Summary

Closes the bug class where gbrain sync --strategy code against a 1500-file gstack repo could pin one thread at 99% CPU for hours with zero disk writes and a page_count that stayed at 0. Three real defects, fixed together in one bisectable commit, plus version bump and CHANGELOG.

Bug fix (commit 1):

• Tree-sitter chunker had no wall-clock cap. A single pathological file could wedge the whole sync inside WASM. New parseWithTimeout helper in src/core/chunkers/code.ts wraps parser.parse() with setTimeoutMicros(timeoutMs * 1000), throws ChunkerTimeoutError on null, and the caller's try/finally reaps parser+tree (closes the leak Codex flagged where the catch block returned without delete()). Default 30s, override via GBRAIN_CHUNKER_TIMEOUT_MS. Falls back to fallbackChunks (recursive text chunker) on timeout — degrades search quality on that one file, doesn't wedge sync.
• Code-strategy first-sync silently no-op'd on code files. performFullSync called runImport(repoPath) with no strategy; runImport only ever walked .md/.mdx. Now opts.strategy threads end-to-end (full-sync write path AND dry-run). Code files actually reach the dispatcher, which already routes them to importCodeFile.
• Walker was thrice-redundant. collectMarkdownFiles (lstat-safe) and walkSyncableFiles (statSync, weaker for no good reason) collapsed into one hardened collectSyncableFiles in src/commands/import.ts: lstat + symlink-skip with canonical log line; inode-cycle Map keyed on ${st_dev}:${st_ino} (defense-in-depth for non-symlink loops); MAX_WALK_DEPTH=32 structural backstop with GBRAIN_MAX_WALK_DEPTH override; .sort() output (Codex C8: runImport's checkpoint resume is index-based against a sorted list). Walker-context multimodal carve-out preserved at one site (Codex C5).
• Per-phase diagnostic logging. Structured [gbrain phase] <name> start/done <ms> stderr lines on sync.git_pull, sync.fullsync.import, import.collect_files, and per-file slow path (>5s). The next hang reproduction names the phase that wedged.

Version + docs (commit 2): VERSION 0.30.1 → 0.31.2, package.json synced, CHANGELOG entry under [0.31.2] (preserved master's [0.30.2] entry below), TODOS.md filed gbrain query <common-keyword> 7-day-zombie investigation (PIDs 39429, 46624) and the deferred amarillo-shape PGLite + Postgres E2E as v0.31.3 follow-ups.

Test Coverage

CODE PATHS                                                    USER FLOWS
[+] src/core/chunkers/code.ts
  ├── parseWithTimeout()                                      [+] gbrain sync --strategy code on amarillo-shape repo
  │   ├── [★★★ TESTED] timeout fires → fallback (stub seam)    └── [DEFERRED to v0.31.3]
  │   ├── [★★ TESTED]  default-timeout on real code                       (PGLite + Postgres E2E for the user's
  │   ├── [★★ TESTED]  env override works                                  exact 1500-file self-symlink topology)
  │   ├── [★★★ TESTED] cleanup-on-throw smoke
  │   ├── [★★★ TESTED] setTimeoutMicros API absent → loud
  │   └── [★★★ TESTED] ChunkerTimeoutError shape (filePath/ms)
[+] src/commands/import.ts
  ├── collectSyncableFiles()
  │   ├── [★★★ TESTED] symlink-skip + canonical log
  │   ├── [★★★ TESTED] inode-cycle detection
  │   ├── [★★★ TESTED] max-depth bailout (40-level tree)
  │   ├── [★★★ TESTED] strategy filter (code/markdown/auto)
  │   ├── [★★★ TESTED] dot-dir + node_modules skip
  │   ├── [★★★ TESTED] multimodal preservation under markdown
  │   └── [★★★ TESTED] deterministic ordering (codex C8)
  └── runImport()
      └── opts.strategy threaded (verified via existing import-resume + extract tests)

COVERAGE: 14/14 new code paths tested (100%)  |  GAPS: 0
QUALITY: ★★★:11 ★★:3 ★:0

bun test (unit suite, 109 targeted cases across changed surface): 109 pass / 0 fail. Typecheck clean.

Tests: existing test count + 14 new (test/sync-walker-symlink.test.ts × 7, test/chunker-timeout.test.ts × 7).

Pre-Landing Review

Codex outside-voice review caught 5 bug-grade findings during plan stage (parser cleanup leak, multimodal contradiction, missed dry-run plumbing, deterministic-order requirement, machine-speed-flaky tests). All five folded into the shipped fix without re-litigating the bundle decision. See ~/.claude/plans/system-instruction-you-are-working-jolly-stallman.md for the full decision history (D1-D5 + Codex C1-C11).

No new findings on the diff itself.

Plan Completion

7/8 plan items DONE. Deferred E2E item is filed as P2 follow-up; unit tests already prove walker hardening and chunker timeout behaviors.

TODOS

Filed two v0.31.2 follow-ups:

• P1: Investigate gbrain query <common-keyword> infinite loop (separate bug class, 7-day zombies at 6+GB RES)
• P2: PGLite + Postgres E2E for amarillo-shape topology (deferred from this PR)

Test plan

• bun run typecheck clean
• bun test test/sync-walker-symlink.test.ts test/chunker-timeout.test.ts — 14 pass
• Targeted suite (sync, extract, import-file, import-resume, sync-strategy) — 109 pass / 0 fail
• Manual repro on ~/conductor/workspaces/gstack/amarillo-v2 against the user's actual source — user to verify post-merge (gbrain upgrade + gbrain sync --strategy code --source <id> should complete in ~25-35 min and produce non-zero page_count)

🤖 Generated with Claude Code

^{Need help on this PR? Tag @codesmith with what you need.}

• Let Codesmith autofix CI failures and bot reviews