fix(scanner): consume raw page owner resume oracle (#7391)

Treat persisted raw page owner entries as a validated set, not a read_dir-order prefix, so restart scans can consume committed owner pages without recounting them against the raw enumeration budget. Commit checkpointed partial pages, validate owner parent/generation/digests before the skip oracle, and fail closed on duplicate/corrupt page state.

Extend the real scanner restart driver to enforce fixed raw-entry and object budgets across fresh OS processes, and report owner-index coverage in each worker round.

Co-authored-by: zhi22915 <qiuzgang@gmail.com>

scripts/diagnose_scanner_enumeration_restart.py

@@ -30,15 +30,21 @@ def validate_report(report, *, round_number, pid, objects, budget):
         if type(report.get(key)) is not int or report[key] != value:
             raise ValueError(f"worker report mismatch: {key}")
     for key in ("raw_entries", "raw_name_bytes", "objects_before", "objects_retained",
-                "versions_retained", "bytes_retained", "objects_processed"):
+                "versions_retained", "bytes_retained", "objects_processed",
+                "raw_page_index_committed_entries", "raw_page_index_indexed_entries"):
         if type(report.get(key)) is not int or not 0 <= report[key] <= 1048576:
             raise ValueError(f"invalid bounded counter: {key}")
-    if report["raw_entries"] == 0:
+    made_budgeted_object_progress = report["objects_processed"] > 0
+    if report["raw_entries"] == 0 and not made_budgeted_object_progress:
         raise ValueError("nonempty fixture must observe raw entries; budget hook may not have run")
     if report["raw_entries"] > budget:
         raise ValueError("raw-entry budget exceeded; no unbudgeted tail is permitted")
+    if report["objects_processed"] > budget:
+        raise ValueError("object budget exceeded; no unbudgeted scan tail is permitted")
     for key in ("raw_first_entry", "raw_last_entry"):
         value = report.get(key)
+        if report["raw_entries"] == 0 and made_budgeted_object_progress and value is None:
+            continue
         if type(value) is not str or not 0 < len(value.encode("utf-8")) <= 512:
             raise ValueError(f"invalid raw entry marker: {key}")
     if type(report.get("snapshot_complete")) is not bool:

scripts/test_diagnose_scanner_enumeration_restart.py

@@ -10,6 +10,8 @@ class ReportTests(unittest.TestCase):
         return dict(schema=1, round=0, pid=123, objects_expected=4, raw_entry_budget=16,
                     raw_entries=8, raw_name_bytes=64, objects_before=0, objects_retained=4,
                     versions_retained=4, bytes_retained=4, objects_processed=4,
+                    raw_page_index_committed_entries=4,
+                    raw_page_index_indexed_entries=4,
                     raw_first_entry="bucket/object-0000",
                     raw_last_entry="bucket/object-0003/xl.meta",
                     snapshot_complete=True, outcome="complete")
@@ -44,6 +46,11 @@ class ReportTests(unittest.TestCase):
         with self.assertRaises(ValueError):
             self.validate(report)
 
+        report = self.report()
+        report["objects_processed"] = 17
+        with self.assertRaises(ValueError):
+            self.validate(report)
+
     def test_missing_or_oversized_raw_marker_rejected(self):
         for value in (None, True, "", "x" * 513):
             with self.subTest(value=value):
@@ -55,9 +62,19 @@ class ReportTests(unittest.TestCase):
     def test_complete_coverage_without_entry_observation_rejected(self):
         report = self.report()
         report["raw_entries"] = 0
+        report["objects_processed"] = 0
+        report["raw_page_index_committed_entries"] = 3
         with self.assertRaises(ValueError):
             self.validate(report)
 
+    def test_budgeted_object_progress_can_consume_all_raw_entries(self):
+        report = self.report()
+        report["raw_entries"] = 0
+        report["raw_first_entry"] = None
+        report["raw_last_entry"] = None
+        report["objects_processed"] = 1
+        self.validate(report)
+
     def test_missing_wrong_type_and_negative_counter_rejected(self):
         for value in (None, True, -1, "8", 1048577):
             with self.subTest(value=value):