Fail fast for source cache populate options

Author: apstenku123

scripts/streaming_reindex.py

@@ -1135,6 +1135,18 @@ def main(argv: list[str]) -> int:
         (OUTPUT_ROOT / str(tl)).mkdir(parents=True, exist_ok=True)
     manifest = Manifest.load(MANIFEST_PATH)
     resume = not args.no_resume
+    source_cache_dir = Path(args.source_cache_dir) if args.source_cache_dir else None
+    source_dir_roots = [Path(p) for p in args.source_dir_root]
+    if args.source_cache_only and source_cache_dir is None:
+        raise SystemExit("--source-cache-only requires --source-cache-dir")
+    if args.source_cache_populate_only and source_cache_dir is None:
+        raise SystemExit("--source-cache-populate-only requires --source-cache-dir")
+    if args.source_cache_populate_only and args.source_cache_only:
+        raise SystemExit("--source-cache-populate-only cannot be combined with --source-cache-only")
+    if args.source_cache_populate_only and args.commit_source:
+        raise SystemExit("--source-cache-populate-only is code-only; remove --commit-source")
+    if source_dir_roots and (source_cache_dir is not None or args.source_cache_only):
+        raise SystemExit("--source-dir-root cannot be combined with source cache flags")
 
     # Shared global dedup db (cross-repo + cross-stream). FAIL LOUD up front:
     # open it once here so a bad path / missing datasketch crashes before any
@@ -1223,18 +1235,6 @@ def main(argv: list[str]) -> int:
         def should_process(repo: str) -> bool:
             return is_code_worktree_repo(repo) and not (resume and manifest.is_done(repo))
 
-        source_cache_dir = Path(args.source_cache_dir) if args.source_cache_dir else None
-        source_dir_roots = [Path(p) for p in args.source_dir_root]
-        if args.source_cache_only and source_cache_dir is None:
-            raise SystemExit("--source-cache-only requires --source-cache-dir")
-        if args.source_cache_populate_only and source_cache_dir is None:
-            raise SystemExit("--source-cache-populate-only requires --source-cache-dir")
-        if args.source_cache_populate_only and args.source_cache_only:
-            raise SystemExit("--source-cache-populate-only cannot be combined with --source-cache-only")
-        if args.source_cache_populate_only and args.commit_source:
-            raise SystemExit("--source-cache-populate-only is code-only; remove --commit-source")
-        if source_dir_roots and (source_cache_dir is not None or args.source_cache_only):
-            raise SystemExit("--source-dir-root cannot be combined with source cache flags")
         if args.source_cache_populate_only:
             report = populate_source_cache(
                 work_root,

tests/test_streaming_reindex_run_checked.py

@@ -183,3 +183,51 @@ def fake_stream(work_root, should_process, *, source_cache_dir, source_cache_onl
         "repos": [{"repo": "repo-a", "path": str(cache / "repo-a")}],
         "repo_count": 1,
     }
+
+
+def test_source_cache_populate_rejects_commit_source_before_processing(
+    tmp_path,
+    monkeypatch,
+) -> None:
+    import streaming_reindex
+
+    commit_source = tmp_path / "commits.jsonl"
+    commit_source.write_text("{}\n", encoding="utf-8")
+    calls = []
+
+    def fake_process_one_commit_source(*args, **kwargs):
+        calls.append((args, kwargs))
+        raise AssertionError("commit source should not be processed")
+
+    monkeypatch.setattr(
+        streaming_reindex,
+        "OUTPUT_ROOT",
+        tmp_path / "outputs" / "reindexed",
+    )
+    monkeypatch.setattr(
+        streaming_reindex,
+        "MANIFEST_PATH",
+        tmp_path / "outputs" / "reindexed" / "_done.json",
+    )
+    monkeypatch.setattr(
+        streaming_reindex,
+        "process_one_commit_source",
+        fake_process_one_commit_source,
+    )
+
+    try:
+        streaming_reindex.main(
+            [
+                "--source-cache-dir",
+                str(tmp_path / "cache"),
+                "--source-cache-populate-only",
+                "--commit-source",
+                f"sample={commit_source}",
+            ]
+        )
+    except SystemExit as exc:
+        assert "--source-cache-populate-only is code-only" in str(exc)
+    else:  # pragma: no cover
+        raise AssertionError("expected source-cache populate validation failure")
+
+    assert calls == []