inline-changelog: fix a critical bug in write_pending that delete data

Since a93e52f0b6ff we no longer use inline-revlog for the changelog. The goal there was to
solve the lack of testing for the two variants (inline vs split) and reduce the
complexity of the interaction with "diverted-write" on the changelog level.

However many existing repository still have inline-changelog and we
automatically move them to normal revlog as soon as we have the chances.

Unfortunately This conversion is buggy and can result in the destruction of the
changelog.i if hook triggers the "write pending" mechanism.

The bugs comes from the "revlog splitting" logic and the "write_pending" logic
stepping over each other. Ironically the change in a93e52f0b6ff aims at no
longer having this kind of problem.

This changesets fix this issue and add associated tests.

Fixing this reveal that the transaction hooks end up not seeing the pending
transaction content, because the name is not right ("changelog.i.s.a" instead of
"changelog.i.s") we fix this in the next changeset.

tests/test-split-legacy-inline-changelog.t

@@ -0,0 +1,307 b''
+======================================================
+Test operation on repository with an inlined changelog
+======================================================
+
+Inlined revlog has been a bag of complexity for a long time and the combination
+with special transaction logic on the changelog was a long source of bugs
+poorly covered by the test suites.
+
+We stopped doing any usage of inlined-revlog for changelog in a93e52f0b6ff,
+upgrading legacy inlined version as soon as possible when we see them. However
+this Mercurial does not produce such inlined-changelog that case is very poorly
+covered in the test suites. This test file aims at covering these cases.
+
+Double checking test data
+=========================
+
+We should have a repository around
+
+  $ mkdir sanity-check
+  $ cd sanity-check
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ cd inlined-changelog
+  $ hg root
+  $TESTTMP/sanity-check/inlined-changelog
+
+The repository should not be corrupted initially
+
+  $ hg verify
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 1 changesets with 1 changes to 1 files
+
+The changelog of that repository MUST be inlined
+
+  $ hg debugrevlog -c | grep -E '^flags\b'
+  flags  : inline
+
+Touching that repository MUST split that inlined changelog
+
+  $ hg branch foo --quiet
+  $ hg commit -m foo --quiet
+  $ hg debugrevlog -c | grep -E '^flags\b'
+  flags  : (none)
+
+  $ cd ../..
+
+Test doing a simple commit
+==========================
+
+Simple commit
+-------------
+
+  $ mkdir simple-commit
+  $ cd simple-commit
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ cd inlined-changelog
+  $ hg up --quiet
+  $ hg log -GT '[{rev}] {desc}\n'
+  @  [0] first commit
+  
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  $ hg verify
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -GT '[{rev}] {desc}\n'
+  @  [1] second changeset
+  |
+  o  [0] first commit
+  
+  $ cd ../..
+
+Simple commit with a pretxn hook configured
+-------------------------------------------
+
+Before 6.7.3 this used to delete the changelog index
+
+  $ mkdir pretxnclose-commit
+  $ cd pretxnclose-commit
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ cat >> inlined-changelog/.hg/hgrc <<EOF
+  > [hooks]
+  > pretxnclose=hg log -r tip -T "pre-txn tip rev: {rev}\n"
+  > EOF
+  $ cd inlined-changelog
+  $ hg up --quiet
+  $ hg log -GT '[{rev}] {desc}\n'
+  @  [0] first commit
+  
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  pre-txn tip rev: 1 (missing-correct-output !)
+  warning: ignoring unknown working parent 11b63e930bf2! (known-bad-output !)
+  pre-txn tip rev: 0 (known-bad-output !)
+  $ hg verify
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -GT '[{rev}] {desc}\n'
+  @  [1] second changeset
+  |
+  o  [0] first commit
+  
+  $ cd ../..
+
+Test pushing to a repository with a repository revlog
+=====================================================
+
+Simple local push
+-----------------
+
+  $ mkdir simple-local-push
+  $ cd simple-local-push
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [0] first commit
+
+  $ hg clone --pull inlined-changelog client
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 1 changesets with 1 changes to 1 files
+  new changesets 827f11bfd362
+  updating to branch default
+  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ cd client
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  $ hg push
+  pushing to $TESTTMP/*/inlined-changelog (glob)
+  searching for changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 1 changesets with 1 changes to 1 files
+  $ cd ..
+
+  $ hg verify -R inlined-changelog
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [1] second changeset
+  [0] first commit
+  $ cd ..
+
+Simple local push with a pretxnchangegroup hook
+-----------------------------------------------
+
+Before 6.7.3 this used to delete the server changelog
+
+  $ mkdir pretxnchangegroup-local-push
+  $ cd pretxnchangegroup-local-push
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ cat >> inlined-changelog/.hg/hgrc <<EOF
+  > [hooks]
+  > pretxnchangegroup=hg log -r tip -T "pre-txn tip rev: {rev}\n"
+  > EOF
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [0] first commit
+
+  $ hg clone --pull inlined-changelog client
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 1 changesets with 1 changes to 1 files
+  new changesets 827f11bfd362
+  updating to branch default
+  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ cd client
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  $ hg push
+  pushing to $TESTTMP/*/inlined-changelog (glob)
+  searching for changes
+  adding changesets
+  adding manifests
+  adding file changes
+  pre-txn tip rev: 1 (missing-correct-output !)
+  pre-txn tip rev: 0 (known-bad-output !)
+  added 1 changesets with 1 changes to 1 files
+  $ cd ..
+
+  $ hg verify -R inlined-changelog
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [1] second changeset
+  [0] first commit
+  $ cd ..
+
+Simple ssh push
+-----------------
+
+  $ mkdir simple-ssh-push
+  $ cd simple-ssh-push
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [0] first commit
+
+  $ hg clone ssh://user@dummy/"`pwd`"/inlined-changelog client
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 1 changesets with 1 changes to 1 files
+  new changesets 827f11bfd362
+  updating to branch default
+  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ cd client
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  $ hg push
+  pushing to ssh://user@dummy/$TESTTMP/simple-ssh-push/inlined-changelog
+  searching for changes
+  remote: adding changesets
+  remote: adding manifests
+  remote: adding file changes
+  remote: added 1 changesets with 1 changes to 1 files
+  $ cd ..
+
+  $ hg verify -R inlined-changelog
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [1] second changeset
+  [0] first commit
+  $ cd ..
+
+Simple ssh push with a pretxnchangegroup hook
+-----------------------------------------------
+
+Before 6.7.3 this used to delete the server changelog
+
+  $ mkdir pretxnchangegroup-ssh-push
+  $ cd pretxnchangegroup-ssh-push
+  $ tar xf $TESTDIR/bundles/inlined-changelog.tar
+  $ cat >> inlined-changelog/.hg/hgrc <<EOF
+  > [hooks]
+  > pretxnchangegroup=hg log -r tip -T "pre-txn tip rev: {rev}\n"
+  > EOF
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [0] first commit
+
+  $ hg clone ssh://user@dummy/"`pwd`"/inlined-changelog client
+  requesting all changes
+  adding changesets
+  adding manifests
+  adding file changes
+  added 1 changesets with 1 changes to 1 files
+  new changesets 827f11bfd362
+  updating to branch default
+  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
+  $ cd client
+  $ echo b > b
+  $ hg add b
+  $ hg commit -m "second changeset"
+  $ hg push
+  pushing to ssh://user@dummy/$TESTTMP/pretxnchangegroup-ssh-push/inlined-changelog
+  searching for changes
+  remote: adding changesets
+  remote: adding manifests
+  remote: adding file changes
+  remote: pre-txn tip rev: 1 (missing-correct-output !)
+  remote: pre-txn tip rev: 0 (known-bad-output !)
+  remote: added 1 changesets with 1 changes to 1 files
+  $ cd ..
+
+  $ hg verify -R inlined-changelog
+  checking changesets
+  checking manifests
+  crosschecking files in changesets and manifests
+  checking files
+  checking dirstate
+  checked 2 changesets with 2 changes to 2 files
+  $ hg log -R inlined-changelog -T '[{rev}] {desc}\n'
+  [1] second changeset
+  [0] first commit
+  $ cd ..

mercurial/changelog.py

@@ -1,507 +1,509 b''
 # changelog.py - changelog class for mercurial
 #
 # Copyright 2005-2007 Olivia 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 .i18n import _
 from .node import (
     bin,
     hex,
 )
 from .thirdparty import attr
 
 from . import (
     encoding,
     error,
     metadata,
     pycompat,
     revlog,
 )
 from .utils import (
     dateutil,
     stringutil,
 )
 from .revlogutils import (
     constants as revlog_constants,
     flagutil,
 )
 
 _defaultextra = {b'branch': b'default'}
 
 
 def _string_escape(text):
     """
     >>> from .pycompat import bytechr as chr
     >>> d = {b'nl': chr(10), b'bs': chr(92), b'cr': chr(13), b'nul': chr(0)}
     >>> s = b"ab%(nl)scd%(bs)s%(bs)sn%(nul)s12ab%(cr)scd%(bs)s%(nl)s" % d
     >>> s
     'ab\\ncd\\\\\\\\n\\x0012ab\\rcd\\\\\\n'
     >>> res = _string_escape(s)
     >>> s == _string_unescape(res)
     True
     """
     # subset of the string_escape codec
     text = (
         text.replace(b'\\', b'\\\\')
         .replace(b'\n', b'\\n')
         .replace(b'\r', b'\\r')
     )
     return text.replace(b'\0', b'\\0')
 
 
 def _string_unescape(text):
     if b'\\0' in text:
         # fix up \0 without getting into trouble with \\0
         text = text.replace(b'\\\\', b'\\\\\n')
         text = text.replace(b'\\0', b'\0')
         text = text.replace(b'\n', b'')
     return stringutil.unescapestr(text)
 
 
 def decodeextra(text):
     """
     >>> from .pycompat import bytechr as chr
     >>> sorted(decodeextra(encodeextra({b'foo': b'bar', b'baz': chr(0) + b'2'})
     ...                    ).items())
     [('baz', '\\x002'), ('branch', 'default'), ('foo', 'bar')]
     >>> sorted(decodeextra(encodeextra({b'foo': b'bar',
     ...                                 b'baz': chr(92) + chr(0) + b'2'})
     ...                    ).items())
     [('baz', '\\\\\\x002'), ('branch', 'default'), ('foo', 'bar')]
     """
     extra = _defaultextra.copy()
     for l in text.split(b'\0'):
         if l:
             k, v = _string_unescape(l).split(b':', 1)
             extra[k] = v
     return extra
 
 
 def encodeextra(d):
     # keys must be sorted to produce a deterministic changelog entry
     items = [_string_escape(b'%s:%s' % (k, d[k])) for k in sorted(d)]
     return b"\0".join(items)
 
 
 def stripdesc(desc):
     """strip trailing whitespace and leading and trailing empty lines"""
     return b'\n'.join([l.rstrip() for l in desc.splitlines()]).strip(b'\n')
 
 
 @attr.s
 class _changelogrevision:
     # Extensions might modify _defaultextra, so let the constructor below pass
     # it in
     extra = attr.ib()
     manifest = attr.ib()
     user = attr.ib(default=b'')
     date = attr.ib(default=(0, 0))
     files = attr.ib(default=attr.Factory(list))
     filesadded = attr.ib(default=None)
     filesremoved = attr.ib(default=None)
     p1copies = attr.ib(default=None)
     p2copies = attr.ib(default=None)
     description = attr.ib(default=b'')
     branchinfo = attr.ib(default=(_defaultextra[b'branch'], False))
 
 
 class changelogrevision:
     """Holds results of a parsed changelog revision.
 
     Changelog revisions consist of multiple pieces of data, including
     the manifest node, user, and date. This object exposes a view into
     the parsed object.
     """
 
     __slots__ = (
         '_offsets',
         '_text',
         '_sidedata',
         '_cpsd',
         '_changes',
     )
 
     def __new__(cls, cl, text, sidedata, cpsd):
         if not text:
             return _changelogrevision(extra=_defaultextra, manifest=cl.nullid)
 
         self = super(changelogrevision, cls).__new__(cls)
         # We could return here and implement the following as an __init__.
         # But doing it here is equivalent and saves an extra function call.
 
         # format used:
         # nodeid\n        : manifest node in ascii
         # user\n          : user, no \n or \r allowed
         # time tz extra\n : date (time is int or float, timezone is int)
         #                 : extra is metadata, encoded and separated by '\0'
         #                 : older versions ignore it
         # files\n\n       : files modified by the cset, no \n or \r allowed
         # (.*)            : comment (free text, ideally utf-8)
         #
         # changelog v0 doesn't use extra
 
         nl1 = text.index(b'\n')
         nl2 = text.index(b'\n', nl1 + 1)
         nl3 = text.index(b'\n', nl2 + 1)
 
         # The list of files may be empty. Which means nl3 is the first of the
         # double newline that precedes the description.
         if text[nl3 + 1 : nl3 + 2] == b'\n':
             doublenl = nl3
         else:
             doublenl = text.index(b'\n\n', nl3 + 1)
 
         self._offsets = (nl1, nl2, nl3, doublenl)
         self._text = text
         self._sidedata = sidedata
         self._cpsd = cpsd
         self._changes = None
 
         return self
 
     @property
     def manifest(self):
         return bin(self._text[0 : self._offsets[0]])
 
     @property
     def user(self):
         off = self._offsets
         return encoding.tolocal(self._text[off[0] + 1 : off[1]])
 
     @property
     def _rawdate(self):
         off = self._offsets
         dateextra = self._text[off[1] + 1 : off[2]]
         return dateextra.split(b' ', 2)[0:2]
 
     @property
     def _rawextra(self):
         off = self._offsets
         dateextra = self._text[off[1] + 1 : off[2]]
         fields = dateextra.split(b' ', 2)
         if len(fields) != 3:
             return None
 
         return fields[2]
 
     @property
     def date(self):
         raw = self._rawdate
         time = float(raw[0])
         # Various tools did silly things with the timezone.
         try:
             timezone = int(raw[1])
         except ValueError:
             timezone = 0
 
         return time, timezone
 
     @property
     def extra(self):
         raw = self._rawextra
         if raw is None:
             return _defaultextra
 
         return decodeextra(raw)
 
     @property
     def changes(self):
         if self._changes is not None:
             return self._changes
         if self._cpsd:
             changes = metadata.decode_files_sidedata(self._sidedata)
         else:
             changes = metadata.ChangingFiles(
                 touched=self.files or (),
                 added=self.filesadded or (),
                 removed=self.filesremoved or (),
                 p1_copies=self.p1copies or {},
                 p2_copies=self.p2copies or {},
             )
         self._changes = changes
         return changes
 
     @property
     def files(self):
         if self._cpsd:
             return sorted(self.changes.touched)
         off = self._offsets
         if off[2] == off[3]:
             return []
 
         return self._text[off[2] + 1 : off[3]].split(b'\n')
 
     @property
     def filesadded(self):
         if self._cpsd:
             return self.changes.added
         else:
             rawindices = self.extra.get(b'filesadded')
         if rawindices is None:
             return None
         return metadata.decodefileindices(self.files, rawindices)
 
     @property
     def filesremoved(self):
         if self._cpsd:
             return self.changes.removed
         else:
             rawindices = self.extra.get(b'filesremoved')
         if rawindices is None:
             return None
         return metadata.decodefileindices(self.files, rawindices)
 
     @property
     def p1copies(self):
         if self._cpsd:
             return self.changes.copied_from_p1
         else:
             rawcopies = self.extra.get(b'p1copies')
         if rawcopies is None:
             return None
         return metadata.decodecopies(self.files, rawcopies)
 
     @property
     def p2copies(self):
         if self._cpsd:
             return self.changes.copied_from_p2
         else:
             rawcopies = self.extra.get(b'p2copies')
         if rawcopies is None:
             return None
         return metadata.decodecopies(self.files, rawcopies)
 
     @property
     def description(self):
         return encoding.tolocal(self._text[self._offsets[3] + 2 :])
 
     @property
     def branchinfo(self):
         extra = self.extra
         return encoding.tolocal(extra.get(b"branch")), b'close' in extra
 
 
 class changelog(revlog.revlog):
     def __init__(self, opener, trypending=False, concurrencychecker=None):
         """Load a changelog revlog using an opener.
 
         If ``trypending`` is true, we attempt to load the index from a
         ``00changelog.i.a`` file instead of the default ``00changelog.i``.
         The ``00changelog.i.a`` file contains index (and possibly inline
         revision) data for a transaction that hasn't been finalized yet.
         It exists in a separate file to facilitate readers (such as
         hooks processes) accessing data before a transaction is finalized.
 
         ``concurrencychecker`` will be passed to the revlog init function, see
         the documentation there.
         """
         revlog.revlog.__init__(
             self,
             opener,
             target=(revlog_constants.KIND_CHANGELOG, None),
             radix=b'00changelog',
             checkambig=True,
             mmaplargeindex=True,
             persistentnodemap=opener.options.get(b'persistent-nodemap', False),
             concurrencychecker=concurrencychecker,
             trypending=trypending,
             may_inline=False,
         )
 
         if self._initempty and (self._format_version == revlog.REVLOGV1):
             # changelogs don't benefit from generaldelta.
 
             self._format_flags &= ~revlog.FLAG_GENERALDELTA
             self.delta_config.general_delta = False
 
         # Delta chains for changelogs tend to be very small because entries
         # tend to be small and don't delta well with each. So disable delta
         # chains.
         self._storedeltachains = False
 
         self._v2_delayed = False
         self._filteredrevs = frozenset()
         self._filteredrevs_hashcache = {}
         self._copiesstorage = opener.options.get(b'copies-storage')
 
     @property
     def filteredrevs(self):
         return self._filteredrevs
 
     @filteredrevs.setter
     def filteredrevs(self, val):
         # Ensure all updates go through this function
         assert isinstance(val, frozenset)
         self._filteredrevs = val
         self._filteredrevs_hashcache = {}
 
     def _write_docket(self, tr):
         if not self._v2_delayed:
             super(changelog, self)._write_docket(tr)
 
     def delayupdate(self, tr):
         """delay visibility of index updates to other readers"""
         assert not self._inner.is_open
         assert not self._may_inline
         # enforce that older changelog that are still inline are split at the
         # first opportunity.
         if self._inline:
             self._enforceinlinesize(tr)
         if self._docket is not None:
             self._v2_delayed = True
         else:
             new_index = self._inner.delay()
             if new_index is not None:
                 self._indexfile = new_index
                 tr.registertmp(new_index)
-        tr.addpending(b'cl-%i' % id(self), self._writepending)
-        tr.addfinalize(b'cl-%i' % id(self), self._finalize)
+        # use "000" as prefix to make sure we run before the spliting of legacy
+        # inline changelog..
+        tr.addpending(b'000-cl-%i' % id(self), self._writepending)
+        tr.addfinalize(b'000-cl-%i' % id(self), self._finalize)
 
     def _finalize(self, tr):
         """finalize index updates"""
         assert not self._inner.is_open
         if self._docket is not None:
             self._docket.write(tr)
             self._v2_delayed = False
         else:
             new_index_file = self._inner.finalize_pending()
             self._indexfile = new_index_file
             if self._inline:
                 msg = 'changelog should not be inline at that point'
                 raise error.ProgrammingError(msg)
 
     def _writepending(self, tr):
         """create a file containing the unfinalized state for
         pretxnchangegroup"""
         assert not self._inner.is_open
         if self._docket:
             any_pending = self._docket.write(tr, pending=True)
             self._v2_delayed = False
         else:
             new_index, any_pending = self._inner.write_pending()
             if new_index is not None:
                 self._indexfile = new_index
                 tr.registertmp(new_index)
         return any_pending
 
     def _enforceinlinesize(self, tr):
         if not self.is_delaying:
             revlog.revlog._enforceinlinesize(self, tr)
 
     def read(self, nodeorrev):
         """Obtain data from a parsed changelog revision.
 
         Returns a 6-tuple of:
 
            - manifest node in binary
            - author/user as a localstr
            - date as a 2-tuple of (time, timezone)
            - list of files
            - commit message as a localstr
            - dict of extra metadata
 
         Unless you need to access all fields, consider calling
         ``changelogrevision`` instead, as it is faster for partial object
         access.
         """
         d = self._revisiondata(nodeorrev)
         sidedata = self.sidedata(nodeorrev)
         copy_sd = self._copiesstorage == b'changeset-sidedata'
         c = changelogrevision(self, d, sidedata, copy_sd)
         return (c.manifest, c.user, c.date, c.files, c.description, c.extra)
 
     def changelogrevision(self, nodeorrev):
         """Obtain a ``changelogrevision`` for a node or revision."""
         text = self._revisiondata(nodeorrev)
         sidedata = self.sidedata(nodeorrev)
         return changelogrevision(
             self, text, sidedata, self._copiesstorage == b'changeset-sidedata'
         )
 
     def readfiles(self, nodeorrev):
         """
         short version of read that only returns the files modified by the cset
         """
         text = self.revision(nodeorrev)
         if not text:
             return []
         last = text.index(b"\n\n")
         l = text[:last].split(b'\n')
         return l[3:]
 
     def add(
         self,
         manifest,
         files,
         desc,
         transaction,
         p1,
         p2,
         user,
         date=None,
         extra=None,
     ):
         # Convert to UTF-8 encoded bytestrings as the very first
         # thing: calling any method on a localstr object will turn it
         # into a str object and the cached UTF-8 string is thus lost.
         user, desc = encoding.fromlocal(user), encoding.fromlocal(desc)
 
         user = user.strip()
         # An empty username or a username with a "\n" will make the
         # revision text contain two "\n\n" sequences -> corrupt
         # repository since read cannot unpack the revision.
         if not user:
             raise error.StorageError(_(b"empty username"))
         if b"\n" in user:
             raise error.StorageError(
                 _(b"username %r contains a newline") % pycompat.bytestr(user)
             )
 
         desc = stripdesc(desc)
 
         if date:
             parseddate = b"%d %d" % dateutil.parsedate(date)
         else:
             parseddate = b"%d %d" % dateutil.makedate()
         if extra:
             branch = extra.get(b"branch")
             if branch in (b"default", b""):
                 del extra[b"branch"]
             elif branch in (b".", b"null", b"tip"):
                 raise error.StorageError(
                     _(b'the name \'%s\' is reserved') % branch
                 )
         sortedfiles = sorted(files.touched)
         flags = 0
         sidedata = None
         if self._copiesstorage == b'changeset-sidedata':
             if files.has_copies_info:
                 flags |= flagutil.REVIDX_HASCOPIESINFO
             sidedata = metadata.encode_files_sidedata(files)
 
         if extra:
             extra = encodeextra(extra)
             parseddate = b"%s %s" % (parseddate, extra)
         l = [hex(manifest), user, parseddate] + sortedfiles + [b"", desc]
         text = b"\n".join(l)
         rev = self.addrevision(
             text, transaction, len(self), p1, p2, sidedata=sidedata, flags=flags
         )
         return self.node(rev)
 
     def branchinfo(self, rev):
         """return the branch name and open/close state of a revision
 
         This function exists because creating a changectx object
         just to access this is costly."""
         return self.changelogrevision(rev).branchinfo
 
     def _nodeduplicatecallback(self, transaction, rev):
         # keep track of revisions that got "re-added", eg: unbunde of know rev.
         #
         # We track them in a list to preserve their order from the source bundle
         duplicates = transaction.changes.setdefault(b'revduplicates', [])
         duplicates.append(rev)

mercurial/revlog.py

@@ -1,4078 +1,4081 b''
 # revlog.py - storage back-end for mercurial
 # coding: utf8
 #
 # Copyright 2005-2007 Olivia 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.
 """
 
 
 import binascii
 import collections
 import contextlib
 import functools
 import io
 import os
 import struct
 import weakref
 import zlib
 
 # import stuff from node for others to import from revlog
 from .node import (
     bin,
     hex,
     nullrev,
     sha1nodeconstants,
     short,
     wdirrev,
 )
 from .i18n import _
 from .revlogutils.constants import (
     ALL_KINDS,
     CHANGELOGV2,
     COMP_MODE_DEFAULT,
     COMP_MODE_INLINE,
     COMP_MODE_PLAIN,
     DELTA_BASE_REUSE_NO,
     DELTA_BASE_REUSE_TRY,
     ENTRY_RANK,
     FEATURES_BY_VERSION,
     FLAG_GENERALDELTA,
     FLAG_INLINE_DATA,
     INDEX_HEADER,
     KIND_CHANGELOG,
     KIND_FILELOG,
     RANK_UNKNOWN,
     REVLOGV0,
     REVLOGV1,
     REVLOGV1_FLAGS,
     REVLOGV2,
     REVLOGV2_FLAGS,
     REVLOG_DEFAULT_FLAGS,
     REVLOG_DEFAULT_FORMAT,
     REVLOG_DEFAULT_VERSION,
     SUPPORTED_FLAGS,
 )
 from .revlogutils.flagutil import (
     REVIDX_DEFAULT_FLAGS,
     REVIDX_ELLIPSIS,
     REVIDX_EXTSTORED,
     REVIDX_FLAGS_ORDER,
     REVIDX_HASCOPIESINFO,
     REVIDX_ISCENSORED,
     REVIDX_RAWTEXT_CHANGING_FLAGS,
 )
 from .thirdparty import attr
 from . import (
     ancestor,
     dagop,
     error,
     mdiff,
     policy,
     pycompat,
     revlogutils,
     templatefilters,
     util,
 )
 from .interfaces import (
     repository,
     util as interfaceutil,
 )
 from .revlogutils import (
     deltas as deltautil,
     docket as docketutil,
     flagutil,
     nodemap as nodemaputil,
     randomaccessfile,
     revlogv0,
     rewrite,
     sidedata as sidedatautil,
 )
 from .utils import (
     storageutil,
     stringutil,
 )
 
 # blanked usage of all the name to prevent pyflakes constraints
 # We need these name available in the module for extensions.
 
 REVLOGV0
 REVLOGV1
 REVLOGV2
 CHANGELOGV2
 FLAG_INLINE_DATA
 FLAG_GENERALDELTA
 REVLOG_DEFAULT_FLAGS
 REVLOG_DEFAULT_FORMAT
 REVLOG_DEFAULT_VERSION
 REVLOGV1_FLAGS
 REVLOGV2_FLAGS
 REVIDX_ISCENSORED
 REVIDX_ELLIPSIS
 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 inline data embedded into a revlog
 _maxinline = 131072
 
 
 # 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 _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 hasattr(
     parsers, 'BaseIndexObject'
 )
 
 
 @interfaceutil.implementer(repository.irevisiondelta)
 @attr.s(slots=True)
 class revlogrevisiondelta:
     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()
     protocol_flags = attr.ib()
     linknode = attr.ib(default=None)
 
 
 @interfaceutil.implementer(repository.iverifyproblem)
 @attr.s(frozen=True)
 class revlogproblem:
     warning = attr.ib(default=None)
     error = attr.ib(default=None)
     node = attr.ib(default=None)
 
 
 def parse_index_v1(data, inline):
     # call the C implementation to parse the index data
     index, cache = parsers.parse_index2(data, inline)
     return index, cache
 
 
 def parse_index_v2(data, inline):
     # call the C implementation to parse the index data
     index, cache = parsers.parse_index2(data, inline, format=REVLOGV2)
     return index, cache
 
 
 def parse_index_cl_v2(data, inline):
     # call the C implementation to parse the index data
     index, cache = parsers.parse_index2(data, inline, format=CHANGELOGV2)
     return index, cache
 
 
 if hasattr(parsers, 'parse_index_devel_nodemap'):
 
     def parse_index_v1_nodemap(data, inline):
         index, cache = parsers.parse_index_devel_nodemap(data, inline)
         return index, cache
 
 
 else:
     parse_index_v1_nodemap = None
 
 
 def parse_index_v1_rust(data, inline, default_header):
     cache = (0, data) if inline else None
     return rustrevlog.Index(data, default_header), cache
 
 
 # corresponds to uncompressed length of indexformatng (2 gigs, 4-byte
 # signed integer)
 _maxentrysize = 0x7FFFFFFF
 
 FILE_TOO_SHORT_MSG = _(
     b'cannot read from revlog %s;'
     b'  expected %d bytes from offset %d, data size is %d'
 )
 
 hexdigits = b'0123456789abcdefABCDEF'
 
 
 class _Config:
     def copy(self):
         return self.__class__(**self.__dict__)
 
 
 @attr.s()
 class FeatureConfig(_Config):
     """Hold configuration values about the available revlog features"""
 
     # the default compression engine
     compression_engine = attr.ib(default=b'zlib')
     # compression engines options
     compression_engine_options = attr.ib(default=attr.Factory(dict))
 
     # can we use censor on this revlog
     censorable = attr.ib(default=False)
     # does this revlog use the "side data" feature
     has_side_data = attr.ib(default=False)
     # might remove rank configuration once the computation has no impact
     compute_rank = attr.ib(default=False)
     # parent order is supposed to be semantically irrelevant, so we
     # normally resort parents to ensure that the first parent is non-null,
     # if there is a non-null parent at all.
     # filelog abuses the parent order as flag to mark some instances of
     # meta-encoded files, so allow it to disable this behavior.
     canonical_parent_order = attr.ib(default=False)
     # can ellipsis commit be used
     enable_ellipsis = attr.ib(default=False)
 
     def copy(self):
         new = super().copy()
         new.compression_engine_options = self.compression_engine_options.copy()
         return new
 
 
 @attr.s()
 class DataConfig(_Config):
     """Hold configuration value about how the revlog data are read"""
 
     # should we try to open the "pending" version of the revlog
     try_pending = attr.ib(default=False)
     # should we try to open the "splitted" version of the revlog
     try_split = attr.ib(default=False)
     #  When True, indexfile should be opened with checkambig=True at writing,
     #  to avoid file stat ambiguity.
     check_ambig = attr.ib(default=False)
 
     # If true, use mmap instead of reading to deal with large index
     mmap_large_index = attr.ib(default=False)
     # how much data is large
     mmap_index_threshold = attr.ib(default=None)
     # How much data to read and cache into the raw revlog data cache.
     chunk_cache_size = attr.ib(default=65536)
 
     # The size of the uncompressed cache compared to the largest revision seen.
     uncompressed_cache_factor = attr.ib(default=None)
 
     # The number of chunk cached
     uncompressed_cache_count = attr.ib(default=None)
 
     # Allow sparse reading of the revlog data
     with_sparse_read = attr.ib(default=False)
     # minimal density of a sparse read chunk
     sr_density_threshold = attr.ib(default=0.50)
     # minimal size of data we skip when performing sparse read
     sr_min_gap_size = attr.ib(default=262144)
 
     # are delta encoded against arbitrary bases.
     generaldelta = attr.ib(default=False)
 
 
 @attr.s()
 class DeltaConfig(_Config):
     """Hold configuration value about how new delta are computed
 
     Some attributes are duplicated from DataConfig to help havign each object
     self contained.
     """
 
     # can delta be encoded against arbitrary bases.
     general_delta = attr.ib(default=False)
     # Allow sparse writing of the revlog data
     sparse_revlog = attr.ib(default=False)
     # maximum length of a delta chain
     max_chain_len = attr.ib(default=None)
     # Maximum distance between delta chain base start and end
     max_deltachain_span = attr.ib(default=-1)
     # If `upper_bound_comp` is not None, this is the expected maximal gain from
     # compression for the data content.
     upper_bound_comp = attr.ib(default=None)
     # Should we try a delta against both parent
     delta_both_parents = attr.ib(default=True)
     # Test delta base candidate group by chunk of this maximal size.
     candidate_group_chunk_size = attr.ib(default=0)
     # Should we display debug information about delta computation
     debug_delta = attr.ib(default=False)
     # trust incoming delta by default
     lazy_delta = attr.ib(default=True)
     # trust the base of incoming delta by default
     lazy_delta_base = attr.ib(default=False)
 
 
 class _InnerRevlog:
     """An inner layer of the revlog object
 
     That layer exist to be able to delegate some operation to Rust, its
     boundaries are arbitrary and based on what we can delegate to Rust.
     """
 
     def __init__(
         self,
         opener,
         index,
         index_file,
         data_file,
         sidedata_file,
         inline,
         data_config,
         delta_config,
         feature_config,
         chunk_cache,
         default_compression_header,
     ):
         self.opener = opener
         self.index = index
 
         self.index_file = index_file
         self.data_file = data_file
         self.sidedata_file = sidedata_file
         self.inline = inline
         self.data_config = data_config
         self.delta_config = delta_config
         self.feature_config = feature_config
 
         # used during diverted write.
         self._orig_index_file = None
 
         self._default_compression_header = default_compression_header
 
         # index
 
         # 3-tuple of file handles being used for active writing.
         self._writinghandles = None
 
         self._segmentfile = randomaccessfile.randomaccessfile(
             self.opener,
             (self.index_file if self.inline else self.data_file),
             self.data_config.chunk_cache_size,
             chunk_cache,
         )
         self._segmentfile_sidedata = randomaccessfile.randomaccessfile(
             self.opener,
             self.sidedata_file,
             self.data_config.chunk_cache_size,
         )
 
         # revlog header -> revlog compressor
         self._decompressors = {}
         # 3-tuple of (node, rev, text) for a raw revision.
         self._revisioncache = None
 
         # cache some uncompressed chunks
         # rev → uncompressed_chunk
         #
         # the max cost is dynamically updated to be proportionnal to the
         # size of revision we actually encounter.
         self._uncompressed_chunk_cache = None
         if self.data_config.uncompressed_cache_factor is not None:
             self._uncompressed_chunk_cache = util.lrucachedict(
                 self.data_config.uncompressed_cache_count,
                 maxcost=65536,  # some arbitrary initial value
             )
 
         self._delay_buffer = None
 
     def __len__(self):
         return len(self.index)
 
     def clear_cache(self):
         assert not self.is_delaying
         self._revisioncache = None
         if self._uncompressed_chunk_cache is not None:
             self._uncompressed_chunk_cache.clear()
         self._segmentfile.clear_cache()
         self._segmentfile_sidedata.clear_cache()
 
     @property
     def canonical_index_file(self):
         if self._orig_index_file is not None:
             return self._orig_index_file
         return self.index_file
 
     @property
     def is_delaying(self):
         """is the revlog is currently delaying the visibility of written data?
 
         The delaying mechanism can be either in-memory or written on disk in a
         side-file."""
         return (self._delay_buffer is not None) or (
             self._orig_index_file is not None
         )
 
     # Derived from index values.
 
     def start(self, rev):
         """the offset of the data chunk for this revision"""
         return int(self.index[rev][0] >> 16)
 
     def length(self, rev):
         """the length of the data chunk for this revision"""
         return self.index[rev][1]
 
     def end(self, rev):
         """the end of the data chunk for this revision"""
         return self.start(rev) + self.length(rev)
 
     def deltaparent(self, rev):
         """return deltaparent of the given revision"""
         base = self.index[rev][3]
         if base == rev:
             return nullrev
         elif self.delta_config.general_delta:
             return base
         else:
             return rev - 1
 
     def issnapshot(self, rev):
         """tells whether rev is a snapshot"""
         if not self.delta_config.sparse_revlog:
             return self.deltaparent(rev) == nullrev
         elif hasattr(self.index, '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]
         while self.length(p1) == 0:
             b = self.deltaparent(p1)
             if b == p1:
                 break
             p1 = b
         p2 = entry[6]
         while self.length(p2) == 0:
             b = self.deltaparent(p2)
             if b == p2:
                 break
             p2 = b
         if base == p1 or base == p2:
             return False
         return self.issnapshot(base)
 
     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.
         """
         generaldelta = self.delta_config.general_delta
         # Try C implementation.
         try:
             return self.index.deltachain(rev, stoprev, generaldelta)
         except AttributeError:
             pass
 
         chain = []
 
         # Alias to prevent attribute lookup in tight loop.
         index = self.index
 
         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
 
     @util.propertycache
     def _compressor(self):
         engine = util.compengines[self.feature_config.compression_engine]
         return engine.revlogcompressor(
             self.feature_config.compression_engine_options
         )
 
     @util.propertycache
     def _decompressor(self):
         """the default decompressor"""
         if self._default_compression_header is None:
             return None
         t = self._default_compression_header
         c = self._get_decompressor(t)
         return c.decompress
 
     def _get_decompressor(self, t):
         try:
             compressor = self._decompressors[t]
         except KeyError:
             try:
                 engine = util.compengines.forrevlogheader(t)
                 compressor = engine.revlogcompressor(
                     self.feature_config.compression_engine_options
                 )
                 self._decompressors[t] = compressor
             except KeyError:
                 raise error.RevlogError(
                     _(b'unknown compression type %s') % binascii.hexlify(t)
                 )
         return compressor
 
     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)
 
         compressor = self._get_decompressor(t)
 
         return compressor.decompress(data)
 
     @contextlib.contextmanager
     def reading(self):
         """Context manager that keeps data and sidedata files open for reading"""
         if len(self.index) == 0:
             yield  # nothing to be read
         elif self._delay_buffer is not None and self.inline:
             msg = "revlog with delayed write should not be inline"
             raise error.ProgrammingError(msg)
         else:
             with self._segmentfile.reading():
                 with self._segmentfile_sidedata.reading():
                     yield
 
     @property
     def is_writing(self):
         """True is a writing context is open"""
         return self._writinghandles is not None
 
     @property
     def is_open(self):
         """True if any file handle is being held
 
         Used for assert and debug in the python code"""
         return self._segmentfile.is_open or self._segmentfile_sidedata.is_open
 
     @contextlib.contextmanager
     def writing(self, transaction, data_end=None, sidedata_end=None):
         """Open the revlog files for writing
 
         Add content to a revlog should be done within such context.
         """
         if self.is_writing:
             yield
         else:
             ifh = dfh = sdfh = None
             try:
                 r = len(self.index)
                 # opening the data file.
                 dsize = 0
                 if r:
                     dsize = self.end(r - 1)
                 dfh = None
                 if not self.inline:
                     try:
                         dfh = self.opener(self.data_file, mode=b"r+")
                         if data_end is None:
                             dfh.seek(0, os.SEEK_END)
                         else:
                             dfh.seek(data_end, os.SEEK_SET)
                     except FileNotFoundError:
                         dfh = self.opener(self.data_file, mode=b"w+")
                     transaction.add(self.data_file, dsize)
                 if self.sidedata_file is not None:
                     assert sidedata_end is not None
                     # revlog-v2 does not inline, help Pytype
                     assert dfh is not None
                     try:
                         sdfh = self.opener(self.sidedata_file, mode=b"r+")
                         dfh.seek(sidedata_end, os.SEEK_SET)
                     except FileNotFoundError:
                         sdfh = self.opener(self.sidedata_file, mode=b"w+")
                     transaction.add(self.sidedata_file, sidedata_end)
 
                 # opening the index file.
                 isize = r * self.index.entry_size
                 ifh = self.__index_write_fp()
                 if self.inline:
                     transaction.add(self.index_file, dsize + isize)
                 else:
                     transaction.add(self.index_file, isize)
                 # exposing all file handle for writing.
                 self._writinghandles = (ifh, dfh, sdfh)
                 self._segmentfile.writing_handle = ifh if self.inline else dfh
                 self._segmentfile_sidedata.writing_handle = sdfh
                 yield
             finally:
                 self._writinghandles = None
                 self._segmentfile.writing_handle = None
                 self._segmentfile_sidedata.writing_handle = None
                 if dfh is not None:
                     dfh.close()
                 if sdfh is not None:
                     sdfh.close()
                 # closing the index file last to avoid exposing referent to
                 # potential unflushed data content.
                 if ifh is not None:
                     ifh.close()
 
     def __index_write_fp(self, index_end=None):
         """internal method to open the index file for writing
 
         You should not use this directly and use `_writing` instead
         """
         try:
             if self._delay_buffer is None:
                 f = self.opener(
                     self.index_file,
                     mode=b"r+",
                     checkambig=self.data_config.check_ambig,
                 )
             else:
                 # check_ambig affect we way we open file for writing, however
                 # here, we do not actually open a file for writting as write
                 # will appened to a delay_buffer. So check_ambig is not
                 # meaningful and unneeded here.
                 f = randomaccessfile.appender(
                     self.opener, self.index_file, b"r+", self._delay_buffer
                 )
             if index_end is None:
                 f.seek(0, os.SEEK_END)
             else:
                 f.seek(index_end, os.SEEK_SET)
             return f
         except FileNotFoundError:
             if self._delay_buffer is None:
                 return self.opener(
                     self.index_file,
                     mode=b"w+",
                     checkambig=self.data_config.check_ambig,
                 )
             else:
                 return randomaccessfile.appender(
                     self.opener, self.index_file, b"w+", self._delay_buffer
                 )
 
     def __index_new_fp(self):
         """internal method to create a new index file for writing
 
         You should not use this unless you are upgrading from inline revlog
         """
         return self.opener(
             self.index_file,
             mode=b"w",
             checkambig=self.data_config.check_ambig,
         )
 
     def split_inline(self, tr, header, new_index_file_path=None):
         """split the data of an inline revlog into an index and a data file"""
         assert self._delay_buffer is None
         existing_handles = False
         if self._writinghandles is not None:
             existing_handles = True
             fp = self._writinghandles[0]
             fp.flush()
             fp.close()
             # We can't use the cached file handle after close(). So prevent
             # its usage.
             self._writinghandles = None
             self._segmentfile.writing_handle = None
             # No need to deal with sidedata writing handle as it is only
             # relevant with revlog-v2 which is never inline, not reaching
             # this code
 
         new_dfh = self.opener(self.data_file, mode=b"w+")
         new_dfh.truncate(0)  # drop any potentially existing data
         try:
             with self.reading():
                 for r in range(len(self.index)):
                     new_dfh.write(self.get_segment_for_revs(r, r)[1])
                 new_dfh.flush()
 
             if new_index_file_path is not None:
                 self.index_file = new_index_file_path
             with self.__index_new_fp() as fp:
                 self.inline = False
                 for i in range(len(self.index)):
                     e = self.index.entry_binary(i)
                     if i == 0:
                         packed_header = self.index.pack_header(header)
                         e = packed_header + e
                     fp.write(e)
 
                 # If we don't use side-write, the temp file replace the real
                 # index when we exit the context manager
 
             self._segmentfile = randomaccessfile.randomaccessfile(
                 self.opener,
                 self.data_file,
                 self.data_config.chunk_cache_size,
             )
 
             if existing_handles:
                 # switched from inline to conventional reopen the index
                 ifh = self.__index_write_fp()
                 self._writinghandles = (ifh, new_dfh, None)
                 self._segmentfile.writing_handle = new_dfh
                 new_dfh = None
                 # No need to deal with sidedata writing handle as it is only
                 # relevant with revlog-v2 which is never inline, not reaching
                 # this code
         finally:
             if new_dfh is not None:
                 new_dfh.close()
         return self.index_file
 
     def get_segment_for_revs(self, startrev, endrev):
         """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.
 
         API: we should consider making this a private part of the InnerRevlog
         at some point.
         """
         # 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.index.entry_size
             end += (endrev + 1) * self.index.entry_size
         length = end - start
 
         return start, self._segmentfile.read_chunk(start, length)
 
     def _chunk(self, rev):
         """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.
         """
         if self._uncompressed_chunk_cache is not None:
             uncomp = self._uncompressed_chunk_cache.get(rev)
             if uncomp is not None:
                 return uncomp
 
         compression_mode = self.index[rev][10]
         data = self.get_segment_for_revs(rev, rev)[1]
         if compression_mode == COMP_MODE_PLAIN:
             uncomp = data
         elif compression_mode == COMP_MODE_DEFAULT:
             uncomp = self._decompressor(data)
         elif compression_mode == COMP_MODE_INLINE:
             uncomp = self.decompress(data)
         else:
             msg = b'unknown compression mode %d'
             msg %= compression_mode
             raise error.RevlogError(msg)
         if self._uncompressed_chunk_cache is not None:
             self._uncompressed_chunk_cache.insert(rev, uncomp, cost=len(uncomp))
         return uncomp
 
     def _chunks(self, revs, 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.index.entry_size
         buffer = util.buffer
 
         fetched_revs = []
         fadd = fetched_revs.append
 
         chunks = []
         ladd = chunks.append
 
         if self._uncompressed_chunk_cache is None:
             fetched_revs = revs
         else:
             for rev in revs:
                 cached_value = self._uncompressed_chunk_cache.get(rev)
                 if cached_value is None:
                     fadd(rev)
                 else:
                     ladd((rev, cached_value))
 
         if not fetched_revs:
             slicedchunks = ()
         elif not self.data_config.with_sparse_read:
             slicedchunks = (fetched_revs,)
         else:
             slicedchunks = deltautil.slicechunk(
                 self,
                 fetched_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.get_segment_for_revs(firstrev, lastrev)
             except OverflowError:
                 # issue4215 - we can't cache a run of chunks greater than
                 # 2G on Windows
                 for rev in revschunk:
                     ladd((rev, self._chunk(rev)))
 
             decomp = self.decompress
             # self._decompressor might be None, but will not be used in that case
             def_decomp = self._decompressor
             for rev in revschunk:
                 chunkstart = start(rev)
                 if inline:
                     chunkstart += (rev + 1) * iosize
                 chunklength = length(rev)
                 comp_mode = self.index[rev][10]
                 c = buffer(data, chunkstart - offset, chunklength)
                 if comp_mode == COMP_MODE_PLAIN:
                     c = c
                 elif comp_mode == COMP_MODE_INLINE:
                     c = decomp(c)
                 elif comp_mode == COMP_MODE_DEFAULT:
                     c = def_decomp(c)
                 else:
                     msg = b'unknown compression mode %d'
                     msg %= comp_mode
                     raise error.RevlogError(msg)
                 ladd((rev, c))
                 if self._uncompressed_chunk_cache is not None:
                     self._uncompressed_chunk_cache.insert(rev, c, len(c))
 
         chunks.sort()
         return [x[1] for x in chunks]
 
     def raw_text(self, node, rev):
         """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:
             cachedrev = self._revisioncache[1]
 
         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._inner._revisioncache after validating the text
         self._revisioncache = None
 
         targetsize = None
         rawsize = self.index[rev][2]
         if 0 <= rawsize:
             targetsize = 4 * rawsize
 
         if self._uncompressed_chunk_cache is not None:
             # dynamically update the uncompressed_chunk_cache size to the
             # largest revision we saw in this revlog.
             factor = self.data_config.uncompressed_cache_factor
             candidate_size = rawsize * factor
             if candidate_size > self._uncompressed_chunk_cache.maxcost:
                 self._uncompressed_chunk_cache.maxcost = candidate_size
 
         bins = self._chunks(chain, 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, sidedata_end):
         """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.index.entry_size * (1 + rev)
         if sidedata_size == 0:
             return {}
 
         if sidedata_end < sidedata_offset + sidedata_size:
             filename = self.sidedata_file
             end = sidedata_end
             offset = sidedata_offset
             length = sidedata_size
             m = FILE_TOO_SHORT_MSG % (filename, length, offset, end)
             raise error.RevlogError(m)
 
         comp_segment = self._segmentfile_sidedata.read_chunk(
             sidedata_offset, sidedata_size
         )
 
         comp = self.index[rev][11]
         if comp == COMP_MODE_PLAIN:
             segment = comp_segment
         elif comp == COMP_MODE_DEFAULT:
             segment = self._decompressor(comp_segment)
         elif comp == COMP_MODE_INLINE:
             segment = self.decompress(comp_segment)
         else:
             msg = b'unknown compression mode %d'
             msg %= comp
             raise error.RevlogError(msg)
 
         sidedata = sidedatautil.deserialize_sidedata(segment)
         return sidedata
 
     def write_entry(
         self,
         transaction,
         entry,
         data,
         link,
         offset,
         sidedata,
         sidedata_offset,
         index_end,
         data_end,
         sidedata_end,
     ):
         # 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.
         if self._writinghandles is None:
             msg = b'adding revision outside `revlog._writing` context'
             raise error.ProgrammingError(msg)
         ifh, dfh, sdfh = self._writinghandles
         if index_end is None:
             ifh.seek(0, os.SEEK_END)
         else:
             ifh.seek(index_end, os.SEEK_SET)
         if dfh:
             if data_end is None:
                 dfh.seek(0, os.SEEK_END)
             else:
                 dfh.seek(data_end, os.SEEK_SET)
         if sdfh:
             sdfh.seek(sidedata_end, os.SEEK_SET)
 
         curr = len(self.index) - 1
         if not self.inline:
             transaction.add(self.data_file, offset)
             if self.sidedata_file:
                 transaction.add(self.sidedata_file, sidedata_offset)
             transaction.add(self.canonical_index_file, curr * len(entry))
             if data[0]:
                 dfh.write(data[0])
             dfh.write(data[1])
             if sidedata:
                 sdfh.write(sidedata)
             if self._delay_buffer is None:
                 ifh.write(entry)
             else:
                 self._delay_buffer.append(entry)
         elif self._delay_buffer is not None:
             msg = b'invalid delayed write on inline revlog'
             raise error.ProgrammingError(msg)
         else:
             offset += curr * self.index.entry_size
             transaction.add(self.canonical_index_file, offset)
             assert not sidedata
             ifh.write(entry)
             ifh.write(data[0])
             ifh.write(data[1])
         return (
             ifh.tell(),
             dfh.tell() if dfh else None,
             sdfh.tell() if sdfh else None,
         )
 
     def _divert_index(self):
         return self.index_file + b'.a'
 
     def delay(self):
         assert not self.is_open
         if self.inline:
             msg = "revlog with delayed write should not be inline"
             raise error.ProgrammingError(msg)
         if self._delay_buffer is not None or self._orig_index_file is not None:
             # delay or divert already in place
             return None
         elif len(self.index) == 0:
             self._orig_index_file = self.index_file
             self.index_file = self._divert_index()
             assert self._orig_index_file is not None
             assert self.index_file is not None
             if self.opener.exists(self.index_file):
                 self.opener.unlink(self.index_file)
             return self.index_file
         else:
             self._delay_buffer = []
             return None
 
     def write_pending(self):
         assert not self.is_open
         if self.inline:
             msg = "revlog with delayed write should not be inline"
             raise error.ProgrammingError(msg)
         if self._orig_index_file is not None:
             return None, True
         any_pending = False
         pending_index_file = self._divert_index()
         if self.opener.exists(pending_index_file):
             self.opener.unlink(pending_index_file)
         util.copyfile(
             self.opener.join(self.index_file),
             self.opener.join(pending_index_file),
         )
         if self._delay_buffer:
             with self.opener(pending_index_file, b'r+') as ifh:
                 ifh.seek(0, os.SEEK_END)
                 ifh.write(b"".join(self._delay_buffer))
             any_pending = True
         self._delay_buffer = None
         self._orig_index_file = self.index_file
         self.index_file = pending_index_file
         return self.index_file, any_pending
 
     def finalize_pending(self):
         assert not self.is_open
         if self.inline:
             msg = "revlog with delayed write should not be inline"
             raise error.ProgrammingError(msg)
 
         delay = self._delay_buffer is not None
         divert = self._orig_index_file is not None
 
         if delay and divert:
             assert False, "unreachable"
         elif delay:
             if self._delay_buffer:
                 with self.opener(self.index_file, b'r+') as ifh:
                     ifh.seek(0, os.SEEK_END)
                     ifh.write(b"".join(self._delay_buffer))
             self._delay_buffer = None
         elif divert:
             if self.opener.exists(self.index_file):
                 self.opener.rename(
                     self.index_file,
                     self._orig_index_file,
                     checkambig=True,
                 )
             self.index_file = self._orig_index_file
             self._orig_index_file = None
         else:
             msg = b"not delay or divert found on this revlog"
             raise error.ProgrammingError(msg)
         return self.canonical_index_file
 
 
 class revlog:
     """
     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).
 
     See mercurial/revlogutils/contants.py for details about the content of an
     index entry.
     """
 
     _flagserrorclass = error.RevlogError
 
     @staticmethod
     def is_inline_index(header_bytes):
         """Determine if a revlog is inline from the initial bytes of the index"""
         if len(header_bytes) == 0:
             return True
 
         header = INDEX_HEADER.unpack(header_bytes)[0]
 
         _format_flags = header & ~0xFFFF
         _format_version = header & 0xFFFF
 
         features = FEATURES_BY_VERSION[_format_version]
         return features[b'inline'](_format_flags)
 
     def __init__(
         self,
         opener,
         target,
         radix,
         postfix=None,  # only exist for `tmpcensored` now
         checkambig=False,
         mmaplargeindex=False,
         censorable=False,
         upperboundcomp=None,
         persistentnodemap=False,
         concurrencychecker=None,
         trypending=False,
         try_split=False,
         canonical_parent_order=True,
         data_config=None,
         delta_config=None,
         feature_config=None,
         may_inline=True,  # may inline new revlog
     ):
         """
         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.
 
         `target`: a (KIND, ID) tuple that identify the content stored in
         this revlog. It help the rest of the code to understand what the revlog
         is about without having to resort to heuristic and index filename
         analysis. Note: that this must be reliably be set by normal code, but
         that test, debug, or performance measurement code might not set this to
         accurate value.
         """
 
         self.radix = radix
 
         self._docket_file = None
         self._indexfile = None
         self._datafile = None
         self._sidedatafile = None
         self._nodemap_file = None
         self.postfix = postfix
         self._trypending = trypending
         self._try_split = try_split
         self._may_inline = may_inline
         self.opener = opener
         if persistentnodemap:
             self._nodemap_file = nodemaputil.get_nodemap_file(self)
 
         assert target[0] in ALL_KINDS
         assert len(target) == 2
         self.target = target
         if feature_config is not None:
             self.feature_config = feature_config.copy()
         elif b'feature-config' in self.opener.options:
             self.feature_config = self.opener.options[b'feature-config'].copy()
         else:
             self.feature_config = FeatureConfig()
         self.feature_config.censorable = censorable
         self.feature_config.canonical_parent_order = canonical_parent_order
         if data_config is not None:
             self.data_config = data_config.copy()
         elif b'data-config' in self.opener.options:
             self.data_config = self.opener.options[b'data-config'].copy()
         else:
             self.data_config = DataConfig()
         self.data_config.check_ambig = checkambig
         self.data_config.mmap_large_index = mmaplargeindex
         if delta_config is not None:
             self.delta_config = delta_config.copy()
         elif b'delta-config' in self.opener.options:
             self.delta_config = self.opener.options[b'delta-config'].copy()
         else:
             self.delta_config = DeltaConfig()
         self.delta_config.upper_bound_comp = upperboundcomp
 
         # Maps rev to chain base rev.
         self._chainbasecache = util.lrucachedict(100)
 
         self.index = None
         self._docket = None
         self._nodemap_docket = None
         # Mapping of partial identifiers to full nodes.
         self._pcache = {}
 
         # other optionnals features
 
         # Make copy of flag processors so each revlog instance can support
         # custom flags.
         self._flagprocessors = dict(flagutil.flagprocessors)
         # prevent nesting of addgroup
         self._adding_group = None
 
         chunk_cache = self._loadindex()
         self._load_inner(chunk_cache)
         self._concurrencychecker = concurrencychecker
 
     def _init_opts(self):
         """process options (from above/config) to setup associated default revlog mode
 
         These values might be affected when actually reading on disk information.
 
         The relevant values are returned for use in _loadindex().
 
         * newversionflags:
             version header to use if we need to create a new revlog
 
         * mmapindexthreshold:
             minimal index size for start to use mmap
 
         * force_nodemap:
             force the usage of a "development" version of the nodemap code
         """
         opts = self.opener.options
 
         if b'changelogv2' in opts and self.revlog_kind == KIND_CHANGELOG:
             new_header = CHANGELOGV2
             compute_rank = opts.get(b'changelogv2.compute-rank', True)
             self.feature_config.compute_rank = compute_rank
         elif b'revlogv2' in opts:
             new_header = REVLOGV2
         elif b'revlogv1' in opts:
             new_header = REVLOGV1
             if self._may_inline:
                 new_header |= FLAG_INLINE_DATA
             if b'generaldelta' in opts:
                 new_header |= FLAG_GENERALDELTA
         elif b'revlogv0' in self.opener.options:
             new_header = REVLOGV0
         else:
             new_header = REVLOG_DEFAULT_VERSION
 
         mmapindexthreshold = None
         if self.data_config.mmap_large_index:
             mmapindexthreshold = self.data_config.mmap_index_threshold
         if self.feature_config.enable_ellipsis:
             self._flagprocessors[REVIDX_ELLIPSIS] = ellipsisprocessor
 
         # revlog v0 doesn't have flag processors
         for flag, processor in opts.get(b'flagprocessors', {}).items():
             flagutil.insertflagprocessor(flag, processor, self._flagprocessors)
 
         chunk_cache_size = self.data_config.chunk_cache_size
         if chunk_cache_size <= 0:
             raise error.RevlogError(
                 _(b'revlog chunk cache size %r is not greater than 0')
                 % chunk_cache_size
             )
         elif chunk_cache_size & (chunk_cache_size - 1):
             raise error.RevlogError(
                 _(b'revlog chunk cache size %r is not a power of 2')
                 % chunk_cache_size
             )
         force_nodemap = opts.get(b'devel-force-nodemap', False)
         return new_header, mmapindexthreshold, force_nodemap
 
     def _get_data(self, filepath, mmap_threshold, size=None):
         """return a file content with or without mmap
 
         If the file is missing return the empty string"""
         try:
             with self.opener(filepath) as fp:
                 if mmap_threshold is not None:
                     file_size = self.opener.fstat(fp).st_size
                     if file_size >= mmap_threshold:
                         if size is not None:
                             # avoid potentiel mmap crash
                             size = min(file_size, size)
                         # TODO: should .close() to release resources without
                         # relying on Python GC
                         if size is None:
                             return util.buffer(util.mmapread(fp))
                         else:
                             return util.buffer(util.mmapread(fp, size))
                 if size is None:
                     return fp.read()
                 else:
                     return fp.read(size)
         except FileNotFoundError:
             return b''
 
     def get_streams(self, max_linkrev, force_inline=False):
         """return a list of streams that represent this revlog
 
         This is used by stream-clone to do bytes to bytes copies of a repository.
 
         This streams data for all revisions that refer to a changelog revision up
         to `max_linkrev`.
 
         If `force_inline` is set, it enforces that the stream will represent an inline revlog.
 
         It returns is a list of three-tuple:
 
             [
                 (filename, bytes_stream, stream_size),
                 …
             ]
         """
         n = len(self)
         index = self.index
         while n > 0:
             linkrev = index[n - 1][4]
             if linkrev < max_linkrev:
                 break
             # note: this loop will rarely go through multiple iterations, since
             # it only traverses commits created during the current streaming
             # pull operation.
             #
             # If this become a problem, using a binary search should cap the
             # runtime of this.
             n = n - 1
         if n == 0:
             # no data to send
             return []
         index_size = n * index.entry_size
         data_size = self.end(n - 1)
 
         # XXX we might have been split (or stripped) since the object
         # initialization, We need to close this race too, but having a way to
         # pre-open the file we feed to the revlog and never closing them before
         # we are done streaming.
 
         if self._inline:
 
             def get_stream():
                 with self.opener(self._indexfile, mode=b"r") as fp:
                     yield None
                     size = index_size + data_size
                     if size <= 65536:
                         yield fp.read(size)
                     else:
                         yield from util.filechunkiter(fp, limit=size)
 
             inline_stream = get_stream()
             next(inline_stream)
             return [
                 (self._indexfile, inline_stream, index_size + data_size),
             ]
         elif force_inline:
 
             def get_stream():
                 with self.reading():
                     yield None
 
                     for rev in range(n):
                         idx = self.index.entry_binary(rev)
                         if rev == 0 and self._docket is None:
                             # re-inject the inline flag
                             header = self._format_flags
                             header |= self._format_version
                             header |= FLAG_INLINE_DATA
                             header = self.index.pack_header(header)
                             idx = header + idx
                         yield idx
                         yield self._inner.get_segment_for_revs(rev, rev)[1]
 
             inline_stream = get_stream()
             next(inline_stream)
             return [
                 (self._indexfile, inline_stream, index_size + data_size),
             ]
         else:
 
             def get_index_stream():
                 with self.opener(self._indexfile, mode=b"r") as fp:
                     yield None
                     if index_size <= 65536:
                         yield fp.read(index_size)
                     else:
                         yield from util.filechunkiter(fp, limit=index_size)
 
             def get_data_stream():
                 with self._datafp() as fp:
                     yield None
                     if data_size <= 65536:
                         yield fp.read(data_size)
                     else:
                         yield from util.filechunkiter(fp, limit=data_size)
 
             index_stream = get_index_stream()
             next(index_stream)
             data_stream = get_data_stream()
             next(data_stream)
             return [
                 (self._datafile, data_stream, data_size),
                 (self._indexfile, index_stream, index_size),
             ]
 
     def _loadindex(self, docket=None):
         new_header, mmapindexthreshold, force_nodemap = self._init_opts()
 
         if self.postfix is not None:
             entry_point = b'%s.i.%s' % (self.radix, self.postfix)
         elif self._trypending and self.opener.exists(b'%s.i.a' % self.radix):
             entry_point = b'%s.i.a' % self.radix
         elif self._try_split and self.opener.exists(self._split_index_file):
             entry_point = self._split_index_file
         else:
             entry_point = b'%s.i' % self.radix
 
         if docket is not None:
             self._docket = docket
             self._docket_file = entry_point
         else:
             self._initempty = True
             entry_data = self._get_data(entry_point, mmapindexthreshold)
             if len(entry_data) > 0:
                 header = INDEX_HEADER.unpack(entry_data[:4])[0]
                 self._initempty = False
             else:
                 header = new_header
 
             self._format_flags = header & ~0xFFFF
             self._format_version = header & 0xFFFF
 
             supported_flags = SUPPORTED_FLAGS.get(self._format_version)
             if supported_flags is None:
                 msg = _(b'unknown version (%d) in revlog %s')
                 msg %= (self._format_version, self.display_id)
                 raise error.RevlogError(msg)
             elif self._format_flags & ~supported_flags:
                 msg = _(b'unknown flags (%#04x) in version %d revlog %s')
                 display_flag = self._format_flags >> 16
                 msg %= (display_flag, self._format_version, self.display_id)
                 raise error.RevlogError(msg)
 
             features = FEATURES_BY_VERSION[self._format_version]
             self._inline = features[b'inline'](self._format_flags)
             self.delta_config.general_delta = features[b'generaldelta'](
                 self._format_flags
             )
             self.feature_config.has_side_data = features[b'sidedata']
 
             if not features[b'docket']:
                 self._indexfile = entry_point
                 index_data = entry_data
             else:
                 self._docket_file = entry_point
                 if self._initempty:
                     self._docket = docketutil.default_docket(self, header)
                 else:
                     self._docket = docketutil.parse_docket(
                         self, entry_data, use_pending=self._trypending
                     )
 
         if self._docket is not None:
             self._indexfile = self._docket.index_filepath()
             index_data = b''
             index_size = self._docket.index_end
             if index_size > 0:
                 index_data = self._get_data(
                     self._indexfile, mmapindexthreshold, size=index_size
                 )
                 if len(index_data) < index_size:
                     msg = _(b'too few index data for %s: got %d, expected %d')
                     msg %= (self.display_id, len(index_data), index_size)
                     raise error.RevlogError(msg)
 
             self._inline = False
             # generaldelta implied by version 2 revlogs.
             self.delta_config.general_delta = True
             # the logic for persistent nodemap will be dealt with within the
             # main docket, so disable it for now.
             self._nodemap_file = None
 
         if self._docket is not None:
             self._datafile = self._docket.data_filepath()
             self._sidedatafile = self._docket.sidedata_filepath()
         elif self.postfix is None:
             self._datafile = b'%s.d' % self.radix
         else:
             self._datafile = b'%s.d.%s' % (self.radix, self.postfix)
 
         self.nodeconstants = sha1nodeconstants
         self.nullid = self.nodeconstants.nullid
 
         # sparse-revlog can't be on without general-delta (issue6056)
         if not self.delta_config.general_delta:
             self.delta_config.sparse_revlog = False
 
         self._storedeltachains = True
 
         devel_nodemap = (
             self._nodemap_file
             and force_nodemap
             and parse_index_v1_nodemap is not None
         )
 
         use_rust_index = False
         if rustrevlog is not None and self._nodemap_file is not None:
             # we would like to use the rust_index in all case, especially
             # because it is necessary for AncestorsIterator and LazyAncestors
             # since the 6.7 cycle.
             #
             # However, the performance impact of inconditionnaly building the
             # nodemap is currently a problem for non-persistent nodemap
             # repository.
             use_rust_index = True
 
         self._parse_index = parse_index_v1
         if self._format_version == REVLOGV0:
             self._parse_index = revlogv0.parse_index_v0
         elif self._format_version == REVLOGV2:
             self._parse_index = parse_index_v2
         elif self._format_version == CHANGELOGV2:
             self._parse_index = parse_index_cl_v2
         elif devel_nodemap:
             self._parse_index = parse_index_v1_nodemap
         elif use_rust_index:
             self._parse_index = functools.partial(
                 parse_index_v1_rust, default_header=new_header
             )
         try:
             d = self._parse_index(index_data, self._inline)
             index, chunkcache = d
             use_nodemap = (
                 not self._inline
                 and self._nodemap_file is not None
                 and hasattr(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.display_id
             )
         self.index = index
         # revnum -> (chain-length, sum-delta-length)
         self._chaininfocache = util.lrucachedict(500)
 
         return chunkcache
 
     def _load_inner(self, chunk_cache):
         if self._docket is None:
             default_compression_header = None
         else:
             default_compression_header = self._docket.default_compression_header
 
         self._inner = _InnerRevlog(
             opener=self.opener,
             index=self.index,
             index_file=self._indexfile,
             data_file=self._datafile,
             sidedata_file=self._sidedatafile,
             inline=self._inline,
             data_config=self.data_config,
             delta_config=self.delta_config,
             feature_config=self.feature_config,
             chunk_cache=chunk_cache,
             default_compression_header=default_compression_header,
         )
 
     def get_revlog(self):
         """simple function to mirror API of other not-really-revlog API"""
         return self
 
     @util.propertycache
     def revlog_kind(self):
         return self.target[0]
 
     @util.propertycache
     def display_id(self):
         """The public facing "ID" of the revlog that we use in message"""
         if self.revlog_kind == KIND_FILELOG:
             # Reference the file without the "data/" prefix, so it is familiar
             # to the user.
             return self.target[1]
         else:
             return self.radix
 
     def _datafp(self, mode=b'r'):
         """file object for the revlog's data file"""
         return self.opener(self._datafile, mode=mode)
 
     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(range(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)
 
     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):
         """update on disk cache
 
         If a transaction is passed, the update may be delayed to transaction
         commit."""
         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):
         """Clear in-memory caches"""
         self._chainbasecache.clear()
         self._inner.clear_cache()
         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 hasattr(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):
         """return the revision number associated with a <nodeid>"""
         try:
             return self.index.rev(node)
         except TypeError:
             raise
         except error.RevlogError:
             # parsers.c radix tree lookup failed
             if (
                 node == self.nodeconstants.wdirid
                 or node in self.nodeconstants.wdirfilenodeids
             ):
                 raise error.WdirUnsupported
             raise error.LookupError(node, self.display_id, _(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 sidedata_cut_off(self, rev):
         sd_cut_off = self.index[rev][8]
         if sd_cut_off != 0:
             return sd_cut_off
         # This is some annoying dance, because entries without sidedata
         # currently use 0 as their ofsset. (instead of previous-offset +
         # previous-size)
         #
         # We should reconsider this sidedata → 0 sidata_offset policy.
         # In the meantime, we need this.
         while 0 <= rev:
             e = self.index[rev]
             if e[9] != 0:
                 return e[8] + e[9]
             rev -= 1
         return 0
 
     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 not self.feature_config.has_side_data:
             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))
 
     def fast_rank(self, rev):
         """Return the rank of a revision if already known, or None otherwise.
 
         The rank of a revision is the size of the sub-graph it defines as a
         head. Equivalently, the rank of a revision `r` is the size of the set
         `ancestors(r)`, `r` included.
 
         This method returns the rank retrieved from the revlog in constant
         time. It makes no attempt at computing unknown values for versions of
         the revlog which do not persist the rank.
         """
         rank = self.index[rev][ENTRY_RANK]
         if self._format_version != CHANGELOGV2 or rank == RANK_UNKNOWN:
             return None
         if rev == nullrev:
             return 0  # convention
         return rank
 
     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
 
         if self.feature_config.canonical_parent_order and entry[5] == nullrev:
             return entry[6], entry[5]
         else:
             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)]
         # inline node() to avoid function call overhead
         if self.feature_config.canonical_parent_order and d[5] == self.nullid:
             return i[d[6]][7], i[d[5]][7]
         else:
             return i[d[5]][7], i[d[6]][7]
 
     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.delta_config.general_delta
         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):
         return self._inner._deltachain(rev, stoprev=stoprev)
 
     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 and self.index.rust_ext_compat:
             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 = [self.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:
             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 and self.index.rust_ext_compat:
             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 = [self.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 = [self.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],
                 [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 == self.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 != self.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 = [self.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 heads.items() 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 and self.index.rust_ext_compat:
             return rustdagop.headrevs(self.index, revs)
         return dagop.headrevs(revs, self._uncheckedparentrevs)
 
     def headrevsdiff(self, start, stop):
         try:
             return self.index.headrevsdiff(start, stop)
         except AttributeError:
             return dagop.headrevsdiff(self._uncheckedparentrevs, start, stop)
 
     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 _head_node_ids(self):
         try:
             return self.index.head_node_ids()
         except AttributeError:
             return [self.node(r) for r in self.headrevs()]
 
     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 [self.nullid]
             return self._head_node_ids()
         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 diffheads(self, start, stop):
         """return the nodes that make up the difference between
         heads of revs before `start` and heads of revs before `stop`"""
         removed, added = self.headrevsdiff(start, stop)
         return [self.node(r) for r in removed], [self.node(r) for r in added]
 
     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 self.nullid
 
     def _match(self, id):
         if isinstance(id, int):
             # rev
             return self.node(id)
         if len(id) == self.nodeconstants.nodelen:
             # 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) == 2 * self.nodeconstants.nodelen:
             try:
                 # a full hex nodeid?
                 node = bin(id)
                 self.rev(node)
                 return node
             except (binascii.Error, error.LookupError):
                 pass
 
     def _partialmatch(self, id):
         # we don't care wdirfilenodeids as they should be always full hash
         maybewdir = self.nodeconstants.wdirhex.startswith(id)
         ambiguous = False
         try:
             partial = self.index.partialmatch(id)
             if partial and self.hasnode(partial):
                 if maybewdir:
                     # single 'ff...' match in radix tree, ambiguous with wdir
                     ambiguous = True
                 else:
                     return partial
             elif maybewdir:
                 # no 'ff...' match in radix tree, wdir identified
                 raise error.WdirUnsupported
             else:
                 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):
                 ambiguous = True
             # fall through to slow path that filters hidden revisions
         except (AttributeError, ValueError):
             # we are pure python, or key is not hex
             pass
         if ambiguous:
             raise error.AmbiguousPrefixLookupError(
                 id, self.display_id, _(b'ambiguous identifier')
             )
 
         if id in self._pcache:
             return self._pcache[id]
 
         if len(id) <= 40:
             # hex(node)[:...]
             l = len(id) // 2 * 2  # grab an even number of digits
             try:
                 # we're dropping the last digit, so let's check that it's hex,
                 # to avoid the expensive computation below if it's not
                 if len(id) % 2 > 0:
                     if not (id[-1] in hexdigits):
                         return None
                 prefix = bin(id[:l])
             except binascii.Error:
                 pass
             else:
                 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 self.nodeconstants.nullhex.startswith(id):
                     nl.append(self.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.display_id, _(b'ambiguous identifier')
                     )
                 if maybewdir:
                     raise error.WdirUnsupported
                 return None
 
     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.display_id, _(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.display_id, _(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 != self.nodeconstants.wdirid:
                     raise error.LookupError(
                         node, self.display_id, _(b'no node')
                     )
             except AttributeError:
                 # Fall through to pure code
                 pass
 
         if node == self.nodeconstants.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 deltaparent(self, rev):
         """return deltaparent of the given revision"""
         base = self.index[rev][3]
         if base == rev:
             return nullrev
         elif self.delta_config.general_delta:
             return base
         else:
             return rev - 1
 
     def issnapshot(self, rev):
         """tells whether rev is a snapshot"""
         ret = self._inner.issnapshot(rev)
         self.issnapshot = self._inner.issnapshot
         return ret
 
     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._inner._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._inner._chunk(rev2))
 
         return mdiff.textdiff(self.rawdata(rev1), self.rawdata(rev2))
 
     def revision(self, nodeorrev):
         """return an uncompressed revision of a given node or revision
         number.
         """
         return self._revisiondata(nodeorrev)
 
     def sidedata(self, nodeorrev):
         """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.
         """
         # deal with <nodeorrev> argument type
         if isinstance(nodeorrev, int):
             rev = nodeorrev
         else:
             rev = self.rev(nodeorrev)
         return self._sidedata(rev)
 
     def _rawtext(self, node, rev):
         """return the possibly unvalidated rawtext for a revision
 
         returns (rev, rawtext, validated)
         """
         # Check if we have the entry in cache
         # The cache entry looks like (node, rev, rawtext)
         if self._inner._revisioncache:
             if self._inner._revisioncache[0] == node:
                 return (rev, self._inner._revisioncache[2], True)
 
         if rev is None:
             rev = self.rev(node)
 
         return self._inner.raw_text(node, rev)
 
     def _revisiondata(self, nodeorrev, 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 == self.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)
 
         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
         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
 
         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._inner._revisioncache = (node, rev, rawtext)
 
         return text
 
     def _sidedata(self, rev):
         """Return the sidedata for a given revision number."""
         sidedata_end = None
         if self._docket is not None:
             sidedata_end = self._docket.sidedata_end
         return self._inner.sidedata(rev, sidedata_end)
 
     def rawdata(self, nodeorrev):
         """return an uncompressed raw data of a given node or revision number."""
         return self._revisiondata(nodeorrev, raw=True)
 
     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._inner._revisioncache
                     and self._inner._revisioncache[0] == node
                 ):
                     self._inner._revisioncache = None
 
                 revornode = rev
                 if revornode is None:
                     revornode = templatefilters.short(hex(node))
                 raise error.RevlogError(
                     _(b"integrity check failed on %s:%s")
                     % (self.display_id, pycompat.bytestr(revornode))
                 )
         except error.RevlogError:
             if self.feature_config.censorable and storageutil.iscensoredtext(
                 text
             ):
                 raise error.CensoredNodeError(self.display_id, node, text)
             raise
 
     @property
     def _split_index_file(self):
         """the path where to expect the index of an ongoing splitting operation
 
         The file will only exist if a splitting operation is in progress, but
         it is always expected at the same location."""
         parts = self.radix.split(b'/')
         if len(parts) > 1:
             # adds a '-s' prefix to the ``data/` or `meta/` base
             head = parts[0] + b'-s'
             mids = parts[1:-1]
             tail = parts[-1] + b'.i'
             pieces = [head] + mids + [tail]
             return b'/'.join(pieces)
         else:
             # the revlog is stored at the root of the store (changelog or
             # manifest), no risk of collision.
             return self.radix + b'.i.s'
 
     def _enforceinlinesize(self, tr):
         """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
         total_size = self.start(tiprev) + self.length(tiprev)
         if not self._inline or (self._may_inline and total_size < _maxinline):
             return
 
         if self._docket is not None:
             msg = b"inline revlog should not have a docket"
             raise error.ProgrammingError(msg)
 
         # In the common case, we enforce inline size because the revlog has
         # been appened too. And in such case, it must have an initial offset
         # recorded in the transaction.
         troffset = tr.findoffset(self._inner.canonical_index_file)
         pre_touched = troffset is not None
         if not pre_touched and self.target[0] != KIND_CHANGELOG:
             raise error.RevlogError(
                 _(b"%s not found in the transaction") % self._indexfile
             )
 
         tr.addbackup(self._inner.canonical_index_file, for_offset=pre_touched)
         tr.add(self._datafile, 0)
 
         new_index_file_path = None
         old_index_file_path = self._indexfile
         new_index_file_path = self._split_index_file
         opener = self.opener
         weak_self = weakref.ref(self)
 
         # the "split" index replace the real index when the transaction is
         # finalized
         def finalize_callback(tr):
             opener.rename(
                 new_index_file_path,
                 old_index_file_path,
                 checkambig=True,
             )
             maybe_self = weak_self()
             if maybe_self is not None:
                 maybe_self._indexfile = old_index_file_path
                 maybe_self._inner.index_file = maybe_self._indexfile
 
         def abort_callback(tr):
             maybe_self = weak_self()
             if maybe_self is not None:
                 maybe_self._indexfile = old_index_file_path
                 maybe_self._inner.inline = True
                 maybe_self._inner.index_file = old_index_file_path
 
         tr.registertmp(new_index_file_path)
+        # we use 001 here to make this this happens after the finalisation of
+        # pending changelog write (using 000). Otherwise the two finalizer
+        # would step over each other and delete the changelog.i file.
         if self.target[1] is not None:
-            callback_id = b'000-revlog-split-%d-%s' % self.target
+            callback_id = b'001-revlog-split-%d-%s' % self.target
         else:
-            callback_id = b'000-revlog-split-%d' % self.target[0]
+            callback_id = b'001-revlog-split-%d' % self.target[0]
         tr.addfinalize(callback_id, finalize_callback)
         tr.addabort(callback_id, abort_callback)
 
         self._format_flags &= ~FLAG_INLINE_DATA
         self._inner.split_inline(
             tr,
             self._format_flags | self._format_version,
             new_index_file_path=new_index_file_path,
         )
 
         self._inline = False
         if new_index_file_path is not None:
             self._indexfile = new_index_file_path
 
         nodemaputil.setup_persistent_nodemap(tr, self)
 
     def _nodeduplicatecallback(self, transaction, node):
         """called when trying to add a node already stored."""
 
     @contextlib.contextmanager
     def reading(self):
         with self._inner.reading():
             yield
 
     @contextlib.contextmanager
     def _writing(self, transaction):
         if self._trypending:
             msg = b'try to write in a `trypending` revlog: %s'
             msg %= self.display_id
             raise error.ProgrammingError(msg)
         if self._inner.is_writing:
             yield
         else:
             data_end = None
             sidedata_end = None
             if self._docket is not None:
                 data_end = self._docket.data_end
                 sidedata_end = self._docket.sidedata_end
             with self._inner.writing(
                 transaction,
                 data_end=data_end,
                 sidedata_end=sidedata_end,
             ):
                 yield
                 if self._docket is not None:
                     self._write_docket(transaction)
 
     @property
     def is_delaying(self):
         return self._inner.is_delaying
 
     def _write_docket(self, transaction):
         """write the current docket on disk
 
         Exist as a method to help changelog to implement transaction logic
 
         We could also imagine using the same transaction logic for all revlog
         since docket are cheap."""
         self._docket.write(transaction)
 
     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.display_id
             )
 
         if sidedata is None:
             sidedata = {}
         elif sidedata and not self.feature_config.has_side_data:
             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.display_id, 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).
         """
         with self._writing(transaction):
             return self._addrevision(
                 node,
                 rawtext,
                 transaction,
                 link,
                 p1,
                 p2,
                 flags,
                 cachedelta,
                 deltacomputer=deltacomputer,
                 sidedata=sidedata,
             )
 
     def compress(self, data):
         return self._inner.compress(data)
 
     def decompress(self, data):
         return self._inner.decompress(data)
 
     def _addrevision(
         self,
         node,
         rawtext,
         transaction,
         link,
         p1,
         p2,
         flags,
         cachedelta,
         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 == self.nullid:
             raise error.RevlogError(
                 _(b"%s: attempt to add null revision") % self.display_id
             )
         if (
             node == self.nodeconstants.wdirid
             or node in self.nodeconstants.wdirfilenodeids
         ):
             raise error.RevlogError(
                 _(b"%s: attempt to add wdir revision") % self.display_id
             )
         if self._inner._writinghandles is None:
             msg = b'adding revision outside `revlog._writing` context'
             raise error.ProgrammingError(msg)
 
         btext = [rawtext]
 
         curr = len(self)
         prev = curr - 1
 
         offset = self._get_data_offset(prev)
 
         if self._concurrencychecker:
             ifh, dfh, sdfh = self._inner._writinghandles
             # XXX no checking for the sidedata file
             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.index.entry_size
                 )
             else:
                 # Entries in the .i are a consistent size.
                 self._concurrencychecker(
                     ifh, self._indexfile, curr * self.index.entry_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:
             write_debug = None
             if self.delta_config.debug_delta:
                 write_debug = transaction._report
             deltacomputer = deltautil.deltacomputer(
                 self, write_debug=write_debug
             )
 
         if cachedelta is not None and len(cachedelta) == 2:
             # If the cached delta has no information about how it should be
             # reused, add the default reuse instruction according to the
             # revlog's configuration.
             if (
                 self.delta_config.general_delta
                 and self.delta_config.lazy_delta_base
             ):
                 delta_base_reuse = DELTA_BASE_REUSE_TRY
             else:
                 delta_base_reuse = DELTA_BASE_REUSE_NO
             cachedelta = (cachedelta[0], cachedelta[1], delta_base_reuse)
 
         revinfo = revlogutils.revisioninfo(
             node,
             p1,
             p2,
             btext,
             textlen,
             cachedelta,
             flags,
         )
 
         deltainfo = deltacomputer.finddeltainfo(revinfo)
 
         compression_mode = COMP_MODE_INLINE
         if self._docket is not None:
             default_comp = self._docket.default_compression_header
             r = deltautil.delta_compression(default_comp, deltainfo)
             compression_mode, deltainfo = r
 
         sidedata_compression_mode = COMP_MODE_INLINE
         if sidedata and self.feature_config.has_side_data:
             sidedata_compression_mode = COMP_MODE_PLAIN
             serialized_sidedata = sidedatautil.serialize_sidedata(sidedata)
             sidedata_offset = self._docket.sidedata_end
             h, comp_sidedata = self._inner.compress(serialized_sidedata)
             if (
                 h != b'u'
                 and comp_sidedata[0:1] != b'\0'
                 and len(comp_sidedata) < len(serialized_sidedata)
             ):
                 assert not h
                 if (
                     comp_sidedata[0:1]
                     == self._docket.default_compression_header
                 ):
                     sidedata_compression_mode = COMP_MODE_DEFAULT
                     serialized_sidedata = comp_sidedata
                 else:
                     sidedata_compression_mode = COMP_MODE_INLINE
                     serialized_sidedata = comp_sidedata
         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
 
         rank = RANK_UNKNOWN
         if self.feature_config.compute_rank:
             if (p1r, p2r) == (nullrev, nullrev):
                 rank = 1
             elif p1r != nullrev and p2r == nullrev:
                 rank = 1 + self.fast_rank(p1r)
             elif p1r == nullrev and p2r != nullrev:
                 rank = 1 + self.fast_rank(p2r)
             else:  # merge node
                 if rustdagop is not None and self.index.rust_ext_compat:
                     rank = rustdagop.rank(self.index, p1r, p2r)
                 else:
                     pmin, pmax = sorted((p1r, p2r))
                     rank = 1 + self.fast_rank(pmax)
                     rank += sum(1 for _ in self.findmissingrevs([pmax], [pmin]))
 
         e = revlogutils.entry(
             flags=flags,
             data_offset=offset,
             data_compressed_length=deltainfo.deltalen,
             data_uncompressed_length=textlen,
             data_compression_mode=compression_mode,
             data_delta_base=deltainfo.base,
             link_rev=link,
             parent_rev_1=p1r,
             parent_rev_2=p2r,
             node_id=node,
             sidedata_offset=sidedata_offset,
             sidedata_compressed_length=len(serialized_sidedata),
             sidedata_compression_mode=sidedata_compression_mode,
             rank=rank,
         )
 
         self.index.append(e)
         entry = self.index.entry_binary(curr)
         if curr == 0 and self._docket is None:
             header = self._format_flags | self._format_version
             header = self.index.pack_header(header)
             entry = header + entry
         self._writeentry(
             transaction,
             entry,
             deltainfo.data,
             link,
             offset,
             serialized_sidedata,
             sidedata_offset,
         )
 
         rawtext = btext[0]
 
         if alwayscache and rawtext is None:
             rawtext = deltacomputer.buildtext(revinfo)
 
         if type(rawtext) == bytes:  # only accept immutable objects
             self._inner._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._docket is None:
             return self.end(prev)
         else:
             return self._docket.data_end
 
     def _writeentry(
         self,
         transaction,
         entry,
         data,
         link,
         offset,
         sidedata,
         sidedata_offset,
     ):
         # 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.
         index_end = data_end = sidedata_end = None
         if self._docket is not None:
             index_end = self._docket.index_end
             data_end = self._docket.data_end
             sidedata_end = self._docket.sidedata_end
 
         files_end = self._inner.write_entry(
             transaction,
             entry,
             data,
             link,
             offset,
             sidedata,
             sidedata_offset,
             index_end,
             data_end,
             sidedata_end,
         )
         self._enforceinlinesize(transaction)
         if self._docket is not None:
             self._docket.index_end = files_end[0]
             self._docket.data_end = files_end[1]
             self._docket.sidedata_end = files_end[2]
 
         nodemaputil.setup_persistent_nodemap(transaction, self)
 
     def addgroup(
         self,
         deltas,
         linkmapper,
         transaction,
         alwayscache=False,
         addrevisioncb=None,
         duplicaterevisioncb=None,
         debug_info=None,
         delta_base_reuse_policy=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._adding_group:
             raise error.ProgrammingError(b'cannot nest addgroup() calls')
 
         # read the default delta-base reuse policy from revlog config if the
         # group did not specify one.
         if delta_base_reuse_policy is None:
             if (
                 self.delta_config.general_delta
                 and self.delta_config.lazy_delta_base
             ):
                 delta_base_reuse_policy = DELTA_BASE_REUSE_TRY
             else:
                 delta_base_reuse_policy = DELTA_BASE_REUSE_NO
 
         self._adding_group = True
         empty = True
         try:
             with self._writing(transaction):
                 write_debug = None
                 if self.delta_config.debug_delta:
                     write_debug = transaction._report
                 deltacomputer = deltautil.deltacomputer(
                     self,
                     write_debug=write_debug,
                     debug_info=debug_info,
                 )
                 # 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.radix, _(b'unknown parent')
                             )
 
                     if not self.index.has_node(deltabase):
                         raise error.LookupError(
                             deltabase, self.display_id, _(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.display_id, self.node(baserev)
                             )
 
                     if not flags and self._peek_iscensored(baserev, delta):
                         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, delta_base_reuse_policy),
                         alwayscache=alwayscache,
                         deltacomputer=deltacomputer,
                         sidedata=sidedata,
                     )
 
                     if addrevisioncb:
                         addrevisioncb(self, rev)
                     empty = False
         finally:
             self._adding_group = False
         return not empty
 
     def iscensored(self, rev):
         """Check if a file revision is censored."""
         if not self.feature_config.censorable:
             return False
 
         return self.flags(rev) & REVIDX_ISCENSORED
 
     def _peek_iscensored(self, baserev, delta):
         """Quickly check if a delta produces a censored revision."""
         if not self.feature_config.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
         data_end = self.start(rev)
         if not self._inline:
             transaction.add(self._datafile, data_end)
             end = rev * self.index.entry_size
         else:
             end = data_end + (rev * self.index.entry_size)
 
         if self._sidedatafile:
             sidedata_end = self.sidedata_cut_off(rev)
             transaction.add(self._sidedatafile, sidedata_end)
 
         transaction.add(self._indexfile, end)
         if self._docket is not None:
             # XXX we could, leverage the docket while stripping. However it is
             # not powerfull enough at the time of this comment
             self._docket.index_end = end
             self._docket.data_end = data_end
             self._docket.sidedata_end = sidedata_end
             self._docket.write(transaction, stripping=True)
 
         # then reset internal state in memory to forget those revisions
         self._chaininfocache = util.lrucachedict(500)
         self._inner.clear_cache()
 
         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 FileNotFoundError:
             dd = 0
 
         try:
             f = self.opener(self._indexfile)
             f.seek(0, io.SEEK_END)
             actual = f.tell()
             f.close()
             s = self.index.entry_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 FileNotFoundError:
             di = 0
 
         return (dd, di)
 
     def files(self):
         """return list of files that compose this revlog"""
         res = [self._indexfile]
         if self._docket_file is None:
             if not self._inline:
                 res.append(self._datafile)
         else:
             res.append(self._docket_file)
             res.extend(self._docket.old_index_filepaths(include_empty=False))
             if self._docket.data_end:
                 res.append(self._datafile)
             res.extend(self._docket.old_data_filepaths(include_empty=False))
             if self._docket.sidedata_end:
                 res.append(self._sidedatafile)
             res.extend(self._docket.old_sidedata_filepaths(include_empty=False))
         return res
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
         sidedata_helpers=None,
         debug_info=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.delta_config.general_delta:
             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,
             debug_info=debug_info,
         )
 
     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,
         sidedata_helpers=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.
 
         See `revlogutil.sidedata.get_sidedata_helpers` for the doc on
         `sidedata_helpers`.
         """
         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.
         old_delta_config = destrevlog.delta_config
         destrevlog.delta_config = destrevlog.delta_config.copy()
 
         try:
             if deltareuse == self.DELTAREUSEALWAYS:
                 destrevlog.delta_config.lazy_delta_base = True
                 destrevlog.delta_config.lazy_delta = True
             elif deltareuse == self.DELTAREUSESAMEREVS:
                 destrevlog.delta_config.lazy_delta_base = False
                 destrevlog.delta_config.lazy_delta = True
             elif deltareuse == self.DELTAREUSENEVER:
                 destrevlog.delta_config.lazy_delta_base = False
                 destrevlog.delta_config.lazy_delta = False
 
             delta_both_parents = (
                 forcedeltabothparents or old_delta_config.delta_both_parents
             )
             destrevlog.delta_config.delta_both_parents = delta_both_parents
 
             with self.reading(), destrevlog._writing(tr):
                 self._clone(
                     tr,
                     destrevlog,
                     addrevisioncb,
                     deltareuse,
                     forcedeltabothparents,
                     sidedata_helpers,
                 )
 
         finally:
             destrevlog.delta_config = old_delta_config
 
     def _clone(
         self,
         tr,
         destrevlog,
         addrevisioncb,
         deltareuse,
         forcedeltabothparents,
         sidedata_helpers,
     ):
         """perform the core duty of `revlog.clone` after parameter processing"""
         write_debug = None
         if self.delta_config.debug_delta:
             write_debug = tr._report
         deltacomputer = deltautil.deltacomputer(
             destrevlog,
             write_debug=write_debug,
         )
         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]
 
             # (Possibly) reuse the delta from the revlog if allowed and
             # the revlog chunk is a delta.
             cachedelta = None
             rawtext = None
             if deltareuse == self.DELTAREUSEFULLADD:
                 text = self._revisiondata(rev)
                 sidedata = self.sidedata(rev)
 
                 if sidedata_helpers is not None:
                     (sidedata, new_flags) = sidedatautil.run_sidedata_helpers(
                         self, sidedata_helpers, sidedata, rev
                     )
                     flags = flags | new_flags[0] & ~new_flags[1]
 
                 destrevlog.addrevision(
                     text,
                     tr,
                     linkrev,
                     p1,
                     p2,
                     cachedelta=cachedelta,
                     node=node,
                     flags=flags,
                     deltacomputer=deltacomputer,
                     sidedata=sidedata,
                 )
             else:
                 if destrevlog.delta_config.lazy_delta:
                     dp = self.deltaparent(rev)
                     if dp != nullrev:
                         cachedelta = (dp, bytes(self._inner._chunk(rev)))
 
                 sidedata = None
                 if not cachedelta:
                     try:
                         rawtext = self._revisiondata(rev)
                     except error.CensoredNodeError as censored:
                         assert flags & REVIDX_ISCENSORED
                         rawtext = censored.tombstone
                     sidedata = self.sidedata(rev)
                 if sidedata is None:
                     sidedata = self.sidedata(rev)
 
                 if sidedata_helpers is not None:
                     (sidedata, new_flags) = sidedatautil.run_sidedata_helpers(
                         self, sidedata_helpers, sidedata, rev
                     )
                     flags = flags | new_flags[0] & ~new_flags[1]
 
                 destrevlog._addrevision(
                     node,
                     rawtext,
                     tr,
                     linkrev,
                     p1,
                     p2,
                     flags,
                     cachedelta,
                     deltacomputer=deltacomputer,
                     sidedata=sidedata,
                 )
 
             if addrevisioncb:
                 addrevisioncb(self, rev, node)
 
     def censorrevision(self, tr, censor_nodes, tombstone=b''):
         if self._format_version == REVLOGV0:
             raise error.RevlogError(
                 _(b'cannot censor with version %d revlogs')
                 % self._format_version
             )
         elif self._format_version == REVLOGV1:
             rewrite.v1_censor(self, tr, censor_nodes, tombstone)
         else:
             rewrite.v2_censor(self, tr, censor_nodes, tombstone)
 
     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._format_version
 
         # 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.display_id, 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
 
     def rewrite_sidedata(self, transaction, helpers, startrev, endrev):
         if not self.feature_config.has_side_data:
             return
         # revlog formats with sidedata support does not support inline
         assert not self._inline
         if not helpers[1] and not helpers[2]:
             # Nothing to generate or remove
             return
 
         new_entries = []
         # append the new sidedata
         with self._writing(transaction):
             ifh, dfh, sdfh = self._inner._writinghandles
             dfh.seek(self._docket.sidedata_end, os.SEEK_SET)
 
             current_offset = sdfh.tell()
             for rev in range(startrev, endrev + 1):
                 entry = self.index[rev]
                 new_sidedata, flags = sidedatautil.run_sidedata_helpers(
                     store=self,
                     sidedata_helpers=helpers,
                     sidedata={},
                     rev=rev,
                 )
 
                 serialized_sidedata = sidedatautil.serialize_sidedata(
                     new_sidedata
                 )
 
                 sidedata_compression_mode = COMP_MODE_INLINE
                 if serialized_sidedata and self.feature_config.has_side_data:
                     sidedata_compression_mode = COMP_MODE_PLAIN
                     h, comp_sidedata = self._inner.compress(serialized_sidedata)
                     if (
                         h != b'u'
                         and comp_sidedata[0] != b'\0'
                         and len(comp_sidedata) < len(serialized_sidedata)
                     ):
                         assert not h
                         if (
                             comp_sidedata[0]
                             == self._docket.default_compression_header
                         ):
                             sidedata_compression_mode = COMP_MODE_DEFAULT
                             serialized_sidedata = comp_sidedata
                         else:
                             sidedata_compression_mode = COMP_MODE_INLINE
                             serialized_sidedata = comp_sidedata
                 if entry[8] != 0 or entry[9] != 0:
                     # rewriting entries that already have sidedata is not
                     # supported yet, because it introduces garbage data in the
                     # revlog.
                     msg = b"rewriting existing sidedata is not supported yet"
                     raise error.Abort(msg)
 
                 # Apply (potential) flags to add and to remove after running
                 # the sidedata helpers
                 new_offset_flags = entry[0] | flags[0] & ~flags[1]
                 entry_update = (
                     current_offset,
                     len(serialized_sidedata),
                     new_offset_flags,
                     sidedata_compression_mode,
                 )
 
                 # the sidedata computation might have move the file cursors around
                 sdfh.seek(current_offset, os.SEEK_SET)
                 sdfh.write(serialized_sidedata)
                 new_entries.append(entry_update)
                 current_offset += len(serialized_sidedata)
                 self._docket.sidedata_end = sdfh.tell()
 
             # rewrite the new index entries
             ifh.seek(startrev * self.index.entry_size)
             for i, e in enumerate(new_entries):
                 rev = startrev + i
                 self.index.replace_sidedata_info(rev, *e)
                 packed = self.index.entry_binary(rev)
                 if rev == 0 and self._docket is None:
                     header = self._format_flags | self._format_version
                     header = self.index.pack_header(header)
                     packed = header + packed
                 ifh.write(packed)

tests/test-bookmarks.t

@@ -1,1253 +1,1253 b''
 
   $ hg init repo
   $ cd repo
 
   $ cat > $TESTTMP/hook.sh <<'EOF'
   > echo "test-hook-bookmark: $HG_BOOKMARK:  $HG_OLDNODE -> $HG_NODE"
   > EOF
   $ TESTHOOK="hooks.txnclose-bookmark.test=sh $TESTTMP/hook.sh"
 
 no bookmarks
 
   $ hg bookmarks
   no bookmarks set
 
   $ hg bookmarks -Tjson
   [
   ]
 
 bookmark rev -1
 
   $ hg bookmark X --config "$TESTHOOK"
   test-hook-bookmark: X:   -> 0000000000000000000000000000000000000000
 
 list bookmarks
 
   $ hg bookmarks
    * X                         -1:000000000000
 
 list bookmarks with color
 
   $ hg --config extensions.color= --config color.mode=ansi \
   >    bookmarks --color=always
   \x1b[0;32m * \x1b[0m\x1b[0;32mX\x1b[0m\x1b[0;32m                         -1:000000000000\x1b[0m (esc)
 
   $ echo a > a
   $ hg add a
   $ hg commit -m 0 --config "$TESTHOOK"
   test-hook-bookmark: X:  0000000000000000000000000000000000000000 -> f7b1eb17ad24730a1651fccd46c43826d1bbc2ac
 
 bookmark X moved to rev 0
 
   $ hg bookmarks
    * X                         0:f7b1eb17ad24
 
 look up bookmark
 
   $ hg log -r X
   changeset:   0:f7b1eb17ad24
   bookmark:    X
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     0
   
 
 second bookmark for rev 0, command should work even with ui.strict on
 
   $ hg --config ui.strict=1 bookmark X2 --config "$TESTHOOK"
   test-hook-bookmark: X2:   -> f7b1eb17ad24730a1651fccd46c43826d1bbc2ac
 
 bookmark rev -1 again
 
   $ hg bookmark -r null Y
 
 list bookmarks
 
   $ hg bookmarks
      X                         0:f7b1eb17ad24
    * X2                        0:f7b1eb17ad24
      Y                         -1:000000000000
   $ hg bookmarks -l
      X                         0:f7b1eb17ad24
    * X2                        0:f7b1eb17ad24
      Y                         -1:000000000000
   $ hg bookmarks -l X Y
      X                         0:f7b1eb17ad24
      Y                         -1:000000000000
   $ hg bookmarks -l .
    * X2                        0:f7b1eb17ad24
   $ hg bookmarks -l X A Y
   abort: bookmark 'A' does not exist
   [10]
   $ hg bookmarks -l -r0
   abort: cannot specify both --list and --rev
   [10]
   $ hg bookmarks -l --inactive
   abort: cannot specify both --inactive and --list
   [10]
 
   $ hg log -T '{bookmarks % "{rev} {bookmark}\n"}'
   0 X
   0 X2
 
   $ echo b > b
   $ hg add b
   $ hg commit -m 1 --config "$TESTHOOK"
   test-hook-bookmark: X2:  f7b1eb17ad24730a1651fccd46c43826d1bbc2ac -> 925d80f479bb026b0fb3deb27503780b13f74123
 
   $ hg bookmarks -T '{rev}:{node|shortest} {bookmark} {desc|firstline}\n'
   0:f7b1 X 0
   1:925d X2 1
   -1:0000 Y 
 
   $ hg bookmarks -Tjson
   [
    {
     "active": false,
     "bookmark": "X",
     "node": "f7b1eb17ad24730a1651fccd46c43826d1bbc2ac",
     "rev": 0
    },
    {
     "active": true,
     "bookmark": "X2",
     "node": "925d80f479bb026b0fb3deb27503780b13f74123",
     "rev": 1
    },
    {
     "active": false,
     "bookmark": "Y",
     "node": "0000000000000000000000000000000000000000",
     "rev": -1
    }
   ]
 
 bookmarks revset
 
   $ hg log -r 'bookmark()'
   changeset:   0:f7b1eb17ad24
   bookmark:    X
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     0
   
   changeset:   1:925d80f479bb
   bookmark:    X2
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     1
   
   $ hg log -r 'bookmark(Y)'
   $ hg log -r 'bookmark(X2)'
   changeset:   1:925d80f479bb
   bookmark:    X2
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     1
   
   $ hg log -r 'bookmark("re:X")'
   changeset:   0:f7b1eb17ad24
   bookmark:    X
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     0
   
   changeset:   1:925d80f479bb
   bookmark:    X2
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     1
   
   $ hg log -r 'bookmark("literal:X")'
   changeset:   0:f7b1eb17ad24
   bookmark:    X
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     0
   
 
 "." is expanded to the active bookmark:
 
   $ hg log -r 'bookmark(.)'
   changeset:   1:925d80f479bb
   bookmark:    X2
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     1
   
 
 but "literal:." is not since "." seems not a literal bookmark:
 
   $ hg log -r 'bookmark("literal:.")'
   abort: bookmark '.' does not exist
   [10]
 
 "." should fail if there's no active bookmark:
 
   $ hg bookmark --inactive
   $ hg log -r 'bookmark(.)'
   abort: no active bookmark
   [10]
   $ hg log -r 'present(bookmark(.))'
 
   $ hg log -r 'bookmark(unknown)'
   abort: bookmark 'unknown' does not exist
   [10]
   $ hg log -r 'bookmark("literal:unknown")'
   abort: bookmark 'unknown' does not exist
   [10]
   $ hg log -r 'bookmark("re:unknown")'
   $ hg log -r 'present(bookmark("literal:unknown"))'
   $ hg log -r 'present(bookmark("re:unknown"))'
 
   $ hg help revsets | grep 'bookmark('
       "bookmark([name])"
 
 reactivate "X2"
 
   $ hg update X2
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (activating bookmark X2)
 
 bookmarks X and X2 moved to rev 1, Y at rev -1
 
   $ hg bookmarks
      X                         0:f7b1eb17ad24
    * X2                        1:925d80f479bb
      Y                         -1:000000000000
 
 bookmark rev 0 again
 
   $ hg bookmark -r 0 Z
 
   $ hg update X
   0 files updated, 0 files merged, 1 files removed, 0 files unresolved
   (activating bookmark X)
   $ echo c > c
   $ hg add c
   $ hg commit -m 2
   created new head
 
 bookmarks X moved to rev 2, Y at rev -1, Z at rev 0
 
   $ hg bookmarks
    * X                         2:db815d6d32e6
      X2                        1:925d80f479bb
      Y                         -1:000000000000
      Z                         0:f7b1eb17ad24
 
 rename nonexistent bookmark
 
   $ hg bookmark -m A B
   abort: bookmark 'A' does not exist
   [10]
 
 rename to existent bookmark
 
   $ hg bookmark -m X Y
   abort: bookmark 'Y' already exists (use -f to force)
   [255]
 
 force rename to existent bookmark
 
   $ hg bookmark -f -m X Y
 
 rename bookmark using .
 
   $ hg book rename-me
   $ hg book -m . renamed --config "$TESTHOOK"
   test-hook-bookmark: rename-me:  db815d6d32e69058eadefc8cffbad37675707975 -> 
   test-hook-bookmark: renamed:   -> db815d6d32e69058eadefc8cffbad37675707975
   $ hg bookmark
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
    * renamed                   2:db815d6d32e6
   $ hg up -q Y
   $ hg book -d renamed --config "$TESTHOOK"
   test-hook-bookmark: renamed:  db815d6d32e69058eadefc8cffbad37675707975 -> 
 
 rename bookmark using . with no active bookmark
 
   $ hg book rename-me
   $ hg book -i rename-me
   $ hg book -m . renamed
   abort: no active bookmark
   [10]
   $ hg up -q Y
   $ hg book -d rename-me
 
 delete bookmark using .
 
   $ hg book delete-me
   $ hg book -d .
   $ hg bookmark
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
   $ hg up -q Y
 
 delete bookmark using . with no active bookmark
 
   $ hg book delete-me
   $ hg book -i delete-me
   $ hg book -d .
   abort: no active bookmark
   [10]
   $ hg up -q Y
   $ hg book -d delete-me
 
 list bookmarks
 
   $ hg bookmark
      X2                        1:925d80f479bb
    * Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
 
 bookmarks from a revset
   $ hg bookmark -r '.^1' REVSET
   $ hg bookmark -r ':tip' TIP
   $ hg up -q TIP
   $ hg bookmarks
      REVSET                    0:f7b1eb17ad24
    * TIP                       2:db815d6d32e6
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
 
   $ hg bookmark -d REVSET
   $ hg bookmark -d TIP
 
 rename without new name or multiple names
 
   $ hg bookmark -m Y
   abort: new bookmark name required
   [10]
   $ hg bookmark -m Y Y2 Y3
   abort: only one new bookmark name allowed
   [10]
 
 delete without name
 
   $ hg bookmark -d
   abort: bookmark name required
   [10]
 
 delete nonexistent bookmark
 
   $ hg bookmark -d A
   abort: bookmark 'A' does not exist
   [10]
 
 delete with --inactive
 
   $ hg bookmark -d --inactive Y
   abort: cannot specify both --inactive and --delete
   [10]
 
 bookmark name with spaces should be stripped
 
   $ hg bookmark ' x  y '
 
 list bookmarks
 
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
    * x  y                      2:db815d6d32e6
   $ hg log -T '{bookmarks % "{rev} {bookmark}\n"}'
   2 Y
   2 x  y
   1 X2
   0 Z
 
 look up stripped bookmark name
 
   $ hg log -r '"x  y"'
   changeset:   2:db815d6d32e6
   bookmark:    Y
   bookmark:    x  y
   tag:         tip
   parent:      0:f7b1eb17ad24
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     2
   
 
 reject bookmark name with newline
 
   $ hg bookmark '
   > '
   abort: bookmark names cannot consist entirely of whitespace
   [10]
 
   $ hg bookmark -m Z '
   > '
   abort: bookmark names cannot consist entirely of whitespace
   [10]
 
 bookmark with reserved name
 
   $ hg bookmark tip
   abort: the name 'tip' is reserved
   [10]
 
   $ hg bookmark .
   abort: the name '.' is reserved
   [10]
 
   $ hg bookmark null
   abort: the name 'null' is reserved
   [10]
 
 
 bookmark with existing name
 
   $ hg bookmark X2
   abort: bookmark 'X2' already exists (use -f to force)
   [255]
 
   $ hg bookmark -m Y Z
   abort: bookmark 'Z' already exists (use -f to force)
   [255]
 
 bookmark with name of branch
 
   $ hg bookmark default
   abort: a bookmark cannot have the name of an existing branch
   [255]
 
   $ hg bookmark -m Y default
   abort: a bookmark cannot have the name of an existing branch
   [255]
 
 bookmark with integer name
 
   $ hg bookmark 10
   abort: cannot use an integer as a name
   [10]
 
 bookmark with a name that matches a node id
   $ hg bookmark 925d80f479bb db815d6d32e6 --config "$TESTHOOK"
   bookmark 925d80f479bb matches a changeset hash
   (did you leave a -r out of an 'hg bookmark' command?)
   bookmark db815d6d32e6 matches a changeset hash
   (did you leave a -r out of an 'hg bookmark' command?)
   test-hook-bookmark: 925d80f479bb:   -> db815d6d32e69058eadefc8cffbad37675707975
   test-hook-bookmark: db815d6d32e6:   -> db815d6d32e69058eadefc8cffbad37675707975
   $ hg bookmark -d 925d80f479bb
   $ hg bookmark -d db815d6d32e6
 
   $ cd ..
 
 bookmark with a name that matches an ambiguous node id
 
   $ hg init ambiguous
   $ cd ambiguous
   $ echo 0 > a
   $ hg ci -qAm 0
   $ for i in 1057 2857 4025; do
   >   hg up -q 0
   >   echo $i > a
   >   hg ci -qm $i
   > done
   $ hg up -q null
   $ hg log -r0: -T '{rev}:{node}\n'
   0:b4e73ffab476aa0ee32ed81ca51e07169844bc6a
   1:c56256a09cd28e5764f32e8e2810d0f01e2e357a
   2:c5623987d205cd6d9d8389bfc40fff9dbb670b48
   3:c562ddd9c94164376c20b86b0b4991636a3bf84f
 
   $ hg bookmark -r0 c562
   $ hg bookmarks
      c562                      0:b4e73ffab476
 
   $ cd ..
 
 incompatible options
 
   $ cd repo
 
   $ hg bookmark -m Y -d Z
   abort: cannot specify both --delete and --rename
   [10]
 
   $ hg bookmark -r 1 -d Z
   abort: cannot specify both --delete and --rev
   [10]
 
   $ hg bookmark -r 1 -m Z Y
   abort: cannot specify both --rename and --rev
   [10]
 
 force bookmark with existing name
 
   $ hg bookmark -f X2 --config "$TESTHOOK"
   test-hook-bookmark: X2:  925d80f479bb026b0fb3deb27503780b13f74123 -> db815d6d32e69058eadefc8cffbad37675707975
 
 force bookmark back to where it was, should deactivate it
 
   $ hg bookmark -fr1 X2
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         0:f7b1eb17ad24
      x  y                      2:db815d6d32e6
 
 forward bookmark to descendant without --force
 
   $ hg bookmark Z
   moving bookmark 'Z' forward from f7b1eb17ad24
 
 list bookmarks
 
   $ hg bookmark
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg log -T '{bookmarks % "{rev} {bookmark}\n"}'
   2 Y
   2 Z
   2 x  y
   1 X2
 
 revision but no bookmark name
 
   $ hg bookmark -r .
   abort: bookmark name required
   [10]
 
 bookmark name with whitespace only
 
   $ hg bookmark ' '
   abort: bookmark names cannot consist entirely of whitespace
   [10]
 
   $ hg bookmark -m Y ' '
   abort: bookmark names cannot consist entirely of whitespace
   [10]
 
 invalid bookmark
 
   $ hg bookmark 'foo:bar'
   abort: ':' cannot be used in a name
   [10]
 
   $ hg bookmark 'foo
   > bar'
   abort: '\n' cannot be used in a name
   [10]
 
 the bookmark extension should be ignored now that it is part of core
 
   $ echo "[extensions]" >> $HGRCPATH
   $ echo "bookmarks=" >> $HGRCPATH
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
 
 test summary
 
   $ hg summary
   parent: 2:db815d6d32e6 tip
    2
   branch: default
   bookmarks: *Z Y x  y
   commit: (clean)
   update: 1 new changesets, 2 branch heads (merge)
   phases: 3 draft
 
 test id
 
   $ hg id
   db815d6d32e6 tip Y/Z/x  y
 
 test rollback
 
   $ echo foo > f1
   $ hg bookmark tmp-rollback
   $ hg add .
   adding f1
   $ hg ci -mr
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
    * tmp-rollback              3:2bf5cfec5864
      x  y                      2:db815d6d32e6
   $ hg rollback
   repository tip rolled back to revision 2 (undo commit)
   working directory now based on revision 2
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
    * tmp-rollback              2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg bookmark -f Z -r 1
   $ hg rollback
   repository tip rolled back to revision 2 (undo bookmark)
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
    * tmp-rollback              2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg bookmark -d tmp-rollback
 
 activate bookmark on working dir parent without --force
 
   $ hg bookmark --inactive Z
   $ hg bookmark Z
 
 deactivate current 'Z', but also add 'Y'
 
   $ hg bookmark -d Y
   $ hg bookmark --inactive Z Y
   $ hg bookmark -l
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg bookmark Z
 
 bookmark wdir to activate it (issue6218)
 
   $ hg bookmark -d Z
   $ hg bookmark -r 'wdir()' Z
   $ hg bookmark -l
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
 
 test clone
 
   $ hg bookmark -r 2 -i @
   $ hg bookmark -r 2 -i a@
   $ hg bookmarks
      @                         2:db815d6d32e6
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         2:db815d6d32e6
      a@                        2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg clone . cloned-bookmarks
   updating to bookmark @
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R cloned-bookmarks bookmarks
    * @                         2:db815d6d32e6
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
      a@                        2:db815d6d32e6
      x  y                      2:db815d6d32e6
 
 test clone with pull protocol
 
   $ hg clone --pull . cloned-bookmarks-pull
   requesting all changes
   adding changesets
   adding manifests
   adding file changes
   added 3 changesets with 3 changes to 3 files (+1 heads)
   new changesets f7b1eb17ad24:db815d6d32e6
   updating to bookmark @
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R cloned-bookmarks-pull bookmarks
    * @                         2:db815d6d32e6
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
      a@                        2:db815d6d32e6
      x  y                      2:db815d6d32e6
 
 delete multiple bookmarks at once
 
   $ hg bookmark -d @ a@
 
 test clone with a bookmark named "default" (issue3677)
 
   $ hg bookmark -r 1 -f -i default
   $ hg clone . cloned-bookmark-default
   updating to branch default
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R cloned-bookmark-default bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
      Z                         2:db815d6d32e6
      default                   1:925d80f479bb
      x  y                      2:db815d6d32e6
   $ hg -R cloned-bookmark-default parents -q
   2:db815d6d32e6
   $ hg bookmark -d default
 
 test clone with a specific revision
 
   $ hg clone -r 925d80 . cloned-bookmarks-rev
   adding changesets
   adding manifests
   adding file changes
   added 2 changesets with 2 changes to 2 files
   new changesets f7b1eb17ad24:925d80f479bb
   updating to branch default
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R cloned-bookmarks-rev bookmarks
      X2                        1:925d80f479bb
 
 test clone with update to a bookmark
 
   $ hg clone -u Z . ../cloned-bookmarks-update
   updating to branch default
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R ../cloned-bookmarks-update bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
 
 create bundle with two heads
 
   $ hg clone . tobundle
   updating to branch default
   2 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ echo x > tobundle/x
   $ hg -R tobundle add tobundle/x
   $ hg -R tobundle commit -m'x'
   $ hg -R tobundle update -r -2
   0 files updated, 0 files merged, 1 files removed, 0 files unresolved
   $ echo y > tobundle/y
   $ hg -R tobundle branch test
   marked working directory as branch test
   (branches are permanent and global, did you want a bookmark?)
   $ hg -R tobundle add tobundle/y
   $ hg -R tobundle commit -m'y'
   $ hg -R tobundle bundle tobundle.hg
   searching for changes
   2 changesets found
   $ hg unbundle tobundle.hg
   adding changesets
   adding manifests
   adding file changes
   added 2 changesets with 2 changes to 2 files (+1 heads)
   new changesets 125c9a1d6df6:9ba5f110a0b3 (2 drafts)
   (run 'hg heads' to see heads, 'hg merge' to merge)
 
 update to active bookmark if it's not the parent
 
 (it is known issue that fsmonitor can't handle nested repositories. In
 this test scenario, cloned-bookmark-default and tobundle exist in the
 working directory of current repository)
 
   $ hg summary
   parent: 2:db815d6d32e6 
    2
   branch: default
   bookmarks: *Z Y x  y
   commit: 1 added, 1 unknown (new branch head) (no-fsmonitor !)
   commit: 1 added, * unknown (new branch head) (glob) (fsmonitor !)
   update: 2 new changesets (update)
   phases: 5 draft
   $ hg update
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   updating bookmark Z
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         2:db815d6d32e6
    * Z                         3:125c9a1d6df6
      x  y                      2:db815d6d32e6
 
 pull --update works the same as pull && update
 
   $ hg bookmark -r3 Y
   moving bookmark 'Y' forward from db815d6d32e6
   $ cp -R ../cloned-bookmarks-update ../cloned-bookmarks-manual-update
   $ cp -R ../cloned-bookmarks-update ../cloned-bookmarks-manual-update-with-divergence
 
 (manual version)
 
   $ hg -R ../cloned-bookmarks-manual-update update Y
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (activating bookmark Y)
   $ hg -R ../cloned-bookmarks-manual-update pull .
   pulling from .
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   updating bookmark Y
   updating bookmark Z
   added 2 changesets with 2 changes to 2 files (+1 heads)
   new changesets 125c9a1d6df6:9ba5f110a0b3
   (run 'hg heads' to see heads, 'hg merge' to merge)
 
 (# tests strange but with --date crashing when bookmark have to move)
 
   $ hg -R ../cloned-bookmarks-manual-update update -d 1986
   abort: revision matching date not found
   [10]
   $ hg -R ../cloned-bookmarks-manual-update update
   updating to active bookmark Y
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
 (all in one version)
 
   $ hg -R ../cloned-bookmarks-update update Y
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (activating bookmark Y)
   $ hg -R ../cloned-bookmarks-update pull --update .
   pulling from .
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   updating bookmark Y
   updating bookmark Z
   added 2 changesets with 2 changes to 2 files (+1 heads)
   new changesets 125c9a1d6df6:9ba5f110a0b3
   updating to active bookmark Y
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 
 We warn about divergent during bare update to the active bookmark
 
   $ hg -R ../cloned-bookmarks-manual-update-with-divergence update Y
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   (activating bookmark Y)
   $ hg -R ../cloned-bookmarks-manual-update-with-divergence bookmarks -r X2 Y@1
   $ hg -R ../cloned-bookmarks-manual-update-with-divergence bookmarks
      X2                        1:925d80f479bb
    * Y                         2:db815d6d32e6
      Y@1                       1:925d80f479bb
      Z                         2:db815d6d32e6
      x  y                      2:db815d6d32e6
   $ hg -R ../cloned-bookmarks-manual-update-with-divergence pull
   pulling from $TESTTMP/repo
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   updating bookmark Y
   updating bookmark Z
   added 2 changesets with 2 changes to 2 files (+1 heads)
   new changesets 125c9a1d6df6:9ba5f110a0b3
   (run 'hg heads' to see heads, 'hg merge' to merge)
   $ hg -R ../cloned-bookmarks-manual-update-with-divergence update
   updating to active bookmark Y
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   1 other divergent bookmarks for "Y"
 
 test wrongly formated bookmark
 
   $ echo '' >> .hg/bookmarks
   $ hg bookmarks
      X2                        1:925d80f479bb
      Y                         3:125c9a1d6df6
    * Z                         3:125c9a1d6df6
      x  y                      2:db815d6d32e6
   $ echo "Ican'thasformatedlines" >> .hg/bookmarks
   $ hg bookmarks
   malformed line in .hg/bookmarks: "Ican'thasformatedlines"
      X2                        1:925d80f479bb
      Y                         3:125c9a1d6df6
    * Z                         3:125c9a1d6df6
      x  y                      2:db815d6d32e6
 
 test missing revisions
 
   $ echo "925d80f479b925d80f479bc925d80f479bccabab z" > .hg/bookmarks
   $ hg book
   no bookmarks set
 
 test stripping a non-checked-out but bookmarked revision
 
   $ hg log --graph
   o  changeset:   4:9ba5f110a0b3
   |  branch:      test
   |  tag:         tip
   |  parent:      2:db815d6d32e6
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     y
   |
   | @  changeset:   3:125c9a1d6df6
   |/   user:        test
   |    date:        Thu Jan 01 00:00:00 1970 +0000
   |    summary:     x
   |
   o  changeset:   2:db815d6d32e6
   |  parent:      0:f7b1eb17ad24
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     2
   |
   | o  changeset:   1:925d80f479bb
   |/   user:        test
   |    date:        Thu Jan 01 00:00:00 1970 +0000
   |    summary:     1
   |
   o  changeset:   0:f7b1eb17ad24
      user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     0
   
   $ hg book should-end-on-two
   $ hg co --clean 4
   1 files updated, 0 files merged, 1 files removed, 0 files unresolved
   (leaving bookmark should-end-on-two)
   $ hg book four
   $ hg --config extensions.mq= strip 3
   saved backup bundle to * (glob)
 should-end-on-two should end up pointing to revision 2, as that's the
 tipmost surviving ancestor of the stripped revision.
   $ hg log --graph
   @  changeset:   3:9ba5f110a0b3
   |  branch:      test
   |  bookmark:    four
   |  tag:         tip
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     y
   |
   o  changeset:   2:db815d6d32e6
   |  bookmark:    should-end-on-two
   |  parent:      0:f7b1eb17ad24
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     2
   |
   | o  changeset:   1:925d80f479bb
   |/   user:        test
   |    date:        Thu Jan 01 00:00:00 1970 +0000
   |    summary:     1
   |
   o  changeset:   0:f7b1eb17ad24
      user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     0
   
 
 no-op update doesn't deactivate bookmarks
 
 (it is known issue that fsmonitor can't handle nested repositories. In
 this test scenario, cloned-bookmark-default and tobundle exist in the
 working directory of current repository)
 
   $ hg bookmarks
    * four                      3:9ba5f110a0b3
      should-end-on-two         2:db815d6d32e6
   $ hg up four
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg up
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg sum
   parent: 3:9ba5f110a0b3 tip
    y
   branch: test
   bookmarks: *four
   commit: 2 unknown (clean) (no-fsmonitor !)
   commit: * unknown (clean) (glob) (fsmonitor !)
   update: (current)
   phases: 4 draft
 
 test clearing divergent bookmarks of linear ancestors
 
   $ hg bookmark Z -r 0
   $ hg bookmark Z@1 -r 1
   $ hg bookmark Z@2 -r 2
   $ hg bookmark Z@3 -r 3
   $ hg book
      Z                         0:f7b1eb17ad24
      Z@1                       1:925d80f479bb
      Z@2                       2:db815d6d32e6
      Z@3                       3:9ba5f110a0b3
    * four                      3:9ba5f110a0b3
      should-end-on-two         2:db815d6d32e6
   $ hg bookmark Z
   moving bookmark 'Z' forward from f7b1eb17ad24
   $ hg book
    * Z                         3:9ba5f110a0b3
      Z@1                       1:925d80f479bb
      four                      3:9ba5f110a0b3
      should-end-on-two         2:db815d6d32e6
 
 test clearing only a single divergent bookmark across branches
 
   $ hg book foo -r 1
   $ hg book foo@1 -r 0
   $ hg book foo@2 -r 2
   $ hg book foo@3 -r 3
   $ hg book foo -r foo@3
   $ hg book
    * Z                         3:9ba5f110a0b3
      Z@1                       1:925d80f479bb
      foo                       3:9ba5f110a0b3
      foo@1                     0:f7b1eb17ad24
      foo@2                     2:db815d6d32e6
      four                      3:9ba5f110a0b3
      should-end-on-two         2:db815d6d32e6
 
 pull --update works the same as pull && update (case #2)
 
 It is assumed that "hg pull" itself doesn't update current active
 bookmark ('Y' in tests below).
 
   $ hg pull -q ../cloned-bookmarks-update
   divergent bookmark Z stored as Z@2
 
 (pulling revision on another named branch with --update updates
 neither the working directory nor current active bookmark: "no-op"
 case)
 
   $ echo yy >> y
   $ hg commit -m yy
 
   $ hg -R ../cloned-bookmarks-update bookmarks | grep ' Y '
    * Y                         3:125c9a1d6df6
   $ hg -R ../cloned-bookmarks-update path
   default = $TESTTMP/repo
   $ pwd
   $TESTTMP/repo
   $ hg -R ../cloned-bookmarks-update pull . --update
   pulling from .
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   divergent bookmark Z stored as Z@default
   adding remote bookmark foo
   adding remote bookmark four
   adding remote bookmark should-end-on-two
   added 1 changesets with 1 changes to 1 files
   new changesets 5fb12f0f2d51
   0 files updated, 0 files merged, 0 files removed, 0 files unresolved
   $ hg -R ../cloned-bookmarks-update parents -T "{rev}:{node|short}\n"
   3:125c9a1d6df6
   $ hg -R ../cloned-bookmarks-update bookmarks | grep ' Y '
    * Y                         3:125c9a1d6df6
 
 (pulling revision on current named/topological branch with --update
 updates the working directory and current active bookmark)
 
   $ hg update -C -q 125c9a1d6df6
   $ echo xx >> x
   $ hg commit -m xx
 
   $ hg -R ../cloned-bookmarks-update bookmarks | grep ' Y '
    * Y                         3:125c9a1d6df6
   $ hg -R ../cloned-bookmarks-update pull . --update
   pulling from .
   searching for changes
   adding changesets
   adding manifests
   adding file changes
   divergent bookmark Z stored as Z@default
   added 1 changesets with 1 changes to 1 files
   new changesets 81dcce76aa0b
   1 files updated, 0 files merged, 0 files removed, 0 files unresolved
   updating bookmark Y
   $ hg -R ../cloned-bookmarks-update parents -T "{rev}:{node|short}\n"
   6:81dcce76aa0b
   $ hg -R ../cloned-bookmarks-update bookmarks | grep ' Y '
    * Y                         6:81dcce76aa0b
 
   $ cd ..
 
 ensure changelog is written before bookmarks
   $ hg init orderrepo
   $ cd orderrepo
   $ touch a
   $ hg commit -Aqm one
   $ hg book mybook
   $ echo a > a
 
   $ cat > $TESTTMP/pausefinalize.py <<EOF
   > import os
   > import time
   > from mercurial import extensions, localrepo
   > def transaction(orig, self, desc, report=None):
   >    tr = orig(self, desc, report)
   >    def sleep(*args, **kwargs):
   >        retry = 20
   >        while retry > 0 and not os.path.exists(b"$TESTTMP/unpause"):
   >            retry -= 1
   >            time.sleep(0.5)
   >        if os.path.exists(b"$TESTTMP/unpause"):
   >            os.remove(b"$TESTTMP/unpause")
-  >    # It is important that this finalizer start with 'a', so it runs before
-  >    # the changelog finalizer appends to the changelog.
-  >    tr.addfinalize(b'a-sleep', sleep)
+  >    # It is important that this finalizer start with '000-a', so it runs
+  >    # before the changelog finalizer appends to the changelog.
+  >    tr.addfinalize(b'000-a-sleep', sleep)
   >    return tr
   > 
   > def extsetup(ui):
   >    # This extension inserts an artifical pause during the transaction
   >    # finalizer, so we can run commands mid-transaction-close.
   >    extensions.wrapfunction(localrepo.localrepository, 'transaction',
   >                            transaction)
   > EOF
   $ hg commit -qm two --config extensions.pausefinalize=$TESTTMP/pausefinalize.py &
   $ sleep 2
   $ hg log -r .
   changeset:   0:867bc5792c8c
   bookmark:    mybook
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     one
   
   $ hg bookmarks
    * mybook                    0:867bc5792c8c
   $ touch $TESTTMP/unpause
 
   $ cd ..
 
 check whether HG_PENDING makes pending changes only in related
 repositories visible to an external hook.
 
 (emulate a transaction running concurrently by copied
 .hg/bookmarks.pending in subsequent test)
 
   $ cat > $TESTTMP/savepending.sh <<EOF
   > cp .hg/bookmarks.pending .hg/bookmarks.pending.saved
   > exit 1 # to avoid adding new bookmark for subsequent tests
   > EOF
 
   $ hg init unrelated
   $ cd unrelated
   $ echo a > a
   $ hg add a
   $ hg commit -m '#0'
   $ hg --config hooks.pretxnclose="sh $TESTTMP/savepending.sh" bookmarks INVISIBLE
   abort: pretxnclose hook exited with status 1
   [40]
   $ cp .hg/bookmarks.pending.saved .hg/bookmarks.pending
 
 (check visible bookmarks while transaction running in repo)
 
   $ cat > $TESTTMP/checkpending.sh <<EOF
   > echo "@repo"
   > hg -R "$TESTTMP/repo" bookmarks
   > echo "@unrelated"
   > hg -R "$TESTTMP/unrelated" bookmarks
   > exit 1 # to avoid adding new bookmark for subsequent tests
   > EOF
 
   $ cd ../repo
   $ hg --config hooks.pretxnclose="sh $TESTTMP/checkpending.sh" bookmarks NEW
   @repo
    * NEW                       6:81dcce76aa0b
      X2                        1:925d80f479bb
      Y                         4:125c9a1d6df6
      Z                         5:5fb12f0f2d51
      Z@1                       1:925d80f479bb
      Z@2                       4:125c9a1d6df6
      foo                       3:9ba5f110a0b3
      foo@1                     0:f7b1eb17ad24
      foo@2                     2:db815d6d32e6
      four                      3:9ba5f110a0b3
      should-end-on-two         2:db815d6d32e6
      x  y                      2:db815d6d32e6
   @unrelated
   no bookmarks set
   abort: pretxnclose hook exited with status 1
   [40]
 
 Check pretxnclose-bookmark can abort a transaction
 --------------------------------------------------
 
 add hooks:
 
 * to prevent NEW bookmark on a non-public changeset
 * to prevent non-forward move of NEW bookmark
 
   $ cat << EOF >> .hg/hgrc
   > [hooks]
   > pretxnclose-bookmark.force-public  = sh -c "(echo \$HG_BOOKMARK| grep -v NEW > /dev/null) || [ -z \"\$HG_NODE\" ] || (hg log -r \"\$HG_NODE\" -T '{phase}' | grep public > /dev/null)"
   > pretxnclose-bookmark.force-forward = sh -c "(echo \$HG_BOOKMARK| grep -v NEW > /dev/null) || [ -z \"\$HG_NODE\" ] || (hg log -r \"max(\$HG_OLDNODE::\$HG_NODE)\" -T 'MATCH' | grep MATCH > /dev/null)"
   > EOF
 
   $ hg log -G -T phases
   @  changeset:   6:81dcce76aa0b
   |  tag:         tip
   |  phase:       draft
   |  parent:      4:125c9a1d6df6
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     xx
   |
   | o  changeset:   5:5fb12f0f2d51
   | |  branch:      test
   | |  bookmark:    Z
   | |  phase:       draft
   | |  parent:      3:9ba5f110a0b3
   | |  user:        test
   | |  date:        Thu Jan 01 00:00:00 1970 +0000
   | |  summary:     yy
   | |
   o |  changeset:   4:125c9a1d6df6
   | |  bookmark:    Y
   | |  bookmark:    Z@2
   | |  phase:       public
   | |  parent:      2:db815d6d32e6
   | |  user:        test
   | |  date:        Thu Jan 01 00:00:00 1970 +0000
   | |  summary:     x
   | |
   | o  changeset:   3:9ba5f110a0b3
   |/   branch:      test
   |    bookmark:    foo
   |    bookmark:    four
   |    phase:       public
   |    user:        test
   |    date:        Thu Jan 01 00:00:00 1970 +0000
   |    summary:     y
   |
   o  changeset:   2:db815d6d32e6
   |  bookmark:    foo@2
   |  bookmark:    should-end-on-two
   |  bookmark:    x  y
   |  phase:       public
   |  parent:      0:f7b1eb17ad24
   |  user:        test
   |  date:        Thu Jan 01 00:00:00 1970 +0000
   |  summary:     2
   |
   | o  changeset:   1:925d80f479bb
   |/   bookmark:    X2
   |    bookmark:    Z@1
   |    phase:       public
   |    user:        test
   |    date:        Thu Jan 01 00:00:00 1970 +0000
   |    summary:     1
   |
   o  changeset:   0:f7b1eb17ad24
      bookmark:    foo@1
      phase:       public
      user:        test
      date:        Thu Jan 01 00:00:00 1970 +0000
      summary:     0
   
 
 attempt to create on a default changeset
 
   $ hg bookmark -r 81dcce76aa0b NEW
   abort: pretxnclose-bookmark.force-public hook exited with status 1
   [40]
 
 create on a public changeset
 
   $ hg bookmark -r 9ba5f110a0b3 NEW
 
 move to the other branch
 
   $ hg bookmark -f -r 125c9a1d6df6 NEW
   abort: pretxnclose-bookmark.force-forward hook exited with status 1
   [40]