fix(comments): address code-review BLOCKER + 4 IMPORTANT issues

From the end-of-impl review:

1. BLOCKER: SQLite cross-thread error in stack comments extract --bib.
   _bib_lookup_factory was sharing the bib.connect() connection across
   walker worker threads — sqlite3 defaults to check_same_thread=True
   so every worker would have raised ProgrammingError. Tests didn't
   catch it because all CLI tests pass --no-bib. Fix: open a fresh
   read-only connection with check_same_thread=False directly via
   conf.path("db.bib"); SQLite handles concurrent reads safely.

2. view._iter_index_rows now catches yaml.YAMLError too, so one
   malformed combined.md doesn't crash `stack comments stats`.

3. Image attachments (.jpg/.png/.tif/.tiff/.gif/.bmp) now classify as
   ocr_needed instead of being silently dropped as unsupported. This
   gives the phase-2 OCR command a clean candidate set to query.
   Real CMS comments include scanned-letter images.

4. view.register() defaults to rebuild=True so downstream consumers
   (#254/#255) always see fresh data. Pass rebuild=False to skip.

5. extract_comment writes via tmp file + os.replace() so concurrent
   walkers / restarts never see a half-written combined.md.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/cli/comments.py

@@ -21,14 +21,21 @@ _DEFAULT_ROOT = Path(".state/comments")
 
 
 def _bib_lookup_factory(use_bib: bool):
-    """Return a callable comment_id -> inline body. Empty if --no-bib."""
+    """Return a callable ``comment_id -> inline body``. Empty if --no-bib.
+
+    Opens its own sqlite connection with ``check_same_thread=False`` so
+    walker worker threads can call it directly. Read-only path; SQLite's
+    default thread-safe mode handles the concurrent reads.
+    """
     if not use_bib:
         return lambda _cid: ""
 
-    from bib import connect
+    import sqlite3
 
-    store = connect()
-    con = store._con()  # noqa: SLF001
+    from conf import path
+
+    db = str(path("db.bib"))
+    con = sqlite3.connect(db, check_same_thread=False)
 
     def lookup(comment_id: str) -> str:
         row = con.execute(

src/rex/comments/combine.py

@@ -7,6 +7,7 @@ sections). Resumable: existing combined.md is a no-op unless force=True.
 
 from __future__ import annotations
 
+import os
 from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any
@@ -62,7 +63,11 @@ def extract_comment(
         "attachments": attachments_meta,
     }
 
-    out.write_text(_emit(frontmatter, body_text))
+    # Atomic write: tmp file + rename, so concurrent walkers / restarts
+    # never see a half-written combined.md.
+    tmp = out.with_suffix(out.suffix + ".tmp")
+    tmp.write_text(_emit(frontmatter, body_text))
+    os.replace(tmp, out)
     return out
 
 

src/rex/comments/extract.py

@@ -26,6 +26,9 @@ class ExtractResult:
     chars: int
 
 
+_IMAGE_SUFFIXES = (".jpg", ".jpeg", ".png", ".tif", ".tiff", ".gif", ".bmp")
+
+
 def extract_attachment(path: Path) -> ExtractResult:
     """Extract text from one attachment. See module docstring for status values."""
     suffix = path.suffix.lower()
@@ -35,6 +38,10 @@ def extract_attachment(path: Path) -> ExtractResult:
         return _extract_docx(path)
     if suffix in (".txt", ".html", ".htm"):
         return _extract_text_like(path, suffix)
+    if suffix in _IMAGE_SUFFIXES:
+        # Real CMS comments include scanned-letter images; route to
+        # phase-2 OCR queue, don't drop them as "unsupported".
+        return ExtractResult(text="", status="ocr_needed", chars=0)
     return ExtractResult(text="", status="unsupported", chars=0)
 
 

src/rex/comments/view.py

@@ -12,6 +12,7 @@ import csv
 from pathlib import Path
 
 import duckdb
+import yaml
 
 from rex.comments.combine import parse_combined
 
@@ -40,10 +41,20 @@ def rebuild_index(root: Path) -> Path:
     return out
 
 
-def register(con: duckdb.DuckDBPyConnection, *, root: Path) -> None:
-    """Create the ``comments_index`` view in *con* over *root*/_index.csv."""
+def register(
+    con: duckdb.DuckDBPyConnection,
+    *,
+    root: Path,
+    rebuild: bool = True,
+) -> None:
+    """Create the ``comments_index`` view in *con* over *root*/_index.csv.
+
+    Rebuilds the index by default — downstream queries should see fresh
+    data. Pass ``rebuild=False`` to skip the walk if you've already
+    rebuilt this session and the disk hasn't changed.
+    """
     csv_path = root / _INDEX
-    if not csv_path.is_file():
+    if rebuild or not csv_path.is_file():
         rebuild_index(root)
     con.execute(
         f"CREATE OR REPLACE VIEW {_VIEW} AS "
@@ -69,7 +80,7 @@ def _iter_index_rows(root: Path):
                 continue
             try:
                 fm, _body = parse_combined(md.read_text())
-            except (ValueError, OSError):
+            except (ValueError, OSError, yaml.YAMLError):
                 continue
             atts = fm.get("attachments") or []
             yield (

tests/rex/comments/test_extract_other.py

@@ -33,3 +33,13 @@ def test_extract_unknown_extension(tmp_path: Path):
     result = extract_attachment(p)
     assert result.status == "unsupported"
     assert result.chars == 0
+
+
+def test_extract_image_marked_ocr_needed(tmp_path: Path):
+    """Image attachments (scanned letters) route to phase-2 OCR queue."""
+    for ext in (".jpg", ".jpeg", ".png", ".tif", ".tiff", ".gif", ".bmp"):
+        p = tmp_path / f"scan{ext}"
+        p.write_bytes(b"fake image bytes")
+        result = extract_attachment(p)
+        assert result.status == "ocr_needed", f"{ext}: {result.status}"
+        assert result.chars == 0