nodemap: double check the source docket when doing incremental update

In theory, the index will have the information we expect it to have. However by
security, it seems safer to double check that the incremental data are generated
from the data currently on disk.

Differential Revision: https://phab.mercurial-scm.org/D7890

mercurial/pure/parsers.py

@@ -164,11 +164,13 b' class PersistentNodeMapIndexObject(Index'
         """
         if self._nm_root is None:
             return None
+        docket = self._nm_docket
         changed, data = nodemaputil.update_persistent_data(
-            self, self._nm_root, self._nm_max_idx, self._nm_rev
+            self, self._nm_root, self._nm_max_idx, self._nm_docket.tip_rev
         )
-        self._nm_root = self._nm_max_idx = self._nm_rev = None
-        return changed, data
+
+        self._nm_root = self._nm_max_idx = self._nm_docket = None
+        return docket, changed, data
 
     def update_nodemap_data(self, docket, nm_data):
         """provide full block of persisted binary data for a nodemap
@@ -178,9 +180,9 b' class PersistentNodeMapIndexObject(Index'
         if nm_data is not None:
             self._nm_root, self._nm_max_idx = nodemaputil.parse_data(nm_data)
             if self._nm_root:
-                self._nm_rev = docket.tip_rev
+                self._nm_docket = docket
             else:
-                self._nm_root = self._nm_max_idx = self._nm_rev = None
+                self._nm_root = self._nm_max_idx = self._nm_docket = None
 
 
 class InlinedIndexObject(BaseIndexObject):

mercurial/revlogutils/nodemap.py

@@ -77,18 +77,27 b' def _persist_nodemap(tr, revlog):'
     can_incremental = util.safehasattr(revlog.index, "nodemap_data_incremental")
     ondisk_docket = revlog._nodemap_docket
 
+    data = None
     # first attemp an incremental update of the data
     if can_incremental and ondisk_docket is not None:
         target_docket = revlog._nodemap_docket.copy()
-        data_changed_count, data = revlog.index.nodemap_data_incremental()
-        datafile = _rawdata_filepath(revlog, target_docket)
-        # EXP-TODO: if this is a cache, this should use a cache vfs, not a
-        # store vfs
-        with revlog.opener(datafile, b'a') as fd:
-            fd.write(data)
-        target_docket.data_length += len(data)
-        target_docket.data_unused += data_changed_count
-    else:
+        (
+            src_docket,
+            data_changed_count,
+            data,
+        ) = revlog.index.nodemap_data_incremental()
+        if src_docket != target_docket:
+            data = None
+        else:
+            datafile = _rawdata_filepath(revlog, target_docket)
+            # EXP-TODO: if this is a cache, this should use a cache vfs, not a
+            # store vfs
+            with revlog.opener(datafile, b'a') as fd:
+                fd.write(data)
+            target_docket.data_length += len(data)
+            target_docket.data_unused += data_changed_count
+
+    if data is None:
         # otherwise fallback to a full new export
         target_docket = NodeMapDocket()
         datafile = _rawdata_filepath(revlog, target_docket)
@@ -182,6 +191,20 b' class NodeMapDocket(object):'
         new.data_unused = self.data_unused
         return new
 
+    def __cmp__(self, other):
+        if self.uid < other.uid:
+            return -1
+        if self.uid > other.uid:
+            return 1
+        elif self.data_length < other.data_length:
+            return -1
+        elif self.data_length > other.data_length:
+            return 1
+        return 0
+
+    def __eq__(self, other):
+        return self.uid == other.uid and self.data_length == other.data_length
+
     def serialize(self):
         """return serialized bytes for a docket using the passed uid"""
         data = []