter** — File I/O with timestamps:
- `open()` — create directories and append-mode WriteStreams
- `writeStdout(data)` / `writeStderr(data)` — prefix each line with `[YYYY-MM-DD HH:mm:ss.SSS]`
- Handles backpressure (waits for drain event)
- Emits `log:entry` event via EventBus
### Factory
```typescript
import { createLogger } from './logging/index.js';
const writer = createLogger(processId, eventBus);
await writer.open();
await writer.writeStdout('output data');
await writer.close();
```