changegroupv4: add sidedata helpers

These helpers carry the information and computers needed to rewrite sidedata
when generating/applying patches. We will be making use of them soon.

Differential Revision: https://phab.mercurial-scm.org/D10029

hgext/sqlitestore.py

@@ -1,1315 +1,1317 b''
 # sqlitestore.py - Storage backend that uses SQLite
 #
 # Copyright 2018 Gregory Szorc <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """store repository data in SQLite (EXPERIMENTAL)
 
 The sqlitestore extension enables the storage of repository data in SQLite.
 
 This extension is HIGHLY EXPERIMENTAL. There are NO BACKWARDS COMPATIBILITY
 GUARANTEES. This means that repositories created with this extension may
 only be usable with the exact version of this extension/Mercurial that was
 used. The extension attempts to enforce this in order to prevent repository
 corruption.
 
 In addition, several features are not yet supported or have known bugs:
 
 * Only some data is stored in SQLite. Changeset, manifest, and other repository
   data is not yet stored in SQLite.
 * Transactions are not robust. If the process is aborted at the right time
   during transaction close/rollback, the repository could be in an inconsistent
   state. This problem will diminish once all repository data is tracked by
   SQLite.
 * Bundle repositories do not work (the ability to use e.g.
   `hg -R <bundle-file> log` to automatically overlay a bundle on top of the
   existing repository).
 * Various other features don't work.
 
 This extension should work for basic clone/pull, update, and commit workflows.
 Some history rewriting operations may fail due to lack of support for bundle
 repositories.
 
 To use, activate the extension and set the ``storage.new-repo-backend`` config
 option to ``sqlite`` to enable new repositories to use SQLite for storage.
 """
 
 # To run the test suite with repos using SQLite by default, execute the
 # following:
 #
 # HGREPOFEATURES="sqlitestore" run-tests.py \
 #     --extra-config-opt extensions.sqlitestore= \
 #     --extra-config-opt storage.new-repo-backend=sqlite
 
 from __future__ import absolute_import
 
 import sqlite3
 import struct
 import threading
 import zlib
 
 from mercurial.i18n import _
 from mercurial.node import (
     nullid,
     nullrev,
     short,
 )
 from mercurial.thirdparty import attr
 from mercurial import (
     ancestor,
     dagop,
     encoding,
     error,
     extensions,
     localrepo,
     mdiff,
     pycompat,
     registrar,
     requirements,
     util,
     verify,
 )
 from mercurial.interfaces import (
     repository,
     util as interfaceutil,
 )
 from mercurial.utils import (
     hashutil,
     storageutil,
 )
 
 try:
     from mercurial import zstd
 
     zstd.__version__
 except ImportError:
     zstd = None
 
 configtable = {}
 configitem = registrar.configitem(configtable)
 
 # experimental config: storage.sqlite.compression
 configitem(
     b'storage',
     b'sqlite.compression',
     default=b'zstd' if zstd else b'zlib',
     experimental=True,
 )
 
 # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
 # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
 # be specifying the version(s) of Mercurial they are tested with, or
 # leave the attribute unspecified.
 testedwith = b'ships-with-hg-core'
 
 REQUIREMENT = b'exp-sqlite-001'
 REQUIREMENT_ZSTD = b'exp-sqlite-comp-001=zstd'
 REQUIREMENT_ZLIB = b'exp-sqlite-comp-001=zlib'
 REQUIREMENT_NONE = b'exp-sqlite-comp-001=none'
 REQUIREMENT_SHALLOW_FILES = b'exp-sqlite-shallow-files'
 
 CURRENT_SCHEMA_VERSION = 1
 
 COMPRESSION_NONE = 1
 COMPRESSION_ZSTD = 2
 COMPRESSION_ZLIB = 3
 
 FLAG_CENSORED = 1
 FLAG_MISSING_P1 = 2
 FLAG_MISSING_P2 = 4
 
 CREATE_SCHEMA = [
     # Deltas are stored as content-indexed blobs.
     # compression column holds COMPRESSION_* constant for how the
     # delta is encoded.
     'CREATE TABLE delta ('
     '    id INTEGER PRIMARY KEY, '
     '    compression INTEGER NOT NULL, '
     '    hash BLOB UNIQUE ON CONFLICT ABORT, '
     '    delta BLOB NOT NULL '
     ')',
     # Tracked paths are denormalized to integers to avoid redundant
     # storage of the path name.
     'CREATE TABLE filepath ('
     '    id INTEGER PRIMARY KEY, '
     '    path BLOB NOT NULL '
     ')',
     'CREATE UNIQUE INDEX filepath_path ON filepath (path)',
     # We have a single table for all file revision data.
     # Each file revision is uniquely described by a (path, rev) and
     # (path, node).
     #
     # Revision data is stored as a pointer to the delta producing this
     # revision and the file revision whose delta should be applied before
     # that one. One can reconstruct the delta chain by recursively following
     # the delta base revision pointers until one encounters NULL.
     #
     # flags column holds bitwise integer flags controlling storage options.
     # These flags are defined by the FLAG_* constants.
     'CREATE TABLE fileindex ('
     '    id INTEGER PRIMARY KEY, '
     '    pathid INTEGER REFERENCES filepath(id), '
     '    revnum INTEGER NOT NULL, '
     '    p1rev INTEGER NOT NULL, '
     '    p2rev INTEGER NOT NULL, '
     '    linkrev INTEGER NOT NULL, '
     '    flags INTEGER NOT NULL, '
     '    deltaid INTEGER REFERENCES delta(id), '
     '    deltabaseid INTEGER REFERENCES fileindex(id), '
     '    node BLOB NOT NULL '
     ')',
     'CREATE UNIQUE INDEX fileindex_pathrevnum '
     '    ON fileindex (pathid, revnum)',
     'CREATE UNIQUE INDEX fileindex_pathnode ON fileindex (pathid, node)',
     # Provide a view over all file data for convenience.
     'CREATE VIEW filedata AS '
     'SELECT '
     '    fileindex.id AS id, '
     '    filepath.id AS pathid, '
     '    filepath.path AS path, '
     '    fileindex.revnum AS revnum, '
     '    fileindex.node AS node, '
     '    fileindex.p1rev AS p1rev, '
     '    fileindex.p2rev AS p2rev, '
     '    fileindex.linkrev AS linkrev, '
     '    fileindex.flags AS flags, '
     '    fileindex.deltaid AS deltaid, '
     '    fileindex.deltabaseid AS deltabaseid '
     'FROM filepath, fileindex '
     'WHERE fileindex.pathid=filepath.id',
     'PRAGMA user_version=%d' % CURRENT_SCHEMA_VERSION,
 ]
 
 
 def resolvedeltachain(db, pathid, node, revisioncache, stoprids, zstddctx=None):
     """Resolve a delta chain for a file node."""
 
     # TODO the "not in ({stops})" here is possibly slowing down the query
     # because it needs to perform the lookup on every recursive invocation.
     # This could possibly be faster if we created a temporary query with
     # baseid "poisoned" to null and limited the recursive filter to
     # "is not null".
     res = db.execute(
         'WITH RECURSIVE '
         '    deltachain(deltaid, baseid) AS ('
         '        SELECT deltaid, deltabaseid FROM fileindex '
         '            WHERE pathid=? AND node=? '
         '        UNION ALL '
         '        SELECT fileindex.deltaid, deltabaseid '
         '            FROM fileindex, deltachain '
         '            WHERE '
         '                fileindex.id=deltachain.baseid '
         '                AND deltachain.baseid IS NOT NULL '
         '                AND fileindex.id NOT IN ({stops}) '
         '    ) '
         'SELECT deltachain.baseid, compression, delta '
         'FROM deltachain, delta '
         'WHERE delta.id=deltachain.deltaid'.format(
             stops=','.join(['?'] * len(stoprids))
         ),
         tuple([pathid, node] + list(stoprids.keys())),
     )
 
     deltas = []
     lastdeltabaseid = None
 
     for deltabaseid, compression, delta in res:
         lastdeltabaseid = deltabaseid
 
         if compression == COMPRESSION_ZSTD:
             delta = zstddctx.decompress(delta)
         elif compression == COMPRESSION_NONE:
             delta = delta
         elif compression == COMPRESSION_ZLIB:
             delta = zlib.decompress(delta)
         else:
             raise SQLiteStoreError(
                 b'unhandled compression type: %d' % compression
             )
 
         deltas.append(delta)
 
     if lastdeltabaseid in stoprids:
         basetext = revisioncache[stoprids[lastdeltabaseid]]
     else:
         basetext = deltas.pop()
 
     deltas.reverse()
     fulltext = mdiff.patches(basetext, deltas)
 
     # SQLite returns buffer instances for blob columns on Python 2. This
     # type can propagate through the delta application layer. Because
     # downstream callers assume revisions are bytes, cast as needed.
     if not isinstance(fulltext, bytes):
         fulltext = bytes(delta)
 
     return fulltext
 
 
 def insertdelta(db, compression, hash, delta):
     try:
         return db.execute(
             'INSERT INTO delta (compression, hash, delta) VALUES (?, ?, ?)',
             (compression, hash, delta),
         ).lastrowid
     except sqlite3.IntegrityError:
         return db.execute(
             'SELECT id FROM delta WHERE hash=?', (hash,)
         ).fetchone()[0]
 
 
 class SQLiteStoreError(error.StorageError):
     pass
 
 
 @attr.s
 class revisionentry(object):
     rid = attr.ib()
     rev = attr.ib()
     node = attr.ib()
     p1rev = attr.ib()
     p2rev = attr.ib()
     p1node = attr.ib()
     p2node = attr.ib()
     linkrev = attr.ib()
     flags = attr.ib()
 
 
 @interfaceutil.implementer(repository.irevisiondelta)
 @attr.s(slots=True)
 class sqliterevisiondelta(object):
     node = attr.ib()
     p1node = attr.ib()
     p2node = attr.ib()
     basenode = attr.ib()
     flags = attr.ib()
     baserevisionsize = attr.ib()
     revision = attr.ib()
     delta = attr.ib()
     sidedata = attr.ib()
     linknode = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.iverifyproblem)
 @attr.s(frozen=True)
 class sqliteproblem(object):
     warning = attr.ib(default=None)
     error = attr.ib(default=None)
     node = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.ifilestorage)
 class sqlitefilestore(object):
     """Implements storage for an individual tracked path."""
 
     def __init__(self, db, path, compression):
         self._db = db
         self._path = path
 
         self._pathid = None
 
         # revnum -> node
         self._revtonode = {}
         # node -> revnum
         self._nodetorev = {}
         # node -> data structure
         self._revisions = {}
 
         self._revisioncache = util.lrucachedict(10)
 
         self._compengine = compression
 
         if compression == b'zstd':
             self._cctx = zstd.ZstdCompressor(level=3)
             self._dctx = zstd.ZstdDecompressor()
         else:
             self._cctx = None
             self._dctx = None
 
         self._refreshindex()
 
     def _refreshindex(self):
         self._revtonode = {}
         self._nodetorev = {}
         self._revisions = {}
 
         res = list(
             self._db.execute(
                 'SELECT id FROM filepath WHERE path=?', (self._path,)
             )
         )
 
         if not res:
             self._pathid = None
             return
 
         self._pathid = res[0][0]
 
         res = self._db.execute(
             'SELECT id, revnum, node, p1rev, p2rev, linkrev, flags '
             'FROM fileindex '
             'WHERE pathid=? '
             'ORDER BY revnum ASC',
             (self._pathid,),
         )
 
         for i, row in enumerate(res):
             rid, rev, node, p1rev, p2rev, linkrev, flags = row
 
             if i != rev:
                 raise SQLiteStoreError(
                     _(b'sqlite database has inconsistent revision numbers')
                 )
 
             if p1rev == nullrev:
                 p1node = nullid
             else:
                 p1node = self._revtonode[p1rev]
 
             if p2rev == nullrev:
                 p2node = nullid
             else:
                 p2node = self._revtonode[p2rev]
 
             entry = revisionentry(
                 rid=rid,
                 rev=rev,
                 node=node,
                 p1rev=p1rev,
                 p2rev=p2rev,
                 p1node=p1node,
                 p2node=p2node,
                 linkrev=linkrev,
                 flags=flags,
             )
 
             self._revtonode[rev] = node
             self._nodetorev[node] = rev
             self._revisions[node] = entry
 
     # Start of ifileindex interface.
 
     def __len__(self):
         return len(self._revisions)
 
     def __iter__(self):
         return iter(pycompat.xrange(len(self._revisions)))
 
     def hasnode(self, node):
         if node == nullid:
             return False
 
         return node in self._nodetorev
 
     def revs(self, start=0, stop=None):
         return storageutil.iterrevs(
             len(self._revisions), start=start, stop=stop
         )
 
     def parents(self, node):
         if node == nullid:
             return nullid, nullid
 
         if node not in self._revisions:
             raise error.LookupError(node, self._path, _(b'no node'))
 
         entry = self._revisions[node]
         return entry.p1node, entry.p2node
 
     def parentrevs(self, rev):
         if rev == nullrev:
             return nullrev, nullrev
 
         if rev not in self._revtonode:
             raise IndexError(rev)
 
         entry = self._revisions[self._revtonode[rev]]
         return entry.p1rev, entry.p2rev
 
     def rev(self, node):
         if node == nullid:
             return nullrev
 
         if node not in self._nodetorev:
             raise error.LookupError(node, self._path, _(b'no node'))
 
         return self._nodetorev[node]
 
     def node(self, rev):
         if rev == nullrev:
             return nullid
 
         if rev not in self._revtonode:
             raise IndexError(rev)
 
         return self._revtonode[rev]
 
     def lookup(self, node):
         return storageutil.fileidlookup(self, node, self._path)
 
     def linkrev(self, rev):
         if rev == nullrev:
             return nullrev
 
         if rev not in self._revtonode:
             raise IndexError(rev)
 
         entry = self._revisions[self._revtonode[rev]]
         return entry.linkrev
 
     def iscensored(self, rev):
         if rev == nullrev:
             return False
 
         if rev not in self._revtonode:
             raise IndexError(rev)
 
         return self._revisions[self._revtonode[rev]].flags & FLAG_CENSORED
 
     def commonancestorsheads(self, node1, node2):
         rev1 = self.rev(node1)
         rev2 = self.rev(node2)
 
         ancestors = ancestor.commonancestorsheads(self.parentrevs, rev1, rev2)
         return pycompat.maplist(self.node, ancestors)
 
     def descendants(self, revs):
         # TODO we could implement this using a recursive SQL query, which
         # might be faster.
         return dagop.descendantrevs(revs, self.revs, self.parentrevs)
 
     def heads(self, start=None, stop=None):
         if start is None and stop is None:
             if not len(self):
                 return [nullid]
 
         startrev = self.rev(start) if start is not None else nullrev
         stoprevs = {self.rev(n) for n in stop or []}
 
         revs = dagop.headrevssubset(
             self.revs, self.parentrevs, startrev=startrev, stoprevs=stoprevs
         )
 
         return [self.node(rev) for rev in revs]
 
     def children(self, node):
         rev = self.rev(node)
 
         res = self._db.execute(
             'SELECT'
             '  node '
             '  FROM filedata '
             '  WHERE path=? AND (p1rev=? OR p2rev=?) '
             '  ORDER BY revnum ASC',
             (self._path, rev, rev),
         )
 
         return [row[0] for row in res]
 
     # End of ifileindex interface.
 
     # Start of ifiledata interface.
 
     def size(self, rev):
         if rev == nullrev:
             return 0
 
         if rev not in self._revtonode:
             raise IndexError(rev)
 
         node = self._revtonode[rev]
 
         if self.renamed(node):
             return len(self.read(node))
 
         return len(self.revision(node))
 
     def revision(self, node, raw=False, _verifyhash=True):
         if node in (nullid, nullrev):
             return b''
 
         if isinstance(node, int):
             node = self.node(node)
 
         if node not in self._nodetorev:
             raise error.LookupError(node, self._path, _(b'no node'))
 
         if node in self._revisioncache:
             return self._revisioncache[node]
 
         # Because we have a fulltext revision cache, we are able to
         # short-circuit delta chain traversal and decompression as soon as
         # we encounter a revision in the cache.
 
         stoprids = {self._revisions[n].rid: n for n in self._revisioncache}
 
         if not stoprids:
             stoprids[-1] = None
 
         fulltext = resolvedeltachain(
             self._db,
             self._pathid,
             node,
             self._revisioncache,
             stoprids,
             zstddctx=self._dctx,
         )
 
         # Don't verify hashes if parent nodes were rewritten, as the hash
         # wouldn't verify.
         if self._revisions[node].flags & (FLAG_MISSING_P1 | FLAG_MISSING_P2):
             _verifyhash = False
 
         if _verifyhash:
             self._checkhash(fulltext, node)
             self._revisioncache[node] = fulltext
 
         return fulltext
 
     def rawdata(self, *args, **kwargs):
         return self.revision(*args, **kwargs)
 
     def read(self, node):
         return storageutil.filtermetadata(self.revision(node))
 
     def renamed(self, node):
         return storageutil.filerevisioncopied(self, node)
 
     def cmp(self, node, fulltext):
         return not storageutil.filedataequivalent(self, node, fulltext)
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
+        sidedata_helpers=None,
     ):
         if nodesorder not in (b'nodes', b'storage', b'linear', None):
             raise error.ProgrammingError(
                 b'unhandled value for nodesorder: %s' % nodesorder
             )
 
         nodes = [n for n in nodes if n != nullid]
 
         if not nodes:
             return
 
         # TODO perform in a single query.
         res = self._db.execute(
             'SELECT revnum, deltaid FROM fileindex '
             'WHERE pathid=? '
             '    AND node in (%s)' % (','.join(['?'] * len(nodes))),
             tuple([self._pathid] + nodes),
         )
 
         deltabases = {}
 
         for rev, deltaid in res:
             res = self._db.execute(
                 'SELECT revnum from fileindex WHERE pathid=? AND deltaid=?',
                 (self._pathid, deltaid),
             )
             deltabases[rev] = res.fetchone()[0]
 
         # TODO define revdifffn so we can use delta from storage.
         for delta in storageutil.emitrevisions(
             self,
             nodes,
             nodesorder,
             sqliterevisiondelta,
             deltaparentfn=deltabases.__getitem__,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
             deltamode=deltamode,
+            sidedata_helpers=sidedata_helpers,
         ):
 
             yield delta
 
     # End of ifiledata interface.
 
     # Start of ifilemutation interface.
 
     def add(self, filedata, meta, transaction, linkrev, p1, p2):
         if meta or filedata.startswith(b'\x01\n'):
             filedata = storageutil.packmeta(meta, filedata)
 
         rev = self.addrevision(filedata, transaction, linkrev, p1, p2)
         return self.node(rev)
 
     def addrevision(
         self,
         revisiondata,
         transaction,
         linkrev,
         p1,
         p2,
         node=None,
         flags=0,
         cachedelta=None,
     ):
         if flags:
             raise SQLiteStoreError(_(b'flags not supported on revisions'))
 
         validatehash = node is not None
         node = node or storageutil.hashrevisionsha1(revisiondata, p1, p2)
 
         if validatehash:
             self._checkhash(revisiondata, node, p1, p2)
 
         rev = self._nodetorev.get(node)
         if rev is not None:
             return rev
 
         rev = self._addrawrevision(
             node, revisiondata, transaction, linkrev, p1, p2
         )
 
         self._revisioncache[node] = revisiondata
         return rev
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         addrevisioncb=None,
         duplicaterevisioncb=None,
         maybemissingparents=False,
     ):
         empty = True
 
         for (
             node,
             p1,
             p2,
             linknode,
             deltabase,
             delta,
             wireflags,
             sidedata,
         ) in deltas:
             storeflags = 0
 
             if wireflags & repository.REVISION_FLAG_CENSORED:
                 storeflags |= FLAG_CENSORED
 
             if wireflags & ~repository.REVISION_FLAG_CENSORED:
                 raise SQLiteStoreError(b'unhandled revision flag')
 
             if maybemissingparents:
                 if p1 != nullid and not self.hasnode(p1):
                     p1 = nullid
                     storeflags |= FLAG_MISSING_P1
 
                 if p2 != nullid and not self.hasnode(p2):
                     p2 = nullid
                     storeflags |= FLAG_MISSING_P2
 
             baserev = self.rev(deltabase)
 
             # If base is censored, delta must be full replacement in a single
             # patch operation.
             if baserev != nullrev and self.iscensored(baserev):
                 hlen = struct.calcsize(b'>lll')
                 oldlen = len(self.rawdata(deltabase, _verifyhash=False))
                 newlen = len(delta) - hlen
 
                 if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
                     raise error.CensoredBaseError(self._path, deltabase)
 
             if not (storeflags & FLAG_CENSORED) and storageutil.deltaiscensored(
                 delta, baserev, lambda x: len(self.rawdata(x))
             ):
                 storeflags |= FLAG_CENSORED
 
             linkrev = linkmapper(linknode)
 
             if node in self._revisions:
                 # Possibly reset parents to make them proper.
                 entry = self._revisions[node]
 
                 if entry.flags & FLAG_MISSING_P1 and p1 != nullid:
                     entry.p1node = p1
                     entry.p1rev = self._nodetorev[p1]
                     entry.flags &= ~FLAG_MISSING_P1
 
                     self._db.execute(
                         'UPDATE fileindex SET p1rev=?, flags=? WHERE id=?',
                         (self._nodetorev[p1], entry.flags, entry.rid),
                     )
 
                 if entry.flags & FLAG_MISSING_P2 and p2 != nullid:
                     entry.p2node = p2
                     entry.p2rev = self._nodetorev[p2]
                     entry.flags &= ~FLAG_MISSING_P2
 
                     self._db.execute(
                         'UPDATE fileindex SET p2rev=?, flags=? WHERE id=?',
                         (self._nodetorev[p1], entry.flags, entry.rid),
                     )
 
                 if duplicaterevisioncb:
                     duplicaterevisioncb(self, self.rev(node))
                 empty = False
                 continue
 
             if deltabase == nullid:
                 text = mdiff.patch(b'', delta)
                 storedelta = None
             else:
                 text = None
                 storedelta = (deltabase, delta)
 
             rev = self._addrawrevision(
                 node,
                 text,
                 transaction,
                 linkrev,
                 p1,
                 p2,
                 storedelta=storedelta,
                 flags=storeflags,
             )
 
             if addrevisioncb:
                 addrevisioncb(self, rev)
             empty = False
 
         return not empty
 
     def censorrevision(self, tr, censornode, tombstone=b''):
         tombstone = storageutil.packmeta({b'censored': tombstone}, b'')
 
         # This restriction is cargo culted from revlogs and makes no sense for
         # SQLite, since columns can be resized at will.
         if len(tombstone) > len(self.rawdata(censornode)):
             raise error.Abort(
                 _(b'censor tombstone must be no longer than censored data')
             )
 
         # We need to replace the censored revision's data with the tombstone.
         # But replacing that data will have implications for delta chains that
         # reference it.
         #
         # While "better," more complex strategies are possible, we do something
         # simple: we find delta chain children of the censored revision and we
         # replace those incremental deltas with fulltexts of their corresponding
         # revision. Then we delete the now-unreferenced delta and original
         # revision and insert a replacement.
 
         # Find the delta to be censored.
         censoreddeltaid = self._db.execute(
             'SELECT deltaid FROM fileindex WHERE id=?',
             (self._revisions[censornode].rid,),
         ).fetchone()[0]
 
         # Find all its delta chain children.
         # TODO once we support storing deltas for !files, we'll need to look
         # for those delta chains too.
         rows = list(
             self._db.execute(
                 'SELECT id, pathid, node FROM fileindex '
                 'WHERE deltabaseid=? OR deltaid=?',
                 (censoreddeltaid, censoreddeltaid),
             )
         )
 
         for row in rows:
             rid, pathid, node = row
 
             fulltext = resolvedeltachain(
                 self._db, pathid, node, {}, {-1: None}, zstddctx=self._dctx
             )
 
             deltahash = hashutil.sha1(fulltext).digest()
 
             if self._compengine == b'zstd':
                 deltablob = self._cctx.compress(fulltext)
                 compression = COMPRESSION_ZSTD
             elif self._compengine == b'zlib':
                 deltablob = zlib.compress(fulltext)
                 compression = COMPRESSION_ZLIB
             elif self._compengine == b'none':
                 deltablob = fulltext
                 compression = COMPRESSION_NONE
             else:
                 raise error.ProgrammingError(
                     b'unhandled compression engine: %s' % self._compengine
                 )
 
             if len(deltablob) >= len(fulltext):
                 deltablob = fulltext
                 compression = COMPRESSION_NONE
 
             deltaid = insertdelta(self._db, compression, deltahash, deltablob)
 
             self._db.execute(
                 'UPDATE fileindex SET deltaid=?, deltabaseid=NULL '
                 'WHERE id=?',
                 (deltaid, rid),
             )
 
         # Now create the tombstone delta and replace the delta on the censored
         # node.
         deltahash = hashutil.sha1(tombstone).digest()
         tombstonedeltaid = insertdelta(
             self._db, COMPRESSION_NONE, deltahash, tombstone
         )
 
         flags = self._revisions[censornode].flags
         flags |= FLAG_CENSORED
 
         self._db.execute(
             'UPDATE fileindex SET flags=?, deltaid=?, deltabaseid=NULL '
             'WHERE pathid=? AND node=?',
             (flags, tombstonedeltaid, self._pathid, censornode),
         )
 
         self._db.execute('DELETE FROM delta WHERE id=?', (censoreddeltaid,))
 
         self._refreshindex()
         self._revisioncache.clear()
 
     def getstrippoint(self, minlink):
         return storageutil.resolvestripinfo(
             minlink,
             len(self) - 1,
             [self.rev(n) for n in self.heads()],
             self.linkrev,
             self.parentrevs,
         )
 
     def strip(self, minlink, transaction):
         if not len(self):
             return
 
         rev, _ignored = self.getstrippoint(minlink)
 
         if rev == len(self):
             return
 
         for rev in self.revs(rev):
             self._db.execute(
                 'DELETE FROM fileindex WHERE pathid=? AND node=?',
                 (self._pathid, self.node(rev)),
             )
 
         # TODO how should we garbage collect data in delta table?
 
         self._refreshindex()
 
     # End of ifilemutation interface.
 
     # Start of ifilestorage interface.
 
     def files(self):
         return []
 
     def sidedata(self, nodeorrev, _df=None):
         # Not supported for now
         return {}
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         d = {}
 
         if exclusivefiles:
             d[b'exclusivefiles'] = []
 
         if sharedfiles:
             # TODO list sqlite file(s) here.
             d[b'sharedfiles'] = []
 
         if revisionscount:
             d[b'revisionscount'] = len(self)
 
         if trackedsize:
             d[b'trackedsize'] = sum(
                 len(self.revision(node)) for node in self._nodetorev
             )
 
         if storedsize:
             # TODO implement this?
             d[b'storedsize'] = None
 
         return d
 
     def verifyintegrity(self, state):
         state[b'skipread'] = set()
 
         for rev in self:
             node = self.node(rev)
 
             try:
                 self.revision(node)
             except Exception as e:
                 yield sqliteproblem(
                     error=_(b'unpacking %s: %s') % (short(node), e), node=node
                 )
 
                 state[b'skipread'].add(node)
 
     # End of ifilestorage interface.
 
     def _checkhash(self, fulltext, node, p1=None, p2=None):
         if p1 is None and p2 is None:
             p1, p2 = self.parents(node)
 
         if node == storageutil.hashrevisionsha1(fulltext, p1, p2):
             return
 
         try:
             del self._revisioncache[node]
         except KeyError:
             pass
 
         if storageutil.iscensoredtext(fulltext):
             raise error.CensoredNodeError(self._path, node, fulltext)
 
         raise SQLiteStoreError(_(b'integrity check failed on %s') % self._path)
 
     def _addrawrevision(
         self,
         node,
         revisiondata,
         transaction,
         linkrev,
         p1,
         p2,
         storedelta=None,
         flags=0,
     ):
         if self._pathid is None:
             res = self._db.execute(
                 'INSERT INTO filepath (path) VALUES (?)', (self._path,)
             )
             self._pathid = res.lastrowid
 
         # For simplicity, always store a delta against p1.
         # TODO we need a lot more logic here to make behavior reasonable.
 
         if storedelta:
             deltabase, delta = storedelta
 
             if isinstance(deltabase, int):
                 deltabase = self.node(deltabase)
 
         else:
             assert revisiondata is not None
             deltabase = p1
 
             if deltabase == nullid:
                 delta = revisiondata
             else:
                 delta = mdiff.textdiff(
                     self.revision(self.rev(deltabase)), revisiondata
                 )
 
         # File index stores a pointer to its delta and the parent delta.
         # The parent delta is stored via a pointer to the fileindex PK.
         if deltabase == nullid:
             baseid = None
         else:
             baseid = self._revisions[deltabase].rid
 
         # Deltas are stored with a hash of their content. This allows
         # us to de-duplicate. The table is configured to ignore conflicts
         # and it is faster to just insert and silently noop than to look
         # first.
         deltahash = hashutil.sha1(delta).digest()
 
         if self._compengine == b'zstd':
             deltablob = self._cctx.compress(delta)
             compression = COMPRESSION_ZSTD
         elif self._compengine == b'zlib':
             deltablob = zlib.compress(delta)
             compression = COMPRESSION_ZLIB
         elif self._compengine == b'none':
             deltablob = delta
             compression = COMPRESSION_NONE
         else:
             raise error.ProgrammingError(
                 b'unhandled compression engine: %s' % self._compengine
             )
 
         # Don't store compressed data if it isn't practical.
         if len(deltablob) >= len(delta):
             deltablob = delta
             compression = COMPRESSION_NONE
 
         deltaid = insertdelta(self._db, compression, deltahash, deltablob)
 
         rev = len(self)
 
         if p1 == nullid:
             p1rev = nullrev
         else:
             p1rev = self._nodetorev[p1]
 
         if p2 == nullid:
             p2rev = nullrev
         else:
             p2rev = self._nodetorev[p2]
 
         rid = self._db.execute(
             'INSERT INTO fileindex ('
             '    pathid, revnum, node, p1rev, p2rev, linkrev, flags, '
             '    deltaid, deltabaseid) '
             '    VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)',
             (
                 self._pathid,
                 rev,
                 node,
                 p1rev,
                 p2rev,
                 linkrev,
                 flags,
                 deltaid,
                 baseid,
             ),
         ).lastrowid
 
         entry = revisionentry(
             rid=rid,
             rev=rev,
             node=node,
             p1rev=p1rev,
             p2rev=p2rev,
             p1node=p1,
             p2node=p2,
             linkrev=linkrev,
             flags=flags,
         )
 
         self._nodetorev[node] = rev
         self._revtonode[rev] = node
         self._revisions[node] = entry
 
         return rev
 
 
 class sqliterepository(localrepo.localrepository):
     def cancopy(self):
         return False
 
     def transaction(self, *args, **kwargs):
         current = self.currenttransaction()
 
         tr = super(sqliterepository, self).transaction(*args, **kwargs)
 
         if current:
             return tr
 
         self._dbconn.execute('BEGIN TRANSACTION')
 
         def committransaction(_):
             self._dbconn.commit()
 
         tr.addfinalize(b'sqlitestore', committransaction)
 
         return tr
 
     @property
     def _dbconn(self):
         # SQLite connections can only be used on the thread that created
         # them. In most cases, this "just works." However, hgweb uses
         # multiple threads.
         tid = threading.current_thread().ident
 
         if self._db:
             if self._db[0] == tid:
                 return self._db[1]
 
         db = makedb(self.svfs.join(b'db.sqlite'))
         self._db = (tid, db)
 
         return db
 
 
 def makedb(path):
     """Construct a database handle for a database at path."""
 
     db = sqlite3.connect(encoding.strfromlocal(path))
     db.text_factory = bytes
 
     res = db.execute('PRAGMA user_version').fetchone()[0]
 
     # New database.
     if res == 0:
         for statement in CREATE_SCHEMA:
             db.execute(statement)
 
         db.commit()
 
     elif res == CURRENT_SCHEMA_VERSION:
         pass
 
     else:
         raise error.Abort(_(b'sqlite database has unrecognized version'))
 
     db.execute('PRAGMA journal_mode=WAL')
 
     return db
 
 
 def featuresetup(ui, supported):
     supported.add(REQUIREMENT)
 
     if zstd:
         supported.add(REQUIREMENT_ZSTD)
 
     supported.add(REQUIREMENT_ZLIB)
     supported.add(REQUIREMENT_NONE)
     supported.add(REQUIREMENT_SHALLOW_FILES)
     supported.add(requirements.NARROW_REQUIREMENT)
 
 
 def newreporequirements(orig, ui, createopts):
     if createopts[b'backend'] != b'sqlite':
         return orig(ui, createopts)
 
     # This restriction can be lifted once we have more confidence.
     if b'sharedrepo' in createopts:
         raise error.Abort(
             _(b'shared repositories not supported with SQLite store')
         )
 
     # This filtering is out of an abundance of caution: we want to ensure
     # we honor creation options and we do that by annotating exactly the
     # creation options we recognize.
     known = {
         b'narrowfiles',
         b'backend',
         b'shallowfilestore',
     }
 
     unsupported = set(createopts) - known
     if unsupported:
         raise error.Abort(
             _(b'SQLite store does not support repo creation option: %s')
             % b', '.join(sorted(unsupported))
         )
 
     # Since we're a hybrid store that still relies on revlogs, we fall back
     # to using the revlogv1 backend's storage requirements then adding our
     # own requirement.
     createopts[b'backend'] = b'revlogv1'
     requirements = orig(ui, createopts)
     requirements.add(REQUIREMENT)
 
     compression = ui.config(b'storage', b'sqlite.compression')
 
     if compression == b'zstd' and not zstd:
         raise error.Abort(
             _(
                 b'storage.sqlite.compression set to "zstd" but '
                 b'zstandard compression not available to this '
                 b'Mercurial install'
             )
         )
 
     if compression == b'zstd':
         requirements.add(REQUIREMENT_ZSTD)
     elif compression == b'zlib':
         requirements.add(REQUIREMENT_ZLIB)
     elif compression == b'none':
         requirements.add(REQUIREMENT_NONE)
     else:
         raise error.Abort(
             _(
                 b'unknown compression engine defined in '
                 b'storage.sqlite.compression: %s'
             )
             % compression
         )
 
     if createopts.get(b'shallowfilestore'):
         requirements.add(REQUIREMENT_SHALLOW_FILES)
 
     return requirements
 
 
 @interfaceutil.implementer(repository.ilocalrepositoryfilestorage)
 class sqlitefilestorage(object):
     """Repository file storage backed by SQLite."""
 
     def file(self, path):
         if path[0] == b'/':
             path = path[1:]
 
         if REQUIREMENT_ZSTD in self.requirements:
             compression = b'zstd'
         elif REQUIREMENT_ZLIB in self.requirements:
             compression = b'zlib'
         elif REQUIREMENT_NONE in self.requirements:
             compression = b'none'
         else:
             raise error.Abort(
                 _(
                     b'unable to determine what compression engine '
                     b'to use for SQLite storage'
                 )
             )
 
         return sqlitefilestore(self._dbconn, path, compression)
 
 
 def makefilestorage(orig, requirements, features, **kwargs):
     """Produce a type conforming to ``ilocalrepositoryfilestorage``."""
     if REQUIREMENT in requirements:
         if REQUIREMENT_SHALLOW_FILES in requirements:
             features.add(repository.REPO_FEATURE_SHALLOW_FILE_STORAGE)
 
         return sqlitefilestorage
     else:
         return orig(requirements=requirements, features=features, **kwargs)
 
 
 def makemain(orig, ui, requirements, **kwargs):
     if REQUIREMENT in requirements:
         if REQUIREMENT_ZSTD in requirements and not zstd:
             raise error.Abort(
                 _(
                     b'repository uses zstandard compression, which '
                     b'is not available to this Mercurial install'
                 )
             )
 
         return sqliterepository
 
     return orig(requirements=requirements, **kwargs)
 
 
 def verifierinit(orig, self, *args, **kwargs):
     orig(self, *args, **kwargs)
 
     # We don't care that files in the store don't align with what is
     # advertised. So suppress these warnings.
     self.warnorphanstorefiles = False
 
 
 def extsetup(ui):
     localrepo.featuresetupfuncs.add(featuresetup)
     extensions.wrapfunction(
         localrepo, b'newreporequirements', newreporequirements
     )
     extensions.wrapfunction(localrepo, b'makefilestorage', makefilestorage)
     extensions.wrapfunction(localrepo, b'makemain', makemain)
     extensions.wrapfunction(verify.verifier, b'__init__', verifierinit)
 
 
 def reposetup(ui, repo):
     if isinstance(repo, sqliterepository):
         repo._db = None
 
     # TODO check for bundlerepository?

mercurial/changegroup.py

@@ -1,1794 +1,1861 b''
 # changegroup.py - Mercurial changegroup manipulation functions
 #
 #  Copyright 2006 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import os
 import struct
 import weakref
 
 from .i18n import _
 from .node import (
     hex,
     nullid,
     nullrev,
     short,
 )
 from .pycompat import open
 
 from . import (
     error,
     match as matchmod,
     mdiff,
     phases,
     pycompat,
     requirements,
     scmutil,
     util,
 )
 
 from .interfaces import repository
 from .revlogutils import sidedata as sidedatamod
 
 _CHANGEGROUPV1_DELTA_HEADER = struct.Struct(b"20s20s20s20s")
 _CHANGEGROUPV2_DELTA_HEADER = struct.Struct(b"20s20s20s20s20s")
 _CHANGEGROUPV3_DELTA_HEADER = struct.Struct(b">20s20s20s20s20sH")
 
 LFS_REQUIREMENT = b'lfs'
 
 readexactly = util.readexactly
 
 
 def getchunk(stream):
     """return the next chunk from stream as a string"""
     d = readexactly(stream, 4)
     l = struct.unpack(b">l", d)[0]
     if l <= 4:
         if l:
             raise error.Abort(_(b"invalid chunk length %d") % l)
         return b""
     return readexactly(stream, l - 4)
 
 
 def chunkheader(length):
     """return a changegroup chunk header (string)"""
     return struct.pack(b">l", length + 4)
 
 
 def closechunk():
     """return a changegroup chunk header (string) for a zero-length chunk"""
     return struct.pack(b">l", 0)
 
 
 def _fileheader(path):
     """Obtain a changegroup chunk header for a named path."""
     return chunkheader(len(path)) + path
 
 
 def writechunks(ui, chunks, filename, vfs=None):
     """Write chunks to a file and return its filename.
 
     The stream is assumed to be a bundle file.
     Existing files will not be overwritten.
     If no filename is specified, a temporary file is created.
     """
     fh = None
     cleanup = None
     try:
         if filename:
             if vfs:
                 fh = vfs.open(filename, b"wb")
             else:
                 # Increase default buffer size because default is usually
                 # small (4k is common on Linux).
                 fh = open(filename, b"wb", 131072)
         else:
             fd, filename = pycompat.mkstemp(prefix=b"hg-bundle-", suffix=b".hg")
             fh = os.fdopen(fd, "wb")
         cleanup = filename
         for c in chunks:
             fh.write(c)
         cleanup = None
         return filename
     finally:
         if fh is not None:
             fh.close()
         if cleanup is not None:
             if filename and vfs:
                 vfs.unlink(cleanup)
             else:
                 os.unlink(cleanup)
 
 
 class cg1unpacker(object):
     """Unpacker for cg1 changegroup streams.
 
     A changegroup unpacker handles the framing of the revision data in
     the wire format. Most consumers will want to use the apply()
     method to add the changes from the changegroup to a repository.
 
     If you're forwarding a changegroup unmodified to another consumer,
     use getchunks(), which returns an iterator of changegroup
     chunks. This is mostly useful for cases where you need to know the
     data stream has ended by observing the end of the changegroup.
 
     deltachunk() is useful only if you're applying delta data. Most
     consumers should prefer apply() instead.
 
     A few other public methods exist. Those are used only for
     bundlerepo and some debug commands - their use is discouraged.
     """
 
     deltaheader = _CHANGEGROUPV1_DELTA_HEADER
     deltaheadersize = deltaheader.size
     version = b'01'
     _grouplistcount = 1  # One list of files after the manifests
 
     def __init__(self, fh, alg, extras=None):
         if alg is None:
             alg = b'UN'
         if alg not in util.compengines.supportedbundletypes:
             raise error.Abort(_(b'unknown stream compression type: %s') % alg)
         if alg == b'BZ':
             alg = b'_truncatedBZ'
 
         compengine = util.compengines.forbundletype(alg)
         self._stream = compengine.decompressorreader(fh)
         self._type = alg
         self.extras = extras or {}
         self.callback = None
 
     # These methods (compressed, read, seek, tell) all appear to only
     # be used by bundlerepo, but it's a little hard to tell.
     def compressed(self):
         return self._type is not None and self._type != b'UN'
 
     def read(self, l):
         return self._stream.read(l)
 
     def seek(self, pos):
         return self._stream.seek(pos)
 
     def tell(self):
         return self._stream.tell()
 
     def close(self):
         return self._stream.close()
 
     def _chunklength(self):
         d = readexactly(self._stream, 4)
         l = struct.unpack(b">l", d)[0]
         if l <= 4:
             if l:
                 raise error.Abort(_(b"invalid chunk length %d") % l)
             return 0
         if self.callback:
             self.callback()
         return l - 4
 
     def changelogheader(self):
         """v10 does not have a changelog header chunk"""
         return {}
 
     def manifestheader(self):
         """v10 does not have a manifest header chunk"""
         return {}
 
     def filelogheader(self):
         """return the header of the filelogs chunk, v10 only has the filename"""
         l = self._chunklength()
         if not l:
             return {}
         fname = readexactly(self._stream, l)
         return {b'filename': fname}
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, cs = headertuple
         if prevnode is None:
             deltabase = p1
         else:
             deltabase = prevnode
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
     def deltachunk(self, prevnode):
         l = self._chunklength()
         if not l:
             return {}
         headerdata = readexactly(self._stream, self.deltaheadersize)
         header = self.deltaheader.unpack(headerdata)
         delta = readexactly(self._stream, l - self.deltaheadersize)
         node, p1, p2, deltabase, cs, flags = self._deltaheader(header, prevnode)
         # cg4 forward-compat
         sidedata = {}
         return (node, p1, p2, cs, deltabase, delta, flags, sidedata)
 
     def getchunks(self):
         """returns all the chunks contains in the bundle
 
         Used when you need to forward the binary stream to a file or another
         network API. To do so, it parse the changegroup data, otherwise it will
         block in case of sshrepo because it don't know the end of the stream.
         """
         # For changegroup 1 and 2, we expect 3 parts: changelog, manifestlog,
         # and a list of filelogs. For changegroup 3, we expect 4 parts:
         # changelog, manifestlog, a list of tree manifestlogs, and a list of
         # filelogs.
         #
         # Changelog and manifestlog parts are terminated with empty chunks. The
         # tree and file parts are a list of entry sections. Each entry section
         # is a series of chunks terminating in an empty chunk. The list of these
         # entry sections is terminated in yet another empty chunk, so we know
         # we've reached the end of the tree/file list when we reach an empty
         # chunk that was proceeded by no non-empty chunks.
 
         parts = 0
         while parts < 2 + self._grouplistcount:
             noentries = True
             while True:
                 chunk = getchunk(self)
                 if not chunk:
                     # The first two empty chunks represent the end of the
                     # changelog and the manifestlog portions. The remaining
                     # empty chunks represent either A) the end of individual
                     # tree or file entries in the file list, or B) the end of
                     # the entire list. It's the end of the entire list if there
                     # were no entries (i.e. noentries is True).
                     if parts < 2:
                         parts += 1
                     elif noentries:
                         parts += 1
                     break
                 noentries = False
                 yield chunkheader(len(chunk))
                 pos = 0
                 while pos < len(chunk):
                     next = pos + 2 ** 20
                     yield chunk[pos:next]
                     pos = next
             yield closechunk()
 
     def _unpackmanifests(self, repo, revmap, trp, prog):
         self.callback = prog.increment
         # no need to check for empty manifest group here:
         # if the result of the merge of 1 and 2 is the same in 3 and 4,
         # no new manifest will be created and the manifest group will
         # be empty during the pull
         self.manifestheader()
         deltas = self.deltaiter()
         repo.manifestlog.getstorage(b'').addgroup(deltas, revmap, trp)
         prog.complete()
         self.callback = None
 
     def apply(
         self,
         repo,
         tr,
         srctype,
         url,
         targetphase=phases.draft,
         expectedtotal=None,
+        sidedata_categories=None,
     ):
         """Add the changegroup returned by source.read() to this repo.
         srctype is a string like 'push', 'pull', or 'unbundle'.  url is
         the URL of the repo where this changegroup is coming from.
 
         Return an integer summarizing the change to this repo:
         - nothing changed or no source: 0
         - more heads than before: 1+added heads (2..n)
         - fewer heads than before: -1-removed heads (-2..-n)
         - number of heads stays the same: 1
+
+        `sidedata_categories` is an optional set of the remote's sidedata wanted
+        categories.
         """
         repo = repo.unfiltered()
 
+        # Only useful if we're adding sidedata categories. If both peers have
+        # the same categories, then we simply don't do anything.
+        if self.version == b'04' and srctype == b'pull':
+            sidedata_helpers = get_sidedata_helpers(
+                repo,
+                sidedata_categories or set(),
+                pull=True,
+            )
+        else:
+            sidedata_helpers = None
+
         def csmap(x):
             repo.ui.debug(b"add changeset %s\n" % short(x))
             return len(cl)
 
         def revmap(x):
             return cl.rev(x)
 
         try:
             # The transaction may already carry source information. In this
             # case we use the top level data. We overwrite the argument
             # because we need to use the top level value (if they exist)
             # in this function.
             srctype = tr.hookargs.setdefault(b'source', srctype)
             tr.hookargs.setdefault(b'url', url)
             repo.hook(
                 b'prechangegroup', throw=True, **pycompat.strkwargs(tr.hookargs)
             )
 
             # write changelog data to temp files so concurrent readers
             # will not see an inconsistent view
             cl = repo.changelog
             cl.delayupdate(tr)
             oldheads = set(cl.heads())
 
             trp = weakref.proxy(tr)
             # pull off the changeset group
             repo.ui.status(_(b"adding changesets\n"))
             clstart = len(cl)
             progress = repo.ui.makeprogress(
                 _(b'changesets'), unit=_(b'chunks'), total=expectedtotal
             )
             self.callback = progress.increment
 
             efilesset = set()
             duprevs = []
 
             def ondupchangelog(cl, rev):
                 if rev < clstart:
                     duprevs.append(rev)
 
             def onchangelog(cl, rev):
                 ctx = cl.changelogrevision(rev)
                 efilesset.update(ctx.files)
                 repo.register_changeset(rev, ctx)
 
             self.changelogheader()
             deltas = self.deltaiter()
             if not cl.addgroup(
                 deltas,
                 csmap,
                 trp,
                 alwayscache=True,
                 addrevisioncb=onchangelog,
                 duplicaterevisioncb=ondupchangelog,
             ):
                 repo.ui.develwarn(
                     b'applied empty changelog from changegroup',
                     config=b'warn-empty-changegroup',
                 )
             efiles = len(efilesset)
             clend = len(cl)
             changesets = clend - clstart
             progress.complete()
             del deltas
             # TODO Python 2.7 removal
             # del efilesset
             efilesset = None
             self.callback = None
 
             # pull off the manifest group
             repo.ui.status(_(b"adding manifests\n"))
             # We know that we'll never have more manifests than we had
             # changesets.
             progress = repo.ui.makeprogress(
                 _(b'manifests'), unit=_(b'chunks'), total=changesets
             )
             self._unpackmanifests(repo, revmap, trp, progress)
 
             needfiles = {}
             if repo.ui.configbool(b'server', b'validate'):
                 cl = repo.changelog
                 ml = repo.manifestlog
                 # validate incoming csets have their manifests
                 for cset in pycompat.xrange(clstart, clend):
                     mfnode = cl.changelogrevision(cset).manifest
                     mfest = ml[mfnode].readdelta()
                     # store file nodes we must see
                     for f, n in pycompat.iteritems(mfest):
                         needfiles.setdefault(f, set()).add(n)
 
             # process the files
             repo.ui.status(_(b"adding file changes\n"))
             newrevs, newfiles = _addchangegroupfiles(
                 repo, self, revmap, trp, efiles, needfiles
             )
 
             # making sure the value exists
             tr.changes.setdefault(b'changegroup-count-changesets', 0)
             tr.changes.setdefault(b'changegroup-count-revisions', 0)
             tr.changes.setdefault(b'changegroup-count-files', 0)
             tr.changes.setdefault(b'changegroup-count-heads', 0)
 
             # some code use bundle operation for internal purpose. They usually
             # set `ui.quiet` to do this outside of user sight. Size the report
             # of such operation now happens at the end of the transaction, that
             # ui.quiet has not direct effect on the output.
             #
             # To preserve this intend use an inelegant hack, we fail to report
             # the change if `quiet` is set. We should probably move to
             # something better, but this is a good first step to allow the "end
             # of transaction report" to pass tests.
             if not repo.ui.quiet:
                 tr.changes[b'changegroup-count-changesets'] += changesets
                 tr.changes[b'changegroup-count-revisions'] += newrevs
                 tr.changes[b'changegroup-count-files'] += newfiles
 
             deltaheads = 0
             if oldheads:
                 heads = cl.heads()
                 deltaheads += len(heads) - len(oldheads)
                 for h in heads:
                     if h not in oldheads and repo[h].closesbranch():
                         deltaheads -= 1
 
             # see previous comment about checking ui.quiet
             if not repo.ui.quiet:
                 tr.changes[b'changegroup-count-heads'] += deltaheads
             repo.invalidatevolatilesets()
 
             if changesets > 0:
                 if b'node' not in tr.hookargs:
                     tr.hookargs[b'node'] = hex(cl.node(clstart))
                     tr.hookargs[b'node_last'] = hex(cl.node(clend - 1))
                     hookargs = dict(tr.hookargs)
                 else:
                     hookargs = dict(tr.hookargs)
                     hookargs[b'node'] = hex(cl.node(clstart))
                     hookargs[b'node_last'] = hex(cl.node(clend - 1))
                 repo.hook(
                     b'pretxnchangegroup',
                     throw=True,
                     **pycompat.strkwargs(hookargs)
                 )
 
             added = pycompat.xrange(clstart, clend)
             phaseall = None
             if srctype in (b'push', b'serve'):
                 # Old servers can not push the boundary themselves.
                 # New servers won't push the boundary if changeset already
                 # exists locally as secret
                 #
                 # We should not use added here but the list of all change in
                 # the bundle
                 if repo.publishing():
                     targetphase = phaseall = phases.public
                 else:
                     # closer target phase computation
 
                     # Those changesets have been pushed from the
                     # outside, their phases are going to be pushed
                     # alongside. Therefor `targetphase` is
                     # ignored.
                     targetphase = phaseall = phases.draft
             if added:
                 phases.registernew(repo, tr, targetphase, added)
             if phaseall is not None:
                 if duprevs:
                     duprevs.extend(added)
                 else:
                     duprevs = added
                 phases.advanceboundary(repo, tr, phaseall, [], revs=duprevs)
                 duprevs = []
 
             if changesets > 0:
 
                 def runhooks(unused_success):
                     # These hooks run when the lock releases, not when the
                     # transaction closes. So it's possible for the changelog
                     # to have changed since we last saw it.
                     if clstart >= len(repo):
                         return
 
                     repo.hook(b"changegroup", **pycompat.strkwargs(hookargs))
 
                     for rev in added:
                         args = hookargs.copy()
                         args[b'node'] = hex(cl.node(rev))
                         del args[b'node_last']
                         repo.hook(b"incoming", **pycompat.strkwargs(args))
 
                     newheads = [h for h in repo.heads() if h not in oldheads]
                     repo.ui.log(
                         b"incoming",
                         b"%d incoming changes - new heads: %s\n",
                         len(added),
                         b', '.join([hex(c[:6]) for c in newheads]),
                     )
 
                 tr.addpostclose(
                     b'changegroup-runhooks-%020i' % clstart,
                     lambda tr: repo._afterlock(runhooks),
                 )
         finally:
             repo.ui.flush()
         # never return 0 here:
         if deltaheads < 0:
             ret = deltaheads - 1
         else:
             ret = deltaheads + 1
         return ret
 
     def deltaiter(self):
         """
         returns an iterator of the deltas in this changegroup
 
         Useful for passing to the underlying storage system to be stored.
         """
         chain = None
         for chunkdata in iter(lambda: self.deltachunk(chain), {}):
             # Chunkdata: (node, p1, p2, cs, deltabase, delta, flags)
             yield chunkdata
             chain = chunkdata[0]
 
 
 class cg2unpacker(cg1unpacker):
     """Unpacker for cg2 streams.
 
     cg2 streams add support for generaldelta, so the delta header
     format is slightly different. All other features about the data
     remain the same.
     """
 
     deltaheader = _CHANGEGROUPV2_DELTA_HEADER
     deltaheadersize = deltaheader.size
     version = b'02'
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs = headertuple
         flags = 0
         return node, p1, p2, deltabase, cs, flags
 
 
 class cg3unpacker(cg2unpacker):
     """Unpacker for cg3 streams.
 
     cg3 streams add support for exchanging treemanifests and revlog
     flags. It adds the revlog flags to the delta header and an empty chunk
     separating manifests and files.
     """
 
     deltaheader = _CHANGEGROUPV3_DELTA_HEADER
     deltaheadersize = deltaheader.size
     version = b'03'
     _grouplistcount = 2  # One list of manifests and one list of files
 
     def _deltaheader(self, headertuple, prevnode):
         node, p1, p2, deltabase, cs, flags = headertuple
         return node, p1, p2, deltabase, cs, flags
 
     def _unpackmanifests(self, repo, revmap, trp, prog):
         super(cg3unpacker, self)._unpackmanifests(repo, revmap, trp, prog)
         for chunkdata in iter(self.filelogheader, {}):
             # If we get here, there are directory manifests in the changegroup
             d = chunkdata[b"filename"]
             repo.ui.debug(b"adding %s revisions\n" % d)
             deltas = self.deltaiter()
             if not repo.manifestlog.getstorage(d).addgroup(deltas, revmap, trp):
                 raise error.Abort(_(b"received dir revlog group is empty"))
 
 
 class cg4unpacker(cg3unpacker):
     """Unpacker for cg4 streams.
 
     cg4 streams add support for exchanging sidedata.
     """
 
     version = b'04'
 
     def deltachunk(self, prevnode):
         res = super(cg4unpacker, self).deltachunk(prevnode)
         if not res:
             return res
 
         (node, p1, p2, cs, deltabase, delta, flags, _sidedata) = res
 
         sidedata_raw = getchunk(self._stream)
         sidedata = {}
         if len(sidedata_raw) > 0:
             sidedata = sidedatamod.deserialize_sidedata(sidedata_raw)
 
         return node, p1, p2, cs, deltabase, delta, flags, sidedata
 
 
 class headerlessfixup(object):
     def __init__(self, fh, h):
         self._h = h
         self._fh = fh
 
     def read(self, n):
         if self._h:
             d, self._h = self._h[:n], self._h[n:]
             if len(d) < n:
                 d += readexactly(self._fh, n - len(d))
             return d
         return readexactly(self._fh, n)
 
 
 def _revisiondeltatochunks(delta, headerfn):
     """Serialize a revisiondelta to changegroup chunks."""
 
     # The captured revision delta may be encoded as a delta against
     # a base revision or as a full revision. The changegroup format
     # requires that everything on the wire be deltas. So for full
     # revisions, we need to invent a header that says to rewrite
     # data.
 
     if delta.delta is not None:
         prefix, data = b'', delta.delta
     elif delta.basenode == nullid:
         data = delta.revision
         prefix = mdiff.trivialdiffheader(len(data))
     else:
         data = delta.revision
         prefix = mdiff.replacediffheader(delta.baserevisionsize, len(data))
 
     meta = headerfn(delta)
 
     yield chunkheader(len(meta) + len(prefix) + len(data))
     yield meta
     if prefix:
         yield prefix
     yield data
 
     sidedata = delta.sidedata
     if sidedata is not None:
         # Need a separate chunk for sidedata to be able to differentiate
         # "raw delta" length and sidedata length
         yield chunkheader(len(sidedata))
         yield sidedata
 
 
 def _sortnodesellipsis(store, nodes, cl, lookup):
     """Sort nodes for changegroup generation."""
     # Ellipses serving mode.
     #
     # In a perfect world, we'd generate better ellipsis-ified graphs
     # for non-changelog revlogs. In practice, we haven't started doing
     # that yet, so the resulting DAGs for the manifestlog and filelogs
     # are actually full of bogus parentage on all the ellipsis
     # nodes. This has the side effect that, while the contents are
     # correct, the individual DAGs might be completely out of whack in
     # a case like 882681bc3166 and its ancestors (back about 10
     # revisions or so) in the main hg repo.
     #
     # The one invariant we *know* holds is that the new (potentially
     # bogus) DAG shape will be valid if we order the nodes in the
     # order that they're introduced in dramatis personae by the
     # changelog, so what we do is we sort the non-changelog histories
     # by the order in which they are used by the changelog.
     key = lambda n: cl.rev(lookup(n))
     return sorted(nodes, key=key)
 
 
 def _resolvenarrowrevisioninfo(
     cl,
     store,
     ischangelog,
     rev,
     linkrev,
     linknode,
     clrevtolocalrev,
     fullclnodes,
     precomputedellipsis,
 ):
     linkparents = precomputedellipsis[linkrev]
 
     def local(clrev):
         """Turn a changelog revnum into a local revnum.
 
         The ellipsis dag is stored as revnums on the changelog,
         but when we're producing ellipsis entries for
         non-changelog revlogs, we need to turn those numbers into
         something local. This does that for us, and during the
         changelog sending phase will also expand the stored
         mappings as needed.
         """
         if clrev == nullrev:
             return nullrev
 
         if ischangelog:
             return clrev
 
         # Walk the ellipsis-ized changelog breadth-first looking for a
         # change that has been linked from the current revlog.
         #
         # For a flat manifest revlog only a single step should be necessary
         # as all relevant changelog entries are relevant to the flat
         # manifest.
         #
         # For a filelog or tree manifest dirlog however not every changelog
         # entry will have been relevant, so we need to skip some changelog
         # nodes even after ellipsis-izing.
         walk = [clrev]
         while walk:
             p = walk[0]
             walk = walk[1:]
             if p in clrevtolocalrev:
                 return clrevtolocalrev[p]
             elif p in fullclnodes:
                 walk.extend([pp for pp in cl.parentrevs(p) if pp != nullrev])
             elif p in precomputedellipsis:
                 walk.extend(
                     [pp for pp in precomputedellipsis[p] if pp != nullrev]
                 )
             else:
                 # In this case, we've got an ellipsis with parents
                 # outside the current bundle (likely an
                 # incremental pull). We "know" that we can use the
                 # value of this same revlog at whatever revision
                 # is pointed to by linknode. "Know" is in scare
                 # quotes because I haven't done enough examination
                 # of edge cases to convince myself this is really
                 # a fact - it works for all the (admittedly
                 # thorough) cases in our testsuite, but I would be
                 # somewhat unsurprised to find a case in the wild
                 # where this breaks down a bit. That said, I don't
                 # know if it would hurt anything.
                 for i in pycompat.xrange(rev, 0, -1):
                     if store.linkrev(i) == clrev:
                         return i
                 # We failed to resolve a parent for this node, so
                 # we crash the changegroup construction.
                 raise error.Abort(
                     b"unable to resolve parent while packing '%s' %r"
                     b' for changeset %r' % (store.indexfile, rev, clrev)
                 )
 
         return nullrev
 
     if not linkparents or (store.parentrevs(rev) == (nullrev, nullrev)):
         p1, p2 = nullrev, nullrev
     elif len(linkparents) == 1:
         (p1,) = sorted(local(p) for p in linkparents)
         p2 = nullrev
     else:
         p1, p2 = sorted(local(p) for p in linkparents)
 
     p1node, p2node = store.node(p1), store.node(p2)
 
     return p1node, p2node, linknode
 
 
 def deltagroup(
     repo,
     store,
     nodes,
     ischangelog,
     lookup,
     forcedeltaparentprev,
     topic=None,
     ellipses=False,
     clrevtolocalrev=None,
     fullclnodes=None,
     precomputedellipsis=None,
+    sidedata_helpers=None,
 ):
     """Calculate deltas for a set of revisions.
 
     Is a generator of ``revisiondelta`` instances.
 
     If topic is not None, progress detail will be generated using this
     topic name (e.g. changesets, manifests, etc).
+
+    See `storageutil.emitrevisions` for the doc on `sidedata_helpers`.
     """
     if not nodes:
         return
 
     cl = repo.changelog
 
     if ischangelog:
         # `hg log` shows changesets in storage order. To preserve order
         # across clones, send out changesets in storage order.
         nodesorder = b'storage'
     elif ellipses:
         nodes = _sortnodesellipsis(store, nodes, cl, lookup)
         nodesorder = b'nodes'
     else:
         nodesorder = None
 
     # Perform ellipses filtering and revision massaging. We do this before
     # emitrevisions() because a) filtering out revisions creates less work
     # for emitrevisions() b) dropping revisions would break emitrevisions()'s
     # assumptions about delta choices and we would possibly send a delta
     # referencing a missing base revision.
     #
     # Also, calling lookup() has side-effects with regards to populating
     # data structures. If we don't call lookup() for each node or if we call
     # lookup() after the first pass through each node, things can break -
     # possibly intermittently depending on the python hash seed! For that
     # reason, we store a mapping of all linknodes during the initial node
     # pass rather than use lookup() on the output side.
     if ellipses:
         filtered = []
         adjustedparents = {}
         linknodes = {}
 
         for node in nodes:
             rev = store.rev(node)
             linknode = lookup(node)
             linkrev = cl.rev(linknode)
             clrevtolocalrev[linkrev] = rev
 
             # If linknode is in fullclnodes, it means the corresponding
             # changeset was a full changeset and is being sent unaltered.
             if linknode in fullclnodes:
                 linknodes[node] = linknode
 
             # If the corresponding changeset wasn't in the set computed
             # as relevant to us, it should be dropped outright.
             elif linkrev not in precomputedellipsis:
                 continue
 
             else:
                 # We could probably do this later and avoid the dict
                 # holding state. But it likely doesn't matter.
                 p1node, p2node, linknode = _resolvenarrowrevisioninfo(
                     cl,
                     store,
                     ischangelog,
                     rev,
                     linkrev,
                     linknode,
                     clrevtolocalrev,
                     fullclnodes,
                     precomputedellipsis,
                 )
 
                 adjustedparents[node] = (p1node, p2node)
                 linknodes[node] = linknode
 
             filtered.append(node)
 
         nodes = filtered
 
     # We expect the first pass to be fast, so we only engage the progress
     # meter for constructing the revision deltas.
     progress = None
     if topic is not None:
         progress = repo.ui.makeprogress(
             topic, unit=_(b'chunks'), total=len(nodes)
         )
 
     configtarget = repo.ui.config(b'devel', b'bundle.delta')
     if configtarget not in (b'', b'p1', b'full'):
         msg = _("""config "devel.bundle.delta" as unknown value: %s""")
         repo.ui.warn(msg % configtarget)
 
     deltamode = repository.CG_DELTAMODE_STD
     if forcedeltaparentprev:
         deltamode = repository.CG_DELTAMODE_PREV
     elif configtarget == b'p1':
         deltamode = repository.CG_DELTAMODE_P1
     elif configtarget == b'full':
         deltamode = repository.CG_DELTAMODE_FULL
 
     revisions = store.emitrevisions(
         nodes,
         nodesorder=nodesorder,
         revisiondata=True,
         assumehaveparentrevisions=not ellipses,
         deltamode=deltamode,
+        sidedata_helpers=sidedata_helpers,
     )
 
     for i, revision in enumerate(revisions):
         if progress:
             progress.update(i + 1)
 
         if ellipses:
             linknode = linknodes[revision.node]
 
             if revision.node in adjustedparents:
                 p1node, p2node = adjustedparents[revision.node]
                 revision.p1node = p1node
                 revision.p2node = p2node
                 revision.flags |= repository.REVISION_FLAG_ELLIPSIS
 
         else:
             linknode = lookup(revision.node)
 
         revision.linknode = linknode
         yield revision
 
     if progress:
         progress.complete()
 
 
 class cgpacker(object):
     def __init__(
         self,
         repo,
         oldmatcher,
         matcher,
         version,
         builddeltaheader,
         manifestsend,
         forcedeltaparentprev=False,
         bundlecaps=None,
         ellipses=False,
         shallow=False,
         ellipsisroots=None,
         fullnodes=None,
         remote_sidedata=None,
     ):
         """Given a source repo, construct a bundler.
 
         oldmatcher is a matcher that matches on files the client already has.
         These will not be included in the changegroup.
 
         matcher is a matcher that matches on files to include in the
         changegroup. Used to facilitate sparse changegroups.
 
         forcedeltaparentprev indicates whether delta parents must be against
         the previous revision in a delta group. This should only be used for
         compatibility with changegroup version 1.
 
         builddeltaheader is a callable that constructs the header for a group
         delta.
 
         manifestsend is a chunk to send after manifests have been fully emitted.
 
         ellipses indicates whether ellipsis serving mode is enabled.
 
         bundlecaps is optional and can be used to specify the set of
         capabilities which can be used to build the bundle. While bundlecaps is
         unused in core Mercurial, extensions rely on this feature to communicate
         capabilities to customize the changegroup packer.
 
         shallow indicates whether shallow data might be sent. The packer may
         need to pack file contents not introduced by the changes being packed.
 
         fullnodes is the set of changelog nodes which should not be ellipsis
         nodes. We store this rather than the set of nodes that should be
         ellipsis because for very large histories we expect this to be
         significantly smaller.
 
         remote_sidedata is the set of sidedata categories wanted by the remote.
         """
         assert oldmatcher
         assert matcher
         self._oldmatcher = oldmatcher
         self._matcher = matcher
 
         self.version = version
         self._forcedeltaparentprev = forcedeltaparentprev
         self._builddeltaheader = builddeltaheader
         self._manifestsend = manifestsend
         self._ellipses = ellipses
 
         # Set of capabilities we can use to build the bundle.
         if bundlecaps is None:
             bundlecaps = set()
         self._bundlecaps = bundlecaps
         if remote_sidedata is None:
             remote_sidedata = set()
         self._remote_sidedata = remote_sidedata
         self._isshallow = shallow
         self._fullclnodes = fullnodes
 
         # Maps ellipsis revs to their roots at the changelog level.
         self._precomputedellipsis = ellipsisroots
 
         self._repo = repo
 
         if self._repo.ui.verbose and not self._repo.ui.debugflag:
             self._verbosenote = self._repo.ui.note
         else:
             self._verbosenote = lambda s: None
 
     def generate(
         self, commonrevs, clnodes, fastpathlinkrev, source, changelog=True
     ):
         """Yield a sequence of changegroup byte chunks.
         If changelog is False, changelog data won't be added to changegroup
         """
 
         repo = self._repo
         cl = repo.changelog
 
         self._verbosenote(_(b'uncompressed size of bundle content:\n'))
         size = 0
 
+        sidedata_helpers = None
+        if self.version == b'04':
+            remote_sidedata = self._remote_sidedata
+            if source == b'strip':
+                # We're our own remote when stripping, get the no-op helpers
+                # TODO a better approach would be for the strip bundle to
+                # correctly advertise its sidedata categories directly.
+                remote_sidedata = repo._wanted_sidedata
+            sidedata_helpers = get_sidedata_helpers(repo, remote_sidedata)
+
         clstate, deltas = self._generatechangelog(
-            cl, clnodes, generate=changelog
+            cl,
+            clnodes,
+            generate=changelog,
+            sidedata_helpers=sidedata_helpers,
         )
         for delta in deltas:
             for chunk in _revisiondeltatochunks(delta, self._builddeltaheader):
                 size += len(chunk)
                 yield chunk
 
         close = closechunk()
         size += len(close)
         yield closechunk()
 
         self._verbosenote(_(b'%8.i (changelog)\n') % size)
 
         clrevorder = clstate[b'clrevorder']
         manifests = clstate[b'manifests']
         changedfiles = clstate[b'changedfiles']
 
         # We need to make sure that the linkrev in the changegroup refers to
         # the first changeset that introduced the manifest or file revision.
         # The fastpath is usually safer than the slowpath, because the filelogs
         # are walked in revlog order.
         #
         # When taking the slowpath when the manifest revlog uses generaldelta,
         # the manifest may be walked in the "wrong" order. Without 'clrevorder',
         # we would get an incorrect linkrev (see fix in cc0ff93d0c0c).
         #
         # When taking the fastpath, we are only vulnerable to reordering
         # of the changelog itself. The changelog never uses generaldelta and is
         # never reordered. To handle this case, we simply take the slowpath,
         # which already has the 'clrevorder' logic. This was also fixed in
         # cc0ff93d0c0c.
 
         # Treemanifests don't work correctly with fastpathlinkrev
         # either, because we don't discover which directory nodes to
         # send along with files. This could probably be fixed.
         fastpathlinkrev = fastpathlinkrev and not scmutil.istreemanifest(repo)
 
         fnodes = {}  # needed file nodes
 
         size = 0
         it = self.generatemanifests(
             commonrevs,
             clrevorder,
             fastpathlinkrev,
             manifests,
             fnodes,
             source,
             clstate[b'clrevtomanifestrev'],
+            sidedata_helpers=sidedata_helpers,
         )
 
         for tree, deltas in it:
             if tree:
                 assert self.version in (b'03', b'04')
                 chunk = _fileheader(tree)
                 size += len(chunk)
                 yield chunk
 
             for delta in deltas:
                 chunks = _revisiondeltatochunks(delta, self._builddeltaheader)
                 for chunk in chunks:
                     size += len(chunk)
                     yield chunk
 
             close = closechunk()
             size += len(close)
             yield close
 
         self._verbosenote(_(b'%8.i (manifests)\n') % size)
         yield self._manifestsend
 
         mfdicts = None
         if self._ellipses and self._isshallow:
             mfdicts = [
                 (repo.manifestlog[n].read(), lr)
                 for (n, lr) in pycompat.iteritems(manifests)
             ]
 
         manifests.clear()
         clrevs = {cl.rev(x) for x in clnodes}
 
         it = self.generatefiles(
             changedfiles,
             commonrevs,
             source,
             mfdicts,
             fastpathlinkrev,
             fnodes,
             clrevs,
+            sidedata_helpers=sidedata_helpers,
         )
 
         for path, deltas in it:
             h = _fileheader(path)
             size = len(h)
             yield h
 
             for delta in deltas:
                 chunks = _revisiondeltatochunks(delta, self._builddeltaheader)
                 for chunk in chunks:
                     size += len(chunk)
                     yield chunk
 
             close = closechunk()
             size += len(close)
             yield close
 
             self._verbosenote(_(b'%8.i  %s\n') % (size, path))
 
         yield closechunk()
 
         if clnodes:
             repo.hook(b'outgoing', node=hex(clnodes[0]), source=source)
 
-    def _generatechangelog(self, cl, nodes, generate=True):
+    def _generatechangelog(
+        self, cl, nodes, generate=True, sidedata_helpers=None
+    ):
         """Generate data for changelog chunks.
 
         Returns a 2-tuple of a dict containing state and an iterable of
         byte chunks. The state will not be fully populated until the
         chunk stream has been fully consumed.
 
         if generate is False, the state will be fully populated and no chunk
         stream will be yielded
+
+        See `storageutil.emitrevisions` for the doc on `sidedata_helpers`.
         """
         clrevorder = {}
         manifests = {}
         mfl = self._repo.manifestlog
         changedfiles = set()
         clrevtomanifestrev = {}
 
         state = {
             b'clrevorder': clrevorder,
             b'manifests': manifests,
             b'changedfiles': changedfiles,
             b'clrevtomanifestrev': clrevtomanifestrev,
         }
 
         if not (generate or self._ellipses):
             # sort the nodes in storage order
             nodes = sorted(nodes, key=cl.rev)
             for node in nodes:
                 c = cl.changelogrevision(node)
                 clrevorder[node] = len(clrevorder)
                 # record the first changeset introducing this manifest version
                 manifests.setdefault(c.manifest, node)
                 # Record a complete list of potentially-changed files in
                 # this manifest.
                 changedfiles.update(c.files)
 
             return state, ()
 
         # Callback for the changelog, used to collect changed files and
         # manifest nodes.
         # Returns the linkrev node (identity in the changelog case).
         def lookupcl(x):
             c = cl.changelogrevision(x)
             clrevorder[x] = len(clrevorder)
 
             if self._ellipses:
                 # Only update manifests if x is going to be sent. Otherwise we
                 # end up with bogus linkrevs specified for manifests and
                 # we skip some manifest nodes that we should otherwise
                 # have sent.
                 if (
                     x in self._fullclnodes
                     or cl.rev(x) in self._precomputedellipsis
                 ):
 
                     manifestnode = c.manifest
                     # Record the first changeset introducing this manifest
                     # version.
                     manifests.setdefault(manifestnode, x)
                     # Set this narrow-specific dict so we have the lowest
                     # manifest revnum to look up for this cl revnum. (Part of
                     # mapping changelog ellipsis parents to manifest ellipsis
                     # parents)
                     clrevtomanifestrev.setdefault(
                         cl.rev(x), mfl.rev(manifestnode)
                     )
                 # We can't trust the changed files list in the changeset if the
                 # client requested a shallow clone.
                 if self._isshallow:
                     changedfiles.update(mfl[c.manifest].read().keys())
                 else:
                     changedfiles.update(c.files)
             else:
                 # record the first changeset introducing this manifest version
                 manifests.setdefault(c.manifest, x)
                 # Record a complete list of potentially-changed files in
                 # this manifest.
                 changedfiles.update(c.files)
 
             return x
 
         gen = deltagroup(
             self._repo,
             cl,
             nodes,
             True,
             lookupcl,
             self._forcedeltaparentprev,
             ellipses=self._ellipses,
             topic=_(b'changesets'),
             clrevtolocalrev={},
             fullclnodes=self._fullclnodes,
             precomputedellipsis=self._precomputedellipsis,
+            sidedata_helpers=sidedata_helpers,
         )
 
         return state, gen
 
     def generatemanifests(
         self,
         commonrevs,
         clrevorder,
         fastpathlinkrev,
         manifests,
         fnodes,
         source,
         clrevtolocalrev,
+        sidedata_helpers=None,
     ):
         """Returns an iterator of changegroup chunks containing manifests.
 
         `source` is unused here, but is used by extensions like remotefilelog to
         change what is sent based in pulls vs pushes, etc.
+
+        See `storageutil.emitrevisions` for the doc on `sidedata_helpers`.
         """
         repo = self._repo
         mfl = repo.manifestlog
         tmfnodes = {b'': manifests}
 
         # Callback for the manifest, used to collect linkrevs for filelog
         # revisions.
         # Returns the linkrev node (collected in lookupcl).
         def makelookupmflinknode(tree, nodes):
             if fastpathlinkrev:
                 assert not tree
                 return (
                     manifests.__getitem__
                 )  # pytype: disable=unsupported-operands
 
             def lookupmflinknode(x):
                 """Callback for looking up the linknode for manifests.
 
                 Returns the linkrev node for the specified manifest.
 
                 SIDE EFFECT:
 
                 1) fclnodes gets populated with the list of relevant
                    file nodes if we're not using fastpathlinkrev
                 2) When treemanifests are in use, collects treemanifest nodes
                    to send
 
                 Note that this means manifests must be completely sent to
                 the client before you can trust the list of files and
                 treemanifests to send.
                 """
                 clnode = nodes[x]
                 mdata = mfl.get(tree, x).readfast(shallow=True)
                 for p, n, fl in mdata.iterentries():
                     if fl == b't':  # subdirectory manifest
                         subtree = tree + p + b'/'
                         tmfclnodes = tmfnodes.setdefault(subtree, {})
                         tmfclnode = tmfclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[tmfclnode]:
                             tmfclnodes[n] = clnode
                     else:
                         f = tree + p
                         fclnodes = fnodes.setdefault(f, {})
                         fclnode = fclnodes.setdefault(n, clnode)
                         if clrevorder[clnode] < clrevorder[fclnode]:
                             fclnodes[n] = clnode
                 return clnode
 
             return lookupmflinknode
 
         while tmfnodes:
             tree, nodes = tmfnodes.popitem()
 
             should_visit = self._matcher.visitdir(tree[:-1])
             if tree and not should_visit:
                 continue
 
             store = mfl.getstorage(tree)
 
             if not should_visit:
                 # No nodes to send because this directory is out of
                 # the client's view of the repository (probably
                 # because of narrow clones). Do this even for the root
                 # directory (tree=='')
                 prunednodes = []
             else:
                 # Avoid sending any manifest nodes we can prove the
                 # client already has by checking linkrevs. See the
                 # related comment in generatefiles().
                 prunednodes = self._prunemanifests(store, nodes, commonrevs)
 
             if tree and not prunednodes:
                 continue
 
             lookupfn = makelookupmflinknode(tree, nodes)
 
             deltas = deltagroup(
                 self._repo,
                 store,
                 prunednodes,
                 False,
                 lookupfn,
                 self._forcedeltaparentprev,
                 ellipses=self._ellipses,
                 topic=_(b'manifests'),
                 clrevtolocalrev=clrevtolocalrev,
                 fullclnodes=self._fullclnodes,
                 precomputedellipsis=self._precomputedellipsis,
+                sidedata_helpers=sidedata_helpers,
             )
 
             if not self._oldmatcher.visitdir(store.tree[:-1]):
                 yield tree, deltas
             else:
                 # 'deltas' is a generator and we need to consume it even if
                 # we are not going to send it because a side-effect is that
                 # it updates tmdnodes (via lookupfn)
                 for d in deltas:
                     pass
                 if not tree:
                     yield tree, []
 
     def _prunemanifests(self, store, nodes, commonrevs):
         if not self._ellipses:
             # In non-ellipses case and large repositories, it is better to
             # prevent calling of store.rev and store.linkrev on a lot of
             # nodes as compared to sending some extra data
             return nodes.copy()
         # This is split out as a separate method to allow filtering
         # commonrevs in extension code.
         #
         # TODO(augie): this shouldn't be required, instead we should
         # make filtering of revisions to send delegated to the store
         # layer.
         frev, flr = store.rev, store.linkrev
         return [n for n in nodes if flr(frev(n)) not in commonrevs]
 
     # The 'source' parameter is useful for extensions
     def generatefiles(
         self,
         changedfiles,
         commonrevs,
         source,
         mfdicts,
         fastpathlinkrev,
         fnodes,
         clrevs,
+        sidedata_helpers=None,
     ):
         changedfiles = [
             f
             for f in changedfiles
             if self._matcher(f) and not self._oldmatcher(f)
         ]
 
         if not fastpathlinkrev:
 
             def normallinknodes(unused, fname):
                 return fnodes.get(fname, {})
 
         else:
             cln = self._repo.changelog.node
 
             def normallinknodes(store, fname):
                 flinkrev = store.linkrev
                 fnode = store.node
                 revs = ((r, flinkrev(r)) for r in store)
                 return {fnode(r): cln(lr) for r, lr in revs if lr in clrevs}
 
         clrevtolocalrev = {}
 
         if self._isshallow:
             # In a shallow clone, the linknodes callback needs to also include
             # those file nodes that are in the manifests we sent but weren't
             # introduced by those manifests.
             commonctxs = [self._repo[c] for c in commonrevs]
             clrev = self._repo.changelog.rev
 
             def linknodes(flog, fname):
                 for c in commonctxs:
                     try:
                         fnode = c.filenode(fname)
                         clrevtolocalrev[c.rev()] = flog.rev(fnode)
                     except error.ManifestLookupError:
                         pass
                 links = normallinknodes(flog, fname)
                 if len(links) != len(mfdicts):
                     for mf, lr in mfdicts:
                         fnode = mf.get(fname, None)
                         if fnode in links:
                             links[fnode] = min(links[fnode], lr, key=clrev)
                         elif fnode:
                             links[fnode] = lr
                 return links
 
         else:
             linknodes = normallinknodes
 
         repo = self._repo
         progress = repo.ui.makeprogress(
             _(b'files'), unit=_(b'files'), total=len(changedfiles)
         )
         for i, fname in enumerate(sorted(changedfiles)):
             filerevlog = repo.file(fname)
             if not filerevlog:
                 raise error.Abort(
                     _(b"empty or missing file data for %s") % fname
                 )
 
             clrevtolocalrev.clear()
 
             linkrevnodes = linknodes(filerevlog, fname)
             # Lookup for filenodes, we collected the linkrev nodes above in the
             # fastpath case and with lookupmf in the slowpath case.
             def lookupfilelog(x):
                 return linkrevnodes[x]
 
             frev, flr = filerevlog.rev, filerevlog.linkrev
             # Skip sending any filenode we know the client already
             # has. This avoids over-sending files relatively
             # inexpensively, so it's not a problem if we under-filter
             # here.
             filenodes = [
                 n for n in linkrevnodes if flr(frev(n)) not in commonrevs
             ]
 
             if not filenodes:
                 continue
 
             progress.update(i + 1, item=fname)
 
             deltas = deltagroup(
                 self._repo,
                 filerevlog,
                 filenodes,
                 False,
                 lookupfilelog,
                 self._forcedeltaparentprev,
                 ellipses=self._ellipses,
                 clrevtolocalrev=clrevtolocalrev,
                 fullclnodes=self._fullclnodes,
                 precomputedellipsis=self._precomputedellipsis,
+                sidedata_helpers=sidedata_helpers,
             )
 
             yield fname, deltas
 
         progress.complete()
 
 
 def _makecg1packer(
     repo,
     oldmatcher,
     matcher,
     bundlecaps,
     ellipses=False,
     shallow=False,
     ellipsisroots=None,
     fullnodes=None,
     remote_sidedata=None,
 ):
     builddeltaheader = lambda d: _CHANGEGROUPV1_DELTA_HEADER.pack(
         d.node, d.p1node, d.p2node, d.linknode
     )
 
     return cgpacker(
         repo,
         oldmatcher,
         matcher,
         b'01',
         builddeltaheader=builddeltaheader,
         manifestsend=b'',
         forcedeltaparentprev=True,
         bundlecaps=bundlecaps,
         ellipses=ellipses,
         shallow=shallow,
         ellipsisroots=ellipsisroots,
         fullnodes=fullnodes,
     )
 
 
 def _makecg2packer(
     repo,
     oldmatcher,
     matcher,
     bundlecaps,
     ellipses=False,
     shallow=False,
     ellipsisroots=None,
     fullnodes=None,
     remote_sidedata=None,
 ):
     builddeltaheader = lambda d: _CHANGEGROUPV2_DELTA_HEADER.pack(
         d.node, d.p1node, d.p2node, d.basenode, d.linknode
     )
 
     return cgpacker(
         repo,
         oldmatcher,
         matcher,
         b'02',
         builddeltaheader=builddeltaheader,
         manifestsend=b'',
         bundlecaps=bundlecaps,
         ellipses=ellipses,
         shallow=shallow,
         ellipsisroots=ellipsisroots,
         fullnodes=fullnodes,
     )
 
 
 def _makecg3packer(
     repo,
     oldmatcher,
     matcher,
     bundlecaps,
     ellipses=False,
     shallow=False,
     ellipsisroots=None,
     fullnodes=None,
     remote_sidedata=None,
 ):
     builddeltaheader = lambda d: _CHANGEGROUPV3_DELTA_HEADER.pack(
         d.node, d.p1node, d.p2node, d.basenode, d.linknode, d.flags
     )
 
     return cgpacker(
         repo,
         oldmatcher,
         matcher,
         b'03',
         builddeltaheader=builddeltaheader,
         manifestsend=closechunk(),
         bundlecaps=bundlecaps,
         ellipses=ellipses,
         shallow=shallow,
         ellipsisroots=ellipsisroots,
         fullnodes=fullnodes,
     )
 
 
 def _makecg4packer(
     repo,
     oldmatcher,
     matcher,
     bundlecaps,
     ellipses=False,
     shallow=False,
     ellipsisroots=None,
     fullnodes=None,
     remote_sidedata=None,
 ):
     # Same header func as cg3. Sidedata is in a separate chunk from the delta to
     # differenciate "raw delta" and sidedata.
     builddeltaheader = lambda d: _CHANGEGROUPV3_DELTA_HEADER.pack(
         d.node, d.p1node, d.p2node, d.basenode, d.linknode, d.flags
     )
 
     return cgpacker(
         repo,
         oldmatcher,
         matcher,
         b'04',
         builddeltaheader=builddeltaheader,
         manifestsend=closechunk(),
         bundlecaps=bundlecaps,
         ellipses=ellipses,
         shallow=shallow,
         ellipsisroots=ellipsisroots,
         fullnodes=fullnodes,
         remote_sidedata=remote_sidedata,
     )
 
 
 _packermap = {
     b'01': (_makecg1packer, cg1unpacker),
     # cg2 adds support for exchanging generaldelta
     b'02': (_makecg2packer, cg2unpacker),
     # cg3 adds support for exchanging revlog flags and treemanifests
     b'03': (_makecg3packer, cg3unpacker),
     # ch4 adds support for exchanging sidedata
     b'04': (_makecg4packer, cg4unpacker),
 }
 
 
 def allsupportedversions(repo):
     versions = set(_packermap.keys())
     needv03 = False
     if (
         repo.ui.configbool(b'experimental', b'changegroup3')
         or repo.ui.configbool(b'experimental', b'treemanifest')
         or scmutil.istreemanifest(repo)
     ):
         # we keep version 03 because we need to to exchange treemanifest data
         #
         # we also keep vresion 01 and 02, because it is possible for repo to
         # contains both normal and tree manifest at the same time. so using
         # older version to pull data is viable
         #
         # (or even to push subset of history)
         needv03 = True
     has_revlogv2 = requirements.REVLOGV2_REQUIREMENT in repo.requirements
     if not has_revlogv2:
         versions.discard(b'04')
     if not needv03:
         versions.discard(b'03')
     return versions
 
 
 # Changegroup versions that can be applied to the repo
 def supportedincomingversions(repo):
     return allsupportedversions(repo)
 
 
 # Changegroup versions that can be created from the repo
 def supportedoutgoingversions(repo):
     versions = allsupportedversions(repo)
     if scmutil.istreemanifest(repo):
         # Versions 01 and 02 support only flat manifests and it's just too
         # expensive to convert between the flat manifest and tree manifest on
         # the fly. Since tree manifests are hashed differently, all of history
         # would have to be converted. Instead, we simply don't even pretend to
         # support versions 01 and 02.
         versions.discard(b'01')
         versions.discard(b'02')
     if requirements.NARROW_REQUIREMENT in repo.requirements:
         # Versions 01 and 02 don't support revlog flags, and we need to
         # support that for stripping and unbundling to work.
         versions.discard(b'01')
         versions.discard(b'02')
     if LFS_REQUIREMENT in repo.requirements:
         # Versions 01 and 02 don't support revlog flags, and we need to
         # mark LFS entries with REVIDX_EXTSTORED.
         versions.discard(b'01')
         versions.discard(b'02')
 
     return versions
 
 
 def localversion(repo):
     # Finds the best version to use for bundles that are meant to be used
     # locally, such as those from strip and shelve, and temporary bundles.
     return max(supportedoutgoingversions(repo))
 
 
 def safeversion(repo):
     # Finds the smallest version that it's safe to assume clients of the repo
     # will support. For example, all hg versions that support generaldelta also
     # support changegroup 02.
     versions = supportedoutgoingversions(repo)
     if requirements.GENERALDELTA_REQUIREMENT in repo.requirements:
         versions.discard(b'01')
     assert versions
     return min(versions)
 
 
 def getbundler(
     version,
     repo,
     bundlecaps=None,
     oldmatcher=None,
     matcher=None,
     ellipses=False,
     shallow=False,
     ellipsisroots=None,
     fullnodes=None,
     remote_sidedata=None,
 ):
     assert version in supportedoutgoingversions(repo)
 
     if matcher is None:
         matcher = matchmod.always()
     if oldmatcher is None:
         oldmatcher = matchmod.never()
 
     if version == b'01' and not matcher.always():
         raise error.ProgrammingError(
             b'version 01 changegroups do not support sparse file matchers'
         )
 
     if ellipses and version in (b'01', b'02'):
         raise error.Abort(
             _(
                 b'ellipsis nodes require at least cg3 on client and server, '
                 b'but negotiated version %s'
             )
             % version
         )
 
     # Requested files could include files not in the local store. So
     # filter those out.
     matcher = repo.narrowmatch(matcher)
 
     fn = _packermap[version][0]
     return fn(
         repo,
         oldmatcher,
         matcher,
         bundlecaps,
         ellipses=ellipses,
         shallow=shallow,
         ellipsisroots=ellipsisroots,
         fullnodes=fullnodes,
         remote_sidedata=remote_sidedata,
     )
 
 
 def getunbundler(version, fh, alg, extras=None):
     return _packermap[version][1](fh, alg, extras=extras)
 
 
 def _changegroupinfo(repo, nodes, source):
     if repo.ui.verbose or source == b'bundle':
         repo.ui.status(_(b"%d changesets found\n") % len(nodes))
     if repo.ui.debugflag:
         repo.ui.debug(b"list of changesets:\n")
         for node in nodes:
             repo.ui.debug(b"%s\n" % hex(node))
 
 
 def makechangegroup(
     repo, outgoing, version, source, fastpath=False, bundlecaps=None
 ):
     cgstream = makestream(
         repo,
         outgoing,
         version,
         source,
         fastpath=fastpath,
         bundlecaps=bundlecaps,
     )
     return getunbundler(
         version,
         util.chunkbuffer(cgstream),
         None,
         {b'clcount': len(outgoing.missing)},
     )
 
 
 def makestream(
     repo,
     outgoing,
     version,
     source,
     fastpath=False,
     bundlecaps=None,
     matcher=None,
     remote_sidedata=None,
 ):
     bundler = getbundler(
         version,
         repo,
         bundlecaps=bundlecaps,
         matcher=matcher,
         remote_sidedata=remote_sidedata,
     )
 
     repo = repo.unfiltered()
     commonrevs = outgoing.common
     csets = outgoing.missing
     heads = outgoing.ancestorsof
     # We go through the fast path if we get told to, or if all (unfiltered
     # heads have been requested (since we then know there all linkrevs will
     # be pulled by the client).
     heads.sort()
     fastpathlinkrev = fastpath or (
         repo.filtername is None and heads == sorted(repo.heads())
     )
 
     repo.hook(b'preoutgoing', throw=True, source=source)
     _changegroupinfo(repo, csets, source)
     return bundler.generate(commonrevs, csets, fastpathlinkrev, source)
 
 
 def _addchangegroupfiles(repo, source, revmap, trp, expectedfiles, needfiles):
     revisions = 0
     files = 0
     progress = repo.ui.makeprogress(
         _(b'files'), unit=_(b'files'), total=expectedfiles
     )
     for chunkdata in iter(source.filelogheader, {}):
         files += 1
         f = chunkdata[b"filename"]
         repo.ui.debug(b"adding %s revisions\n" % f)
         progress.increment()
         fl = repo.file(f)
         o = len(fl)
         try:
             deltas = source.deltaiter()
             if not fl.addgroup(deltas, revmap, trp):
                 raise error.Abort(_(b"received file revlog group is empty"))
         except error.CensoredBaseError as e:
             raise error.Abort(_(b"received delta base is censored: %s") % e)
         revisions += len(fl) - o
         if f in needfiles:
             needs = needfiles[f]
             for new in pycompat.xrange(o, len(fl)):
                 n = fl.node(new)
                 if n in needs:
                     needs.remove(n)
                 else:
                     raise error.Abort(_(b"received spurious file revlog entry"))
             if not needs:
                 del needfiles[f]
     progress.complete()
 
     for f, needs in pycompat.iteritems(needfiles):
         fl = repo.file(f)
         for n in needs:
             try:
                 fl.rev(n)
             except error.LookupError:
                 raise error.Abort(
                     _(b'missing file data for %s:%s - run hg verify')
                     % (f, hex(n))
                 )
 
     return revisions, files
+
+
+def get_sidedata_helpers(repo, remote_sd_categories, pull=False):
+    # Computers for computing sidedata on-the-fly
+    sd_computers = collections.defaultdict(list)
+    # Computers for categories to remove from sidedata
+    sd_removers = collections.defaultdict(list)
+
+    to_generate = remote_sd_categories - repo._wanted_sidedata
+    to_remove = repo._wanted_sidedata - remote_sd_categories
+    if pull:
+        to_generate, to_remove = to_remove, to_generate
+
+    for revlog_kind, computers in repo._sidedata_computers.items():
+        for category, computer in computers.items():
+            if category in to_generate:
+                sd_computers[revlog_kind].append(computer)
+            if category in to_remove:
+                sd_removers[revlog_kind].append(computer)
+
+    sidedata_helpers = (repo, sd_computers, sd_removers)
+    return sidedata_helpers

mercurial/filelog.py

@@ -1,292 +1,294 b''
 # filelog.py - file history class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 from .i18n import _
 from .node import (
     nullid,
     nullrev,
 )
 from . import (
     error,
     revlog,
 )
 from .interfaces import (
     repository,
     util as interfaceutil,
 )
 from .utils import storageutil
 
 
 @interfaceutil.implementer(repository.ifilestorage)
 class filelog(object):
     def __init__(self, opener, path):
         self._revlog = revlog.revlog(
             opener, b'/'.join((b'data', path + b'.i')), censorable=True
         )
         # Full name of the user visible file, relative to the repository root.
         # Used by LFS.
         self._revlog.filename = path
         self._revlog.revlog_kind = b'filelog'
 
     def __len__(self):
         return len(self._revlog)
 
     def __iter__(self):
         return self._revlog.__iter__()
 
     def hasnode(self, node):
         if node in (nullid, nullrev):
             return False
 
         try:
             self._revlog.rev(node)
             return True
         except (TypeError, ValueError, IndexError, error.LookupError):
             return False
 
     def revs(self, start=0, stop=None):
         return self._revlog.revs(start=start, stop=stop)
 
     def parents(self, node):
         return self._revlog.parents(node)
 
     def parentrevs(self, rev):
         return self._revlog.parentrevs(rev)
 
     def rev(self, node):
         return self._revlog.rev(node)
 
     def node(self, rev):
         return self._revlog.node(rev)
 
     def lookup(self, node):
         return storageutil.fileidlookup(
             self._revlog, node, self._revlog.indexfile
         )
 
     def linkrev(self, rev):
         return self._revlog.linkrev(rev)
 
     def commonancestorsheads(self, node1, node2):
         return self._revlog.commonancestorsheads(node1, node2)
 
     # Used by dagop.blockdescendants().
     def descendants(self, revs):
         return self._revlog.descendants(revs)
 
     def heads(self, start=None, stop=None):
         return self._revlog.heads(start, stop)
 
     # Used by hgweb, children extension.
     def children(self, node):
         return self._revlog.children(node)
 
     def iscensored(self, rev):
         return self._revlog.iscensored(rev)
 
     def revision(self, node, _df=None, raw=False):
         return self._revlog.revision(node, _df=_df, raw=raw)
 
     def rawdata(self, node, _df=None):
         return self._revlog.rawdata(node, _df=_df)
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
+        sidedata_helpers=None,
     ):
         return self._revlog.emitrevisions(
             nodes,
             nodesorder=nodesorder,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
             deltamode=deltamode,
+            sidedata_helpers=sidedata_helpers,
         )
 
     def addrevision(
         self,
         revisiondata,
         transaction,
         linkrev,
         p1,
         p2,
         node=None,
         flags=revlog.REVIDX_DEFAULT_FLAGS,
         cachedelta=None,
     ):
         return self._revlog.addrevision(
             revisiondata,
             transaction,
             linkrev,
             p1,
             p2,
             node=node,
             flags=flags,
             cachedelta=cachedelta,
         )
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         addrevisioncb=None,
         duplicaterevisioncb=None,
         maybemissingparents=False,
     ):
         if maybemissingparents:
             raise error.Abort(
                 _(
                     b'revlog storage does not support missing '
                     b'parents write mode'
                 )
             )
 
         return self._revlog.addgroup(
             deltas,
             linkmapper,
             transaction,
             addrevisioncb=addrevisioncb,
             duplicaterevisioncb=duplicaterevisioncb,
         )
 
     def getstrippoint(self, minlink):
         return self._revlog.getstrippoint(minlink)
 
     def strip(self, minlink, transaction):
         return self._revlog.strip(minlink, transaction)
 
     def censorrevision(self, tr, node, tombstone=b''):
         return self._revlog.censorrevision(tr, node, tombstone=tombstone)
 
     def files(self):
         return self._revlog.files()
 
     def read(self, node):
         return storageutil.filtermetadata(self.revision(node))
 
     def add(self, text, meta, transaction, link, p1=None, p2=None):
         if meta or text.startswith(b'\1\n'):
             text = storageutil.packmeta(meta, text)
         rev = self.addrevision(text, transaction, link, p1, p2)
         return self.node(rev)
 
     def renamed(self, node):
         return storageutil.filerevisioncopied(self, node)
 
     def size(self, rev):
         """return the size of a given revision"""
 
         # for revisions with renames, we have to go the slow way
         node = self.node(rev)
         if self.renamed(node):
             return len(self.read(node))
         if self.iscensored(rev):
             return 0
 
         # XXX if self.read(node).startswith("\1\n"), this returns (size+4)
         return self._revlog.size(rev)
 
     def cmp(self, node, text):
         """compare text with a given file revision
 
         returns True if text is different than what is stored.
         """
         return not storageutil.filedataequivalent(self, node, text)
 
     def verifyintegrity(self, state):
         return self._revlog.verifyintegrity(state)
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         return self._revlog.storageinfo(
             exclusivefiles=exclusivefiles,
             sharedfiles=sharedfiles,
             revisionscount=revisionscount,
             trackedsize=trackedsize,
             storedsize=storedsize,
         )
 
     # TODO these aren't part of the interface and aren't internal methods.
     # Callers should be fixed to not use them.
 
     # Used by bundlefilelog, unionfilelog.
     @property
     def indexfile(self):
         return self._revlog.indexfile
 
     @indexfile.setter
     def indexfile(self, value):
         self._revlog.indexfile = value
 
     # Used by repo upgrade.
     def clone(self, tr, destrevlog, **kwargs):
         if not isinstance(destrevlog, filelog):
             raise error.ProgrammingError(b'expected filelog to clone()')
 
         return self._revlog.clone(tr, destrevlog._revlog, **kwargs)
 
 
 class narrowfilelog(filelog):
     """Filelog variation to be used with narrow stores."""
 
     def __init__(self, opener, path, narrowmatch):
         super(narrowfilelog, self).__init__(opener, path)
         self._narrowmatch = narrowmatch
 
     def renamed(self, node):
         res = super(narrowfilelog, self).renamed(node)
 
         # Renames that come from outside the narrowspec are problematic
         # because we may lack the base text for the rename. This can result
         # in code attempting to walk the ancestry or compute a diff
         # encountering a missing revision. We address this by silently
         # removing rename metadata if the source file is outside the
         # narrow spec.
         #
         # A better solution would be to see if the base revision is available,
         # rather than assuming it isn't.
         #
         # An even better solution would be to teach all consumers of rename
         # metadata that the base revision may not be available.
         #
         # TODO consider better ways of doing this.
         if res and not self._narrowmatch(res[0]):
             return None
 
         return res
 
     def size(self, rev):
         # Because we have a custom renamed() that may lie, we need to call
         # the base renamed() to report accurate results.
         node = self.node(rev)
         if super(narrowfilelog, self).renamed(node):
             return len(self.read(node))
         else:
             return super(narrowfilelog, self).size(rev)
 
     def cmp(self, node, text):
         # We don't call `super` because narrow parents can be buggy in case of a
         # ambiguous dirstate. Always take the slow path until there is a better
         # fix, see issue6150.
 
         # Censored files compare against the empty file.
         if self.iscensored(self.rev(node)):
             return text != b''
 
         return self.read(node) != text

mercurial/manifest.py

@@ -1,2355 +1,2357 b''
 # manifest.py - manifest revision class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import heapq
 import itertools
 import struct
 import weakref
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
 )
 from .pycompat import getattr
 from . import (
     encoding,
     error,
     match as matchmod,
     mdiff,
     pathutil,
     policy,
     pycompat,
     revlog,
     util,
 )
 from .interfaces import (
     repository,
     util as interfaceutil,
 )
 
 parsers = policy.importmod('parsers')
 propertycache = util.propertycache
 
 # Allow tests to more easily test the alternate path in manifestdict.fastdelta()
 FASTDELTA_TEXTDIFF_THRESHOLD = 1000
 
 
 def _parse(data):
     # This method does a little bit of excessive-looking
     # precondition checking. This is so that the behavior of this
     # class exactly matches its C counterpart to try and help
     # prevent surprise breakage for anyone that develops against
     # the pure version.
     if data and data[-1:] != b'\n':
         raise ValueError(b'Manifest did not end in a newline.')
     prev = None
     for l in data.splitlines():
         if prev is not None and prev > l:
             raise ValueError(b'Manifest lines not in sorted order.')
         prev = l
         f, n = l.split(b'\0')
         nl = len(n)
         flags = n[-1:]
         if flags in _manifestflags:
             n = n[:-1]
             nl -= 1
         else:
             flags = b''
         if nl not in (40, 64):
             raise ValueError(b'Invalid manifest line')
 
         yield f, bin(n), flags
 
 
 def _text(it):
     files = []
     lines = []
     for f, n, fl in it:
         files.append(f)
         # if this is changed to support newlines in filenames,
         # be sure to check the templates/ dir again (especially *-raw.tmpl)
         lines.append(b"%s\0%s%s\n" % (f, hex(n), fl))
 
     _checkforbidden(files)
     return b''.join(lines)
 
 
 class lazymanifestiter(object):
     def __init__(self, lm):
         self.pos = 0
         self.lm = lm
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data[0]
         self.pos += 1
         zeropos = data.find(b'\x00', pos)
         return data[pos:zeropos]
 
     __next__ = next
 
 
 class lazymanifestiterentries(object):
     def __init__(self, lm):
         self.lm = lm
         self.pos = 0
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data
         zeropos = data.find(b'\x00', pos)
         nlpos = data.find(b'\n', pos)
         if zeropos == -1 or nlpos == -1 or nlpos < zeropos:
             raise error.StorageError(b'Invalid manifest line')
         flags = data[nlpos - 1 : nlpos]
         if flags in _manifestflags:
             hlen = nlpos - zeropos - 2
         else:
             hlen = nlpos - zeropos - 1
             flags = b''
         if hlen not in (40, 64):
             raise error.StorageError(b'Invalid manifest line')
         hashval = unhexlify(
             data, self.lm.extrainfo[self.pos], zeropos + 1, hlen
         )
         self.pos += 1
         return (data[pos:zeropos], hashval, flags)
 
     __next__ = next
 
 
 def unhexlify(data, extra, pos, length):
     s = bin(data[pos : pos + length])
     if extra:
         s += chr(extra & 0xFF)
     return s
 
 
 def _cmp(a, b):
     return (a > b) - (a < b)
 
 
 _manifestflags = {b'', b'l', b't', b'x'}
 
 
 class _lazymanifest(object):
     """A pure python manifest backed by a byte string.  It is supplimented with
     internal lists as it is modified, until it is compacted back to a pure byte
     string.
 
     ``data`` is the initial manifest data.
 
     ``positions`` is a list of offsets, one per manifest entry.  Positive
     values are offsets into ``data``, negative values are offsets into the
     ``extradata`` list.  When an entry is removed, its entry is dropped from
     ``positions``.  The values are encoded such that when walking the list and
     indexing into ``data`` or ``extradata`` as appropriate, the entries are
     sorted by filename.
 
     ``extradata`` is a list of (key, hash, flags) for entries that were added or
     modified since the manifest was created or compacted.
     """
 
     def __init__(
         self,
         data,
         positions=None,
         extrainfo=None,
         extradata=None,
         hasremovals=False,
     ):
         if positions is None:
             self.positions = self.findlines(data)
             self.extrainfo = [0] * len(self.positions)
             self.data = data
             self.extradata = []
             self.hasremovals = False
         else:
             self.positions = positions[:]
             self.extrainfo = extrainfo[:]
             self.extradata = extradata[:]
             self.data = data
             self.hasremovals = hasremovals
 
     def findlines(self, data):
         if not data:
             return []
         pos = data.find(b"\n")
         if pos == -1 or data[-1:] != b'\n':
             raise ValueError(b"Manifest did not end in a newline.")
         positions = [0]
         prev = data[: data.find(b'\x00')]
         while pos < len(data) - 1 and pos != -1:
             positions.append(pos + 1)
             nexts = data[pos + 1 : data.find(b'\x00', pos + 1)]
             if nexts < prev:
                 raise ValueError(b"Manifest lines not in sorted order.")
             prev = nexts
             pos = data.find(b"\n", pos + 1)
         return positions
 
     def _get(self, index):
         # get the position encoded in pos:
         #   positive number is an index in 'data'
         #   negative number is in extrapieces
         pos = self.positions[index]
         if pos >= 0:
             return self.data, pos
         return self.extradata[-pos - 1], -1
 
     def _getkey(self, pos):
         if pos >= 0:
             return self.data[pos : self.data.find(b'\x00', pos + 1)]
         return self.extradata[-pos - 1][0]
 
     def bsearch(self, key):
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last) // 2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return midpoint
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return -1
 
     def bsearch2(self, key):
         # same as the above, but will always return the position
         # done for performance reasons
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last) // 2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return (midpoint, True)
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return (first, False)
 
     def __contains__(self, key):
         return self.bsearch(key) != -1
 
     def __getitem__(self, key):
         if not isinstance(key, bytes):
             raise TypeError(b"getitem: manifest keys must be a bytes.")
         needle = self.bsearch(key)
         if needle == -1:
             raise KeyError
         data, pos = self._get(needle)
         if pos == -1:
             return (data[1], data[2])
         zeropos = data.find(b'\x00', pos)
         nlpos = data.find(b'\n', zeropos)
         assert 0 <= needle <= len(self.positions)
         assert len(self.extrainfo) == len(self.positions)
         if zeropos == -1 or nlpos == -1 or nlpos < zeropos:
             raise error.StorageError(b'Invalid manifest line')
         hlen = nlpos - zeropos - 1
         flags = data[nlpos - 1 : nlpos]
         if flags in _manifestflags:
             hlen -= 1
         else:
             flags = b''
         if hlen not in (40, 64):
             raise error.StorageError(b'Invalid manifest line')
         hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, hlen)
         return (hashval, flags)
 
     def __delitem__(self, key):
         needle, found = self.bsearch2(key)
         if not found:
             raise KeyError
         cur = self.positions[needle]
         self.positions = self.positions[:needle] + self.positions[needle + 1 :]
         self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1 :]
         if cur >= 0:
             # This does NOT unsort the list as far as the search functions are
             # concerned, as they only examine lines mapped by self.positions.
             self.data = self.data[:cur] + b'\x00' + self.data[cur + 1 :]
             self.hasremovals = True
 
     def __setitem__(self, key, value):
         if not isinstance(key, bytes):
             raise TypeError(b"setitem: manifest keys must be a byte string.")
         if not isinstance(value, tuple) or len(value) != 2:
             raise TypeError(
                 b"Manifest values must be a tuple of (node, flags)."
             )
         hashval = value[0]
         if not isinstance(hashval, bytes) or len(hashval) not in (20, 32):
             raise TypeError(b"node must be a 20-byte or 32-byte byte string")
         flags = value[1]
         if not isinstance(flags, bytes) or len(flags) > 1:
             raise TypeError(b"flags must a 0 or 1 byte string, got %r", flags)
         needle, found = self.bsearch2(key)
         if found:
             # put the item
             pos = self.positions[needle]
             if pos < 0:
                 self.extradata[-pos - 1] = (key, hashval, value[1])
             else:
                 # just don't bother
                 self.extradata.append((key, hashval, value[1]))
                 self.positions[needle] = -len(self.extradata)
         else:
             # not found, put it in with extra positions
             self.extradata.append((key, hashval, value[1]))
             self.positions = (
                 self.positions[:needle]
                 + [-len(self.extradata)]
                 + self.positions[needle:]
             )
             self.extrainfo = (
                 self.extrainfo[:needle] + [0] + self.extrainfo[needle:]
             )
 
     def copy(self):
         # XXX call _compact like in C?
         return _lazymanifest(
             self.data,
             self.positions,
             self.extrainfo,
             self.extradata,
             self.hasremovals,
         )
 
     def _compact(self):
         # hopefully not called TOO often
         if len(self.extradata) == 0 and not self.hasremovals:
             return
         l = []
         i = 0
         offset = 0
         self.extrainfo = [0] * len(self.positions)
         while i < len(self.positions):
             if self.positions[i] >= 0:
                 cur = self.positions[i]
                 last_cut = cur
 
                 # Collect all contiguous entries in the buffer at the current
                 # offset, breaking out only for added/modified items held in
                 # extradata, or a deleted line prior to the next position.
                 while True:
                     self.positions[i] = offset
                     i += 1
                     if i == len(self.positions) or self.positions[i] < 0:
                         break
 
                     # A removed file has no positions[] entry, but does have an
                     # overwritten first byte.  Break out and find the end of the
                     # current good entry/entries if there is a removed file
                     # before the next position.
                     if (
                         self.hasremovals
                         and self.data.find(b'\n\x00', cur, self.positions[i])
                         != -1
                     ):
                         break
 
                     offset += self.positions[i] - cur
                     cur = self.positions[i]
                 end_cut = self.data.find(b'\n', cur)
                 if end_cut != -1:
                     end_cut += 1
                 offset += end_cut - cur
                 l.append(self.data[last_cut:end_cut])
             else:
                 while i < len(self.positions) and self.positions[i] < 0:
                     cur = self.positions[i]
                     t = self.extradata[-cur - 1]
                     l.append(self._pack(t))
                     self.positions[i] = offset
                     # Hashes are either 20 bytes (old sha1s) or 32
                     # bytes (new non-sha1).
                     hlen = 20
                     if len(t[1]) > 25:
                         hlen = 32
                     if len(t[1]) > hlen:
                         self.extrainfo[i] = ord(t[1][hlen + 1])
                     offset += len(l[-1])
                     i += 1
         self.data = b''.join(l)
         self.hasremovals = False
         self.extradata = []
 
     def _pack(self, d):
         n = d[1]
         assert len(n) in (20, 32)
         return d[0] + b'\x00' + hex(n) + d[2] + b'\n'
 
     def text(self):
         self._compact()
         return self.data
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.'''
         # XXX think whether efficiency matters here
         diff = {}
 
         for fn, e1, flags in self.iterentries():
             if fn not in m2:
                 diff[fn] = (e1, flags), (None, b'')
             else:
                 e2 = m2[fn]
                 if (e1, flags) != e2:
                     diff[fn] = (e1, flags), e2
                 elif clean:
                     diff[fn] = None
 
         for fn, e2, flags in m2.iterentries():
             if fn not in self:
                 diff[fn] = (None, b''), (e2, flags)
 
         return diff
 
     def iterentries(self):
         return lazymanifestiterentries(self)
 
     def iterkeys(self):
         return lazymanifestiter(self)
 
     def __iter__(self):
         return lazymanifestiter(self)
 
     def __len__(self):
         return len(self.positions)
 
     def filtercopy(self, filterfn):
         # XXX should be optimized
         c = _lazymanifest(b'')
         for f, n, fl in self.iterentries():
             if filterfn(f):
                 c[f] = n, fl
         return c
 
 
 try:
     _lazymanifest = parsers.lazymanifest
 except AttributeError:
     pass
 
 
 @interfaceutil.implementer(repository.imanifestdict)
 class manifestdict(object):
     def __init__(self, data=b''):
         self._lm = _lazymanifest(data)
 
     def __getitem__(self, key):
         return self._lm[key][0]
 
     def find(self, key):
         return self._lm[key]
 
     def __len__(self):
         return len(self._lm)
 
     def __nonzero__(self):
         # nonzero is covered by the __len__ function, but implementing it here
         # makes it easier for extensions to override.
         return len(self._lm) != 0
 
     __bool__ = __nonzero__
 
     def __setitem__(self, key, node):
         self._lm[key] = node, self.flags(key)
 
     def __contains__(self, key):
         if key is None:
             return False
         return key in self._lm
 
     def __delitem__(self, key):
         del self._lm[key]
 
     def __iter__(self):
         return self._lm.__iter__()
 
     def iterkeys(self):
         return self._lm.iterkeys()
 
     def keys(self):
         return list(self.iterkeys())
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match is not None:
             match = matchmod.badmatch(match, lambda path, msg: None)
             sm2 = set(m2.walk(match))
             return {f for f in self.walk(match) if f not in sm2}
         return {f for f in self if f not in m2}
 
     @propertycache
     def _dirs(self):
         return pathutil.dirs(self)
 
     def dirs(self):
         return self._dirs
 
     def hasdir(self, dir):
         return dir in self._dirs
 
     def _filesfastpath(self, match):
         """Checks whether we can correctly and quickly iterate over matcher
         files instead of over manifest files."""
         files = match.files()
         return len(files) < 100 and (
             match.isexact()
             or (match.prefix() and all(fn in self for fn in files))
         )
 
     def walk(self, match):
         """Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         """
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         # avoid the entire walk if we're only looking for specific files
         if self._filesfastpath(match):
             for fn in sorted(fset):
                 if fn in self:
                     yield fn
             return
 
         for fn in self:
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             if match(fn):
                 yield fn
 
         # for dirstate.walk, files=[''] means "walk the whole tree".
         # follow that here, too
         fset.discard(b'')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         if self._filesfastpath(match):
             m = manifestdict()
             lm = self._lm
             for fn in match.files():
                 if fn in lm:
                     m._lm[fn] = lm[fn]
             return m
 
         m = manifestdict()
         m._lm = self._lm.filtercopy(match)
         return m
 
     def diff(self, m2, match=None, clean=False):
         """Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         """
         if match:
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.diff(m2, clean=clean)
         return self._lm.diff(m2._lm, clean)
 
     def setflag(self, key, flag):
         if flag not in _manifestflags:
             raise TypeError(b"Invalid manifest flag set.")
         self._lm[key] = self[key], flag
 
     def get(self, key, default=None):
         try:
             return self._lm[key][0]
         except KeyError:
             return default
 
     def flags(self, key):
         try:
             return self._lm[key][1]
         except KeyError:
             return b''
 
     def copy(self):
         c = manifestdict()
         c._lm = self._lm.copy()
         return c
 
     def items(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iteritems(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iterentries(self):
         return self._lm.iterentries()
 
     def text(self):
         # most likely uses native version
         return self._lm.text()
 
     def fastdelta(self, base, changes):
         """Given a base manifest text as a bytearray and a list of changes
         relative to that text, compute a delta that can be used by revlog.
         """
         delta = []
         dstart = None
         dend = None
         dline = [b""]
         start = 0
         # zero copy representation of base as a buffer
         addbuf = util.buffer(base)
 
         changes = list(changes)
         if len(changes) < FASTDELTA_TEXTDIFF_THRESHOLD:
             # start with a readonly loop that finds the offset of
             # each line and creates the deltas
             for f, todelete in changes:
                 # bs will either be the index of the item or the insert point
                 start, end = _msearch(addbuf, f, start)
                 if not todelete:
                     h, fl = self._lm[f]
                     l = b"%s\0%s%s\n" % (f, hex(h), fl)
                 else:
                     if start == end:
                         # item we want to delete was not found, error out
                         raise AssertionError(
                             _(b"failed to remove %s from manifest") % f
                         )
                     l = b""
                 if dstart is not None and dstart <= start and dend >= start:
                     if dend < end:
                         dend = end
                     if l:
                         dline.append(l)
                 else:
                     if dstart is not None:
                         delta.append([dstart, dend, b"".join(dline)])
                     dstart = start
                     dend = end
                     dline = [l]
 
             if dstart is not None:
                 delta.append([dstart, dend, b"".join(dline)])
             # apply the delta to the base, and get a delta for addrevision
             deltatext, arraytext = _addlistdelta(base, delta)
         else:
             # For large changes, it's much cheaper to just build the text and
             # diff it.
             arraytext = bytearray(self.text())
             deltatext = mdiff.textdiff(
                 util.buffer(base), util.buffer(arraytext)
             )
 
         return arraytext, deltatext
 
 
 def _msearch(m, s, lo=0, hi=None):
     """return a tuple (start, end) that says where to find s within m.
 
     If the string is found m[start:end] are the line containing
     that string.  If start == end the string was not found and
     they indicate the proper sorted insertion point.
 
     m should be a buffer, a memoryview or a byte string.
     s is a byte string"""
 
     def advance(i, c):
         while i < lenm and m[i : i + 1] != c:
             i += 1
         return i
 
     if not s:
         return (lo, lo)
     lenm = len(m)
     if not hi:
         hi = lenm
     while lo < hi:
         mid = (lo + hi) // 2
         start = mid
         while start > 0 and m[start - 1 : start] != b'\n':
             start -= 1
         end = advance(start, b'\0')
         if bytes(m[start:end]) < s:
             # we know that after the null there are 40 bytes of sha1
             # this translates to the bisect lo = mid + 1
             lo = advance(end + 40, b'\n') + 1
         else:
             # this translates to the bisect hi = mid
             hi = start
     end = advance(lo, b'\0')
     found = m[lo:end]
     if s == found:
         # we know that after the null there are 40 bytes of sha1
         end = advance(end + 40, b'\n')
         return (lo, end + 1)
     else:
         return (lo, lo)
 
 
 def _checkforbidden(l):
     """Check filenames for illegal characters."""
     for f in l:
         if b'\n' in f or b'\r' in f:
             raise error.StorageError(
                 _(b"'\\n' and '\\r' disallowed in filenames: %r")
                 % pycompat.bytestr(f)
             )
 
 
 # apply the changes collected during the bisect loop to our addlist
 # return a delta suitable for addrevision
 def _addlistdelta(addlist, x):
     # for large addlist arrays, building a new array is cheaper
     # than repeatedly modifying the existing one
     currentposition = 0
     newaddlist = bytearray()
 
     for start, end, content in x:
         newaddlist += addlist[currentposition:start]
         if content:
             newaddlist += bytearray(content)
 
         currentposition = end
 
     newaddlist += addlist[currentposition:]
 
     deltatext = b"".join(
         struct.pack(b">lll", start, end, len(content)) + content
         for start, end, content in x
     )
     return deltatext, newaddlist
 
 
 def _splittopdir(f):
     if b'/' in f:
         dir, subpath = f.split(b'/', 1)
         return dir + b'/', subpath
     else:
         return b'', f
 
 
 _noop = lambda s: None
 
 
 @interfaceutil.implementer(repository.imanifestdict)
 class treemanifest(object):
     def __init__(self, dir=b'', text=b''):
         self._dir = dir
         self._node = nullid
         self._loadfunc = _noop
         self._copyfunc = _noop
         self._dirty = False
         self._dirs = {}
         self._lazydirs = {}
         # Using _lazymanifest here is a little slower than plain old dicts
         self._files = {}
         self._flags = {}
         if text:
 
             def readsubtree(subdir, subm):
                 raise AssertionError(
                     b'treemanifest constructor only accepts flat manifests'
                 )
 
             self.parse(text, readsubtree)
             self._dirty = True  # Mark flat manifest dirty after parsing
 
     def _subpath(self, path):
         return self._dir + path
 
     def _loadalllazy(self):
         selfdirs = self._dirs
         subpath = self._subpath
         for d, (node, readsubtree, docopy) in pycompat.iteritems(
             self._lazydirs
         ):
             if docopy:
                 selfdirs[d] = readsubtree(subpath(d), node).copy()
             else:
                 selfdirs[d] = readsubtree(subpath(d), node)
         self._lazydirs = {}
 
     def _loadlazy(self, d):
         v = self._lazydirs.get(d)
         if v:
             node, readsubtree, docopy = v
             if docopy:
                 self._dirs[d] = readsubtree(self._subpath(d), node).copy()
             else:
                 self._dirs[d] = readsubtree(self._subpath(d), node)
             del self._lazydirs[d]
 
     def _loadchildrensetlazy(self, visit):
         if not visit:
             return None
         if visit == b'all' or visit == b'this':
             self._loadalllazy()
             return None
 
         loadlazy = self._loadlazy
         for k in visit:
             loadlazy(k + b'/')
         return visit
 
     def _loaddifflazy(self, t1, t2):
         """load items in t1 and t2 if they're needed for diffing.
 
         The criteria currently is:
         - if it's not present in _lazydirs in either t1 or t2, load it in the
           other (it may already be loaded or it may not exist, doesn't matter)
         - if it's present in _lazydirs in both, compare the nodeid; if it
           differs, load it in both
         """
         toloadlazy = []
         for d, v1 in pycompat.iteritems(t1._lazydirs):
             v2 = t2._lazydirs.get(d)
             if not v2 or v2[0] != v1[0]:
                 toloadlazy.append(d)
         for d, v1 in pycompat.iteritems(t2._lazydirs):
             if d not in t1._lazydirs:
                 toloadlazy.append(d)
 
         for d in toloadlazy:
             t1._loadlazy(d)
             t2._loadlazy(d)
 
     def __len__(self):
         self._load()
         size = len(self._files)
         self._loadalllazy()
         for m in self._dirs.values():
             size += m.__len__()
         return size
 
     def __nonzero__(self):
         # Faster than "__len() != 0" since it avoids loading sub-manifests
         return not self._isempty()
 
     __bool__ = __nonzero__
 
     def _isempty(self):
         self._load()  # for consistency; already loaded by all callers
         # See if we can skip loading everything.
         if self._files or (
             self._dirs and any(not m._isempty() for m in self._dirs.values())
         ):
             return False
         self._loadalllazy()
         return not self._dirs or all(m._isempty() for m in self._dirs.values())
 
     @encoding.strmethod
     def __repr__(self):
         return (
             b'<treemanifest dir=%s, node=%s, loaded=%r, dirty=%r at 0x%x>'
             % (
                 self._dir,
                 hex(self._node),
                 bool(self._loadfunc is _noop),
                 self._dirty,
                 id(self),
             )
         )
 
     def dir(self):
         """The directory that this tree manifest represents, including a
         trailing '/'. Empty string for the repo root directory."""
         return self._dir
 
     def node(self):
         """This node of this instance. nullid for unsaved instances. Should
         be updated when the instance is read or written from a revlog.
         """
         assert not self._dirty
         return self._node
 
     def setnode(self, node):
         self._node = node
         self._dirty = False
 
     def iterentries(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(
             itertools.chain(self._dirs.items(), self._files.items())
         ):
             if p in self._files:
                 yield self._subpath(p), n, self._flags.get(p, b'')
             else:
                 for x in n.iterentries():
                     yield x
 
     def items(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(
             itertools.chain(self._dirs.items(), self._files.items())
         ):
             if p in self._files:
                 yield self._subpath(p), n
             else:
                 for f, sn in pycompat.iteritems(n):
                     yield f, sn
 
     iteritems = items
 
     def iterkeys(self):
         self._load()
         self._loadalllazy()
         for p in sorted(itertools.chain(self._dirs, self._files)):
             if p in self._files:
                 yield self._subpath(p)
             else:
                 for f in self._dirs[p]:
                     yield f
 
     def keys(self):
         return list(self.iterkeys())
 
     def __iter__(self):
         return self.iterkeys()
 
     def __contains__(self, f):
         if f is None:
             return False
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return False
 
             return self._dirs[dir].__contains__(subpath)
         else:
             return f in self._files
 
     def get(self, f, default=None):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return default
             return self._dirs[dir].get(subpath, default)
         else:
             return self._files.get(f, default)
 
     def __getitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             return self._dirs[dir].__getitem__(subpath)
         else:
             return self._files[f]
 
     def flags(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return b''
             return self._dirs[dir].flags(subpath)
         else:
             if f in self._lazydirs or f in self._dirs:
                 return b''
             return self._flags.get(f, b'')
 
     def find(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             return self._dirs[dir].find(subpath)
         else:
             return self._files[f], self._flags.get(f, b'')
 
     def __delitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             self._dirs[dir].__delitem__(subpath)
             # If the directory is now empty, remove it
             if self._dirs[dir]._isempty():
                 del self._dirs[dir]
         else:
             del self._files[f]
             if f in self._flags:
                 del self._flags[f]
         self._dirty = True
 
     def __setitem__(self, f, n):
         assert n is not None
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].__setitem__(subpath, n)
         else:
             # manifest nodes are either 20 bytes or 32 bytes,
             # depending on the hash in use. Assert this as historically
             # sometimes extra bytes were added.
             assert len(n) in (20, 32)
             self._files[f] = n
         self._dirty = True
 
     def _load(self):
         if self._loadfunc is not _noop:
             lf, self._loadfunc = self._loadfunc, _noop
             lf(self)
         elif self._copyfunc is not _noop:
             cf, self._copyfunc = self._copyfunc, _noop
             cf(self)
 
     def setflag(self, f, flags):
         """Set the flags (symlink, executable) for path f."""
         if flags not in _manifestflags:
             raise TypeError(b"Invalid manifest flag set.")
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].setflag(subpath, flags)
         else:
             self._flags[f] = flags
         self._dirty = True
 
     def copy(self):
         copy = treemanifest(self._dir)
         copy._node = self._node
         copy._dirty = self._dirty
         if self._copyfunc is _noop:
 
             def _copyfunc(s):
                 self._load()
                 s._lazydirs = {
                     d: (n, r, True)
                     for d, (n, r, c) in pycompat.iteritems(self._lazydirs)
                 }
                 sdirs = s._dirs
                 for d, v in pycompat.iteritems(self._dirs):
                     sdirs[d] = v.copy()
                 s._files = dict.copy(self._files)
                 s._flags = dict.copy(self._flags)
 
             if self._loadfunc is _noop:
                 _copyfunc(copy)
             else:
                 copy._copyfunc = _copyfunc
         else:
             copy._copyfunc = self._copyfunc
         return copy
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match and not match.always():
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.filesnotin(m2)
 
         files = set()
 
         def _filesnotin(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             self._loaddifflazy(t1, t2)
             for d, m1 in pycompat.iteritems(t1._dirs):
                 if d in t2._dirs:
                     m2 = t2._dirs[d]
                     _filesnotin(m1, m2)
                 else:
                     files.update(m1.iterkeys())
 
             for fn in t1._files:
                 if fn not in t2._files:
                     files.add(t1._subpath(fn))
 
         _filesnotin(self, m2)
         return files
 
     @propertycache
     def _alldirs(self):
         return pathutil.dirs(self)
 
     def dirs(self):
         return self._alldirs
 
     def hasdir(self, dir):
         self._load()
         topdir, subdir = _splittopdir(dir)
         if topdir:
             self._loadlazy(topdir)
             if topdir in self._dirs:
                 return self._dirs[topdir].hasdir(subdir)
             return False
         dirslash = dir + b'/'
         return dirslash in self._dirs or dirslash in self._lazydirs
 
     def walk(self, match):
         """Generates matching file names.
 
         It also reports nonexistent files by marking them bad with match.bad().
         """
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         for fn in self._walk(match):
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             yield fn
 
         # for dirstate.walk, files=[''] means "walk the whole tree".
         # follow that here, too
         fset.discard(b'')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _walk(self, match):
         '''Recursively generates matching file names for walk().'''
         visit = match.visitchildrenset(self._dir[:-1])
         if not visit:
             return
 
         # yield this dir's files and walk its submanifests
         self._load()
         visit = self._loadchildrensetlazy(visit)
         for p in sorted(list(self._dirs) + list(self._files)):
             if p in self._files:
                 fullp = self._subpath(p)
                 if match(fullp):
                     yield fullp
             else:
                 if not visit or p[:-1] in visit:
                     for f in self._dirs[p]._walk(match):
                         yield f
 
     def _matches(self, match):
         """recursively generate a new manifest filtered by the match argument."""
         if match.always():
             return self.copy()
         return self._matches_inner(match)
 
     def _matches_inner(self, match):
         if match.always():
             return self.copy()
 
         visit = match.visitchildrenset(self._dir[:-1])
         if visit == b'all':
             return self.copy()
         ret = treemanifest(self._dir)
         if not visit:
             return ret
 
         self._load()
         for fn in self._files:
             # While visitchildrenset *usually* lists only subdirs, this is
             # actually up to the matcher and may have some files in the set().
             # If visit == 'this', we should obviously look at the files in this
             # directory; if visit is a set, and fn is in it, we should inspect
             # fn (but no need to inspect things not in the set).
             if visit != b'this' and fn not in visit:
                 continue
             fullp = self._subpath(fn)
             # visitchildrenset isn't perfect, we still need to call the regular
             # matcher code to further filter results.
             if not match(fullp):
                 continue
             ret._files[fn] = self._files[fn]
             if fn in self._flags:
                 ret._flags[fn] = self._flags[fn]
 
         visit = self._loadchildrensetlazy(visit)
         for dir, subm in pycompat.iteritems(self._dirs):
             if visit and dir[:-1] not in visit:
                 continue
             m = subm._matches_inner(match)
             if not m._isempty():
                 ret._dirs[dir] = m
 
         if not ret._isempty():
             ret._dirty = True
         return ret
 
     def fastdelta(self, base, changes):
         raise FastdeltaUnavailable()
 
     def diff(self, m2, match=None, clean=False):
         """Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         """
         if match and not match.always():
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.diff(m2, clean=clean)
         result = {}
         emptytree = treemanifest()
 
         def _iterativediff(t1, t2, stack):
             """compares two tree manifests and append new tree-manifests which
             needs to be compared to stack"""
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             self._loaddifflazy(t1, t2)
 
             for d, m1 in pycompat.iteritems(t1._dirs):
                 m2 = t2._dirs.get(d, emptytree)
                 stack.append((m1, m2))
 
             for d, m2 in pycompat.iteritems(t2._dirs):
                 if d not in t1._dirs:
                     stack.append((emptytree, m2))
 
             for fn, n1 in pycompat.iteritems(t1._files):
                 fl1 = t1._flags.get(fn, b'')
                 n2 = t2._files.get(fn, None)
                 fl2 = t2._flags.get(fn, b'')
                 if n1 != n2 or fl1 != fl2:
                     result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
                 elif clean:
                     result[t1._subpath(fn)] = None
 
             for fn, n2 in pycompat.iteritems(t2._files):
                 if fn not in t1._files:
                     fl2 = t2._flags.get(fn, b'')
                     result[t2._subpath(fn)] = ((None, b''), (n2, fl2))
 
         stackls = []
         _iterativediff(self, m2, stackls)
         while stackls:
             t1, t2 = stackls.pop()
             # stackls is populated in the function call
             _iterativediff(t1, t2, stackls)
         return result
 
     def unmodifiedsince(self, m2):
         return not self._dirty and not m2._dirty and self._node == m2._node
 
     def parse(self, text, readsubtree):
         selflazy = self._lazydirs
         for f, n, fl in _parse(text):
             if fl == b't':
                 f = f + b'/'
                 # False below means "doesn't need to be copied" and can use the
                 # cached value from readsubtree directly.
                 selflazy[f] = (n, readsubtree, False)
             elif b'/' in f:
                 # This is a flat manifest, so use __setitem__ and setflag rather
                 # than assigning directly to _files and _flags, so we can
                 # assign a path in a subdirectory, and to mark dirty (compared
                 # to nullid).
                 self[f] = n
                 if fl:
                     self.setflag(f, fl)
             else:
                 # Assigning to _files and _flags avoids marking as dirty,
                 # and should be a little faster.
                 self._files[f] = n
                 if fl:
                     self._flags[f] = fl
 
     def text(self):
         """Get the full data of this manifest as a bytestring."""
         self._load()
         return _text(self.iterentries())
 
     def dirtext(self):
         """Get the full data of this directory as a bytestring. Make sure that
         any submanifests have been written first, so their nodeids are correct.
         """
         self._load()
         flags = self.flags
         lazydirs = [
             (d[:-1], v[0], b't') for d, v in pycompat.iteritems(self._lazydirs)
         ]
         dirs = [(d[:-1], self._dirs[d]._node, b't') for d in self._dirs]
         files = [(f, self._files[f], flags(f)) for f in self._files]
         return _text(sorted(dirs + files + lazydirs))
 
     def read(self, gettext, readsubtree):
         def _load_for_read(s):
             s.parse(gettext(), readsubtree)
             s._dirty = False
 
         self._loadfunc = _load_for_read
 
     def writesubtrees(self, m1, m2, writesubtree, match):
         self._load()  # for consistency; should never have any effect here
         m1._load()
         m2._load()
         emptytree = treemanifest()
 
         def getnode(m, d):
             ld = m._lazydirs.get(d)
             if ld:
                 return ld[0]
             return m._dirs.get(d, emptytree)._node
 
         # let's skip investigating things that `match` says we do not need.
         visit = match.visitchildrenset(self._dir[:-1])
         visit = self._loadchildrensetlazy(visit)
         if visit == b'this' or visit == b'all':
             visit = None
         for d, subm in pycompat.iteritems(self._dirs):
             if visit and d[:-1] not in visit:
                 continue
             subp1 = getnode(m1, d)
             subp2 = getnode(m2, d)
             if subp1 == nullid:
                 subp1, subp2 = subp2, subp1
             writesubtree(subm, subp1, subp2, match)
 
     def walksubtrees(self, matcher=None):
         """Returns an iterator of the subtrees of this manifest, including this
         manifest itself.
 
         If `matcher` is provided, it only returns subtrees that match.
         """
         if matcher and not matcher.visitdir(self._dir[:-1]):
             return
         if not matcher or matcher(self._dir[:-1]):
             yield self
 
         self._load()
         # OPT: use visitchildrenset to avoid loading everything.
         self._loadalllazy()
         for d, subm in pycompat.iteritems(self._dirs):
             for subtree in subm.walksubtrees(matcher=matcher):
                 yield subtree
 
 
 class manifestfulltextcache(util.lrucachedict):
     """File-backed LRU cache for the manifest cache
 
     File consists of entries, up to EOF:
 
     - 20 bytes node, 4 bytes length, <length> manifest data
 
     These are written in reverse cache order (oldest to newest).
 
     """
 
     _file = b'manifestfulltextcache'
 
     def __init__(self, max):
         super(manifestfulltextcache, self).__init__(max)
         self._dirty = False
         self._read = False
         self._opener = None
 
     def read(self):
         if self._read or self._opener is None:
             return
 
         try:
             with self._opener(self._file) as fp:
                 set = super(manifestfulltextcache, self).__setitem__
                 # ignore trailing data, this is a cache, corruption is skipped
                 while True:
                     # TODO do we need to do work here for sha1 portability?
                     node = fp.read(20)
                     if len(node) < 20:
                         break
                     try:
                         size = struct.unpack(b'>L', fp.read(4))[0]
                     except struct.error:
                         break
                     value = bytearray(fp.read(size))
                     if len(value) != size:
                         break
                     set(node, value)
         except IOError:
             # the file is allowed to be missing
             pass
 
         self._read = True
         self._dirty = False
 
     def write(self):
         if not self._dirty or self._opener is None:
             return
         # rotate backwards to the first used node
         try:
             with self._opener(
                 self._file, b'w', atomictemp=True, checkambig=True
             ) as fp:
                 node = self._head.prev
                 while True:
                     if node.key in self._cache:
                         fp.write(node.key)
                         fp.write(struct.pack(b'>L', len(node.value)))
                         fp.write(node.value)
                     if node is self._head:
                         break
                     node = node.prev
         except IOError:
             # We could not write the cache (eg: permission error)
             # the content can be missing.
             #
             # We could try harder and see if we could recreate a wcache
             # directory were we coudl write too.
             #
             # XXX the error pass silently, having some way to issue an error
             # log `ui.log` would be nice.
             pass
 
     def __len__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__len__()
 
     def __contains__(self, k):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__contains__(k)
 
     def __iter__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__iter__()
 
     def __getitem__(self, k):
         if not self._read:
             self.read()
         # the cache lru order can change on read
         setdirty = self._cache.get(k) is not self._head
         value = super(manifestfulltextcache, self).__getitem__(k)
         if setdirty:
             self._dirty = True
         return value
 
     def __setitem__(self, k, v):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__setitem__(k, v)
         self._dirty = True
 
     def __delitem__(self, k):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__delitem__(k)
         self._dirty = True
 
     def get(self, k, default=None):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).get(k, default=default)
 
     def clear(self, clear_persisted_data=False):
         super(manifestfulltextcache, self).clear()
         if clear_persisted_data:
             self._dirty = True
             self.write()
         self._read = False
 
 
 # and upper bound of what we expect from compression
 # (real live value seems to be "3")
 MAXCOMPRESSION = 3
 
 
 class FastdeltaUnavailable(Exception):
     """Exception raised when fastdelta isn't usable on a manifest."""
 
 
 @interfaceutil.implementer(repository.imanifeststorage)
 class manifestrevlog(object):
     """A revlog that stores manifest texts. This is responsible for caching the
     full-text manifest contents.
     """
 
     def __init__(
         self,
         opener,
         tree=b'',
         dirlogcache=None,
         indexfile=None,
         treemanifest=False,
     ):
         """Constructs a new manifest revlog
 
         `indexfile` - used by extensions to have two manifests at once, like
         when transitioning between flatmanifeset and treemanifests.
 
         `treemanifest` - used to indicate this is a tree manifest revlog. Opener
         options can also be used to make this a tree manifest revlog. The opener
         option takes precedence, so if it is set to True, we ignore whatever
         value is passed in to the constructor.
         """
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         optiontreemanifest = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get(b'manifestcachesize', cachesize)
             optiontreemanifest = opts.get(b'treemanifest', False)
 
         self._treeondisk = optiontreemanifest or treemanifest
 
         self._fulltextcache = manifestfulltextcache(cachesize)
 
         if tree:
             assert self._treeondisk, b'opts is %r' % opts
 
         if indexfile is None:
             indexfile = b'00manifest.i'
             if tree:
                 indexfile = b"meta/" + tree + indexfile
 
         self.tree = tree
 
         # The dirlogcache is kept on the root manifest log
         if tree:
             self._dirlogcache = dirlogcache
         else:
             self._dirlogcache = {b'': self}
 
         self._revlog = revlog.revlog(
             opener,
             indexfile,
             # only root indexfile is cached
             checkambig=not bool(tree),
             mmaplargeindex=True,
             upperboundcomp=MAXCOMPRESSION,
             persistentnodemap=opener.options.get(b'persistent-nodemap', False),
         )
 
         self.index = self._revlog.index
         self.version = self._revlog.version
         self._generaldelta = self._revlog._generaldelta
         self._revlog.revlog_kind = b'manifest'
 
     def _setupmanifestcachehooks(self, repo):
         """Persist the manifestfulltextcache on lock release"""
         if not util.safehasattr(repo, b'_wlockref'):
             return
 
         self._fulltextcache._opener = repo.wcachevfs
         if repo._currentlock(repo._wlockref) is None:
             return
 
         reporef = weakref.ref(repo)
         manifestrevlogref = weakref.ref(self)
 
         def persistmanifestcache(success):
             # Repo is in an unknown state, do not persist.
             if not success:
                 return
 
             repo = reporef()
             self = manifestrevlogref()
             if repo is None or self is None:
                 return
             if repo.manifestlog.getstorage(b'') is not self:
                 # there's a different manifest in play now, abort
                 return
             self._fulltextcache.write()
 
         repo._afterlock(persistmanifestcache)
 
     @property
     def fulltextcache(self):
         return self._fulltextcache
 
     def clearcaches(self, clear_persisted_data=False):
         self._revlog.clearcaches()
         self._fulltextcache.clear(clear_persisted_data=clear_persisted_data)
         self._dirlogcache = {self.tree: self}
 
     def dirlog(self, d):
         if d:
             assert self._treeondisk
         if d not in self._dirlogcache:
             mfrevlog = manifestrevlog(
                 self.opener, d, self._dirlogcache, treemanifest=self._treeondisk
             )
             self._dirlogcache[d] = mfrevlog
         return self._dirlogcache[d]
 
     def add(
         self,
         m,
         transaction,
         link,
         p1,
         p2,
         added,
         removed,
         readtree=None,
         match=None,
     ):
         """add some manifest entry in to the manifest log
 
         input:
 
           m:           the manifest dict we want to store
           transaction: the open transaction
           p1:          manifest-node of p1
           p2:          manifest-node of p2
           added:       file added/changed compared to parent
           removed:     file removed compared to parent
 
         tree manifest input:
 
           readtree:    a function to read a subtree
           match:       a filematcher for the subpart of the tree manifest
         """
         try:
             if p1 not in self.fulltextcache:
                 raise FastdeltaUnavailable()
             # If our first parent is in the manifest cache, we can
             # compute a delta here using properties we know about the
             # manifest up-front, which may save time later for the
             # revlog layer.
 
             _checkforbidden(added)
             # combine the changed lists into one sorted iterator
             work = heapq.merge(
                 [(x, False) for x in sorted(added)],
                 [(x, True) for x in sorted(removed)],
             )
 
             arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
             cachedelta = self._revlog.rev(p1), deltatext
             text = util.buffer(arraytext)
             rev = self._revlog.addrevision(
                 text, transaction, link, p1, p2, cachedelta
             )
             n = self._revlog.node(rev)
         except FastdeltaUnavailable:
             # The first parent manifest isn't already loaded or the
             # manifest implementation doesn't support fastdelta, so
             # we'll just encode a fulltext of the manifest and pass
             # that through to the revlog layer, and let it handle the
             # delta process.
             if self._treeondisk:
                 assert readtree, b"readtree must be set for treemanifest writes"
                 assert match, b"match must be specified for treemanifest writes"
                 m1 = readtree(self.tree, p1)
                 m2 = readtree(self.tree, p2)
                 n = self._addtree(
                     m, transaction, link, m1, m2, readtree, match=match
                 )
                 arraytext = None
             else:
                 text = m.text()
                 rev = self._revlog.addrevision(text, transaction, link, p1, p2)
                 n = self._revlog.node(rev)
                 arraytext = bytearray(text)
 
         if arraytext is not None:
             self.fulltextcache[n] = arraytext
 
         return n
 
     def _addtree(self, m, transaction, link, m1, m2, readtree, match):
         # If the manifest is unchanged compared to one parent,
         # don't write a new revision
         if self.tree != b'' and (
             m.unmodifiedsince(m1) or m.unmodifiedsince(m2)
         ):
             return m.node()
 
         def writesubtree(subm, subp1, subp2, match):
             sublog = self.dirlog(subm.dir())
             sublog.add(
                 subm,
                 transaction,
                 link,
                 subp1,
                 subp2,
                 None,
                 None,
                 readtree=readtree,
                 match=match,
             )
 
         m.writesubtrees(m1, m2, writesubtree, match)
         text = m.dirtext()
         n = None
         if self.tree != b'':
             # Double-check whether contents are unchanged to one parent
             if text == m1.dirtext():
                 n = m1.node()
             elif text == m2.dirtext():
                 n = m2.node()
 
         if not n:
             rev = self._revlog.addrevision(
                 text, transaction, link, m1.node(), m2.node()
             )
             n = self._revlog.node(rev)
 
         # Save nodeid so parent manifest can calculate its nodeid
         m.setnode(n)
         return n
 
     def __len__(self):
         return len(self._revlog)
 
     def __iter__(self):
         return self._revlog.__iter__()
 
     def rev(self, node):
         return self._revlog.rev(node)
 
     def node(self, rev):
         return self._revlog.node(rev)
 
     def lookup(self, value):
         return self._revlog.lookup(value)
 
     def parentrevs(self, rev):
         return self._revlog.parentrevs(rev)
 
     def parents(self, node):
         return self._revlog.parents(node)
 
     def linkrev(self, rev):
         return self._revlog.linkrev(rev)
 
     def checksize(self):
         return self._revlog.checksize()
 
     def revision(self, node, _df=None, raw=False):
         return self._revlog.revision(node, _df=_df, raw=raw)
 
     def rawdata(self, node, _df=None):
         return self._revlog.rawdata(node, _df=_df)
 
     def revdiff(self, rev1, rev2):
         return self._revlog.revdiff(rev1, rev2)
 
     def cmp(self, node, text):
         return self._revlog.cmp(node, text)
 
     def deltaparent(self, rev):
         return self._revlog.deltaparent(rev)
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
+        sidedata_helpers=None,
     ):
         return self._revlog.emitrevisions(
             nodes,
             nodesorder=nodesorder,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
             deltamode=deltamode,
+            sidedata_helpers=sidedata_helpers,
         )
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         alwayscache=False,
         addrevisioncb=None,
         duplicaterevisioncb=None,
     ):
         return self._revlog.addgroup(
             deltas,
             linkmapper,
             transaction,
             alwayscache=alwayscache,
             addrevisioncb=addrevisioncb,
             duplicaterevisioncb=duplicaterevisioncb,
         )
 
     def rawsize(self, rev):
         return self._revlog.rawsize(rev)
 
     def getstrippoint(self, minlink):
         return self._revlog.getstrippoint(minlink)
 
     def strip(self, minlink, transaction):
         return self._revlog.strip(minlink, transaction)
 
     def files(self):
         return self._revlog.files()
 
     def clone(self, tr, destrevlog, **kwargs):
         if not isinstance(destrevlog, manifestrevlog):
             raise error.ProgrammingError(b'expected manifestrevlog to clone()')
 
         return self._revlog.clone(tr, destrevlog._revlog, **kwargs)
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         return self._revlog.storageinfo(
             exclusivefiles=exclusivefiles,
             sharedfiles=sharedfiles,
             revisionscount=revisionscount,
             trackedsize=trackedsize,
             storedsize=storedsize,
         )
 
     @property
     def indexfile(self):
         return self._revlog.indexfile
 
     @indexfile.setter
     def indexfile(self, value):
         self._revlog.indexfile = value
 
     @property
     def opener(self):
         return self._revlog.opener
 
     @opener.setter
     def opener(self, value):
         self._revlog.opener = value
 
 
 @interfaceutil.implementer(repository.imanifestlog)
 class manifestlog(object):
     """A collection class representing the collection of manifest snapshots
     referenced by commits in the repository.
 
     In this situation, 'manifest' refers to the abstract concept of a snapshot
     of the list of files in the given commit. Consumers of the output of this
     class do not care about the implementation details of the actual manifests
     they receive (i.e. tree or flat or lazily loaded, etc)."""
 
     def __init__(self, opener, repo, rootstore, narrowmatch):
         usetreemanifest = False
         cachesize = 4
 
         opts = getattr(opener, 'options', None)
         if opts is not None:
             usetreemanifest = opts.get(b'treemanifest', usetreemanifest)
             cachesize = opts.get(b'manifestcachesize', cachesize)
 
         self._treemanifests = usetreemanifest
 
         self._rootstore = rootstore
         self._rootstore._setupmanifestcachehooks(repo)
         self._narrowmatch = narrowmatch
 
         # A cache of the manifestctx or treemanifestctx for each directory
         self._dirmancache = {}
         self._dirmancache[b''] = util.lrucachedict(cachesize)
 
         self._cachesize = cachesize
 
     def __getitem__(self, node):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
         """
         return self.get(b'', node)
 
     def get(self, tree, node, verify=True):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
 
         `verify` - if True an exception will be thrown if the node is not in
                    the revlog
         """
         if node in self._dirmancache.get(tree, ()):
             return self._dirmancache[tree][node]
 
         if not self._narrowmatch.always():
             if not self._narrowmatch.visitdir(tree[:-1]):
                 return excludeddirmanifestctx(tree, node)
         if tree:
             if self._rootstore._treeondisk:
                 if verify:
                     # Side-effect is LookupError is raised if node doesn't
                     # exist.
                     self.getstorage(tree).rev(node)
 
                 m = treemanifestctx(self, tree, node)
             else:
                 raise error.Abort(
                     _(
                         b"cannot ask for manifest directory '%s' in a flat "
                         b"manifest"
                     )
                     % tree
                 )
         else:
             if verify:
                 # Side-effect is LookupError is raised if node doesn't exist.
                 self._rootstore.rev(node)
 
             if self._treemanifests:
                 m = treemanifestctx(self, b'', node)
             else:
                 m = manifestctx(self, node)
 
         if node != nullid:
             mancache = self._dirmancache.get(tree)
             if not mancache:
                 mancache = util.lrucachedict(self._cachesize)
                 self._dirmancache[tree] = mancache
             mancache[node] = m
         return m
 
     def getstorage(self, tree):
         return self._rootstore.dirlog(tree)
 
     def clearcaches(self, clear_persisted_data=False):
         self._dirmancache.clear()
         self._rootstore.clearcaches(clear_persisted_data=clear_persisted_data)
 
     def rev(self, node):
         return self._rootstore.rev(node)
 
     def update_caches(self, transaction):
         return self._rootstore._revlog.update_caches(transaction=transaction)
 
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memmanifestctx(object):
     def __init__(self, manifestlog):
         self._manifestlog = manifestlog
         self._manifestdict = manifestdict()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     def read(self):
         return self._manifestdict
 
     def write(self, transaction, link, p1, p2, added, removed, match=None):
         return self._storage().add(
             self._manifestdict,
             transaction,
             link,
             p1,
             p2,
             added,
             removed,
             match=match,
         )
 
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class manifestctx(object):
     """A class representing a single revision of a manifest, including its
     contents, its parent revs, and its linkrev.
     """
 
     def __init__(self, manifestlog, node):
         self._manifestlog = manifestlog
         self._data = None
 
         self._node = node
 
         # TODO: We eventually want p1, p2, and linkrev exposed on this class,
         # but let's add it later when something needs it and we can load it
         # lazily.
         # self.p1, self.p2 = store.parents(node)
         # rev = store.rev(node)
         # self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def node(self):
         return self._node
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def read(self):
         if self._data is None:
             if self._node == nullid:
                 self._data = manifestdict()
             else:
                 store = self._storage()
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = manifestdict(text)
         return self._data
 
     def readfast(self, shallow=False):
         """Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, nothing changes since this is a flat manifest.
         """
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if deltaparent != nullrev and deltaparent in store.parentrevs(r):
             return self.readdelta()
         return self.read()
 
     def readdelta(self, shallow=False):
         """Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         Changing the value of `shallow` has no effect on flat manifests.
         """
         store = self._storage()
         r = store.rev(self._node)
         d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
         return manifestdict(d)
 
     def find(self, key):
         return self.read().find(key)
 
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memtreemanifestctx(object):
     def __init__(self, manifestlog, dir=b''):
         self._manifestlog = manifestlog
         self._dir = dir
         self._treemanifest = treemanifest()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self._treemanifest.copy()
         return memmf
 
     def read(self):
         return self._treemanifest
 
     def write(self, transaction, link, p1, p2, added, removed, match=None):
         def readtree(dir, node):
             return self._manifestlog.get(dir, node).read()
 
         return self._storage().add(
             self._treemanifest,
             transaction,
             link,
             p1,
             p2,
             added,
             removed,
             readtree=readtree,
             match=match,
         )
 
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class treemanifestctx(object):
     def __init__(self, manifestlog, dir, node):
         self._manifestlog = manifestlog
         self._dir = dir
         self._data = None
 
         self._node = node
 
         # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
         # we can instantiate treemanifestctx objects for directories we don't
         # have on disk.
         # self.p1, self.p2 = store.parents(node)
         # rev = store.rev(node)
         # self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         narrowmatch = self._manifestlog._narrowmatch
         if not narrowmatch.always():
             if not narrowmatch.visitdir(self._dir[:-1]):
                 return excludedmanifestrevlog(self._dir)
         return self._manifestlog.getstorage(self._dir)
 
     def read(self):
         if self._data is None:
             store = self._storage()
             if self._node == nullid:
                 self._data = treemanifest()
             # TODO accessing non-public API
             elif store._treeondisk:
                 m = treemanifest(dir=self._dir)
 
                 def gettext():
                     return store.revision(self._node)
 
                 def readsubtree(dir, subm):
                     # Set verify to False since we need to be able to create
                     # subtrees for trees that don't exist on disk.
                     return self._manifestlog.get(dir, subm, verify=False).read()
 
                 m.read(gettext, readsubtree)
                 m.setnode(self._node)
                 self._data = m
             else:
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = treemanifest(dir=self._dir, text=text)
 
         return self._data
 
     def node(self):
         return self._node
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def readdelta(self, shallow=False):
         """Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         If `shallow` is True, this will read the delta for this directory,
         without recursively reading subdirectory manifests. Instead, any
         subdirectory entry will be reported as it appears in the manifest, i.e.
         the subdirectory will be reported among files and distinguished only by
         its 't' flag.
         """
         store = self._storage()
         if shallow:
             r = store.rev(self._node)
             d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
             return manifestdict(d)
         else:
             # Need to perform a slow delta
             r0 = store.deltaparent(store.rev(self._node))
             m0 = self._manifestlog.get(self._dir, store.node(r0)).read()
             m1 = self.read()
             md = treemanifest(dir=self._dir)
             for f, ((n0, fl0), (n1, fl1)) in pycompat.iteritems(m0.diff(m1)):
                 if n1:
                     md[f] = n1
                     if fl1:
                         md.setflag(f, fl1)
             return md
 
     def readfast(self, shallow=False):
         """Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, it only returns the entries from this manifest,
         and not any submanifests.
         """
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if deltaparent != nullrev and deltaparent in store.parentrevs(r):
             return self.readdelta(shallow=shallow)
 
         if shallow:
             return manifestdict(store.revision(self._node))
         else:
             return self.read()
 
     def find(self, key):
         return self.read().find(key)
 
 
 class excludeddir(treemanifest):
     """Stand-in for a directory that is excluded from the repository.
 
     With narrowing active on a repository that uses treemanifests,
     some of the directory revlogs will be excluded from the resulting
     clone. This is a huge storage win for clients, but means we need
     some sort of pseudo-manifest to surface to internals so we can
     detect a merge conflict outside the narrowspec. That's what this
     class is: it stands in for a directory whose node is known, but
     whose contents are unknown.
     """
 
     def __init__(self, dir, node):
         super(excludeddir, self).__init__(dir)
         self._node = node
         # Add an empty file, which will be included by iterators and such,
         # appearing as the directory itself (i.e. something like "dir/")
         self._files[b''] = node
         self._flags[b''] = b't'
 
     # Manifests outside the narrowspec should never be modified, so avoid
     # copying. This makes a noticeable difference when there are very many
     # directories outside the narrowspec. Also, it makes sense for the copy to
     # be of the same type as the original, which would not happen with the
     # super type's copy().
     def copy(self):
         return self
 
 
 class excludeddirmanifestctx(treemanifestctx):
     """context wrapper for excludeddir - see that docstring for rationale"""
 
     def __init__(self, dir, node):
         self._dir = dir
         self._node = node
 
     def read(self):
         return excludeddir(self._dir, self._node)
 
     def readfast(self, shallow=False):
         # special version of readfast since we don't have underlying storage
         return self.read()
 
     def write(self, *args):
         raise error.ProgrammingError(
             b'attempt to write manifest from excluded dir %s' % self._dir
         )
 
 
 class excludedmanifestrevlog(manifestrevlog):
     """Stand-in for excluded treemanifest revlogs.
 
     When narrowing is active on a treemanifest repository, we'll have
     references to directories we can't see due to the revlog being
     skipped. This class exists to conform to the manifestrevlog
     interface for those directories and proactively prevent writes to
     outside the narrowspec.
     """
 
     def __init__(self, dir):
         self._dir = dir
 
     def __len__(self):
         raise error.ProgrammingError(
             b'attempt to get length of excluded dir %s' % self._dir
         )
 
     def rev(self, node):
         raise error.ProgrammingError(
             b'attempt to get rev from excluded dir %s' % self._dir
         )
 
     def linkrev(self, node):
         raise error.ProgrammingError(
             b'attempt to get linkrev from excluded dir %s' % self._dir
         )
 
     def node(self, rev):
         raise error.ProgrammingError(
             b'attempt to get node from excluded dir %s' % self._dir
         )
 
     def add(self, *args, **kwargs):
         # We should never write entries in dirlogs outside the narrow clone.
         # However, the method still gets called from writesubtree() in
         # _addtree(), so we need to handle it. We should possibly make that
         # avoid calling add() with a clean manifest (_dirty is always False
         # in excludeddir instances).
         pass

mercurial/revlog.py

@@ -1,3201 +1,3203 b''
 # revlog.py - storage back-end for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 """Storage back-end for Mercurial.
 
 This provides efficient delta storage with O(1) retrieve and append
 and O(changes) merge between branches.
 """
 
 from __future__ import absolute_import
 
 import collections
 import contextlib
 import errno
 import io
 import os
 import struct
 import zlib
 
 # import stuff from node for others to import from revlog
 from .node import (
     bin,
     hex,
     nullhex,
     nullid,
     nullrev,
     short,
     wdirfilenodeids,
     wdirhex,
     wdirid,
     wdirrev,
 )
 from .i18n import _
 from .pycompat import getattr
 from .revlogutils.constants import (
     FLAG_GENERALDELTA,
     FLAG_INLINE_DATA,
     REVLOGV0,
     REVLOGV1,
     REVLOGV1_FLAGS,
     REVLOGV2,
     REVLOGV2_FLAGS,
     REVLOG_DEFAULT_FLAGS,
     REVLOG_DEFAULT_FORMAT,
     REVLOG_DEFAULT_VERSION,
 )
 from .revlogutils.flagutil import (
     REVIDX_DEFAULT_FLAGS,
     REVIDX_ELLIPSIS,
     REVIDX_EXTSTORED,
     REVIDX_FLAGS_ORDER,
     REVIDX_HASCOPIESINFO,
     REVIDX_ISCENSORED,
     REVIDX_RAWTEXT_CHANGING_FLAGS,
     REVIDX_SIDEDATA,
 )
 from .thirdparty import attr
 from . import (
     ancestor,
     dagop,
     error,
     mdiff,
     policy,
     pycompat,
     templatefilters,
     util,
 )
 from .interfaces import (
     repository,
     util as interfaceutil,
 )
 from .revlogutils import (
     deltas as deltautil,
     flagutil,
     nodemap as nodemaputil,
     sidedata as sidedatautil,
 )
 from .utils import (
     storageutil,
     stringutil,
 )
 from .pure import parsers as pureparsers
 
 # blanked usage of all the name to prevent pyflakes constraints
 # We need these name available in the module for extensions.
 REVLOGV0
 REVLOGV1
 REVLOGV2
 FLAG_INLINE_DATA
 FLAG_GENERALDELTA
 REVLOG_DEFAULT_FLAGS
 REVLOG_DEFAULT_FORMAT
 REVLOG_DEFAULT_VERSION
 REVLOGV1_FLAGS
 REVLOGV2_FLAGS
 REVIDX_ISCENSORED
 REVIDX_ELLIPSIS
 REVIDX_SIDEDATA
 REVIDX_HASCOPIESINFO
 REVIDX_EXTSTORED
 REVIDX_DEFAULT_FLAGS
 REVIDX_FLAGS_ORDER
 REVIDX_RAWTEXT_CHANGING_FLAGS
 
 parsers = policy.importmod('parsers')
 rustancestor = policy.importrust('ancestor')
 rustdagop = policy.importrust('dagop')
 rustrevlog = policy.importrust('revlog')
 
 # Aliased for performance.
 _zlibdecompress = zlib.decompress
 
 # max size of revlog with inline data
 _maxinline = 131072
 _chunksize = 1048576
 
 # Flag processors for REVIDX_ELLIPSIS.
 def ellipsisreadprocessor(rl, text):
     return text, False
 
 
 def ellipsiswriteprocessor(rl, text):
     return text, False
 
 
 def ellipsisrawprocessor(rl, text):
     return False
 
 
 ellipsisprocessor = (
     ellipsisreadprocessor,
     ellipsiswriteprocessor,
     ellipsisrawprocessor,
 )
 
 
 def getoffset(q):
     return int(q >> 16)
 
 
 def gettype(q):
     return int(q & 0xFFFF)
 
 
 def offset_type(offset, type):
     if (type & ~flagutil.REVIDX_KNOWN_FLAGS) != 0:
         raise ValueError(b'unknown revlog index flags')
     return int(int(offset) << 16 | type)
 
 
 def _verify_revision(rl, skipflags, state, node):
     """Verify the integrity of the given revlog ``node`` while providing a hook
     point for extensions to influence the operation."""
     if skipflags:
         state[b'skipread'].add(node)
     else:
         # Side-effect: read content and verify hash.
         rl.revision(node)
 
 
 # True if a fast implementation for persistent-nodemap is available
 #
 # We also consider we have a "fast" implementation in "pure" python because
 # people using pure don't really have performance consideration (and a
 # wheelbarrow of other slowness source)
 HAS_FAST_PERSISTENT_NODEMAP = rustrevlog is not None or util.safehasattr(
     parsers, 'BaseIndexObject'
 )
 
 
 @attr.s(slots=True, frozen=True)
 class _revisioninfo(object):
     """Information about a revision that allows building its fulltext
     node:       expected hash of the revision
     p1, p2:     parent revs of the revision
     btext:      built text cache consisting of a one-element list
     cachedelta: (baserev, uncompressed_delta) or None
     flags:      flags associated to the revision storage
 
     One of btext[0] or cachedelta must be set.
     """
 
     node = attr.ib()
     p1 = attr.ib()
     p2 = attr.ib()
     btext = attr.ib()
     textlen = attr.ib()
     cachedelta = attr.ib()
     flags = attr.ib()
 
 
 @interfaceutil.implementer(repository.irevisiondelta)
 @attr.s(slots=True)
 class revlogrevisiondelta(object):
     node = attr.ib()
     p1node = attr.ib()
     p2node = attr.ib()
     basenode = attr.ib()
     flags = attr.ib()
     baserevisionsize = attr.ib()
     revision = attr.ib()
     delta = attr.ib()
     sidedata = attr.ib()
     linknode = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.iverifyproblem)
 @attr.s(frozen=True)
 class revlogproblem(object):
     warning = attr.ib(default=None)
     error = attr.ib(default=None)
     node = attr.ib(default=None)
 
 
 # index v0:
 #  4 bytes: offset
 #  4 bytes: compressed length
 #  4 bytes: base rev
 #  4 bytes: link rev
 # 20 bytes: parent 1 nodeid
 # 20 bytes: parent 2 nodeid
 # 20 bytes: nodeid
 indexformatv0 = struct.Struct(b">4l20s20s20s")
 indexformatv0_pack = indexformatv0.pack
 indexformatv0_unpack = indexformatv0.unpack
 
 
 class revlogoldindex(list):
     @property
     def nodemap(self):
         msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]"
         util.nouideprecwarn(msg, b'5.3', stacklevel=2)
         return self._nodemap
 
     @util.propertycache
     def _nodemap(self):
         nodemap = nodemaputil.NodeMap({nullid: nullrev})
         for r in range(0, len(self)):
             n = self[r][7]
             nodemap[n] = r
         return nodemap
 
     def has_node(self, node):
         """return True if the node exist in the index"""
         return node in self._nodemap
 
     def rev(self, node):
         """return a revision for a node
 
         If the node is unknown, raise a RevlogError"""
         return self._nodemap[node]
 
     def get_rev(self, node):
         """return a revision for a node
 
         If the node is unknown, return None"""
         return self._nodemap.get(node)
 
     def append(self, tup):
         self._nodemap[tup[7]] = len(self)
         super(revlogoldindex, self).append(tup)
 
     def __delitem__(self, i):
         if not isinstance(i, slice) or not i.stop == -1 or i.step is not None:
             raise ValueError(b"deleting slices only supports a:-1 with step 1")
         for r in pycompat.xrange(i.start, len(self)):
             del self._nodemap[self[r][7]]
         super(revlogoldindex, self).__delitem__(i)
 
     def clearcaches(self):
         self.__dict__.pop('_nodemap', None)
 
     def __getitem__(self, i):
         if i == -1:
             return (0, 0, 0, -1, -1, -1, -1, nullid)
         return list.__getitem__(self, i)
 
 
 class revlogoldio(object):
     def __init__(self):
         self.size = indexformatv0.size
 
     def parseindex(self, data, inline):
         s = self.size
         index = []
         nodemap = nodemaputil.NodeMap({nullid: nullrev})
         n = off = 0
         l = len(data)
         while off + s <= l:
             cur = data[off : off + s]
             off += s
             e = indexformatv0_unpack(cur)
             # transform to revlogv1 format
             e2 = (
                 offset_type(e[0], 0),
                 e[1],
                 -1,
                 e[2],
                 e[3],
                 nodemap.get(e[4], nullrev),
                 nodemap.get(e[5], nullrev),
                 e[6],
             )
             index.append(e2)
             nodemap[e[6]] = n
             n += 1
 
         index = revlogoldindex(index)
         return index, None
 
     def packentry(self, entry, node, version, rev):
         if gettype(entry[0]):
             raise error.RevlogError(
                 _(b'index entry flags need revlog version 1')
             )
         e2 = (
             getoffset(entry[0]),
             entry[1],
             entry[3],
             entry[4],
             node(entry[5]),
             node(entry[6]),
             entry[7],
         )
         return indexformatv0_pack(*e2)
 
 
 # index ng:
 #  6 bytes: offset
 #  2 bytes: flags
 #  4 bytes: compressed length
 #  4 bytes: uncompressed length
 #  4 bytes: base rev
 #  4 bytes: link rev
 #  4 bytes: parent 1 rev
 #  4 bytes: parent 2 rev
 # 32 bytes: nodeid
 indexformatng = struct.Struct(b">Qiiiiii20s12x")
 indexformatng_pack = indexformatng.pack
 versionformat = struct.Struct(b">I")
 versionformat_pack = versionformat.pack
 versionformat_unpack = versionformat.unpack
 
 # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte
 # signed integer)
 _maxentrysize = 0x7FFFFFFF
 
 
 class revlogio(object):
     def __init__(self):
         self.size = indexformatng.size
 
     def parseindex(self, data, inline):
         # call the C implementation to parse the index data
         index, cache = parsers.parse_index2(data, inline)
         return index, cache
 
     def packentry(self, entry, node, version, rev):
         p = indexformatng_pack(*entry)
         if rev == 0:
             p = versionformat_pack(version) + p[4:]
         return p
 
 
 indexformatv2 = struct.Struct(pureparsers.Index2Mixin.index_format)
 indexformatv2_pack = indexformatv2.pack
 
 
 class revlogv2io(object):
     def __init__(self):
         self.size = indexformatv2.size
 
     def parseindex(self, data, inline):
         index, cache = parsers.parse_index2(data, inline, revlogv2=True)
         return index, cache
 
     def packentry(self, entry, node, version, rev):
         p = indexformatv2_pack(*entry)
         if rev == 0:
             p = versionformat_pack(version) + p[4:]
         return p
 
 
 NodemapRevlogIO = None
 
 if util.safehasattr(parsers, 'parse_index_devel_nodemap'):
 
     class NodemapRevlogIO(revlogio):
         """A debug oriented IO class that return a PersistentNodeMapIndexObject
 
         The PersistentNodeMapIndexObject object is meant to test the persistent nodemap feature.
         """
 
         def parseindex(self, data, inline):
             index, cache = parsers.parse_index_devel_nodemap(data, inline)
             return index, cache
 
 
 class rustrevlogio(revlogio):
     def parseindex(self, data, inline):
         index, cache = super(rustrevlogio, self).parseindex(data, inline)
         return rustrevlog.MixedIndex(index), cache
 
 
 class revlog(object):
     """
     the underlying revision storage object
 
     A revlog consists of two parts, an index and the revision data.
 
     The index is a file with a fixed record size containing
     information on each revision, including its nodeid (hash), the
     nodeids of its parents, the position and offset of its data within
     the data file, and the revision it's based on. Finally, each entry
     contains a linkrev entry that can serve as a pointer to external
     data.
 
     The revision data itself is a linear collection of data chunks.
     Each chunk represents a revision and is usually represented as a
     delta against the previous chunk. To bound lookup time, runs of
     deltas are limited to about 2 times the length of the original
     version data. This makes retrieval of a version proportional to
     its size, or O(1) relative to the number of revisions.
 
     Both pieces of the revlog are written to in an append-only
     fashion, which means we never need to rewrite a file to insert or
     remove data, and can use some simple techniques to avoid the need
     for locking while reading.
 
     If checkambig, indexfile is opened with checkambig=True at
     writing, to avoid file stat ambiguity.
 
     If mmaplargeindex is True, and an mmapindexthreshold is set, the
     index will be mmapped rather than read if it is larger than the
     configured threshold.
 
     If censorable is True, the revlog can have censored revisions.
 
     If `upperboundcomp` is not None, this is the expected maximal gain from
     compression for the data content.
 
     `concurrencychecker` is an optional function that receives 3 arguments: a
     file handle, a filename, and an expected position. It should check whether
     the current position in the file handle is valid, and log/warn/fail (by
     raising).
     """
 
     _flagserrorclass = error.RevlogError
 
     def __init__(
         self,
         opener,
         indexfile,
         datafile=None,
         checkambig=False,
         mmaplargeindex=False,
         censorable=False,
         upperboundcomp=None,
         persistentnodemap=False,
         concurrencychecker=None,
     ):
         """
         create a revlog object
 
         opener is a function that abstracts the file opening operation
         and can be used to implement COW semantics or the like.
 
         """
         self.upperboundcomp = upperboundcomp
         self.indexfile = indexfile
         self.datafile = datafile or (indexfile[:-2] + b".d")
         self.nodemap_file = None
         if persistentnodemap:
             self.nodemap_file = nodemaputil.get_nodemap_file(
                 opener, self.indexfile
             )
 
         self.opener = opener
         #  When True, indexfile is opened with checkambig=True at writing, to
         #  avoid file stat ambiguity.
         self._checkambig = checkambig
         self._mmaplargeindex = mmaplargeindex
         self._censorable = censorable
         # 3-tuple of (node, rev, text) for a raw revision.
         self._revisioncache = None
         # Maps rev to chain base rev.
         self._chainbasecache = util.lrucachedict(100)
         # 2-tuple of (offset, data) of raw data from the revlog at an offset.
         self._chunkcache = (0, b'')
         # How much data to read and cache into the raw revlog data cache.
         self._chunkcachesize = 65536
         self._maxchainlen = None
         self._deltabothparents = True
         self.index = None
         self._nodemap_docket = None
         # Mapping of partial identifiers to full nodes.
         self._pcache = {}
         # Mapping of revision integer to full node.
         self._compengine = b'zlib'
         self._compengineopts = {}
         self._maxdeltachainspan = -1
         self._withsparseread = False
         self._sparserevlog = False
         self._srdensitythreshold = 0.50
         self._srmingapsize = 262144
 
         # Make copy of flag processors so each revlog instance can support
         # custom flags.
         self._flagprocessors = dict(flagutil.flagprocessors)
 
         # 2-tuple of file handles being used for active writing.
         self._writinghandles = None
 
         self._loadindex()
 
         self._concurrencychecker = concurrencychecker
 
     def _loadindex(self):
         mmapindexthreshold = None
         opts = self.opener.options
 
         if b'revlogv2' in opts:
             newversionflags = REVLOGV2 | FLAG_INLINE_DATA
         elif b'revlogv1' in opts:
             newversionflags = REVLOGV1 | FLAG_INLINE_DATA
             if b'generaldelta' in opts:
                 newversionflags |= FLAG_GENERALDELTA
         elif b'revlogv0' in self.opener.options:
             newversionflags = REVLOGV0
         else:
             newversionflags = REVLOG_DEFAULT_VERSION
 
         if b'chunkcachesize' in opts:
             self._chunkcachesize = opts[b'chunkcachesize']
         if b'maxchainlen' in opts:
             self._maxchainlen = opts[b'maxchainlen']
         if b'deltabothparents' in opts:
             self._deltabothparents = opts[b'deltabothparents']
         self._lazydelta = bool(opts.get(b'lazydelta', True))
         self._lazydeltabase = False
         if self._lazydelta:
             self._lazydeltabase = bool(opts.get(b'lazydeltabase', False))
         if b'compengine' in opts:
             self._compengine = opts[b'compengine']
         if b'zlib.level' in opts:
             self._compengineopts[b'zlib.level'] = opts[b'zlib.level']
         if b'zstd.level' in opts:
             self._compengineopts[b'zstd.level'] = opts[b'zstd.level']
         if b'maxdeltachainspan' in opts:
             self._maxdeltachainspan = opts[b'maxdeltachainspan']
         if self._mmaplargeindex and b'mmapindexthreshold' in opts:
             mmapindexthreshold = opts[b'mmapindexthreshold']
         self.hassidedata = bool(opts.get(b'side-data', False))
         self._sparserevlog = bool(opts.get(b'sparse-revlog', False))
         withsparseread = bool(opts.get(b'with-sparse-read', False))
         # sparse-revlog forces sparse-read
         self._withsparseread = self._sparserevlog or withsparseread
         if b'sparse-read-density-threshold' in opts:
             self._srdensitythreshold = opts[b'sparse-read-density-threshold']
         if b'sparse-read-min-gap-size' in opts:
             self._srmingapsize = opts[b'sparse-read-min-gap-size']
         if opts.get(b'enableellipsis'):
             self._flagprocessors[REVIDX_ELLIPSIS] = ellipsisprocessor
 
         # revlog v0 doesn't have flag processors
         for flag, processor in pycompat.iteritems(
             opts.get(b'flagprocessors', {})
         ):
             flagutil.insertflagprocessor(flag, processor, self._flagprocessors)
 
         if self._chunkcachesize <= 0:
             raise error.RevlogError(
                 _(b'revlog chunk cache size %r is not greater than 0')
                 % self._chunkcachesize
             )
         elif self._chunkcachesize & (self._chunkcachesize - 1):
             raise error.RevlogError(
                 _(b'revlog chunk cache size %r is not a power of 2')
                 % self._chunkcachesize
             )
 
         indexdata = b''
         self._initempty = True
         try:
             with self._indexfp() as f:
                 if (
                     mmapindexthreshold is not None
                     and self.opener.fstat(f).st_size >= mmapindexthreshold
                 ):
                     # TODO: should .close() to release resources without
                     # relying on Python GC
                     indexdata = util.buffer(util.mmapread(f))
                 else:
                     indexdata = f.read()
             if len(indexdata) > 0:
                 versionflags = versionformat_unpack(indexdata[:4])[0]
                 self._initempty = False
             else:
                 versionflags = newversionflags
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
 
             versionflags = newversionflags
 
         self.version = versionflags
 
         flags = versionflags & ~0xFFFF
         fmt = versionflags & 0xFFFF
 
         if fmt == REVLOGV0:
             if flags:
                 raise error.RevlogError(
                     _(b'unknown flags (%#04x) in version %d revlog %s')
                     % (flags >> 16, fmt, self.indexfile)
                 )
 
             self._inline = False
             self._generaldelta = False
 
         elif fmt == REVLOGV1:
             if flags & ~REVLOGV1_FLAGS:
                 raise error.RevlogError(
                     _(b'unknown flags (%#04x) in version %d revlog %s')
                     % (flags >> 16, fmt, self.indexfile)
                 )
 
             self._inline = versionflags & FLAG_INLINE_DATA
             self._generaldelta = versionflags & FLAG_GENERALDELTA
 
         elif fmt == REVLOGV2:
             if flags & ~REVLOGV2_FLAGS:
                 raise error.RevlogError(
                     _(b'unknown flags (%#04x) in version %d revlog %s')
                     % (flags >> 16, fmt, self.indexfile)
                 )
 
             self._inline = versionflags & FLAG_INLINE_DATA
             # generaldelta implied by version 2 revlogs.
             self._generaldelta = True
 
         else:
             raise error.RevlogError(
                 _(b'unknown version (%d) in revlog %s') % (fmt, self.indexfile)
             )
         # sparse-revlog can't be on without general-delta (issue6056)
         if not self._generaldelta:
             self._sparserevlog = False
 
         self._storedeltachains = True
 
         devel_nodemap = (
             self.nodemap_file
             and opts.get(b'devel-force-nodemap', False)
             and NodemapRevlogIO is not None
         )
 
         use_rust_index = False
         if rustrevlog is not None:
             if self.nodemap_file is not None:
                 use_rust_index = True
             else:
                 use_rust_index = self.opener.options.get(b'rust.index')
 
         self._io = revlogio()
         if self.version == REVLOGV0:
             self._io = revlogoldio()
         elif fmt == REVLOGV2:
             self._io = revlogv2io()
         elif devel_nodemap:
             self._io = NodemapRevlogIO()
         elif use_rust_index:
             self._io = rustrevlogio()
         try:
             d = self._io.parseindex(indexdata, self._inline)
             index, _chunkcache = d
             use_nodemap = (
                 not self._inline
                 and self.nodemap_file is not None
                 and util.safehasattr(index, 'update_nodemap_data')
             )
             if use_nodemap:
                 nodemap_data = nodemaputil.persisted_data(self)
                 if nodemap_data is not None:
                     docket = nodemap_data[0]
                     if (
                         len(d[0]) > docket.tip_rev
                         and d[0][docket.tip_rev][7] == docket.tip_node
                     ):
                         # no changelog tampering
                         self._nodemap_docket = docket
                         index.update_nodemap_data(*nodemap_data)
         except (ValueError, IndexError):
             raise error.RevlogError(
                 _(b"index %s is corrupted") % self.indexfile
             )
         self.index, self._chunkcache = d
         if not self._chunkcache:
             self._chunkclear()
         # revnum -> (chain-length, sum-delta-length)
         self._chaininfocache = util.lrucachedict(500)
         # revlog header -> revlog compressor
         self._decompressors = {}
 
     @util.propertycache
     def _compressor(self):
         engine = util.compengines[self._compengine]
         return engine.revlogcompressor(self._compengineopts)
 
     def _indexfp(self, mode=b'r'):
         """file object for the revlog's index file"""
         args = {'mode': mode}
         if mode != b'r':
             args['checkambig'] = self._checkambig
         if mode == b'w':
             args['atomictemp'] = True
         return self.opener(self.indexfile, **args)
 
     def _datafp(self, mode=b'r'):
         """file object for the revlog's data file"""
         return self.opener(self.datafile, mode=mode)
 
     @contextlib.contextmanager
     def _datareadfp(self, existingfp=None):
         """file object suitable to read data"""
         # Use explicit file handle, if given.
         if existingfp is not None:
             yield existingfp
 
         # Use a file handle being actively used for writes, if available.
         # There is some danger to doing this because reads will seek the
         # file. However, _writeentry() performs a SEEK_END before all writes,
         # so we should be safe.
         elif self._writinghandles:
             if self._inline:
                 yield self._writinghandles[0]
             else:
                 yield self._writinghandles[1]
 
         # Otherwise open a new file handle.
         else:
             if self._inline:
                 func = self._indexfp
             else:
                 func = self._datafp
             with func() as fp:
                 yield fp
 
     def tiprev(self):
         return len(self.index) - 1
 
     def tip(self):
         return self.node(self.tiprev())
 
     def __contains__(self, rev):
         return 0 <= rev < len(self)
 
     def __len__(self):
         return len(self.index)
 
     def __iter__(self):
         return iter(pycompat.xrange(len(self)))
 
     def revs(self, start=0, stop=None):
         """iterate over all rev in this revlog (from start to stop)"""
         return storageutil.iterrevs(len(self), start=start, stop=stop)
 
     @property
     def nodemap(self):
         msg = (
             b"revlog.nodemap is deprecated, "
             b"use revlog.index.[has_node|rev|get_rev]"
         )
         util.nouideprecwarn(msg, b'5.3', stacklevel=2)
         return self.index.nodemap
 
     @property
     def _nodecache(self):
         msg = b"revlog._nodecache is deprecated, use revlog.index.nodemap"
         util.nouideprecwarn(msg, b'5.3', stacklevel=2)
         return self.index.nodemap
 
     def hasnode(self, node):
         try:
             self.rev(node)
             return True
         except KeyError:
             return False
 
     def candelta(self, baserev, rev):
         """whether two revisions (baserev, rev) can be delta-ed or not"""
         # Disable delta if either rev requires a content-changing flag
         # processor (ex. LFS). This is because such flag processor can alter
         # the rawtext content that the delta will be based on, and two clients
         # could have a same revlog node with different flags (i.e. different
         # rawtext contents) and the delta could be incompatible.
         if (self.flags(baserev) & REVIDX_RAWTEXT_CHANGING_FLAGS) or (
             self.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS
         ):
             return False
         return True
 
     def update_caches(self, transaction):
         if self.nodemap_file is not None:
             if transaction is None:
                 nodemaputil.update_persistent_nodemap(self)
             else:
                 nodemaputil.setup_persistent_nodemap(transaction, self)
 
     def clearcaches(self):
         self._revisioncache = None
         self._chainbasecache.clear()
         self._chunkcache = (0, b'')
         self._pcache = {}
         self._nodemap_docket = None
         self.index.clearcaches()
         # The python code is the one responsible for validating the docket, we
         # end up having to refresh it here.
         use_nodemap = (
             not self._inline
             and self.nodemap_file is not None
             and util.safehasattr(self.index, 'update_nodemap_data')
         )
         if use_nodemap:
             nodemap_data = nodemaputil.persisted_data(self)
             if nodemap_data is not None:
                 self._nodemap_docket = nodemap_data[0]
                 self.index.update_nodemap_data(*nodemap_data)
 
     def rev(self, node):
         try:
             return self.index.rev(node)
         except TypeError:
             raise
         except error.RevlogError:
             # parsers.c radix tree lookup failed
             if node == wdirid or node in wdirfilenodeids:
                 raise error.WdirUnsupported
             raise error.LookupError(node, self.indexfile, _(b'no node'))
 
     # Accessors for index entries.
 
     # First tuple entry is 8 bytes. First 6 bytes are offset. Last 2 bytes
     # are flags.
     def start(self, rev):
         return int(self.index[rev][0] >> 16)
 
     def flags(self, rev):
         return self.index[rev][0] & 0xFFFF
 
     def length(self, rev):
         return self.index[rev][1]
 
     def sidedata_length(self, rev):
         if self.version & 0xFFFF != REVLOGV2:
             return 0
         return self.index[rev][9]
 
     def rawsize(self, rev):
         """return the length of the uncompressed text for a given revision"""
         l = self.index[rev][2]
         if l >= 0:
             return l
 
         t = self.rawdata(rev)
         return len(t)
 
     def size(self, rev):
         """length of non-raw text (processed by a "read" flag processor)"""
         # fast path: if no "read" flag processor could change the content,
         # size is rawsize. note: ELLIPSIS is known to not change the content.
         flags = self.flags(rev)
         if flags & (flagutil.REVIDX_KNOWN_FLAGS ^ REVIDX_ELLIPSIS) == 0:
             return self.rawsize(rev)
 
         return len(self.revision(rev, raw=False))
 
     def chainbase(self, rev):
         base = self._chainbasecache.get(rev)
         if base is not None:
             return base
 
         index = self.index
         iterrev = rev
         base = index[iterrev][3]
         while base != iterrev:
             iterrev = base
             base = index[iterrev][3]
 
         self._chainbasecache[rev] = base
         return base
 
     def linkrev(self, rev):
         return self.index[rev][4]
 
     def parentrevs(self, rev):
         try:
             entry = self.index[rev]
         except IndexError:
             if rev == wdirrev:
                 raise error.WdirUnsupported
             raise
 
         return entry[5], entry[6]
 
     # fast parentrevs(rev) where rev isn't filtered
     _uncheckedparentrevs = parentrevs
 
     def node(self, rev):
         try:
             return self.index[rev][7]
         except IndexError:
             if rev == wdirrev:
                 raise error.WdirUnsupported
             raise
 
     # Derived from index values.
 
     def end(self, rev):
         return self.start(rev) + self.length(rev)
 
     def parents(self, node):
         i = self.index
         d = i[self.rev(node)]
         return i[d[5]][7], i[d[6]][7]  # map revisions to nodes inline
 
     def chainlen(self, rev):
         return self._chaininfo(rev)[0]
 
     def _chaininfo(self, rev):
         chaininfocache = self._chaininfocache
         if rev in chaininfocache:
             return chaininfocache[rev]
         index = self.index
         generaldelta = self._generaldelta
         iterrev = rev
         e = index[iterrev]
         clen = 0
         compresseddeltalen = 0
         while iterrev != e[3]:
             clen += 1
             compresseddeltalen += e[1]
             if generaldelta:
                 iterrev = e[3]
             else:
                 iterrev -= 1
             if iterrev in chaininfocache:
                 t = chaininfocache[iterrev]
                 clen += t[0]
                 compresseddeltalen += t[1]
                 break
             e = index[iterrev]
         else:
             # Add text length of base since decompressing that also takes
             # work. For cache hits the length is already included.
             compresseddeltalen += e[1]
         r = (clen, compresseddeltalen)
         chaininfocache[rev] = r
         return r
 
     def _deltachain(self, rev, stoprev=None):
         """Obtain the delta chain for a revision.
 
         ``stoprev`` specifies a revision to stop at. If not specified, we
         stop at the base of the chain.
 
         Returns a 2-tuple of (chain, stopped) where ``chain`` is a list of
         revs in ascending order and ``stopped`` is a bool indicating whether
         ``stoprev`` was hit.
         """
         # Try C implementation.
         try:
             return self.index.deltachain(rev, stoprev, self._generaldelta)
         except AttributeError:
             pass
 
         chain = []
 
         # Alias to prevent attribute lookup in tight loop.
         index = self.index
         generaldelta = self._generaldelta
 
         iterrev = rev
         e = index[iterrev]
         while iterrev != e[3] and iterrev != stoprev:
             chain.append(iterrev)
             if generaldelta:
                 iterrev = e[3]
             else:
                 iterrev -= 1
             e = index[iterrev]
 
         if iterrev == stoprev:
             stopped = True
         else:
             chain.append(iterrev)
             stopped = False
 
         chain.reverse()
         return chain, stopped
 
     def ancestors(self, revs, stoprev=0, inclusive=False):
         """Generate the ancestors of 'revs' in reverse revision order.
         Does not generate revs lower than stoprev.
 
         See the documentation for ancestor.lazyancestors for more details."""
 
         # first, make sure start revisions aren't filtered
         revs = list(revs)
         checkrev = self.node
         for r in revs:
             checkrev(r)
         # and we're sure ancestors aren't filtered as well
 
         if rustancestor is not None:
             lazyancestors = rustancestor.LazyAncestors
             arg = self.index
         else:
             lazyancestors = ancestor.lazyancestors
             arg = self._uncheckedparentrevs
         return lazyancestors(arg, revs, stoprev=stoprev, inclusive=inclusive)
 
     def descendants(self, revs):
         return dagop.descendantrevs(revs, self.revs, self.parentrevs)
 
     def findcommonmissing(self, common=None, heads=None):
         """Return a tuple of the ancestors of common and the ancestors of heads
         that are not ancestors of common. In revset terminology, we return the
         tuple:
 
           ::common, (::heads) - (::common)
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of node IDs.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullid]
         if heads is None:
             heads = self.heads()
 
         common = [self.rev(n) for n in common]
         heads = [self.rev(n) for n in heads]
 
         # we want the ancestors, but inclusive
         class lazyset(object):
             def __init__(self, lazyvalues):
                 self.addedvalues = set()
                 self.lazyvalues = lazyvalues
 
             def __contains__(self, value):
                 return value in self.addedvalues or value in self.lazyvalues
 
             def __iter__(self):
                 added = self.addedvalues
                 for r in added:
                     yield r
                 for r in self.lazyvalues:
                     if not r in added:
                         yield r
 
             def add(self, value):
                 self.addedvalues.add(value)
 
             def update(self, values):
                 self.addedvalues.update(values)
 
         has = lazyset(self.ancestors(common))
         has.add(nullrev)
         has.update(common)
 
         # take all ancestors from heads that aren't in has
         missing = set()
         visit = collections.deque(r for r in heads if r not in has)
         while visit:
             r = visit.popleft()
             if r in missing:
                 continue
             else:
                 missing.add(r)
                 for p in self.parentrevs(r):
                     if p not in has:
                         visit.append(p)
         missing = list(missing)
         missing.sort()
         return has, [self.node(miss) for miss in missing]
 
     def incrementalmissingrevs(self, common=None):
         """Return an object that can be used to incrementally compute the
         revision numbers of the ancestors of arbitrary sets that are not
         ancestors of common. This is an ancestor.incrementalmissingancestors
         object.
 
         'common' is a list of revision numbers. If common is not supplied, uses
         nullrev.
         """
         if common is None:
             common = [nullrev]
 
         if rustancestor is not None:
             return rustancestor.MissingAncestors(self.index, common)
         return ancestor.incrementalmissingancestors(self.parentrevs, common)
 
     def findmissingrevs(self, common=None, heads=None):
         """Return the revision numbers of the ancestors of heads that
         are not ancestors of common.
 
         More specifically, return a list of revision numbers corresponding to
         nodes N such that every N satisfies the following constraints:
 
           1. N is an ancestor of some node in 'heads'
           2. N is not an ancestor of any node in 'common'
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of revision numbers.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullrev]
         if heads is None:
             heads = self.headrevs()
 
         inc = self.incrementalmissingrevs(common=common)
         return inc.missingancestors(heads)
 
     def findmissing(self, common=None, heads=None):
         """Return the ancestors of heads that are not ancestors of common.
 
         More specifically, return a list of nodes N such that every N
         satisfies the following constraints:
 
           1. N is an ancestor of some node in 'heads'
           2. N is not an ancestor of any node in 'common'
 
         The list is sorted by revision number, meaning it is
         topologically sorted.
 
         'heads' and 'common' are both lists of node IDs.  If heads is
         not supplied, uses all of the revlog's heads.  If common is not
         supplied, uses nullid."""
         if common is None:
             common = [nullid]
         if heads is None:
             heads = self.heads()
 
         common = [self.rev(n) for n in common]
         heads = [self.rev(n) for n in heads]
 
         inc = self.incrementalmissingrevs(common=common)
         return [self.node(r) for r in inc.missingancestors(heads)]
 
     def nodesbetween(self, roots=None, heads=None):
         """Return a topological path from 'roots' to 'heads'.
 
         Return a tuple (nodes, outroots, outheads) where 'nodes' is a
         topologically sorted list of all nodes N that satisfy both of
         these constraints:
 
           1. N is a descendant of some node in 'roots'
           2. N is an ancestor of some node in 'heads'
 
         Every node is considered to be both a descendant and an ancestor
         of itself, so every reachable node in 'roots' and 'heads' will be
         included in 'nodes'.
 
         'outroots' is the list of reachable nodes in 'roots', i.e., the
         subset of 'roots' that is returned in 'nodes'.  Likewise,
         'outheads' is the subset of 'heads' that is also in 'nodes'.
 
         'roots' and 'heads' are both lists of node IDs.  If 'roots' is
         unspecified, uses nullid as the only root.  If 'heads' is
         unspecified, uses list of all of the revlog's heads."""
         nonodes = ([], [], [])
         if roots is not None:
             roots = list(roots)
             if not roots:
                 return nonodes
             lowestrev = min([self.rev(n) for n in roots])
         else:
             roots = [nullid]  # Everybody's a descendant of nullid
             lowestrev = nullrev
         if (lowestrev == nullrev) and (heads is None):
             # We want _all_ the nodes!
             return ([self.node(r) for r in self], [nullid], list(self.heads()))
         if heads is None:
             # All nodes are ancestors, so the latest ancestor is the last
             # node.
             highestrev = len(self) - 1
             # Set ancestors to None to signal that every node is an ancestor.
             ancestors = None
             # Set heads to an empty dictionary for later discovery of heads
             heads = {}
         else:
             heads = list(heads)
             if not heads:
                 return nonodes
             ancestors = set()
             # Turn heads into a dictionary so we can remove 'fake' heads.
             # Also, later we will be using it to filter out the heads we can't
             # find from roots.
             heads = dict.fromkeys(heads, False)
             # Start at the top and keep marking parents until we're done.
             nodestotag = set(heads)
             # Remember where the top was so we can use it as a limit later.
             highestrev = max([self.rev(n) for n in nodestotag])
             while nodestotag:
                 # grab a node to tag
                 n = nodestotag.pop()
                 # Never tag nullid
                 if n == nullid:
                     continue
                 # A node's revision number represents its place in a
                 # topologically sorted list of nodes.
                 r = self.rev(n)
                 if r >= lowestrev:
                     if n not in ancestors:
                         # If we are possibly a descendant of one of the roots
                         # and we haven't already been marked as an ancestor
                         ancestors.add(n)  # Mark as ancestor
                         # Add non-nullid parents to list of nodes to tag.
                         nodestotag.update(
                             [p for p in self.parents(n) if p != nullid]
                         )
                     elif n in heads:  # We've seen it before, is it a fake head?
                         # So it is, real heads should not be the ancestors of
                         # any other heads.
                         heads.pop(n)
             if not ancestors:
                 return nonodes
             # Now that we have our set of ancestors, we want to remove any
             # roots that are not ancestors.
 
             # If one of the roots was nullid, everything is included anyway.
             if lowestrev > nullrev:
                 # But, since we weren't, let's recompute the lowest rev to not
                 # include roots that aren't ancestors.
 
                 # Filter out roots that aren't ancestors of heads
                 roots = [root for root in roots if root in ancestors]
                 # Recompute the lowest revision
                 if roots:
                     lowestrev = min([self.rev(root) for root in roots])
                 else:
                     # No more roots?  Return empty list
                     return nonodes
             else:
                 # We are descending from nullid, and don't need to care about
                 # any other roots.
                 lowestrev = nullrev
                 roots = [nullid]
         # Transform our roots list into a set.
         descendants = set(roots)
         # Also, keep the original roots so we can filter out roots that aren't
         # 'real' roots (i.e. are descended from other roots).
         roots = descendants.copy()
         # Our topologically sorted list of output nodes.
         orderedout = []
         # Don't start at nullid since we don't want nullid in our output list,
         # and if nullid shows up in descendants, empty parents will look like
         # they're descendants.
         for r in self.revs(start=max(lowestrev, 0), stop=highestrev + 1):
             n = self.node(r)
             isdescendant = False
             if lowestrev == nullrev:  # Everybody is a descendant of nullid
                 isdescendant = True
             elif n in descendants:
                 # n is already a descendant
                 isdescendant = True
                 # This check only needs to be done here because all the roots
                 # will start being marked is descendants before the loop.
                 if n in roots:
                     # If n was a root, check if it's a 'real' root.
                     p = tuple(self.parents(n))
                     # If any of its parents are descendants, it's not a root.
                     if (p[0] in descendants) or (p[1] in descendants):
                         roots.remove(n)
             else:
                 p = tuple(self.parents(n))
                 # A node is a descendant if either of its parents are
                 # descendants.  (We seeded the dependents list with the roots
                 # up there, remember?)
                 if (p[0] in descendants) or (p[1] in descendants):
                     descendants.add(n)
                     isdescendant = True
             if isdescendant and ((ancestors is None) or (n in ancestors)):
                 # Only include nodes that are both descendants and ancestors.
                 orderedout.append(n)
                 if (ancestors is not None) and (n in heads):
                     # We're trying to figure out which heads are reachable
                     # from roots.
                     # Mark this head as having been reached
                     heads[n] = True
                 elif ancestors is None:
                     # Otherwise, we're trying to discover the heads.
                     # Assume this is a head because if it isn't, the next step
                     # will eventually remove it.
                     heads[n] = True
                     # But, obviously its parents aren't.
                     for p in self.parents(n):
                         heads.pop(p, None)
         heads = [head for head, flag in pycompat.iteritems(heads) if flag]
         roots = list(roots)
         assert orderedout
         assert roots
         assert heads
         return (orderedout, roots, heads)
 
     def headrevs(self, revs=None):
         if revs is None:
             try:
                 return self.index.headrevs()
             except AttributeError:
                 return self._headrevs()
         if rustdagop is not None:
             return rustdagop.headrevs(self.index, revs)
         return dagop.headrevs(revs, self._uncheckedparentrevs)
 
     def computephases(self, roots):
         return self.index.computephasesmapsets(roots)
 
     def _headrevs(self):
         count = len(self)
         if not count:
             return [nullrev]
         # we won't iter over filtered rev so nobody is a head at start
         ishead = [0] * (count + 1)
         index = self.index
         for r in self:
             ishead[r] = 1  # I may be an head
             e = index[r]
             ishead[e[5]] = ishead[e[6]] = 0  # my parent are not
         return [r for r, val in enumerate(ishead) if val]
 
     def heads(self, start=None, stop=None):
         """return the list of all nodes that have no children
 
         if start is specified, only heads that are descendants of
         start will be returned
         if stop is specified, it will consider all the revs from stop
         as if they had no children
         """
         if start is None and stop is None:
             if not len(self):
                 return [nullid]
             return [self.node(r) for r in self.headrevs()]
 
         if start is None:
             start = nullrev
         else:
             start = self.rev(start)
 
         stoprevs = {self.rev(n) for n in stop or []}
 
         revs = dagop.headrevssubset(
             self.revs, self.parentrevs, startrev=start, stoprevs=stoprevs
         )
 
         return [self.node(rev) for rev in revs]
 
     def children(self, node):
         """find the children of a given node"""
         c = []
         p = self.rev(node)
         for r in self.revs(start=p + 1):
             prevs = [pr for pr in self.parentrevs(r) if pr != nullrev]
             if prevs:
                 for pr in prevs:
                     if pr == p:
                         c.append(self.node(r))
             elif p == nullrev:
                 c.append(self.node(r))
         return c
 
     def commonancestorsheads(self, a, b):
         """calculate all the heads of the common ancestors of nodes a and b"""
         a, b = self.rev(a), self.rev(b)
         ancs = self._commonancestorsheads(a, b)
         return pycompat.maplist(self.node, ancs)
 
     def _commonancestorsheads(self, *revs):
         """calculate all the heads of the common ancestors of revs"""
         try:
             ancs = self.index.commonancestorsheads(*revs)
         except (AttributeError, OverflowError):  # C implementation failed
             ancs = ancestor.commonancestorsheads(self.parentrevs, *revs)
         return ancs
 
     def isancestor(self, a, b):
         """return True if node a is an ancestor of node b
 
         A revision is considered an ancestor of itself."""
         a, b = self.rev(a), self.rev(b)
         return self.isancestorrev(a, b)
 
     def isancestorrev(self, a, b):
         """return True if revision a is an ancestor of revision b
 
         A revision is considered an ancestor of itself.
 
         The implementation of this is trivial but the use of
         reachableroots is not."""
         if a == nullrev:
             return True
         elif a == b:
             return True
         elif a > b:
             return False
         return bool(self.reachableroots(a, [b], [a], includepath=False))
 
     def reachableroots(self, minroot, heads, roots, includepath=False):
         """return (heads(::(<roots> and <roots>::<heads>)))
 
         If includepath is True, return (<roots>::<heads>)."""
         try:
             return self.index.reachableroots2(
                 minroot, heads, roots, includepath
             )
         except AttributeError:
             return dagop._reachablerootspure(
                 self.parentrevs, minroot, roots, heads, includepath
             )
 
     def ancestor(self, a, b):
         """calculate the "best" common ancestor of nodes a and b"""
 
         a, b = self.rev(a), self.rev(b)
         try:
             ancs = self.index.ancestors(a, b)
         except (AttributeError, OverflowError):
             ancs = ancestor.ancestors(self.parentrevs, a, b)
         if ancs:
             # choose a consistent winner when there's a tie
             return min(map(self.node, ancs))
         return nullid
 
     def _match(self, id):
         if isinstance(id, int):
             # rev
             return self.node(id)
         if len(id) == 20:
             # possibly a binary node
             # odds of a binary node being all hex in ASCII are 1 in 10**25
             try:
                 node = id
                 self.rev(node)  # quick search the index
                 return node
             except error.LookupError:
                 pass  # may be partial hex id
         try:
             # str(rev)
             rev = int(id)
             if b"%d" % rev != id:
                 raise ValueError
             if rev < 0:
                 rev = len(self) + rev
             if rev < 0 or rev >= len(self):
                 raise ValueError
             return self.node(rev)
         except (ValueError, OverflowError):
             pass
         if len(id) == 40:
             try:
                 # a full hex nodeid?
                 node = bin(id)
                 self.rev(node)
                 return node
             except (TypeError, error.LookupError):
                 pass
 
     def _partialmatch(self, id):
         # we don't care wdirfilenodeids as they should be always full hash
         maybewdir = wdirhex.startswith(id)
         try:
             partial = self.index.partialmatch(id)
             if partial and self.hasnode(partial):
                 if maybewdir:
                     # single 'ff...' match in radix tree, ambiguous with wdir
                     raise error.RevlogError
                 return partial
             if maybewdir:
                 # no 'ff...' match in radix tree, wdir identified
                 raise error.WdirUnsupported
             return None
         except error.RevlogError:
             # parsers.c radix tree lookup gave multiple matches
             # fast path: for unfiltered changelog, radix tree is accurate
             if not getattr(self, 'filteredrevs', None):
                 raise error.AmbiguousPrefixLookupError(
                     id, self.indexfile, _(b'ambiguous identifier')
                 )
             # fall through to slow path that filters hidden revisions
         except (AttributeError, ValueError):
             # we are pure python, or key was too short to search radix tree
             pass
 
         if id in self._pcache:
             return self._pcache[id]
 
         if len(id) <= 40:
             try:
                 # hex(node)[:...]
                 l = len(id) // 2  # grab an even number of digits
                 prefix = bin(id[: l * 2])
                 nl = [e[7] for e in self.index if e[7].startswith(prefix)]
                 nl = [
                     n for n in nl if hex(n).startswith(id) and self.hasnode(n)
                 ]
                 if nullhex.startswith(id):
                     nl.append(nullid)
                 if len(nl) > 0:
                     if len(nl) == 1 and not maybewdir:
                         self._pcache[id] = nl[0]
                         return nl[0]
                     raise error.AmbiguousPrefixLookupError(
                         id, self.indexfile, _(b'ambiguous identifier')
                     )
                 if maybewdir:
                     raise error.WdirUnsupported
                 return None
             except TypeError:
                 pass
 
     def lookup(self, id):
         """locate a node based on:
         - revision number or str(revision number)
         - nodeid or subset of hex nodeid
         """
         n = self._match(id)
         if n is not None:
             return n
         n = self._partialmatch(id)
         if n:
             return n
 
         raise error.LookupError(id, self.indexfile, _(b'no match found'))
 
     def shortest(self, node, minlength=1):
         """Find the shortest unambiguous prefix that matches node."""
 
         def isvalid(prefix):
             try:
                 matchednode = self._partialmatch(prefix)
             except error.AmbiguousPrefixLookupError:
                 return False
             except error.WdirUnsupported:
                 # single 'ff...' match
                 return True
             if matchednode is None:
                 raise error.LookupError(node, self.indexfile, _(b'no node'))
             return True
 
         def maybewdir(prefix):
             return all(c == b'f' for c in pycompat.iterbytestr(prefix))
 
         hexnode = hex(node)
 
         def disambiguate(hexnode, minlength):
             """Disambiguate against wdirid."""
             for length in range(minlength, len(hexnode) + 1):
                 prefix = hexnode[:length]
                 if not maybewdir(prefix):
                     return prefix
 
         if not getattr(self, 'filteredrevs', None):
             try:
                 length = max(self.index.shortest(node), minlength)
                 return disambiguate(hexnode, length)
             except error.RevlogError:
                 if node != wdirid:
                     raise error.LookupError(node, self.indexfile, _(b'no node'))
             except AttributeError:
                 # Fall through to pure code
                 pass
 
         if node == wdirid:
             for length in range(minlength, len(hexnode) + 1):
                 prefix = hexnode[:length]
                 if isvalid(prefix):
                     return prefix
 
         for length in range(minlength, len(hexnode) + 1):
             prefix = hexnode[:length]
             if isvalid(prefix):
                 return disambiguate(hexnode, length)
 
     def cmp(self, node, text):
         """compare text with a given file revision
 
         returns True if text is different than what is stored.
         """
         p1, p2 = self.parents(node)
         return storageutil.hashrevisionsha1(text, p1, p2) != node
 
     def _cachesegment(self, offset, data):
         """Add a segment to the revlog cache.
 
         Accepts an absolute offset and the data that is at that location.
         """
         o, d = self._chunkcache
         # try to add to existing cache
         if o + len(d) == offset and len(d) + len(data) < _chunksize:
             self._chunkcache = o, d + data
         else:
             self._chunkcache = offset, data
 
     def _readsegment(self, offset, length, df=None):
         """Load a segment of raw data from the revlog.
 
         Accepts an absolute offset, length to read, and an optional existing
         file handle to read from.
 
         If an existing file handle is passed, it will be seeked and the
         original seek position will NOT be restored.
 
         Returns a str or buffer of raw byte data.
 
         Raises if the requested number of bytes could not be read.
         """
         # Cache data both forward and backward around the requested
         # data, in a fixed size window. This helps speed up operations
         # involving reading the revlog backwards.
         cachesize = self._chunkcachesize
         realoffset = offset & ~(cachesize - 1)
         reallength = (
             (offset + length + cachesize) & ~(cachesize - 1)
         ) - realoffset
         with self._datareadfp(df) as df:
             df.seek(realoffset)
             d = df.read(reallength)
 
         self._cachesegment(realoffset, d)
         if offset != realoffset or reallength != length:
             startoffset = offset - realoffset
             if len(d) - startoffset < length:
                 raise error.RevlogError(
                     _(
                         b'partial read of revlog %s; expected %d bytes from '
                         b'offset %d, got %d'
                     )
                     % (
                         self.indexfile if self._inline else self.datafile,
                         length,
                         realoffset,
                         len(d) - startoffset,
                     )
                 )
 
             return util.buffer(d, startoffset, length)
 
         if len(d) < length:
             raise error.RevlogError(
                 _(
                     b'partial read of revlog %s; expected %d bytes from offset '
                     b'%d, got %d'
                 )
                 % (
                     self.indexfile if self._inline else self.datafile,
                     length,
                     offset,
                     len(d),
                 )
             )
 
         return d
 
     def _getsegment(self, offset, length, df=None):
         """Obtain a segment of raw data from the revlog.
 
         Accepts an absolute offset, length of bytes to obtain, and an
         optional file handle to the already-opened revlog. If the file
         handle is used, it's original seek position will not be preserved.
 
         Requests for data may be returned from a cache.
 
         Returns a str or a buffer instance of raw byte data.
         """
         o, d = self._chunkcache
         l = len(d)
 
         # is it in the cache?
         cachestart = offset - o
         cacheend = cachestart + length
         if cachestart >= 0 and cacheend <= l:
             if cachestart == 0 and cacheend == l:
                 return d  # avoid a copy
             return util.buffer(d, cachestart, cacheend - cachestart)
 
         return self._readsegment(offset, length, df=df)
 
     def _getsegmentforrevs(self, startrev, endrev, df=None):
         """Obtain a segment of raw data corresponding to a range of revisions.
 
         Accepts the start and end revisions and an optional already-open
         file handle to be used for reading. If the file handle is read, its
         seek position will not be preserved.
 
         Requests for data may be satisfied by a cache.
 
         Returns a 2-tuple of (offset, data) for the requested range of
         revisions. Offset is the integer offset from the beginning of the
         revlog and data is a str or buffer of the raw byte data.
 
         Callers will need to call ``self.start(rev)`` and ``self.length(rev)``
         to determine where each revision's data begins and ends.
         """
         # Inlined self.start(startrev) & self.end(endrev) for perf reasons
         # (functions are expensive).
         index = self.index
         istart = index[startrev]
         start = int(istart[0] >> 16)
         if startrev == endrev:
             end = start + istart[1]
         else:
             iend = index[endrev]
             end = int(iend[0] >> 16) + iend[1]
 
         if self._inline:
             start += (startrev + 1) * self._io.size
             end += (endrev + 1) * self._io.size
         length = end - start
 
         return start, self._getsegment(start, length, df=df)
 
     def _chunk(self, rev, df=None):
         """Obtain a single decompressed chunk for a revision.
 
         Accepts an integer revision and an optional already-open file handle
         to be used for reading. If used, the seek position of the file will not
         be preserved.
 
         Returns a str holding uncompressed data for the requested revision.
         """
         return self.decompress(self._getsegmentforrevs(rev, rev, df=df)[1])
 
     def _chunks(self, revs, df=None, targetsize=None):
         """Obtain decompressed chunks for the specified revisions.
 
         Accepts an iterable of numeric revisions that are assumed to be in
         ascending order. Also accepts an optional already-open file handle
         to be used for reading. If used, the seek position of the file will
         not be preserved.
 
         This function is similar to calling ``self._chunk()`` multiple times,
         but is faster.
 
         Returns a list with decompressed data for each requested revision.
         """
         if not revs:
             return []
         start = self.start
         length = self.length
         inline = self._inline
         iosize = self._io.size
         buffer = util.buffer
 
         l = []
         ladd = l.append
 
         if not self._withsparseread:
             slicedchunks = (revs,)
         else:
             slicedchunks = deltautil.slicechunk(
                 self, revs, targetsize=targetsize
             )
 
         for revschunk in slicedchunks:
             firstrev = revschunk[0]
             # Skip trailing revisions with empty diff
             for lastrev in revschunk[::-1]:
                 if length(lastrev) != 0:
                     break
 
             try:
                 offset, data = self._getsegmentforrevs(firstrev, lastrev, df=df)
             except OverflowError:
                 # issue4215 - we can't cache a run of chunks greater than
                 # 2G on Windows
                 return [self._chunk(rev, df=df) for rev in revschunk]
 
             decomp = self.decompress
             for rev in revschunk:
                 chunkstart = start(rev)
                 if inline:
                     chunkstart += (rev + 1) * iosize
                 chunklength = length(rev)
                 ladd(decomp(buffer(data, chunkstart - offset, chunklength)))
 
         return l
 
     def _chunkclear(self):
         """Clear the raw chunk cache."""
         self._chunkcache = (0, b'')
 
     def deltaparent(self, rev):
         """return deltaparent of the given revision"""
         base = self.index[rev][3]
         if base == rev:
             return nullrev
         elif self._generaldelta:
             return base
         else:
             return rev - 1
 
     def issnapshot(self, rev):
         """tells whether rev is a snapshot"""
         if not self._sparserevlog:
             return self.deltaparent(rev) == nullrev
         elif util.safehasattr(self.index, b'issnapshot'):
             # directly assign the method to cache the testing and access
             self.issnapshot = self.index.issnapshot
             return self.issnapshot(rev)
         if rev == nullrev:
             return True
         entry = self.index[rev]
         base = entry[3]
         if base == rev:
             return True
         if base == nullrev:
             return True
         p1 = entry[5]
         p2 = entry[6]
         if base == p1 or base == p2:
             return False
         return self.issnapshot(base)
 
     def snapshotdepth(self, rev):
         """number of snapshot in the chain before this one"""
         if not self.issnapshot(rev):
             raise error.ProgrammingError(b'revision %d not a snapshot')
         return len(self._deltachain(rev)[0]) - 1
 
     def revdiff(self, rev1, rev2):
         """return or calculate a delta between two revisions
 
         The delta calculated is in binary form and is intended to be written to
         revlog data directly. So this function needs raw revision data.
         """
         if rev1 != nullrev and self.deltaparent(rev2) == rev1:
             return bytes(self._chunk(rev2))
 
         return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2))
 
     def _processflags(self, text, flags, operation, raw=False):
         """deprecated entry point to access flag processors"""
         msg = b'_processflag(...) use the specialized variant'
         util.nouideprecwarn(msg, b'5.2', stacklevel=2)
         if raw:
             return text, flagutil.processflagsraw(self, text, flags)
         elif operation == b'read':
             return flagutil.processflagsread(self, text, flags)
         else:  # write operation
             return flagutil.processflagswrite(self, text, flags)
 
     def revision(self, nodeorrev, _df=None, raw=False):
         """return an uncompressed revision of a given node or revision
         number.
 
         _df - an existing file handle to read from. (internal-only)
         raw - an optional argument specifying if the revision data is to be
         treated as raw data when applying flag transforms. 'raw' should be set
         to True when generating changegroups or in debug commands.
         """
         if raw:
             msg = (
                 b'revlog.revision(..., raw=True) is deprecated, '
                 b'use revlog.rawdata(...)'
             )
             util.nouideprecwarn(msg, b'5.2', stacklevel=2)
         return self._revisiondata(nodeorrev, _df, raw=raw)[0]
 
     def sidedata(self, nodeorrev, _df=None):
         """a map of extra data related to the changeset but not part of the hash
 
         This function currently return a dictionary. However, more advanced
         mapping object will likely be used in the future for a more
         efficient/lazy code.
         """
         return self._revisiondata(nodeorrev, _df)[1]
 
     def _revisiondata(self, nodeorrev, _df=None, raw=False):
         # deal with <nodeorrev> argument type
         if isinstance(nodeorrev, int):
             rev = nodeorrev
             node = self.node(rev)
         else:
             node = nodeorrev
             rev = None
 
         # fast path the special `nullid` rev
         if node == nullid:
             return b"", {}
 
         # ``rawtext`` is the text as stored inside the revlog. Might be the
         # revision or might need to be processed to retrieve the revision.
         rev, rawtext, validated = self._rawtext(node, rev, _df=_df)
 
         if self.version & 0xFFFF == REVLOGV2:
             if rev is None:
                 rev = self.rev(node)
             sidedata = self._sidedata(rev)
         else:
             sidedata = {}
 
         if raw and validated:
             # if we don't want to process the raw text and that raw
             # text is cached, we can exit early.
             return rawtext, sidedata
         if rev is None:
             rev = self.rev(node)
         # the revlog's flag for this revision
         # (usually alter its state or content)
         flags = self.flags(rev)
 
         if validated and flags == REVIDX_DEFAULT_FLAGS:
             # no extra flags set, no flag processor runs, text = rawtext
             return rawtext, sidedata
 
         if raw:
             validatehash = flagutil.processflagsraw(self, rawtext, flags)
             text = rawtext
         else:
             r = flagutil.processflagsread(self, rawtext, flags)
             text, validatehash = r
         if validatehash:
             self.checkhash(text, node, rev=rev)
         if not validated:
             self._revisioncache = (node, rev, rawtext)
 
         return text, sidedata
 
     def _rawtext(self, node, rev, _df=None):
         """return the possibly unvalidated rawtext for a revision
 
         returns (rev, rawtext, validated)
         """
 
         # revision in the cache (could be useful to apply delta)
         cachedrev = None
         # An intermediate text to apply deltas to
         basetext = None
 
         # Check if we have the entry in cache
         # The cache entry looks like (node, rev, rawtext)
         if self._revisioncache:
             if self._revisioncache[0] == node:
                 return (rev, self._revisioncache[2], True)
             cachedrev = self._revisioncache[1]
 
         if rev is None:
             rev = self.rev(node)
 
         chain, stopped = self._deltachain(rev, stoprev=cachedrev)
         if stopped:
             basetext = self._revisioncache[2]
 
         # drop cache to save memory, the caller is expected to
         # update self._revisioncache after validating the text
         self._revisioncache = None
 
         targetsize = None
         rawsize = self.index[rev][2]
         if 0 <= rawsize:
             targetsize = 4 * rawsize
 
         bins = self._chunks(chain, df=_df, targetsize=targetsize)
         if basetext is None:
             basetext = bytes(bins[0])
             bins = bins[1:]
 
         rawtext = mdiff.patches(basetext, bins)
         del basetext  # let us have a chance to free memory early
         return (rev, rawtext, False)
 
     def _sidedata(self, rev):
         """Return the sidedata for a given revision number."""
         index_entry = self.index[rev]
         sidedata_offset = index_entry[8]
         sidedata_size = index_entry[9]
 
         if self._inline:
             sidedata_offset += self._io.size * (1 + rev)
         if sidedata_size == 0:
             return {}
 
         segment = self._getsegment(sidedata_offset, sidedata_size)
         sidedata = sidedatautil.deserialize_sidedata(segment)
         return sidedata
 
     def rawdata(self, nodeorrev, _df=None):
         """return an uncompressed raw data of a given node or revision number.
 
         _df - an existing file handle to read from. (internal-only)
         """
         return self._revisiondata(nodeorrev, _df, raw=True)[0]
 
     def hash(self, text, p1, p2):
         """Compute a node hash.
 
         Available as a function so that subclasses can replace the hash
         as needed.
         """
         return storageutil.hashrevisionsha1(text, p1, p2)
 
     def checkhash(self, text, node, p1=None, p2=None, rev=None):
         """Check node hash integrity.
 
         Available as a function so that subclasses can extend hash mismatch
         behaviors as needed.
         """
         try:
             if p1 is None and p2 is None:
                 p1, p2 = self.parents(node)
             if node != self.hash(text, p1, p2):
                 # Clear the revision cache on hash failure. The revision cache
                 # only stores the raw revision and clearing the cache does have
                 # the side-effect that we won't have a cache hit when the raw
                 # revision data is accessed. But this case should be rare and
                 # it is extra work to teach the cache about the hash
                 # verification state.
                 if self._revisioncache and self._revisioncache[0] == node:
                     self._revisioncache = None
 
                 revornode = rev
                 if revornode is None:
                     revornode = templatefilters.short(hex(node))
                 raise error.RevlogError(
                     _(b"integrity check failed on %s:%s")
                     % (self.indexfile, pycompat.bytestr(revornode))
                 )
         except error.RevlogError:
             if self._censorable and storageutil.iscensoredtext(text):
                 raise error.CensoredNodeError(self.indexfile, node, text)
             raise
 
     def _enforceinlinesize(self, tr, fp=None):
         """Check if the revlog is too big for inline and convert if so.
 
         This should be called after revisions are added to the revlog. If the
         revlog has grown too large to be an inline revlog, it will convert it
         to use multiple index and data files.
         """
         tiprev = len(self) - 1
         if (
             not self._inline
             or (self.start(tiprev) + self.length(tiprev)) < _maxinline
         ):
             return
 
         troffset = tr.findoffset(self.indexfile)
         if troffset is None:
             raise error.RevlogError(
                 _(b"%s not found in the transaction") % self.indexfile
             )
         trindex = 0
         tr.add(self.datafile, 0)
 
         if fp:
             fp.flush()
             fp.close()
             # We can't use the cached file handle after close(). So prevent
             # its usage.
             self._writinghandles = None
 
         with self._indexfp(b'r') as ifh, self._datafp(b'w') as dfh:
             for r in self:
                 dfh.write(self._getsegmentforrevs(r, r, df=ifh)[1])
                 if troffset <= self.start(r):
                     trindex = r
 
         with self._indexfp(b'w') as fp:
             self.version &= ~FLAG_INLINE_DATA
             self._inline = False
             io = self._io
             for i in self:
                 e = io.packentry(self.index[i], self.node, self.version, i)
                 fp.write(e)
 
             # the temp file replace the real index when we exit the context
             # manager
 
         tr.replace(self.indexfile, trindex * self._io.size)
         nodemaputil.setup_persistent_nodemap(tr, self)
         self._chunkclear()
 
     def _nodeduplicatecallback(self, transaction, node):
         """called when trying to add a node already stored."""
 
     def addrevision(
         self,
         text,
         transaction,
         link,
         p1,
         p2,
         cachedelta=None,
         node=None,
         flags=REVIDX_DEFAULT_FLAGS,
         deltacomputer=None,
         sidedata=None,
     ):
         """add a revision to the log
 
         text - the revision data to add
         transaction - the transaction object used for rollback
         link - the linkrev data to add
         p1, p2 - the parent nodeids of the revision
         cachedelta - an optional precomputed delta
         node - nodeid of revision; typically node is not specified, and it is
             computed by default as hash(text, p1, p2), however subclasses might
             use different hashing method (and override checkhash() in such case)
         flags - the known flags to set on the revision
         deltacomputer - an optional deltacomputer instance shared between
             multiple calls
         """
         if link == nullrev:
             raise error.RevlogError(
                 _(b"attempted to add linkrev -1 to %s") % self.indexfile
             )
 
         if sidedata is None:
             sidedata = {}
         elif not self.hassidedata:
             raise error.ProgrammingError(
                 _(b"trying to add sidedata to a revlog who don't support them")
             )
 
         if flags:
             node = node or self.hash(text, p1, p2)
 
         rawtext, validatehash = flagutil.processflagswrite(self, text, flags)
 
         # If the flag processor modifies the revision data, ignore any provided
         # cachedelta.
         if rawtext != text:
             cachedelta = None
 
         if len(rawtext) > _maxentrysize:
             raise error.RevlogError(
                 _(
                     b"%s: size of %d bytes exceeds maximum revlog storage of 2GiB"
                 )
                 % (self.indexfile, len(rawtext))
             )
 
         node = node or self.hash(rawtext, p1, p2)
         rev = self.index.get_rev(node)
         if rev is not None:
             return rev
 
         if validatehash:
             self.checkhash(rawtext, node, p1=p1, p2=p2)
 
         return self.addrawrevision(
             rawtext,
             transaction,
             link,
             p1,
             p2,
             node,
             flags,
             cachedelta=cachedelta,
             deltacomputer=deltacomputer,
             sidedata=sidedata,
         )
 
     def addrawrevision(
         self,
         rawtext,
         transaction,
         link,
         p1,
         p2,
         node,
         flags,
         cachedelta=None,
         deltacomputer=None,
         sidedata=None,
     ):
         """add a raw revision with known flags, node and parents
         useful when reusing a revision not stored in this revlog (ex: received
         over wire, or read from an external bundle).
         """
         dfh = None
         if not self._inline:
             dfh = self._datafp(b"a+")
         ifh = self._indexfp(b"a+")
         try:
             return self._addrevision(
                 node,
                 rawtext,
                 transaction,
                 link,
                 p1,
                 p2,
                 flags,
                 cachedelta,
                 ifh,
                 dfh,
                 deltacomputer=deltacomputer,
                 sidedata=sidedata,
             )
         finally:
             if dfh:
                 dfh.close()
             ifh.close()
 
     def compress(self, data):
         """Generate a possibly-compressed representation of data."""
         if not data:
             return b'', data
 
         compressed = self._compressor.compress(data)
 
         if compressed:
             # The revlog compressor added the header in the returned data.
             return b'', compressed
 
         if data[0:1] == b'\0':
             return b'', data
         return b'u', data
 
     def decompress(self, data):
         """Decompress a revlog chunk.
 
         The chunk is expected to begin with a header identifying the
         format type so it can be routed to an appropriate decompressor.
         """
         if not data:
             return data
 
         # Revlogs are read much more frequently than they are written and many
         # chunks only take microseconds to decompress, so performance is
         # important here.
         #
         # We can make a few assumptions about revlogs:
         #
         # 1) the majority of chunks will be compressed (as opposed to inline
         #    raw data).
         # 2) decompressing *any* data will likely by at least 10x slower than
         #    returning raw inline data.
         # 3) we want to prioritize common and officially supported compression
         #    engines
         #
         # It follows that we want to optimize for "decompress compressed data
         # when encoded with common and officially supported compression engines"
         # case over "raw data" and "data encoded by less common or non-official
         # compression engines." That is why we have the inline lookup first
         # followed by the compengines lookup.
         #
         # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib
         # compressed chunks. And this matters for changelog and manifest reads.
         t = data[0:1]
 
         if t == b'x':
             try:
                 return _zlibdecompress(data)
             except zlib.error as e:
                 raise error.RevlogError(
                     _(b'revlog decompress error: %s')
                     % stringutil.forcebytestr(e)
                 )
         # '\0' is more common than 'u' so it goes first.
         elif t == b'\0':
             return data
         elif t == b'u':
             return util.buffer(data, 1)
 
         try:
             compressor = self._decompressors[t]
         except KeyError:
             try:
                 engine = util.compengines.forrevlogheader(t)
                 compressor = engine.revlogcompressor(self._compengineopts)
                 self._decompressors[t] = compressor
             except KeyError:
                 raise error.RevlogError(_(b'unknown compression type %r') % t)
 
         return compressor.decompress(data)
 
     def _addrevision(
         self,
         node,
         rawtext,
         transaction,
         link,
         p1,
         p2,
         flags,
         cachedelta,
         ifh,
         dfh,
         alwayscache=False,
         deltacomputer=None,
         sidedata=None,
     ):
         """internal function to add revisions to the log
 
         see addrevision for argument descriptions.
 
         note: "addrevision" takes non-raw text, "_addrevision" takes raw text.
 
         if "deltacomputer" is not provided or None, a defaultdeltacomputer will
         be used.
 
         invariants:
         - rawtext is optional (can be None); if not set, cachedelta must be set.
           if both are set, they must correspond to each other.
         """
         if node == nullid:
             raise error.RevlogError(
                 _(b"%s: attempt to add null revision") % self.indexfile
             )
         if node == wdirid or node in wdirfilenodeids:
             raise error.RevlogError(
                 _(b"%s: attempt to add wdir revision") % self.indexfile
             )
 
         if self._inline:
             fh = ifh
         else:
             fh = dfh
 
         btext = [rawtext]
 
         curr = len(self)
         prev = curr - 1
 
         offset = self._get_data_offset(prev)
 
         if self._concurrencychecker:
             if self._inline:
                 # offset is "as if" it were in the .d file, so we need to add on
                 # the size of the entry metadata.
                 self._concurrencychecker(
                     ifh, self.indexfile, offset + curr * self._io.size
                 )
             else:
                 # Entries in the .i are a consistent size.
                 self._concurrencychecker(
                     ifh, self.indexfile, curr * self._io.size
                 )
                 self._concurrencychecker(dfh, self.datafile, offset)
 
         p1r, p2r = self.rev(p1), self.rev(p2)
 
         # full versions are inserted when the needed deltas
         # become comparable to the uncompressed text
         if rawtext is None:
             # need rawtext size, before changed by flag processors, which is
             # the non-raw size. use revlog explicitly to avoid filelog's extra
             # logic that might remove metadata size.
             textlen = mdiff.patchedsize(
                 revlog.size(self, cachedelta[0]), cachedelta[1]
             )
         else:
             textlen = len(rawtext)
 
         if deltacomputer is None:
             deltacomputer = deltautil.deltacomputer(self)
 
         revinfo = _revisioninfo(node, p1, p2, btext, textlen, cachedelta, flags)
 
         deltainfo = deltacomputer.finddeltainfo(revinfo, fh)
 
         if sidedata:
             serialized_sidedata = sidedatautil.serialize_sidedata(sidedata)
             sidedata_offset = offset + deltainfo.deltalen
         else:
             serialized_sidedata = b""
             # Don't store the offset if the sidedata is empty, that way
             # we can easily detect empty sidedata and they will be no different
             # than ones we manually add.
             sidedata_offset = 0
 
         e = (
             offset_type(offset, flags),
             deltainfo.deltalen,
             textlen,
             deltainfo.base,
             link,
             p1r,
             p2r,
             node,
             sidedata_offset,
             len(serialized_sidedata),
         )
 
         if self.version & 0xFFFF != REVLOGV2:
             e = e[:8]
 
         self.index.append(e)
         entry = self._io.packentry(e, self.node, self.version, curr)
         self._writeentry(
             transaction,
             ifh,
             dfh,
             entry,
             deltainfo.data,
             link,
             offset,
             serialized_sidedata,
         )
 
         rawtext = btext[0]
 
         if alwayscache and rawtext is None:
             rawtext = deltacomputer.buildtext(revinfo, fh)
 
         if type(rawtext) == bytes:  # only accept immutable objects
             self._revisioncache = (node, curr, rawtext)
         self._chainbasecache[curr] = deltainfo.chainbase
         return curr
 
     def _get_data_offset(self, prev):
         """Returns the current offset in the (in-transaction) data file.
         Versions < 2 of the revlog can get this 0(1), revlog v2 needs a docket
         file to store that information: since sidedata can be rewritten to the
         end of the data file within a transaction, you can have cases where, for
         example, rev `n` does not have sidedata while rev `n - 1` does, leading
         to `n - 1`'s sidedata being written after `n`'s data.
 
         TODO cache this in a docket file before getting out of experimental."""
         if self.version & 0xFFFF != REVLOGV2:
             return self.end(prev)
 
         offset = 0
         for rev, entry in enumerate(self.index):
             sidedata_end = entry[8] + entry[9]
             # Sidedata for a previous rev has potentially been written after
             # this rev's end, so take the max.
             offset = max(self.end(rev), offset, sidedata_end)
         return offset
 
     def _writeentry(
         self, transaction, ifh, dfh, entry, data, link, offset, sidedata
     ):
         # Files opened in a+ mode have inconsistent behavior on various
         # platforms. Windows requires that a file positioning call be made
         # when the file handle transitions between reads and writes. See
         # 3686fa2b8eee and the mixedfilemodewrapper in windows.py. On other
         # platforms, Python or the platform itself can be buggy. Some versions
         # of Solaris have been observed to not append at the end of the file
         # if the file was seeked to before the end. See issue4943 for more.
         #
         # We work around this issue by inserting a seek() before writing.
         # Note: This is likely not necessary on Python 3. However, because
         # the file handle is reused for reads and may be seeked there, we need
         # to be careful before changing this.
         ifh.seek(0, os.SEEK_END)
         if dfh:
             dfh.seek(0, os.SEEK_END)
 
         curr = len(self) - 1
         if not self._inline:
             transaction.add(self.datafile, offset)
             transaction.add(self.indexfile, curr * len(entry))
             if data[0]:
                 dfh.write(data[0])
             dfh.write(data[1])
             if sidedata:
                 dfh.write(sidedata)
             ifh.write(entry)
         else:
             offset += curr * self._io.size
             transaction.add(self.indexfile, offset)
             ifh.write(entry)
             ifh.write(data[0])
             ifh.write(data[1])
             if sidedata:
                 ifh.write(sidedata)
             self._enforceinlinesize(transaction, ifh)
         nodemaputil.setup_persistent_nodemap(transaction, self)
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         alwayscache=False,
         addrevisioncb=None,
         duplicaterevisioncb=None,
     ):
         """
         add a delta group
 
         given a set of deltas, add them to the revision log. the
         first delta is against its parent, which should be in our
         log, the rest are against the previous delta.
 
         If ``addrevisioncb`` is defined, it will be called with arguments of
         this revlog and the node that was added.
         """
 
         if self._writinghandles:
             raise error.ProgrammingError(b'cannot nest addgroup() calls')
 
         r = len(self)
         end = 0
         if r:
             end = self.end(r - 1)
         ifh = self._indexfp(b"a+")
         isize = r * self._io.size
         if self._inline:
             transaction.add(self.indexfile, end + isize)
             dfh = None
         else:
             transaction.add(self.indexfile, isize)
             transaction.add(self.datafile, end)
             dfh = self._datafp(b"a+")
 
         def flush():
             if dfh:
                 dfh.flush()
             ifh.flush()
 
         self._writinghandles = (ifh, dfh)
         empty = True
 
         try:
             deltacomputer = deltautil.deltacomputer(self)
             # loop through our set of deltas
             for data in deltas:
                 node, p1, p2, linknode, deltabase, delta, flags, sidedata = data
                 link = linkmapper(linknode)
                 flags = flags or REVIDX_DEFAULT_FLAGS
 
                 rev = self.index.get_rev(node)
                 if rev is not None:
                     # this can happen if two branches make the same change
                     self._nodeduplicatecallback(transaction, rev)
                     if duplicaterevisioncb:
                         duplicaterevisioncb(self, rev)
                     empty = False
                     continue
 
                 for p in (p1, p2):
                     if not self.index.has_node(p):
                         raise error.LookupError(
                             p, self.indexfile, _(b'unknown parent')
                         )
 
                 if not self.index.has_node(deltabase):
                     raise error.LookupError(
                         deltabase, self.indexfile, _(b'unknown delta base')
                     )
 
                 baserev = self.rev(deltabase)
 
                 if baserev != nullrev and self.iscensored(baserev):
                     # if base is censored, delta must be full replacement in a
                     # single patch operation
                     hlen = struct.calcsize(b">lll")
                     oldlen = self.rawsize(baserev)
                     newlen = len(delta) - hlen
                     if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
                         raise error.CensoredBaseError(
                             self.indexfile, self.node(baserev)
                         )
 
                 if not flags and self._peek_iscensored(baserev, delta, flush):
                     flags |= REVIDX_ISCENSORED
 
                 # We assume consumers of addrevisioncb will want to retrieve
                 # the added revision, which will require a call to
                 # revision(). revision() will fast path if there is a cache
                 # hit. So, we tell _addrevision() to always cache in this case.
                 # We're only using addgroup() in the context of changegroup
                 # generation so the revision data can always be handled as raw
                 # by the flagprocessor.
                 rev = self._addrevision(
                     node,
                     None,
                     transaction,
                     link,
                     p1,
                     p2,
                     flags,
                     (baserev, delta),
                     ifh,
                     dfh,
                     alwayscache=alwayscache,
                     deltacomputer=deltacomputer,
                     sidedata=sidedata,
                 )
 
                 if addrevisioncb:
                     addrevisioncb(self, rev)
                 empty = False
 
                 if not dfh and not self._inline:
                     # addrevision switched from inline to conventional
                     # reopen the index
                     ifh.close()
                     dfh = self._datafp(b"a+")
                     ifh = self._indexfp(b"a+")
                     self._writinghandles = (ifh, dfh)
         finally:
             self._writinghandles = None
 
             if dfh:
                 dfh.close()
             ifh.close()
         return not empty
 
     def iscensored(self, rev):
         """Check if a file revision is censored."""
         if not self._censorable:
             return False
 
         return self.flags(rev) & REVIDX_ISCENSORED
 
     def _peek_iscensored(self, baserev, delta, flush):
         """Quickly check if a delta produces a censored revision."""
         if not self._censorable:
             return False
 
         return storageutil.deltaiscensored(delta, baserev, self.rawsize)
 
     def getstrippoint(self, minlink):
         """find the minimum rev that must be stripped to strip the linkrev
 
         Returns a tuple containing the minimum rev and a set of all revs that
         have linkrevs that will be broken by this strip.
         """
         return storageutil.resolvestripinfo(
             minlink,
             len(self) - 1,
             self.headrevs(),
             self.linkrev,
             self.parentrevs,
         )
 
     def strip(self, minlink, transaction):
         """truncate the revlog on the first revision with a linkrev >= minlink
 
         This function is called when we're stripping revision minlink and
         its descendants from the repository.
 
         We have to remove all revisions with linkrev >= minlink, because
         the equivalent changelog revisions will be renumbered after the
         strip.
 
         So we truncate the revlog on the first of these revisions, and
         trust that the caller has saved the revisions that shouldn't be
         removed and that it'll re-add them after this truncation.
         """
         if len(self) == 0:
             return
 
         rev, _ = self.getstrippoint(minlink)
         if rev == len(self):
             return
 
         # first truncate the files on disk
         end = self.start(rev)
         if not self._inline:
             transaction.add(self.datafile, end)
             end = rev * self._io.size
         else:
             end += rev * self._io.size
 
         transaction.add(self.indexfile, end)
 
         # then reset internal state in memory to forget those revisions
         self._revisioncache = None
         self._chaininfocache = util.lrucachedict(500)
         self._chunkclear()
 
         del self.index[rev:-1]
 
     def checksize(self):
         """Check size of index and data files
 
         return a (dd, di) tuple.
         - dd: extra bytes for the "data" file
         - di: extra bytes for the "index" file
 
         A healthy revlog will return (0, 0).
         """
         expected = 0
         if len(self):
             expected = max(0, self.end(len(self) - 1))
 
         try:
             with self._datafp() as f:
                 f.seek(0, io.SEEK_END)
                 actual = f.tell()
             dd = actual - expected
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
             dd = 0
 
         try:
             f = self.opener(self.indexfile)
             f.seek(0, io.SEEK_END)
             actual = f.tell()
             f.close()
             s = self._io.size
             i = max(0, actual // s)
             di = actual - (i * s)
             if self._inline:
                 databytes = 0
                 for r in self:
                     databytes += max(0, self.length(r))
                 dd = 0
                 di = actual - len(self) * s - databytes
         except IOError as inst:
             if inst.errno != errno.ENOENT:
                 raise
             di = 0
 
         return (dd, di)
 
     def files(self):
         res = [self.indexfile]
         if not self._inline:
             res.append(self.datafile)
         return res
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
+        sidedata_helpers=None,
     ):
         if nodesorder not in (b'nodes', b'storage', b'linear', None):
             raise error.ProgrammingError(
                 b'unhandled value for nodesorder: %s' % nodesorder
             )
 
         if nodesorder is None and not self._generaldelta:
             nodesorder = b'storage'
 
         if (
             not self._storedeltachains
             and deltamode != repository.CG_DELTAMODE_PREV
         ):
             deltamode = repository.CG_DELTAMODE_FULL
 
         return storageutil.emitrevisions(
             self,
             nodes,
             nodesorder,
             revlogrevisiondelta,
             deltaparentfn=self.deltaparent,
             candeltafn=self.candelta,
             rawsizefn=self.rawsize,
             revdifffn=self.revdiff,
             flagsfn=self.flags,
             deltamode=deltamode,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
+            sidedata_helpers=sidedata_helpers,
         )
 
     DELTAREUSEALWAYS = b'always'
     DELTAREUSESAMEREVS = b'samerevs'
     DELTAREUSENEVER = b'never'
 
     DELTAREUSEFULLADD = b'fulladd'
 
     DELTAREUSEALL = {b'always', b'samerevs', b'never', b'fulladd'}
 
     def clone(
         self,
         tr,
         destrevlog,
         addrevisioncb=None,
         deltareuse=DELTAREUSESAMEREVS,
         forcedeltabothparents=None,
         sidedatacompanion=None,
     ):
         """Copy this revlog to another, possibly with format changes.
 
         The destination revlog will contain the same revisions and nodes.
         However, it may not be bit-for-bit identical due to e.g. delta encoding
         differences.
 
         The ``deltareuse`` argument control how deltas from the existing revlog
         are preserved in the destination revlog. The argument can have the
         following values:
 
         DELTAREUSEALWAYS
            Deltas will always be reused (if possible), even if the destination
            revlog would not select the same revisions for the delta. This is the
            fastest mode of operation.
         DELTAREUSESAMEREVS
            Deltas will be reused if the destination revlog would pick the same
            revisions for the delta. This mode strikes a balance between speed
            and optimization.
         DELTAREUSENEVER
            Deltas will never be reused. This is the slowest mode of execution.
            This mode can be used to recompute deltas (e.g. if the diff/delta
            algorithm changes).
         DELTAREUSEFULLADD
            Revision will be re-added as if their were new content. This is
            slower than DELTAREUSEALWAYS but allow more mechanism to kicks in.
            eg: large file detection and handling.
 
         Delta computation can be slow, so the choice of delta reuse policy can
         significantly affect run time.
 
         The default policy (``DELTAREUSESAMEREVS``) strikes a balance between
         two extremes. Deltas will be reused if they are appropriate. But if the
         delta could choose a better revision, it will do so. This means if you
         are converting a non-generaldelta revlog to a generaldelta revlog,
         deltas will be recomputed if the delta's parent isn't a parent of the
         revision.
 
         In addition to the delta policy, the ``forcedeltabothparents``
         argument controls whether to force compute deltas against both parents
         for merges. By default, the current default is used.
 
         If not None, the `sidedatacompanion` is callable that accept two
         arguments:
 
             (srcrevlog, rev)
 
         and return a quintet that control changes to sidedata content from the
         old revision to the new clone result:
 
             (dropall, filterout, update, new_flags, dropped_flags)
 
         * if `dropall` is True, all sidedata should be dropped
         * `filterout` is a set of sidedata keys that should be dropped
         * `update` is a mapping of additionnal/new key -> value
         * new_flags is a bitfields of new flags that the revision should get
         * dropped_flags is a bitfields of new flags that the revision shoudl not longer have
         """
         if deltareuse not in self.DELTAREUSEALL:
             raise ValueError(
                 _(b'value for deltareuse invalid: %s') % deltareuse
             )
 
         if len(destrevlog):
             raise ValueError(_(b'destination revlog is not empty'))
 
         if getattr(self, 'filteredrevs', None):
             raise ValueError(_(b'source revlog has filtered revisions'))
         if getattr(destrevlog, 'filteredrevs', None):
             raise ValueError(_(b'destination revlog has filtered revisions'))
 
         # lazydelta and lazydeltabase controls whether to reuse a cached delta,
         # if possible.
         oldlazydelta = destrevlog._lazydelta
         oldlazydeltabase = destrevlog._lazydeltabase
         oldamd = destrevlog._deltabothparents
 
         try:
             if deltareuse == self.DELTAREUSEALWAYS:
                 destrevlog._lazydeltabase = True
                 destrevlog._lazydelta = True
             elif deltareuse == self.DELTAREUSESAMEREVS:
                 destrevlog._lazydeltabase = False
                 destrevlog._lazydelta = True
             elif deltareuse == self.DELTAREUSENEVER:
                 destrevlog._lazydeltabase = False
                 destrevlog._lazydelta = False
 
             destrevlog._deltabothparents = forcedeltabothparents or oldamd
 
             self._clone(
                 tr,
                 destrevlog,
                 addrevisioncb,
                 deltareuse,
                 forcedeltabothparents,
                 sidedatacompanion,
             )
 
         finally:
             destrevlog._lazydelta = oldlazydelta
             destrevlog._lazydeltabase = oldlazydeltabase
             destrevlog._deltabothparents = oldamd
 
     def _clone(
         self,
         tr,
         destrevlog,
         addrevisioncb,
         deltareuse,
         forcedeltabothparents,
         sidedatacompanion,
     ):
         """perform the core duty of `revlog.clone` after parameter processing"""
         deltacomputer = deltautil.deltacomputer(destrevlog)
         index = self.index
         for rev in self:
             entry = index[rev]
 
             # Some classes override linkrev to take filtered revs into
             # account. Use raw entry from index.
             flags = entry[0] & 0xFFFF
             linkrev = entry[4]
             p1 = index[entry[5]][7]
             p2 = index[entry[6]][7]
             node = entry[7]
 
             sidedataactions = (False, [], {}, 0, 0)
             if sidedatacompanion is not None:
                 sidedataactions = sidedatacompanion(self, rev)
 
             # (Possibly) reuse the delta from the revlog if allowed and
             # the revlog chunk is a delta.
             cachedelta = None
             rawtext = None
             if any(sidedataactions) or deltareuse == self.DELTAREUSEFULLADD:
                 dropall = sidedataactions[0]
                 filterout = sidedataactions[1]
                 update = sidedataactions[2]
                 new_flags = sidedataactions[3]
                 dropped_flags = sidedataactions[4]
                 text, sidedata = self._revisiondata(rev)
                 if dropall:
                     sidedata = {}
                 for key in filterout:
                     sidedata.pop(key, None)
                 sidedata.update(update)
                 if not sidedata:
                     sidedata = None
 
                 flags |= new_flags
                 flags &= ~dropped_flags
 
                 destrevlog.addrevision(
                     text,
                     tr,
                     linkrev,
                     p1,
                     p2,
                     cachedelta=cachedelta,
                     node=node,
                     flags=flags,
                     deltacomputer=deltacomputer,
                     sidedata=sidedata,
                 )
             else:
                 if destrevlog._lazydelta:
                     dp = self.deltaparent(rev)
                     if dp != nullrev:
                         cachedelta = (dp, bytes(self._chunk(rev)))
 
                 if not cachedelta:
                     rawtext = self.rawdata(rev)
 
                 ifh = destrevlog.opener(
                     destrevlog.indexfile, b'a+', checkambig=False
                 )
                 dfh = None
                 if not destrevlog._inline:
                     dfh = destrevlog.opener(destrevlog.datafile, b'a+')
                 try:
                     destrevlog._addrevision(
                         node,
                         rawtext,
                         tr,
                         linkrev,
                         p1,
                         p2,
                         flags,
                         cachedelta,
                         ifh,
                         dfh,
                         deltacomputer=deltacomputer,
                     )
                 finally:
                     if dfh:
                         dfh.close()
                     ifh.close()
 
             if addrevisioncb:
                 addrevisioncb(self, rev, node)
 
     def censorrevision(self, tr, censornode, tombstone=b''):
         if (self.version & 0xFFFF) == REVLOGV0:
             raise error.RevlogError(
                 _(b'cannot censor with version %d revlogs') % self.version
             )
 
         censorrev = self.rev(censornode)
         tombstone = storageutil.packmeta({b'censored': tombstone}, b'')
 
         if len(tombstone) > self.rawsize(censorrev):
             raise error.Abort(
                 _(b'censor tombstone must be no longer than censored data')
             )
 
         # Rewriting the revlog in place is hard. Our strategy for censoring is
         # to create a new revlog, copy all revisions to it, then replace the
         # revlogs on transaction close.
 
         newindexfile = self.indexfile + b'.tmpcensored'
         newdatafile = self.datafile + b'.tmpcensored'
 
         # This is a bit dangerous. We could easily have a mismatch of state.
         newrl = revlog(self.opener, newindexfile, newdatafile, censorable=True)
         newrl.version = self.version
         newrl._generaldelta = self._generaldelta
         newrl._io = self._io
 
         for rev in self.revs():
             node = self.node(rev)
             p1, p2 = self.parents(node)
 
             if rev == censorrev:
                 newrl.addrawrevision(
                     tombstone,
                     tr,
                     self.linkrev(censorrev),
                     p1,
                     p2,
                     censornode,
                     REVIDX_ISCENSORED,
                 )
 
                 if newrl.deltaparent(rev) != nullrev:
                     raise error.Abort(
                         _(
                             b'censored revision stored as delta; '
                             b'cannot censor'
                         ),
                         hint=_(
                             b'censoring of revlogs is not '
                             b'fully implemented; please report '
                             b'this bug'
                         ),
                     )
                 continue
 
             if self.iscensored(rev):
                 if self.deltaparent(rev) != nullrev:
                     raise error.Abort(
                         _(
                             b'cannot censor due to censored '
                             b'revision having delta stored'
                         )
                     )
                 rawtext = self._chunk(rev)
             else:
                 rawtext = self.rawdata(rev)
 
             newrl.addrawrevision(
                 rawtext, tr, self.linkrev(rev), p1, p2, node, self.flags(rev)
             )
 
         tr.addbackup(self.indexfile, location=b'store')
         if not self._inline:
             tr.addbackup(self.datafile, location=b'store')
 
         self.opener.rename(newrl.indexfile, self.indexfile)
         if not self._inline:
             self.opener.rename(newrl.datafile, self.datafile)
 
         self.clearcaches()
         self._loadindex()
 
     def verifyintegrity(self, state):
         """Verifies the integrity of the revlog.
 
         Yields ``revlogproblem`` instances describing problems that are
         found.
         """
         dd, di = self.checksize()
         if dd:
             yield revlogproblem(error=_(b'data length off by %d bytes') % dd)
         if di:
             yield revlogproblem(error=_(b'index contains %d extra bytes') % di)
 
         version = self.version & 0xFFFF
 
         # The verifier tells us what version revlog we should be.
         if version != state[b'expectedversion']:
             yield revlogproblem(
                 warning=_(b"warning: '%s' uses revlog format %d; expected %d")
                 % (self.indexfile, version, state[b'expectedversion'])
             )
 
         state[b'skipread'] = set()
         state[b'safe_renamed'] = set()
 
         for rev in self:
             node = self.node(rev)
 
             # Verify contents. 4 cases to care about:
             #
             #   common: the most common case
             #   rename: with a rename
             #   meta: file content starts with b'\1\n', the metadata
             #         header defined in filelog.py, but without a rename
             #   ext: content stored externally
             #
             # More formally, their differences are shown below:
             #
             #                       | common | rename | meta  | ext
             #  -------------------------------------------------------
             #   flags()             | 0      | 0      | 0     | not 0
             #   renamed()           | False  | True   | False | ?
             #   rawtext[0:2]=='\1\n'| False  | True   | True  | ?
             #
             # "rawtext" means the raw text stored in revlog data, which
             # could be retrieved by "rawdata(rev)". "text"
             # mentioned below is "revision(rev)".
             #
             # There are 3 different lengths stored physically:
             #  1. L1: rawsize, stored in revlog index
             #  2. L2: len(rawtext), stored in revlog data
             #  3. L3: len(text), stored in revlog data if flags==0, or
             #     possibly somewhere else if flags!=0
             #
             # L1 should be equal to L2. L3 could be different from them.
             # "text" may or may not affect commit hash depending on flag
             # processors (see flagutil.addflagprocessor).
             #
             #              | common  | rename | meta  | ext
             # -------------------------------------------------
             #    rawsize() | L1      | L1     | L1    | L1
             #       size() | L1      | L2-LM  | L1(*) | L1 (?)
             # len(rawtext) | L2      | L2     | L2    | L2
             #    len(text) | L2      | L2     | L2    | L3
             #  len(read()) | L2      | L2-LM  | L2-LM | L3 (?)
             #
             # LM:  length of metadata, depending on rawtext
             # (*): not ideal, see comment in filelog.size
             # (?): could be "- len(meta)" if the resolved content has
             #      rename metadata
             #
             # Checks needed to be done:
             #  1. length check: L1 == L2, in all cases.
             #  2. hash check: depending on flag processor, we may need to
             #     use either "text" (external), or "rawtext" (in revlog).
 
             try:
                 skipflags = state.get(b'skipflags', 0)
                 if skipflags:
                     skipflags &= self.flags(rev)
 
                 _verify_revision(self, skipflags, state, node)
 
                 l1 = self.rawsize(rev)
                 l2 = len(self.rawdata(node))
 
                 if l1 != l2:
                     yield revlogproblem(
                         error=_(b'unpacked size is %d, %d expected') % (l2, l1),
                         node=node,
                     )
 
             except error.CensoredNodeError:
                 if state[b'erroroncensored']:
                     yield revlogproblem(
                         error=_(b'censored file data'), node=node
                     )
                     state[b'skipread'].add(node)
             except Exception as e:
                 yield revlogproblem(
                     error=_(b'unpacking %s: %s')
                     % (short(node), stringutil.forcebytestr(e)),
                     node=node,
                 )
                 state[b'skipread'].add(node)
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         d = {}
 
         if exclusivefiles:
             d[b'exclusivefiles'] = [(self.opener, self.indexfile)]
             if not self._inline:
                 d[b'exclusivefiles'].append((self.opener, self.datafile))
 
         if sharedfiles:
             d[b'sharedfiles'] = []
 
         if revisionscount:
             d[b'revisionscount'] = len(self)
 
         if trackedsize:
             d[b'trackedsize'] = sum(map(self.rawsize, iter(self)))
 
         if storedsize:
             d[b'storedsize'] = sum(
                 self.opener.stat(path).st_size for path in self.files()
             )
 
         return d

mercurial/utils/storageutil.py

@@ -1,513 +1,560 b''
 # storageutil.py - Storage functionality agnostic of backend implementation.
 #
 # Copyright 2018 Gregory Szorc <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import re
 import struct
 
 from ..i18n import _
 from ..node import (
     bin,
     nullid,
     nullrev,
 )
 from .. import (
     dagop,
     error,
     mdiff,
     pycompat,
 )
 from ..interfaces import repository
+from ..revlogutils import sidedata as sidedatamod
 from ..utils import hashutil
 
 _nullhash = hashutil.sha1(nullid)
 
 
 def hashrevisionsha1(text, p1, p2):
     """Compute the SHA-1 for revision data and its parents.
 
     This hash combines both the current file contents and its history
     in a manner that makes it easy to distinguish nodes with the same
     content in the revision graph.
     """
     # As of now, if one of the parent node is null, p2 is null
     if p2 == nullid:
         # deep copy of a hash is faster than creating one
         s = _nullhash.copy()
         s.update(p1)
     else:
         # none of the parent nodes are nullid
         if p1 < p2:
             a = p1
             b = p2
         else:
             a = p2
             b = p1
         s = hashutil.sha1(a)
         s.update(b)
     s.update(text)
     return s.digest()
 
 
 METADATA_RE = re.compile(b'\x01\n')
 
 
 def parsemeta(text):
     """Parse metadata header from revision data.
 
     Returns a 2-tuple of (metadata, offset), where both can be None if there
     is no metadata.
     """
     # text can be buffer, so we can't use .startswith or .index
     if text[:2] != b'\x01\n':
         return None, None
     s = METADATA_RE.search(text, 2).start()
     mtext = text[2:s]
     meta = {}
     for l in mtext.splitlines():
         k, v = l.split(b': ', 1)
         meta[k] = v
     return meta, s + 2
 
 
 def packmeta(meta, text):
     """Add metadata to fulltext to produce revision text."""
     keys = sorted(meta)
     metatext = b''.join(b'%s: %s\n' % (k, meta[k]) for k in keys)
     return b'\x01\n%s\x01\n%s' % (metatext, text)
 
 
 def iscensoredtext(text):
     meta = parsemeta(text)[0]
     return meta and b'censored' in meta
 
 
 def filtermetadata(text):
     """Extract just the revision data from source text.
 
     Returns ``text`` unless it has a metadata header, in which case we return
     a new buffer without hte metadata.
     """
     if not text.startswith(b'\x01\n'):
         return text
 
     offset = text.index(b'\x01\n', 2)
     return text[offset + 2 :]
 
 
 def filerevisioncopied(store, node):
     """Resolve file revision copy metadata.
 
     Returns ``False`` if the file has no copy metadata. Otherwise a
     2-tuple of the source filename and node.
     """
     if store.parents(node)[0] != nullid:
         return False
 
     meta = parsemeta(store.revision(node))[0]
 
     # copy and copyrev occur in pairs. In rare cases due to old bugs,
     # one can occur without the other. So ensure both are present to flag
     # as a copy.
     if meta and b'copy' in meta and b'copyrev' in meta:
         return meta[b'copy'], bin(meta[b'copyrev'])
 
     return False
 
 
 def filedataequivalent(store, node, filedata):
     """Determines whether file data is equivalent to a stored node.
 
     Returns True if the passed file data would hash to the same value
     as a stored revision and False otherwise.
 
     When a stored revision is censored, filedata must be empty to have
     equivalence.
 
     When a stored revision has copy metadata, it is ignored as part
     of the compare.
     """
 
     if filedata.startswith(b'\x01\n'):
         revisiontext = b'\x01\n\x01\n' + filedata
     else:
         revisiontext = filedata
 
     p1, p2 = store.parents(node)
 
     computednode = hashrevisionsha1(revisiontext, p1, p2)
 
     if computednode == node:
         return True
 
     # Censored files compare against the empty file.
     if store.iscensored(store.rev(node)):
         return filedata == b''
 
     # Renaming a file produces a different hash, even if the data
     # remains unchanged. Check if that's the case.
     if store.renamed(node):
         return store.read(node) == filedata
 
     return False
 
 
 def iterrevs(storelen, start=0, stop=None):
     """Iterate over revision numbers in a store."""
     step = 1
 
     if stop is not None:
         if start > stop:
             step = -1
         stop += step
         if stop > storelen:
             stop = storelen
     else:
         stop = storelen
 
     return pycompat.xrange(start, stop, step)
 
 
 def fileidlookup(store, fileid, identifier):
     """Resolve the file node for a value.
 
     ``store`` is an object implementing the ``ifileindex`` interface.
 
     ``fileid`` can be:
 
     * A 20 or 32 byte binary node.
     * An integer revision number
     * A 40 or 64 byte hex node.
     * A bytes that can be parsed as an integer representing a revision number.
 
     ``identifier`` is used to populate ``error.LookupError`` with an identifier
     for the store.
 
     Raises ``error.LookupError`` on failure.
     """
     if isinstance(fileid, int):
         try:
             return store.node(fileid)
         except IndexError:
             raise error.LookupError(
                 b'%d' % fileid, identifier, _(b'no match found')
             )
 
     if len(fileid) in (20, 32):
         try:
             store.rev(fileid)
             return fileid
         except error.LookupError:
             pass
 
     if len(fileid) in (40, 64):
         try:
             rawnode = bin(fileid)
             store.rev(rawnode)
             return rawnode
         except TypeError:
             pass
 
     try:
         rev = int(fileid)
 
         if b'%d' % rev != fileid:
             raise ValueError
 
         try:
             return store.node(rev)
         except (IndexError, TypeError):
             pass
     except (ValueError, OverflowError):
         pass
 
     raise error.LookupError(fileid, identifier, _(b'no match found'))
 
 
 def resolvestripinfo(minlinkrev, tiprev, headrevs, linkrevfn, parentrevsfn):
     """Resolve information needed to strip revisions.
 
     Finds the minimum revision number that must be stripped in order to
     strip ``minlinkrev``.
 
     Returns a 2-tuple of the minimum revision number to do that and a set
     of all revision numbers that have linkrevs that would be broken
     by that strip.
 
     ``tiprev`` is the current tip-most revision. It is ``len(store) - 1``.
     ``headrevs`` is an iterable of head revisions.
     ``linkrevfn`` is a callable that receives a revision and returns a linked
     revision.
     ``parentrevsfn`` is a callable that receives a revision number and returns
     an iterable of its parent revision numbers.
     """
     brokenrevs = set()
     strippoint = tiprev + 1
 
     heads = {}
     futurelargelinkrevs = set()
     for head in headrevs:
         headlinkrev = linkrevfn(head)
         heads[head] = headlinkrev
         if headlinkrev >= minlinkrev:
             futurelargelinkrevs.add(headlinkrev)
 
     # This algorithm involves walking down the rev graph, starting at the
     # heads. Since the revs are topologically sorted according to linkrev,
     # once all head linkrevs are below the minlink, we know there are
     # no more revs that could have a linkrev greater than minlink.
     # So we can stop walking.
     while futurelargelinkrevs:
         strippoint -= 1
         linkrev = heads.pop(strippoint)
 
         if linkrev < minlinkrev:
             brokenrevs.add(strippoint)
         else:
             futurelargelinkrevs.remove(linkrev)
 
         for p in parentrevsfn(strippoint):
             if p != nullrev:
                 plinkrev = linkrevfn(p)
                 heads[p] = plinkrev
                 if plinkrev >= minlinkrev:
                     futurelargelinkrevs.add(plinkrev)
 
     return strippoint, brokenrevs
 
 
 def emitrevisions(
     store,
     nodes,
     nodesorder,
     resultcls,
     deltaparentfn=None,
     candeltafn=None,
     rawsizefn=None,
     revdifffn=None,
     flagsfn=None,
     deltamode=repository.CG_DELTAMODE_STD,
     revisiondata=False,
     assumehaveparentrevisions=False,
+    sidedata_helpers=None,
 ):
     """Generic implementation of ifiledata.emitrevisions().
 
     Emitting revision data is subtly complex. This function attempts to
     encapsulate all the logic for doing so in a backend-agnostic way.
 
     ``store``
        Object conforming to ``ifilestorage`` interface.
 
     ``nodes``
        List of revision nodes whose data to emit.
 
     ``resultcls``
        A type implementing the ``irevisiondelta`` interface that will be
        constructed and returned.
 
     ``deltaparentfn`` (optional)
        Callable receiving a revision number and returning the revision number
        of a revision that the internal delta is stored against. This delta
        will be preferred over computing a new arbitrary delta.
 
        If not defined, a delta will always be computed from raw revision
        data.
 
     ``candeltafn`` (optional)
        Callable receiving a pair of revision numbers that returns a bool
        indicating whether a delta between them can be produced.
 
        If not defined, it is assumed that any two revisions can delta with
        each other.
 
     ``rawsizefn`` (optional)
        Callable receiving a revision number and returning the length of the
        ``store.rawdata(rev)``.
 
        If not defined, ``len(store.rawdata(rev))`` will be called.
 
     ``revdifffn`` (optional)
        Callable receiving a pair of revision numbers that returns a delta
        between them.
 
        If not defined, a delta will be computed by invoking mdiff code
        on ``store.revision()`` results.
 
        Defining this function allows a precomputed or stored delta to be
        used without having to compute on.
 
     ``flagsfn`` (optional)
        Callable receiving a revision number and returns the integer flags
        value for it. If not defined, flags value will be 0.
 
     ``deltamode``
        constaint on delta to be sent:
        * CG_DELTAMODE_STD  - normal mode, try to reuse storage deltas,
        * CG_DELTAMODE_PREV - only delta against "prev",
        * CG_DELTAMODE_FULL - only issue full snapshot.
 
        Whether to send fulltext revisions instead of deltas, if allowed.
 
     ``nodesorder``
     ``revisiondata``
     ``assumehaveparentrevisions``
+    ``sidedata_helpers`` (optional)
+        If not None, means that sidedata should be included.
+        A dictionary of revlog type to tuples of `(repo, computers, removers)`:
+            * `repo` is used as an argument for computers
+            * `computers` is a list of `(category, (keys, computer)` that
+               compute the missing sidedata categories that were asked:
+               * `category` is the sidedata category
+               * `keys` are the sidedata keys to be affected
+               * `computer` is the function `(repo, store, rev, sidedata)` that
+                 returns a new sidedata dict.
+            * `removers` will remove the keys corresponding to the categories
+              that are present, but not needed.
+        If both `computers` and `removers` are empty, sidedata are simply not
+        transformed.
+        Revlog types are `changelog`, `manifest` or `filelog`.
     """
 
     fnode = store.node
     frev = store.rev
 
     if nodesorder == b'nodes':
         revs = [frev(n) for n in nodes]
     elif nodesorder == b'linear':
         revs = {frev(n) for n in nodes}
         revs = dagop.linearize(revs, store.parentrevs)
     else:  # storage and default
         revs = sorted(frev(n) for n in nodes)
 
     prevrev = None
 
     if deltamode == repository.CG_DELTAMODE_PREV or assumehaveparentrevisions:
         prevrev = store.parentrevs(revs[0])[0]
 
     # Set of revs available to delta against.
     available = set()
 
     for rev in revs:
         if rev == nullrev:
             continue
 
         node = fnode(rev)
         p1rev, p2rev = store.parentrevs(rev)
 
         if deltaparentfn:
             deltaparentrev = deltaparentfn(rev)
         else:
             deltaparentrev = nullrev
 
         # Forced delta against previous mode.
         if deltamode == repository.CG_DELTAMODE_PREV:
             baserev = prevrev
 
         # We're instructed to send fulltext. Honor that.
         elif deltamode == repository.CG_DELTAMODE_FULL:
             baserev = nullrev
         # We're instructed to use p1. Honor that
         elif deltamode == repository.CG_DELTAMODE_P1:
             baserev = p1rev
 
         # There is a delta in storage. We try to use that because it
         # amounts to effectively copying data from storage and is
         # therefore the fastest.
         elif deltaparentrev != nullrev:
             # Base revision was already emitted in this group. We can
             # always safely use the delta.
             if deltaparentrev in available:
                 baserev = deltaparentrev
 
             # Base revision is a parent that hasn't been emitted already.
             # Use it if we can assume the receiver has the parent revision.
             elif assumehaveparentrevisions and deltaparentrev in (p1rev, p2rev):
                 baserev = deltaparentrev
 
             # No guarantee the receiver has the delta parent. Send delta
             # against last revision (if possible), which in the common case
             # should be similar enough to this revision that the delta is
             # reasonable.
             elif prevrev is not None:
                 baserev = prevrev
             else:
                 baserev = nullrev
 
         # Storage has a fulltext revision.
 
         # Let's use the previous revision, which is as good a guess as any.
         # There is definitely room to improve this logic.
         elif prevrev is not None:
             baserev = prevrev
         else:
             baserev = nullrev
 
         # But we can't actually use our chosen delta base for whatever
         # reason. Reset to fulltext.
         if baserev != nullrev and (candeltafn and not candeltafn(baserev, rev)):
             baserev = nullrev
 
         revision = None
         delta = None
         baserevisionsize = None
 
         if revisiondata:
             if store.iscensored(baserev) or store.iscensored(rev):
                 try:
                     revision = store.rawdata(node)
                 except error.CensoredNodeError as e:
                     revision = e.tombstone
 
                 if baserev != nullrev:
                     if rawsizefn:
                         baserevisionsize = rawsizefn(baserev)
                     else:
                         baserevisionsize = len(store.rawdata(baserev))
 
             elif (
                 baserev == nullrev and deltamode != repository.CG_DELTAMODE_PREV
             ):
                 revision = store.rawdata(node)
                 available.add(rev)
             else:
                 if revdifffn:
                     delta = revdifffn(baserev, rev)
                 else:
                     delta = mdiff.textdiff(
                         store.rawdata(baserev), store.rawdata(rev)
                     )
 
                 available.add(rev)
 
+        sidedata = None
+        if sidedata_helpers:
+            sidedata = store.sidedata(rev)
+            sidedata = run_sidedata_helpers(
+                store=store,
+                sidedata_helpers=sidedata_helpers,
+                sidedata=sidedata,
+                rev=rev,
+            )
+            sidedata = sidedatamod.serialize_sidedata(sidedata)
+
         yield resultcls(
             node=node,
             p1node=fnode(p1rev),
             p2node=fnode(p2rev),
             basenode=fnode(baserev),
             flags=flagsfn(rev) if flagsfn else 0,
             baserevisionsize=baserevisionsize,
             revision=revision,
             delta=delta,
             sidedata=sidedata,
         )
 
         prevrev = rev
 
 
+def run_sidedata_helpers(store, sidedata_helpers, sidedata, rev):
+    """Returns the sidedata for the given revision after running through
+    the given helpers.
+    - `store`: the revlog this applies to (changelog, manifest, or filelog
+      instance)
+    - `sidedata_helpers`: see `storageutil.emitrevisions`
+    - `sidedata`: previous sidedata at the given rev, if any
+    - `rev`: affected rev of `store`
+    """
+    repo, sd_computers, sd_removers = sidedata_helpers
+    kind = store.revlog_kind
+    for _keys, sd_computer in sd_computers.get(kind, []):
+        sidedata = sd_computer(repo, store, rev, sidedata)
+    for keys, _computer in sd_removers.get(kind, []):
+        for key in keys:
+            sidedata.pop(key, None)
+    return sidedata
+
+
 def deltaiscensored(delta, baserev, baselenfn):
     """Determine if a delta represents censored revision data.
 
     ``baserev`` is the base revision this delta is encoded against.
     ``baselenfn`` is a callable receiving a revision number that resolves the
     length of the revision fulltext.
 
     Returns a bool indicating if the result of the delta represents a censored
     revision.
     """
     # Fragile heuristic: unless new file meta keys are added alphabetically
     # preceding "censored", all censored revisions are prefixed by
     # "\1\ncensored:". A delta producing such a censored revision must be a
     # full-replacement delta, so we inspect the first and only patch in the
     # delta for this prefix.
     hlen = struct.calcsize(b">lll")
     if len(delta) <= hlen:
         return False
 
     oldlen = baselenfn(baserev)
     newlen = len(delta) - hlen
     if delta[:hlen] != mdiff.replacediffheader(oldlen, newlen):
         return False
 
     add = b"\1\ncensored:"
     addlen = len(add)
     return newlen >= addlen and delta[hlen : hlen + addlen] == add

tests/simplestorerepo.py

@@ -1,738 +1,740 b''
 # simplestorerepo.py - Extension that swaps in alternate repository storage.
 #
 # Copyright 2018 Gregory Szorc <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 # To use this with the test suite:
 #
 #   $ HGREPOFEATURES="simplestore" ./run-tests.py \
 #       --extra-config-opt extensions.simplestore=`pwd`/simplestorerepo.py
 
 from __future__ import absolute_import
 
 import stat
 
 from mercurial.i18n import _
 from mercurial.node import (
     bin,
     hex,
     nullid,
     nullrev,
 )
 from mercurial.thirdparty import attr
 from mercurial import (
     ancestor,
     bundlerepo,
     error,
     extensions,
     localrepo,
     mdiff,
     pycompat,
     revlog,
     store,
     verify,
 )
 from mercurial.interfaces import (
     repository,
     util as interfaceutil,
 )
 from mercurial.utils import (
     cborutil,
     storageutil,
 )
 from mercurial.revlogutils import flagutil
 
 # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
 # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
 # be specifying the version(s) of Mercurial they are tested with, or
 # leave the attribute unspecified.
 testedwith = b'ships-with-hg-core'
 
 REQUIREMENT = b'testonly-simplestore'
 
 
 def validatenode(node):
     if isinstance(node, int):
         raise ValueError('expected node; got int')
 
     if len(node) != 20:
         raise ValueError('expected 20 byte node')
 
 
 def validaterev(rev):
     if not isinstance(rev, int):
         raise ValueError('expected int')
 
 
 class simplestoreerror(error.StorageError):
     pass
 
 
 @interfaceutil.implementer(repository.irevisiondelta)
 @attr.s(slots=True)
 class simplestorerevisiondelta(object):
     node = attr.ib()
     p1node = attr.ib()
     p2node = attr.ib()
     basenode = attr.ib()
     flags = attr.ib()
     baserevisionsize = attr.ib()
     revision = attr.ib()
     delta = attr.ib()
     linknode = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.iverifyproblem)
 @attr.s(frozen=True)
 class simplefilestoreproblem(object):
     warning = attr.ib(default=None)
     error = attr.ib(default=None)
     node = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.ifilestorage)
 class filestorage(object):
     """Implements storage for a tracked path.
 
     Data is stored in the VFS in a directory corresponding to the tracked
     path.
 
     Index data is stored in an ``index`` file using CBOR.
 
     Fulltext data is stored in files having names of the node.
     """
 
     _flagserrorclass = simplestoreerror
 
     def __init__(self, svfs, path):
         self._svfs = svfs
         self._path = path
 
         self._storepath = b'/'.join([b'data', path])
         self._indexpath = b'/'.join([self._storepath, b'index'])
 
         indexdata = self._svfs.tryread(self._indexpath)
         if indexdata:
             indexdata = cborutil.decodeall(indexdata)
 
         self._indexdata = indexdata or []
         self._indexbynode = {}
         self._indexbyrev = {}
         self._index = []
         self._refreshindex()
 
         self._flagprocessors = dict(flagutil.flagprocessors)
 
     def _refreshindex(self):
         self._indexbynode.clear()
         self._indexbyrev.clear()
         self._index = []
 
         for i, entry in enumerate(self._indexdata):
             self._indexbynode[entry[b'node']] = entry
             self._indexbyrev[i] = entry
 
         self._indexbynode[nullid] = {
             b'node': nullid,
             b'p1': nullid,
             b'p2': nullid,
             b'linkrev': nullrev,
             b'flags': 0,
         }
 
         self._indexbyrev[nullrev] = {
             b'node': nullid,
             b'p1': nullid,
             b'p2': nullid,
             b'linkrev': nullrev,
             b'flags': 0,
         }
 
         for i, entry in enumerate(self._indexdata):
             p1rev, p2rev = self.parentrevs(self.rev(entry[b'node']))
 
             # start, length, rawsize, chainbase, linkrev, p1, p2, node
             self._index.append(
                 (0, 0, 0, -1, entry[b'linkrev'], p1rev, p2rev, entry[b'node'])
             )
 
         self._index.append((0, 0, 0, -1, -1, -1, -1, nullid))
 
     def __len__(self):
         return len(self._indexdata)
 
     def __iter__(self):
         return iter(range(len(self)))
 
     def revs(self, start=0, stop=None):
         step = 1
         if stop is not None:
             if start > stop:
                 step = -1
 
             stop += step
         else:
             stop = len(self)
 
         return range(start, stop, step)
 
     def parents(self, node):
         validatenode(node)
 
         if node not in self._indexbynode:
             raise KeyError('unknown node')
 
         entry = self._indexbynode[node]
 
         return entry[b'p1'], entry[b'p2']
 
     def parentrevs(self, rev):
         p1, p2 = self.parents(self._indexbyrev[rev][b'node'])
         return self.rev(p1), self.rev(p2)
 
     def rev(self, node):
         validatenode(node)
 
         try:
             self._indexbynode[node]
         except KeyError:
             raise error.LookupError(node, self._indexpath, _('no node'))
 
         for rev, entry in self._indexbyrev.items():
             if entry[b'node'] == node:
                 return rev
 
         raise error.ProgrammingError(b'this should not occur')
 
     def node(self, rev):
         validaterev(rev)
 
         return self._indexbyrev[rev][b'node']
 
     def hasnode(self, node):
         validatenode(node)
         return node in self._indexbynode
 
     def censorrevision(self, tr, censornode, tombstone=b''):
         raise NotImplementedError('TODO')
 
     def lookup(self, node):
         if isinstance(node, int):
             return self.node(node)
 
         if len(node) == 20:
             self.rev(node)
             return node
 
         try:
             rev = int(node)
             if '%d' % rev != node:
                 raise ValueError
 
             if rev < 0:
                 rev = len(self) + rev
             if rev < 0 or rev >= len(self):
                 raise ValueError
 
             return self.node(rev)
         except (ValueError, OverflowError):
             pass
 
         if len(node) == 40:
             try:
                 rawnode = bin(node)
                 self.rev(rawnode)
                 return rawnode
             except TypeError:
                 pass
 
         raise error.LookupError(node, self._path, _('invalid lookup input'))
 
     def linkrev(self, rev):
         validaterev(rev)
 
         return self._indexbyrev[rev][b'linkrev']
 
     def _flags(self, rev):
         validaterev(rev)
 
         return self._indexbyrev[rev][b'flags']
 
     def _candelta(self, baserev, rev):
         validaterev(baserev)
         validaterev(rev)
 
         if (self._flags(baserev) & revlog.REVIDX_RAWTEXT_CHANGING_FLAGS) or (
             self._flags(rev) & revlog.REVIDX_RAWTEXT_CHANGING_FLAGS
         ):
             return False
 
         return True
 
     def checkhash(self, text, node, p1=None, p2=None, rev=None):
         if p1 is None and p2 is None:
             p1, p2 = self.parents(node)
         if node != storageutil.hashrevisionsha1(text, p1, p2):
             raise simplestoreerror(
                 _("integrity check failed on %s") % self._path
             )
 
     def revision(self, nodeorrev, raw=False):
         if isinstance(nodeorrev, int):
             node = self.node(nodeorrev)
         else:
             node = nodeorrev
         validatenode(node)
 
         if node == nullid:
             return b''
 
         rev = self.rev(node)
         flags = self._flags(rev)
 
         path = b'/'.join([self._storepath, hex(node)])
         rawtext = self._svfs.read(path)
 
         if raw:
             validatehash = flagutil.processflagsraw(self, rawtext, flags)
             text = rawtext
         else:
             r = flagutil.processflagsread(self, rawtext, flags)
             text, validatehash = r
         if validatehash:
             self.checkhash(text, node, rev=rev)
 
         return text
 
     def rawdata(self, nodeorrev):
         return self.revision(raw=True)
 
     def read(self, node):
         validatenode(node)
 
         revision = self.revision(node)
 
         if not revision.startswith(b'\1\n'):
             return revision
 
         start = revision.index(b'\1\n', 2)
         return revision[start + 2 :]
 
     def renamed(self, node):
         validatenode(node)
 
         if self.parents(node)[0] != nullid:
             return False
 
         fulltext = self.revision(node)
         m = storageutil.parsemeta(fulltext)[0]
 
         if m and 'copy' in m:
             return m['copy'], bin(m['copyrev'])
 
         return False
 
     def cmp(self, node, text):
         validatenode(node)
 
         t = text
 
         if text.startswith(b'\1\n'):
             t = b'\1\n\1\n' + text
 
         p1, p2 = self.parents(node)
 
         if storageutil.hashrevisionsha1(t, p1, p2) == node:
             return False
 
         if self.iscensored(self.rev(node)):
             return text != b''
 
         if self.renamed(node):
             t2 = self.read(node)
             return t2 != text
 
         return True
 
     def size(self, rev):
         validaterev(rev)
 
         node = self._indexbyrev[rev][b'node']
 
         if self.renamed(node):
             return len(self.read(node))
 
         if self.iscensored(rev):
             return 0
 
         return len(self.revision(node))
 
     def iscensored(self, rev):
         validaterev(rev)
 
         return self._flags(rev) & repository.REVISION_FLAG_CENSORED
 
     def commonancestorsheads(self, a, b):
         validatenode(a)
         validatenode(b)
 
         a = self.rev(a)
         b = self.rev(b)
 
         ancestors = ancestor.commonancestorsheads(self.parentrevs, a, b)
         return pycompat.maplist(self.node, ancestors)
 
     def descendants(self, revs):
         # This is a copy of revlog.descendants()
         first = min(revs)
         if first == nullrev:
             for i in self:
                 yield i
             return
 
         seen = set(revs)
         for i in self.revs(start=first + 1):
             for x in self.parentrevs(i):
                 if x != nullrev and x in seen:
                     seen.add(i)
                     yield i
                     break
 
     # Required by verify.
     def files(self):
         entries = self._svfs.listdir(self._storepath)
 
         # Strip out undo.backup.* files created as part of transaction
         # recording.
         entries = [f for f in entries if not f.startswith('undo.backup.')]
 
         return [b'/'.join((self._storepath, f)) for f in entries]
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         # TODO do a real implementation of this
         return {
             'exclusivefiles': [],
             'sharedfiles': [],
             'revisionscount': len(self),
             'trackedsize': 0,
             'storedsize': None,
         }
 
     def verifyintegrity(self, state):
         state['skipread'] = set()
         for rev in self:
             node = self.node(rev)
             try:
                 self.revision(node)
             except Exception as e:
                 yield simplefilestoreproblem(
                     error='unpacking %s: %s' % (node, e), node=node
                 )
                 state['skipread'].add(node)
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
+        sidedata_helpers=None,
     ):
         # TODO this will probably break on some ordering options.
         nodes = [n for n in nodes if n != nullid]
         if not nodes:
             return
         for delta in storageutil.emitrevisions(
             self,
             nodes,
             nodesorder,
             simplestorerevisiondelta,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
             deltamode=deltamode,
+            sidedata_helpers=sidedata_helpers,
         ):
             yield delta
 
     def add(self, text, meta, transaction, linkrev, p1, p2):
         if meta or text.startswith(b'\1\n'):
             text = storageutil.packmeta(meta, text)
 
         return self.addrevision(text, transaction, linkrev, p1, p2)
 
     def addrevision(
         self,
         text,
         transaction,
         linkrev,
         p1,
         p2,
         node=None,
         flags=revlog.REVIDX_DEFAULT_FLAGS,
         cachedelta=None,
     ):
         validatenode(p1)
         validatenode(p2)
 
         if flags:
             node = node or storageutil.hashrevisionsha1(text, p1, p2)
 
         rawtext, validatehash = flagutil.processflagswrite(self, text, flags)
 
         node = node or storageutil.hashrevisionsha1(text, p1, p2)
 
         if node in self._indexbynode:
             return node
 
         if validatehash:
             self.checkhash(rawtext, node, p1=p1, p2=p2)
 
         return self._addrawrevision(
             node, rawtext, transaction, linkrev, p1, p2, flags
         )
 
     def _addrawrevision(self, node, rawtext, transaction, link, p1, p2, flags):
         transaction.addbackup(self._indexpath)
 
         path = b'/'.join([self._storepath, hex(node)])
 
         self._svfs.write(path, rawtext)
 
         self._indexdata.append(
             {
                 b'node': node,
                 b'p1': p1,
                 b'p2': p2,
                 b'linkrev': link,
                 b'flags': flags,
             }
         )
 
         self._reflectindexupdate()
 
         return node
 
     def _reflectindexupdate(self):
         self._refreshindex()
         self._svfs.write(
             self._indexpath, ''.join(cborutil.streamencode(self._indexdata))
         )
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         addrevisioncb=None,
         duplicaterevisioncb=None,
         maybemissingparents=False,
     ):
         if maybemissingparents:
             raise error.Abort(
                 _('simple store does not support missing parents ' 'write mode')
             )
 
         empty = True
 
         transaction.addbackup(self._indexpath)
 
         for node, p1, p2, linknode, deltabase, delta, flags in deltas:
             linkrev = linkmapper(linknode)
             flags = flags or revlog.REVIDX_DEFAULT_FLAGS
 
             if node in self._indexbynode:
                 if duplicaterevisioncb:
                     duplicaterevisioncb(self, self.rev(node))
                 empty = False
                 continue
 
             # Need to resolve the fulltext from the delta base.
             if deltabase == nullid:
                 text = mdiff.patch(b'', delta)
             else:
                 text = mdiff.patch(self.revision(deltabase), delta)
 
             rev = self._addrawrevision(
                 node, text, transaction, linkrev, p1, p2, flags
             )
 
             if addrevisioncb:
                 addrevisioncb(self, rev)
             empty = False
         return not empty
 
     def _headrevs(self):
         # Assume all revisions are heads by default.
         revishead = {rev: True for rev in self._indexbyrev}
 
         for rev, entry in self._indexbyrev.items():
             # Unset head flag for all seen parents.
             revishead[self.rev(entry[b'p1'])] = False
             revishead[self.rev(entry[b'p2'])] = False
 
         return [rev for rev, ishead in sorted(revishead.items()) if ishead]
 
     def heads(self, start=None, stop=None):
         # This is copied from revlog.py.
         if start is None and stop is None:
             if not len(self):
                 return [nullid]
             return [self.node(r) for r in self._headrevs()]
 
         if start is None:
             start = nullid
         if stop is None:
             stop = []
         stoprevs = {self.rev(n) for n in stop}
         startrev = self.rev(start)
         reachable = {startrev}
         heads = {startrev}
 
         parentrevs = self.parentrevs
         for r in self.revs(start=startrev + 1):
             for p in parentrevs(r):
                 if p in reachable:
                     if r not in stoprevs:
                         reachable.add(r)
                     heads.add(r)
                 if p in heads and p not in stoprevs:
                     heads.remove(p)
 
         return [self.node(r) for r in heads]
 
     def children(self, node):
         validatenode(node)
 
         # This is a copy of revlog.children().
         c = []
         p = self.rev(node)
         for r in self.revs(start=p + 1):
             prevs = [pr for pr in self.parentrevs(r) if pr != nullrev]
             if prevs:
                 for pr in prevs:
                     if pr == p:
                         c.append(self.node(r))
             elif p == nullrev:
                 c.append(self.node(r))
         return c
 
     def getstrippoint(self, minlink):
         return storageutil.resolvestripinfo(
             minlink,
             len(self) - 1,
             self._headrevs(),
             self.linkrev,
             self.parentrevs,
         )
 
     def strip(self, minlink, transaction):
         if not len(self):
             return
 
         rev, _ignored = self.getstrippoint(minlink)
         if rev == len(self):
             return
 
         # Purge index data starting at the requested revision.
         self._indexdata[rev:] = []
         self._reflectindexupdate()
 
 
 def issimplestorefile(f, kind, st):
     if kind != stat.S_IFREG:
         return False
 
     if store.isrevlog(f, kind, st):
         return False
 
     # Ignore transaction undo files.
     if f.startswith('undo.'):
         return False
 
     # Otherwise assume it belongs to the simple store.
     return True
 
 
 class simplestore(store.encodedstore):
     def datafiles(self):
         for x in super(simplestore, self).datafiles():
             yield x
 
         # Supplement with non-revlog files.
         extrafiles = self._walk('data', True, filefilter=issimplestorefile)
 
         for unencoded, encoded, size in extrafiles:
             try:
                 unencoded = store.decodefilename(unencoded)
             except KeyError:
                 unencoded = None
 
             yield unencoded, encoded, size
 
 
 def reposetup(ui, repo):
     if not repo.local():
         return
 
     if isinstance(repo, bundlerepo.bundlerepository):
         raise error.Abort(_('cannot use simple store with bundlerepo'))
 
     class simplestorerepo(repo.__class__):
         def file(self, f):
             return filestorage(self.svfs, f)
 
     repo.__class__ = simplestorerepo
 
 
 def featuresetup(ui, supported):
     supported.add(REQUIREMENT)
 
 
 def newreporequirements(orig, ui, createopts):
     """Modifies default requirements for new repos to use the simple store."""
     requirements = orig(ui, createopts)
 
     # These requirements are only used to affect creation of the store
     # object. We have our own store. So we can remove them.
     # TODO do this once we feel like taking the test hit.
     # if 'fncache' in requirements:
     #    requirements.remove('fncache')
     # if 'dotencode' in requirements:
     #    requirements.remove('dotencode')
 
     requirements.add(REQUIREMENT)
 
     return requirements
 
 
 def makestore(orig, requirements, path, vfstype):
     if REQUIREMENT not in requirements:
         return orig(requirements, path, vfstype)
 
     return simplestore(path, vfstype)
 
 
 def verifierinit(orig, self, *args, **kwargs):
     orig(self, *args, **kwargs)
 
     # We don't care that files in the store don't align with what is
     # advertised. So suppress these warnings.
     self.warnorphanstorefiles = False
 
 
 def extsetup(ui):
     localrepo.featuresetupfuncs.add(featuresetup)
 
     extensions.wrapfunction(
         localrepo, 'newreporequirements', newreporequirements
     )
     extensions.wrapfunction(localrepo, 'makestore', makestore)
     extensions.wrapfunction(verify.verifier, '__init__', verifierinit)