dirstate: remove a update_file's special case for tracked file with p2 data

This case was fishy and can be dealt with by passing more accurate data a higher
level.

This clarify the API and prepare for a larger rework of the data we feeds to
the dirstate.

mercurial/dirstatemap.py

@@ -1,926 +1,917 b''
 # dirstatemap.py
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import errno
 
 from .i18n import _
 
 from . import (
     error,
     pathutil,
     policy,
     pycompat,
     txnutil,
     util,
 )
 
 from .dirstateutils import (
     docket as docketmod,
 )
 
 parsers = policy.importmod('parsers')
 rustmod = policy.importrust('dirstate')
 
 propertycache = util.propertycache
 
 if rustmod is None:
     DirstateItem = parsers.DirstateItem
 else:
     DirstateItem = rustmod.DirstateItem
 
 rangemask = 0x7FFFFFFF
 
 
 class dirstatemap(object):
     """Map encapsulating the dirstate's contents.
 
     The dirstate contains the following state:
 
     - `identity` is the identity of the dirstate file, which can be used to
       detect when changes have occurred to the dirstate file.
 
     - `parents` is a pair containing the parents of the working copy. The
       parents are updated by calling `setparents`.
 
     - the state map maps filenames to tuples of (state, mode, size, mtime),
       where state is a single character representing 'normal', 'added',
       'removed', or 'merged'. It is read by treating the dirstate as a
       dict.  File state is updated by calling various methods (see each
       documentation for details):
 
       - `reset_state`,
       - `set_tracked`
       - `set_untracked`
       - `set_clean`
       - `set_possibly_dirty`
 
     - `copymap` maps destination filenames to their source filename.
 
     The dirstate also provides the following views onto the state:
 
     - `filefoldmap` is a dict mapping normalized filenames to the denormalized
       form that they appear as in the dirstate.
 
     - `dirfoldmap` is a dict mapping normalized directory names to the
       denormalized form that they appear as in the dirstate.
     """
 
     def __init__(self, ui, opener, root, nodeconstants, use_dirstate_v2):
         self._ui = ui
         self._opener = opener
         self._root = root
         self._filename = b'dirstate'
         self._nodelen = 20
         self._nodeconstants = nodeconstants
         assert (
             not use_dirstate_v2
         ), "should have detected unsupported requirement"
 
         self._parents = None
         self._dirtyparents = False
 
         # for consistent view between _pl() and _read() invocations
         self._pendingmode = None
 
     @propertycache
     def _map(self):
         self._map = {}
         self.read()
         return self._map
 
     @propertycache
     def copymap(self):
         self.copymap = {}
         self._map
         return self.copymap
 
     def clear(self):
         self._map.clear()
         self.copymap.clear()
         self.setparents(self._nodeconstants.nullid, self._nodeconstants.nullid)
         util.clearcachedproperty(self, b"_dirs")
         util.clearcachedproperty(self, b"_alldirs")
         util.clearcachedproperty(self, b"filefoldmap")
         util.clearcachedproperty(self, b"dirfoldmap")
 
     def items(self):
         return pycompat.iteritems(self._map)
 
     # forward for python2,3 compat
     iteritems = items
 
     def debug_iter(self, all):
         """
         Return an iterator of (filename, state, mode, size, mtime) tuples
 
         `all` is unused when Rust is not enabled
         """
         for (filename, item) in self.items():
             yield (filename, item.state, item.mode, item.size, item.mtime)
 
     def __len__(self):
         return len(self._map)
 
     def __iter__(self):
         return iter(self._map)
 
     def get(self, key, default=None):
         return self._map.get(key, default)
 
     def __contains__(self, key):
         return key in self._map
 
     def __getitem__(self, key):
         return self._map[key]
 
     def keys(self):
         return self._map.keys()
 
     def preload(self):
         """Loads the underlying data, if it's not already loaded"""
         self._map
 
     def _dirs_incr(self, filename, old_entry=None):
         """incremente the dirstate counter if applicable"""
         if (
             old_entry is None or old_entry.removed
         ) and "_dirs" in self.__dict__:
             self._dirs.addpath(filename)
         if old_entry is None and "_alldirs" in self.__dict__:
             self._alldirs.addpath(filename)
 
     def _dirs_decr(self, filename, old_entry=None, remove_variant=False):
         """decremente the dirstate counter if applicable"""
         if old_entry is not None:
             if "_dirs" in self.__dict__ and not old_entry.removed:
                 self._dirs.delpath(filename)
             if "_alldirs" in self.__dict__ and not remove_variant:
                 self._alldirs.delpath(filename)
         elif remove_variant and "_alldirs" in self.__dict__:
             self._alldirs.addpath(filename)
         if "filefoldmap" in self.__dict__:
             normed = util.normcase(filename)
             self.filefoldmap.pop(normed, None)
 
     def set_possibly_dirty(self, filename):
         """record that the current state of the file on disk is unknown"""
         self[filename].set_possibly_dirty()
 
     def set_clean(self, filename, mode, size, mtime):
         """mark a file as back to a clean state"""
         entry = self[filename]
         mtime = mtime & rangemask
         size = size & rangemask
         entry.set_clean(mode, size, mtime)
         self.copymap.pop(filename, None)
 
     def reset_state(
         self,
         filename,
         wc_tracked=False,
         p1_tracked=False,
         p2_tracked=False,
         merged=False,
         clean_p1=False,
         clean_p2=False,
         possibly_dirty=False,
         parentfiledata=None,
     ):
         """Set a entry to a given state, diregarding all previous state
 
         This is to be used by the part of the dirstate API dedicated to
         adjusting the dirstate after a update/merge.
 
         note: calling this might result to no entry existing at all if the
         dirstate map does not see any point at having one for this file
         anymore.
         """
         if merged and (clean_p1 or clean_p2):
             msg = b'`merged` argument incompatible with `clean_p1`/`clean_p2`'
             raise error.ProgrammingError(msg)
         # copy information are now outdated
         # (maybe new information should be in directly passed to this function)
         self.copymap.pop(filename, None)
 
         if not (p1_tracked or p2_tracked or wc_tracked):
             old_entry = self._map.pop(filename, None)
             self._dirs_decr(filename, old_entry=old_entry)
             self.copymap.pop(filename, None)
             return
         elif merged:
             # XXX might be merged and removed ?
             entry = self.get(filename)
             if entry is None or not entry.tracked:
                 # XXX mostly replicate dirstate.other parent.  We should get
                 # the higher layer to pass us more reliable data where `merged`
                 # actually mean merged. Dropping this clause will show failure
                 # in `test-graft.t`
                 merged = False
                 clean_p2 = True
         elif not (p1_tracked or p2_tracked) and wc_tracked:
             pass  # file is added, nothing special to adjust
         elif (p1_tracked or p2_tracked) and not wc_tracked:
             pass
         elif clean_p2 and wc_tracked:
-            if p1_tracked or self.get(filename) is not None:
-                # XXX the `self.get` call is catching some case in
-                # `test-merge-remove.t` where the file is tracked in p1, the
-                # p1_tracked argument is False.
-                #
-                # In addition, this seems to be a case where the file is marked
-                # as merged without actually being the result of a merge
-                # action. So thing are not ideal here.
-                merged = True
-                clean_p2 = False
+            pass
         elif not p1_tracked and p2_tracked and wc_tracked:
             clean_p2 = True
         elif possibly_dirty:
             pass
         elif wc_tracked:
             # this is a "normal" file
             if parentfiledata is None:
                 msg = b'failed to pass parentfiledata for a normal file: %s'
                 msg %= filename
                 raise error.ProgrammingError(msg)
         else:
             assert False, 'unreachable'
 
         old_entry = self._map.get(filename)
         self._dirs_incr(filename, old_entry)
         entry = DirstateItem(
             wc_tracked=wc_tracked,
             p1_tracked=p1_tracked,
             p2_tracked=p2_tracked,
             merged=merged,
             clean_p1=clean_p1,
             clean_p2=clean_p2,
             possibly_dirty=possibly_dirty,
             parentfiledata=parentfiledata,
         )
         self._map[filename] = entry
 
     def set_tracked(self, filename):
         new = False
         entry = self.get(filename)
         if entry is None:
             self._dirs_incr(filename)
             entry = DirstateItem(
                 p1_tracked=False,
                 p2_tracked=False,
                 wc_tracked=True,
                 merged=False,
                 clean_p1=False,
                 clean_p2=False,
                 possibly_dirty=False,
                 parentfiledata=None,
             )
             self._map[filename] = entry
             new = True
         elif not entry.tracked:
             self._dirs_incr(filename, entry)
             entry.set_tracked()
             new = True
         else:
             # XXX This is probably overkill for more case, but we need this to
             # fully replace the `normallookup` call with `set_tracked` one.
             # Consider smoothing this in the future.
             self.set_possibly_dirty(filename)
         return new
 
     def set_untracked(self, f):
         """Mark a file as no longer tracked in the dirstate map"""
         entry = self.get(f)
         if entry is None:
             return False
         else:
             self._dirs_decr(f, old_entry=entry, remove_variant=not entry.added)
             if not entry.merged:
                 self.copymap.pop(f, None)
             if entry.added:
                 self._map.pop(f, None)
             else:
                 entry.set_untracked()
             return True
 
     @propertycache
     def filefoldmap(self):
         """Returns a dictionary mapping normalized case paths to their
         non-normalized versions.
         """
         try:
             makefilefoldmap = parsers.make_file_foldmap
         except AttributeError:
             pass
         else:
             return makefilefoldmap(
                 self._map, util.normcasespec, util.normcasefallback
             )
 
         f = {}
         normcase = util.normcase
         for name, s in pycompat.iteritems(self._map):
             if not s.removed:
                 f[normcase(name)] = name
         f[b'.'] = b'.'  # prevents useless util.fspath() invocation
         return f
 
     def hastrackeddir(self, d):
         """
         Returns True if the dirstate contains a tracked (not removed) file
         in this directory.
         """
         return d in self._dirs
 
     def hasdir(self, d):
         """
         Returns True if the dirstate contains a file (tracked or removed)
         in this directory.
         """
         return d in self._alldirs
 
     @propertycache
     def _dirs(self):
         return pathutil.dirs(self._map, only_tracked=True)
 
     @propertycache
     def _alldirs(self):
         return pathutil.dirs(self._map)
 
     def _opendirstatefile(self):
         fp, mode = txnutil.trypending(self._root, self._opener, self._filename)
         if self._pendingmode is not None and self._pendingmode != mode:
             fp.close()
             raise error.Abort(
                 _(b'working directory state may be changed parallelly')
             )
         self._pendingmode = mode
         return fp
 
     def parents(self):
         if not self._parents:
             try:
                 fp = self._opendirstatefile()
                 st = fp.read(2 * self._nodelen)
                 fp.close()
             except IOError as err:
                 if err.errno != errno.ENOENT:
                     raise
                 # File doesn't exist, so the current state is empty
                 st = b''
 
             l = len(st)
             if l == self._nodelen * 2:
                 self._parents = (
                     st[: self._nodelen],
                     st[self._nodelen : 2 * self._nodelen],
                 )
             elif l == 0:
                 self._parents = (
                     self._nodeconstants.nullid,
                     self._nodeconstants.nullid,
                 )
             else:
                 raise error.Abort(
                     _(b'working directory state appears damaged!')
                 )
 
         return self._parents
 
     def setparents(self, p1, p2, fold_p2=False):
         self._parents = (p1, p2)
         self._dirtyparents = True
         copies = {}
         if fold_p2:
             for f, s in pycompat.iteritems(self._map):
                 # Discard "merged" markers when moving away from a merge state
                 if s.merged or s.from_p2:
                     source = self.copymap.pop(f, None)
                     if source:
                         copies[f] = source
                     s.drop_merge_data()
         return copies
 
     def read(self):
         # ignore HG_PENDING because identity is used only for writing
         self.identity = util.filestat.frompath(
             self._opener.join(self._filename)
         )
 
         try:
             fp = self._opendirstatefile()
             try:
                 st = fp.read()
             finally:
                 fp.close()
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
             return
         if not st:
             return
 
         if util.safehasattr(parsers, b'dict_new_presized'):
             # Make an estimate of the number of files in the dirstate based on
             # its size. This trades wasting some memory for avoiding costly
             # resizes. Each entry have a prefix of 17 bytes followed by one or
             # two path names. Studies on various large-scale real-world repositories
             # found 54 bytes a reasonable upper limit for the average path names.
             # Copy entries are ignored for the sake of this estimate.
             self._map = parsers.dict_new_presized(len(st) // 71)
 
         # Python's garbage collector triggers a GC each time a certain number
         # of container objects (the number being defined by
         # gc.get_threshold()) are allocated. parse_dirstate creates a tuple
         # for each file in the dirstate. The C version then immediately marks
         # them as not to be tracked by the collector. However, this has no
         # effect on when GCs are triggered, only on what objects the GC looks
         # into. This means that O(number of files) GCs are unavoidable.
         # Depending on when in the process's lifetime the dirstate is parsed,
         # this can get very expensive. As a workaround, disable GC while
         # parsing the dirstate.
         #
         # (we cannot decorate the function directly since it is in a C module)
         parse_dirstate = util.nogc(parsers.parse_dirstate)
         p = parse_dirstate(self._map, self.copymap, st)
         if not self._dirtyparents:
             self.setparents(*p)
 
         # Avoid excess attribute lookups by fast pathing certain checks
         self.__contains__ = self._map.__contains__
         self.__getitem__ = self._map.__getitem__
         self.get = self._map.get
 
     def write(self, _tr, st, now):
         st.write(
             parsers.pack_dirstate(self._map, self.copymap, self.parents(), now)
         )
         st.close()
         self._dirtyparents = False
 
     @propertycache
     def identity(self):
         self._map
         return self.identity
 
     @propertycache
     def dirfoldmap(self):
         f = {}
         normcase = util.normcase
         for name in self._dirs:
             f[normcase(name)] = name
         return f
 
 
 if rustmod is not None:
 
     class dirstatemap(object):
         def __init__(self, ui, opener, root, nodeconstants, use_dirstate_v2):
             self._use_dirstate_v2 = use_dirstate_v2
             self._nodeconstants = nodeconstants
             self._ui = ui
             self._opener = opener
             self._root = root
             self._filename = b'dirstate'
             self._nodelen = 20  # Also update Rust code when changing this!
             self._parents = None
             self._dirtyparents = False
             self._docket = None
 
             # for consistent view between _pl() and _read() invocations
             self._pendingmode = None
 
         def addfile(
             self,
             f,
             mode=0,
             size=None,
             mtime=None,
             added=False,
             merged=False,
             from_p2=False,
             possibly_dirty=False,
         ):
             if added:
                 assert not possibly_dirty
                 assert not from_p2
                 item = DirstateItem.new_added()
             elif merged:
                 assert not possibly_dirty
                 assert not from_p2
                 item = DirstateItem.new_merged()
             elif from_p2:
                 assert not possibly_dirty
                 item = DirstateItem.new_from_p2()
             elif possibly_dirty:
                 item = DirstateItem.new_possibly_dirty()
             else:
                 assert size is not None
                 assert mtime is not None
                 size = size & rangemask
                 mtime = mtime & rangemask
                 item = DirstateItem.new_normal(mode, size, mtime)
             self._rustmap.addfile(f, item)
             if added:
                 self.copymap.pop(f, None)
 
         def reset_state(
             self,
             filename,
             wc_tracked=False,
             p1_tracked=False,
             p2_tracked=False,
             merged=False,
             clean_p1=False,
             clean_p2=False,
             possibly_dirty=False,
             parentfiledata=None,
         ):
             """Set a entry to a given state, disregarding all previous state
 
             This is to be used by the part of the dirstate API dedicated to
             adjusting the dirstate after a update/merge.
 
             note: calling this might result to no entry existing at all if the
             dirstate map does not see any point at having one for this file
             anymore.
             """
             if merged and (clean_p1 or clean_p2):
                 msg = (
                     b'`merged` argument incompatible with `clean_p1`/`clean_p2`'
                 )
                 raise error.ProgrammingError(msg)
             # copy information are now outdated
             # (maybe new information should be in directly passed to this function)
             self.copymap.pop(filename, None)
 
             if not (p1_tracked or p2_tracked or wc_tracked):
                 self._rustmap.drop_item_and_copy_source(filename)
             elif merged:
                 # XXX might be merged and removed ?
                 entry = self.get(filename)
                 if entry is not None and entry.tracked:
                     # XXX mostly replicate dirstate.other parent.  We should get
                     # the higher layer to pass us more reliable data where `merged`
                     # actually mean merged. Dropping the else clause will show
                     # failure in `test-graft.t`
                     self.addfile(filename, merged=True)
                 else:
                     self.addfile(filename, from_p2=True)
             elif not (p1_tracked or p2_tracked) and wc_tracked:
                 self.addfile(
                     filename, added=True, possibly_dirty=possibly_dirty
                 )
             elif (p1_tracked or p2_tracked) and not wc_tracked:
                 # XXX might be merged and removed ?
                 self[filename] = DirstateItem.from_v1_data(b'r', 0, 0, 0)
             elif clean_p2 and wc_tracked:
                 if p1_tracked or self.get(filename) is not None:
                     # XXX the `self.get` call is catching some case in
                     # `test-merge-remove.t` where the file is tracked in p1, the
                     # p1_tracked argument is False.
                     #
                     # In addition, this seems to be a case where the file is marked
                     # as merged without actually being the result of a merge
                     # action. So thing are not ideal here.
                     self.addfile(filename, merged=True)
                 else:
                     self.addfile(filename, from_p2=True)
             elif not p1_tracked and p2_tracked and wc_tracked:
                 self.addfile(
                     filename, from_p2=True, possibly_dirty=possibly_dirty
                 )
             elif possibly_dirty:
                 self.addfile(filename, possibly_dirty=possibly_dirty)
             elif wc_tracked:
                 # this is a "normal" file
                 if parentfiledata is None:
                     msg = b'failed to pass parentfiledata for a normal file: %s'
                     msg %= filename
                     raise error.ProgrammingError(msg)
                 mode, size, mtime = parentfiledata
                 self.addfile(filename, mode=mode, size=size, mtime=mtime)
             else:
                 assert False, 'unreachable'
 
         def set_tracked(self, filename):
             new = False
             entry = self.get(filename)
             if entry is None:
                 self.addfile(filename, added=True)
                 new = True
             elif not entry.tracked:
                 entry.set_tracked()
                 self._rustmap.set_dirstate_item(filename, entry)
                 new = True
             else:
                 # XXX This is probably overkill for more case, but we need this to
                 # fully replace the `normallookup` call with `set_tracked` one.
                 # Consider smoothing this in the future.
                 self.set_possibly_dirty(filename)
             return new
 
         def set_untracked(self, f):
             """Mark a file as no longer tracked in the dirstate map"""
             # in merge is only trigger more logic, so it "fine" to pass it.
             #
             # the inner rust dirstate map code need to be adjusted once the API
             # for dirstate/dirstatemap/DirstateItem is a bit more settled
             entry = self.get(f)
             if entry is None:
                 return False
             else:
                 if entry.added:
                     self._rustmap.drop_item_and_copy_source(f)
                 else:
                     self._rustmap.removefile(f, in_merge=True)
                 return True
 
         def removefile(self, *args, **kwargs):
             return self._rustmap.removefile(*args, **kwargs)
 
         def get(self, *args, **kwargs):
             return self._rustmap.get(*args, **kwargs)
 
         @property
         def copymap(self):
             return self._rustmap.copymap()
 
         def debug_iter(self, all):
             """
             Return an iterator of (filename, state, mode, size, mtime) tuples
 
             `all`: also include with `state == b' '` dirstate tree nodes that
             don't have an associated `DirstateItem`.
 
             """
             return self._rustmap.debug_iter(all)
 
         def preload(self):
             self._rustmap
 
         def clear(self):
             self._rustmap.clear()
             self.setparents(
                 self._nodeconstants.nullid, self._nodeconstants.nullid
             )
             util.clearcachedproperty(self, b"_dirs")
             util.clearcachedproperty(self, b"_alldirs")
             util.clearcachedproperty(self, b"dirfoldmap")
 
         def items(self):
             return self._rustmap.items()
 
         def keys(self):
             return iter(self._rustmap)
 
         def __contains__(self, key):
             return key in self._rustmap
 
         def __getitem__(self, item):
             return self._rustmap[item]
 
         def __len__(self):
             return len(self._rustmap)
 
         def __iter__(self):
             return iter(self._rustmap)
 
         # forward for python2,3 compat
         iteritems = items
 
         def _opendirstatefile(self):
             fp, mode = txnutil.trypending(
                 self._root, self._opener, self._filename
             )
             if self._pendingmode is not None and self._pendingmode != mode:
                 fp.close()
                 raise error.Abort(
                     _(b'working directory state may be changed parallelly')
                 )
             self._pendingmode = mode
             return fp
 
         def _readdirstatefile(self, size=-1):
             try:
                 with self._opendirstatefile() as fp:
                     return fp.read(size)
             except IOError as err:
                 if err.errno != errno.ENOENT:
                     raise
                 # File doesn't exist, so the current state is empty
                 return b''
 
         def setparents(self, p1, p2, fold_p2=False):
             self._parents = (p1, p2)
             self._dirtyparents = True
             copies = {}
             if fold_p2:
                 # Collect into an intermediate list to avoid a `RuntimeError`
                 # exception due to mutation during iteration.
                 # TODO: move this the whole loop to Rust where `iter_mut`
                 # enables in-place mutation of elements of a collection while
                 # iterating it, without mutating the collection itself.
                 candidatefiles = [
                     (f, s)
                     for f, s in self._rustmap.items()
                     if s.merged or s.from_p2
                 ]
                 for f, s in candidatefiles:
                     # Discard "merged" markers when moving away from a merge state
                     if s.merged:
                         source = self.copymap.get(f)
                         if source:
                             copies[f] = source
                         self.reset_state(
                             f,
                             wc_tracked=True,
                             p1_tracked=True,
                             possibly_dirty=True,
                         )
                     # Also fix up otherparent markers
                     elif s.from_p2:
                         source = self.copymap.get(f)
                         if source:
                             copies[f] = source
                         self.reset_state(
                             f,
                             p1_tracked=False,
                             wc_tracked=True,
                         )
             return copies
 
         def parents(self):
             if not self._parents:
                 if self._use_dirstate_v2:
                     self._parents = self.docket.parents
                 else:
                     read_len = self._nodelen * 2
                     st = self._readdirstatefile(read_len)
                     l = len(st)
                     if l == read_len:
                         self._parents = (
                             st[: self._nodelen],
                             st[self._nodelen : 2 * self._nodelen],
                         )
                     elif l == 0:
                         self._parents = (
                             self._nodeconstants.nullid,
                             self._nodeconstants.nullid,
                         )
                     else:
                         raise error.Abort(
                             _(b'working directory state appears damaged!')
                         )
 
             return self._parents
 
         @property
         def docket(self):
             if not self._docket:
                 if not self._use_dirstate_v2:
                     raise error.ProgrammingError(
                         b'dirstate only has a docket in v2 format'
                     )
                 self._docket = docketmod.DirstateDocket.parse(
                     self._readdirstatefile(), self._nodeconstants
                 )
             return self._docket
 
         @propertycache
         def _rustmap(self):
             """
             Fills the Dirstatemap when called.
             """
             # ignore HG_PENDING because identity is used only for writing
             self.identity = util.filestat.frompath(
                 self._opener.join(self._filename)
             )
 
             if self._use_dirstate_v2:
                 if self.docket.uuid:
                     # TODO: use mmap when possible
                     data = self._opener.read(self.docket.data_filename())
                 else:
                     data = b''
                 self._rustmap = rustmod.DirstateMap.new_v2(
                     data, self.docket.data_size, self.docket.tree_metadata
                 )
                 parents = self.docket.parents
             else:
                 self._rustmap, parents = rustmod.DirstateMap.new_v1(
                     self._readdirstatefile()
                 )
 
             if parents and not self._dirtyparents:
                 self.setparents(*parents)
 
             self.__contains__ = self._rustmap.__contains__
             self.__getitem__ = self._rustmap.__getitem__
             self.get = self._rustmap.get
             return self._rustmap
 
         def write(self, tr, st, now):
             if not self._use_dirstate_v2:
                 p1, p2 = self.parents()
                 packed = self._rustmap.write_v1(p1, p2, now)
                 st.write(packed)
                 st.close()
                 self._dirtyparents = False
                 return
 
             # We can only append to an existing data file if there is one
             can_append = self.docket.uuid is not None
             packed, meta, append = self._rustmap.write_v2(now, can_append)
             if append:
                 docket = self.docket
                 data_filename = docket.data_filename()
                 if tr:
                     tr.add(data_filename, docket.data_size)
                 with self._opener(data_filename, b'r+b') as fp:
                     fp.seek(docket.data_size)
                     assert fp.tell() == docket.data_size
                     written = fp.write(packed)
                     if written is not None:  # py2 may return None
                         assert written == len(packed), (written, len(packed))
                 docket.data_size += len(packed)
                 docket.parents = self.parents()
                 docket.tree_metadata = meta
                 st.write(docket.serialize())
                 st.close()
             else:
                 old_docket = self.docket
                 new_docket = docketmod.DirstateDocket.with_new_uuid(
                     self.parents(), len(packed), meta
                 )
                 data_filename = new_docket.data_filename()
                 if tr:
                     tr.add(data_filename, 0)
                 self._opener.write(data_filename, packed)
                 # Write the new docket after the new data file has been
                 # written. Because `st` was opened with `atomictemp=True`,
                 # the actual `.hg/dirstate` file is only affected on close.
                 st.write(new_docket.serialize())
                 st.close()
                 # Remove the old data file after the new docket pointing to
                 # the new data file was written.
                 if old_docket.uuid:
                     data_filename = old_docket.data_filename()
                     unlink = lambda _tr=None: self._opener.unlink(data_filename)
                     if tr:
                         category = b"dirstate-v2-clean-" + old_docket.uuid
                         tr.addpostclose(category, unlink)
                     else:
                         unlink()
                 self._docket = new_docket
             # Reload from the newly-written file
             util.clearcachedproperty(self, b"_rustmap")
             self._dirtyparents = False
 
         @propertycache
         def filefoldmap(self):
             """Returns a dictionary mapping normalized case paths to their
             non-normalized versions.
             """
             return self._rustmap.filefoldmapasdict()
 
         def hastrackeddir(self, d):
             return self._rustmap.hastrackeddir(d)
 
         def hasdir(self, d):
             return self._rustmap.hasdir(d)
 
         @propertycache
         def identity(self):
             self._rustmap
             return self.identity
 
         @propertycache
         def dirfoldmap(self):
             f = {}
             normcase = util.normcase
             for name in self._rustmap.tracked_dirs():
                 f[normcase(name)] = name
             return f
 
         def set_possibly_dirty(self, filename):
             """record that the current state of the file on disk is unknown"""
             entry = self[filename]
             entry.set_possibly_dirty()
             self._rustmap.set_dirstate_item(filename, entry)
 
         def set_clean(self, filename, mode, size, mtime):
             """mark a file as back to a clean state"""
             entry = self[filename]
             mtime = mtime & rangemask
             size = size & rangemask
             entry.set_clean(mode, size, mtime)
             self._rustmap.set_dirstate_item(filename, entry)
             self._rustmap.copymap().pop(filename, None)
 
         def __setitem__(self, key, value):
             assert isinstance(value, DirstateItem)
             self._rustmap.set_dirstate_item(key, value)

mercurial/mergestate.py

@@ -1,866 +1,869 b''
 from __future__ import absolute_import
 
 import collections
 import errno
 import shutil
 import struct
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullrev,
 )
 from . import (
     error,
     filemerge,
     pycompat,
     util,
 )
 from .utils import hashutil
 
 _pack = struct.pack
 _unpack = struct.unpack
 
 
 def _droponode(data):
     # used for compatibility for v1
     bits = data.split(b'\0')
     bits = bits[:-2] + bits[-1:]
     return b'\0'.join(bits)
 
 
 def _filectxorabsent(hexnode, ctx, f):
     if hexnode == ctx.repo().nodeconstants.nullhex:
         return filemerge.absentfilectx(ctx, f)
     else:
         return ctx[f]
 
 
 # Merge state record types. See ``mergestate`` docs for more.
 
 ####
 # merge records which records metadata about a current merge
 # exists only once in a mergestate
 #####
 RECORD_LOCAL = b'L'
 RECORD_OTHER = b'O'
 # record merge labels
 RECORD_LABELS = b'l'
 
 #####
 # record extra information about files, with one entry containing info about one
 # file. Hence, multiple of them can exists
 #####
 RECORD_FILE_VALUES = b'f'
 
 #####
 # merge records which represents state of individual merges of files/folders
 # These are top level records for each entry containing merge related info.
 # Each record of these has info about one file. Hence multiple of them can
 # exists
 #####
 RECORD_MERGED = b'F'
 RECORD_CHANGEDELETE_CONFLICT = b'C'
 # the path was dir on one side of merge and file on another
 RECORD_PATH_CONFLICT = b'P'
 
 #####
 # possible state which a merge entry can have. These are stored inside top-level
 # merge records mentioned just above.
 #####
 MERGE_RECORD_UNRESOLVED = b'u'
 MERGE_RECORD_RESOLVED = b'r'
 MERGE_RECORD_UNRESOLVED_PATH = b'pu'
 MERGE_RECORD_RESOLVED_PATH = b'pr'
 # represents that the file was automatically merged in favor
 # of other version. This info is used on commit.
 # This is now deprecated and commit related information is now
 # stored in RECORD_FILE_VALUES
 MERGE_RECORD_MERGED_OTHER = b'o'
 
 #####
 # top level record which stores other unknown records. Multiple of these can
 # exists
 #####
 RECORD_OVERRIDE = b't'
 
 #####
 # legacy records which are no longer used but kept to prevent breaking BC
 #####
 # This record was release in 5.4 and usage was removed in 5.5
 LEGACY_RECORD_RESOLVED_OTHER = b'R'
 # This record was release in 3.7 and usage was removed in 5.6
 LEGACY_RECORD_DRIVER_RESOLVED = b'd'
 # This record was release in 3.7 and usage was removed in 5.6
 LEGACY_MERGE_DRIVER_STATE = b'm'
 # This record was release in 3.7 and usage was removed in 5.6
 LEGACY_MERGE_DRIVER_MERGE = b'D'
 
 
 ACTION_FORGET = b'f'
 ACTION_REMOVE = b'r'
 ACTION_ADD = b'a'
 ACTION_GET = b'g'
 ACTION_PATH_CONFLICT = b'p'
 ACTION_PATH_CONFLICT_RESOLVE = b'pr'
 ACTION_ADD_MODIFIED = b'am'
 ACTION_CREATED = b'c'
 ACTION_DELETED_CHANGED = b'dc'
 ACTION_CHANGED_DELETED = b'cd'
 ACTION_MERGE = b'm'
 ACTION_LOCAL_DIR_RENAME_GET = b'dg'
 ACTION_DIR_RENAME_MOVE_LOCAL = b'dm'
 ACTION_KEEP = b'k'
 # the file was absent on local side before merge and we should
 # keep it absent (absent means file not present, it can be a result
 # of file deletion, rename etc.)
 ACTION_KEEP_ABSENT = b'ka'
 # the file is absent on the ancestor and remote side of the merge
 # hence this file is new and we should keep it
 ACTION_KEEP_NEW = b'kn'
 ACTION_EXEC = b'e'
 ACTION_CREATED_MERGE = b'cm'
 
 # actions which are no op
 NO_OP_ACTIONS = (
     ACTION_KEEP,
     ACTION_KEEP_ABSENT,
     ACTION_KEEP_NEW,
 )
 
 
 class _mergestate_base(object):
     """track 3-way merge state of individual files
 
     The merge state is stored on disk when needed. Two files are used: one with
     an old format (version 1), and one with a new format (version 2). Version 2
     stores a superset of the data in version 1, including new kinds of records
     in the future. For more about the new format, see the documentation for
     `_readrecordsv2`.
 
     Each record can contain arbitrary content, and has an associated type. This
     `type` should be a letter. If `type` is uppercase, the record is mandatory:
     versions of Mercurial that don't support it should abort. If `type` is
     lowercase, the record can be safely ignored.
 
     Currently known records:
 
     L: the node of the "local" part of the merge (hexified version)
     O: the node of the "other" part of the merge (hexified version)
     F: a file to be merged entry
     C: a change/delete or delete/change conflict
     P: a path conflict (file vs directory)
     f: a (filename, dictionary) tuple of optional values for a given file
     l: the labels for the parts of the merge.
 
     Merge record states (stored in self._state, indexed by filename):
     u: unresolved conflict
     r: resolved conflict
     pu: unresolved path conflict (file conflicts with directory)
     pr: resolved path conflict
     o: file was merged in favor of other parent of merge (DEPRECATED)
 
     The resolve command transitions between 'u' and 'r' for conflicts and
     'pu' and 'pr' for path conflicts.
     """
 
     def __init__(self, repo):
         """Initialize the merge state.
 
         Do not use this directly! Instead call read() or clean()."""
         self._repo = repo
         self._state = {}
         self._stateextras = collections.defaultdict(dict)
         self._local = None
         self._other = None
         self._labels = None
         # contains a mapping of form:
         # {filename : (merge_return_value, action_to_be_performed}
         # these are results of re-running merge process
         # this dict is used to perform actions on dirstate caused by re-running
         # the merge
         self._results = {}
         self._dirty = False
 
     def reset(self):
         pass
 
     def start(self, node, other, labels=None):
         self._local = node
         self._other = other
         self._labels = labels
 
     @util.propertycache
     def local(self):
         if self._local is None:
             msg = b"local accessed but self._local isn't set"
             raise error.ProgrammingError(msg)
         return self._local
 
     @util.propertycache
     def localctx(self):
         return self._repo[self.local]
 
     @util.propertycache
     def other(self):
         if self._other is None:
             msg = b"other accessed but self._other isn't set"
             raise error.ProgrammingError(msg)
         return self._other
 
     @util.propertycache
     def otherctx(self):
         return self._repo[self.other]
 
     def active(self):
         """Whether mergestate is active.
 
         Returns True if there appears to be mergestate. This is a rough proxy
         for "is a merge in progress."
         """
         return bool(self._local) or bool(self._state)
 
     def commit(self):
         """Write current state on disk (if necessary)"""
 
     @staticmethod
     def getlocalkey(path):
         """hash the path of a local file context for storage in the .hg/merge
         directory."""
 
         return hex(hashutil.sha1(path).digest())
 
     def _make_backup(self, fctx, localkey):
         raise NotImplementedError()
 
     def _restore_backup(self, fctx, localkey, flags):
         raise NotImplementedError()
 
     def add(self, fcl, fco, fca, fd):
         """add a new (potentially?) conflicting file the merge state
         fcl: file context for local,
         fco: file context for remote,
         fca: file context for ancestors,
         fd:  file path of the resulting merge.
 
         note: also write the local version to the `.hg/merge` directory.
         """
         if fcl.isabsent():
             localkey = self._repo.nodeconstants.nullhex
         else:
             localkey = mergestate.getlocalkey(fcl.path())
             self._make_backup(fcl, localkey)
         self._state[fd] = [
             MERGE_RECORD_UNRESOLVED,
             localkey,
             fcl.path(),
             fca.path(),
             hex(fca.filenode()),
             fco.path(),
             hex(fco.filenode()),
             fcl.flags(),
         ]
         self._stateextras[fd][b'ancestorlinknode'] = hex(fca.node())
         self._dirty = True
 
     def addpathconflict(self, path, frename, forigin):
         """add a new conflicting path to the merge state
         path:    the path that conflicts
         frename: the filename the conflicting file was renamed to
         forigin: origin of the file ('l' or 'r' for local/remote)
         """
         self._state[path] = [MERGE_RECORD_UNRESOLVED_PATH, frename, forigin]
         self._dirty = True
 
     def addcommitinfo(self, path, data):
         """stores information which is required at commit
         into _stateextras"""
         self._stateextras[path].update(data)
         self._dirty = True
 
     def __contains__(self, dfile):
         return dfile in self._state
 
     def __getitem__(self, dfile):
         return self._state[dfile][0]
 
     def __iter__(self):
         return iter(sorted(self._state))
 
     def files(self):
         return self._state.keys()
 
     def mark(self, dfile, state):
         self._state[dfile][0] = state
         self._dirty = True
 
     def unresolved(self):
         """Obtain the paths of unresolved files."""
 
         for f, entry in pycompat.iteritems(self._state):
             if entry[0] in (
                 MERGE_RECORD_UNRESOLVED,
                 MERGE_RECORD_UNRESOLVED_PATH,
             ):
                 yield f
 
     def allextras(self):
         """return all extras information stored with the mergestate"""
         return self._stateextras
 
     def extras(self, filename):
         """return extras stored with the mergestate for the given filename"""
         return self._stateextras[filename]
 
     def _resolve(self, preresolve, dfile, wctx):
         """rerun merge process for file path `dfile`.
         Returns whether the merge was completed and the return value of merge
         obtained from filemerge._filemerge().
         """
         if self[dfile] in (
             MERGE_RECORD_RESOLVED,
             LEGACY_RECORD_DRIVER_RESOLVED,
         ):
             return True, 0
         stateentry = self._state[dfile]
         state, localkey, lfile, afile, anode, ofile, onode, flags = stateentry
         octx = self._repo[self._other]
         extras = self.extras(dfile)
         anccommitnode = extras.get(b'ancestorlinknode')
         if anccommitnode:
             actx = self._repo[anccommitnode]
         else:
             actx = None
         fcd = _filectxorabsent(localkey, wctx, dfile)
         fco = _filectxorabsent(onode, octx, ofile)
         # TODO: move this to filectxorabsent
         fca = self._repo.filectx(afile, fileid=anode, changectx=actx)
         # "premerge" x flags
         flo = fco.flags()
         fla = fca.flags()
         if b'x' in flags + flo + fla and b'l' not in flags + flo + fla:
             if fca.rev() == nullrev and flags != flo:
                 if preresolve:
                     self._repo.ui.warn(
                         _(
                             b'warning: cannot merge flags for %s '
                             b'without common ancestor - keeping local flags\n'
                         )
                         % afile
                     )
             elif flags == fla:
                 flags = flo
         if preresolve:
             # restore local
             if localkey != self._repo.nodeconstants.nullhex:
                 self._restore_backup(wctx[dfile], localkey, flags)
             else:
                 wctx[dfile].remove(ignoremissing=True)
             complete, merge_ret, deleted = filemerge.premerge(
                 self._repo,
                 wctx,
                 self._local,
                 lfile,
                 fcd,
                 fco,
                 fca,
                 labels=self._labels,
             )
         else:
             complete, merge_ret, deleted = filemerge.filemerge(
                 self._repo,
                 wctx,
                 self._local,
                 lfile,
                 fcd,
                 fco,
                 fca,
                 labels=self._labels,
             )
         if merge_ret is None:
             # If return value of merge is None, then there are no real conflict
             del self._state[dfile]
             self._dirty = True
         elif not merge_ret:
             self.mark(dfile, MERGE_RECORD_RESOLVED)
 
         if complete:
             action = None
             if deleted:
                 if fcd.isabsent():
                     # dc: local picked. Need to drop if present, which may
                     # happen on re-resolves.
                     action = ACTION_FORGET
                 else:
                     # cd: remote picked (or otherwise deleted)
                     action = ACTION_REMOVE
             else:
                 if fcd.isabsent():  # dc: remote picked
                     action = ACTION_GET
                 elif fco.isabsent():  # cd: local picked
                     if dfile in self.localctx:
                         action = ACTION_ADD_MODIFIED
                     else:
                         action = ACTION_ADD
                 # else: regular merges (no action necessary)
             self._results[dfile] = merge_ret, action
 
         return complete, merge_ret
 
     def preresolve(self, dfile, wctx):
         """run premerge process for dfile
 
         Returns whether the merge is complete, and the exit code."""
         return self._resolve(True, dfile, wctx)
 
     def resolve(self, dfile, wctx):
         """run merge process (assuming premerge was run) for dfile
 
         Returns the exit code of the merge."""
         return self._resolve(False, dfile, wctx)[1]
 
     def counts(self):
         """return counts for updated, merged and removed files in this
         session"""
         updated, merged, removed = 0, 0, 0
         for r, action in pycompat.itervalues(self._results):
             if r is None:
                 updated += 1
             elif r == 0:
                 if action == ACTION_REMOVE:
                     removed += 1
                 else:
                     merged += 1
         return updated, merged, removed
 
     def unresolvedcount(self):
         """get unresolved count for this merge (persistent)"""
         return len(list(self.unresolved()))
 
     def actions(self):
         """return lists of actions to perform on the dirstate"""
         actions = {
             ACTION_REMOVE: [],
             ACTION_FORGET: [],
             ACTION_ADD: [],
             ACTION_ADD_MODIFIED: [],
             ACTION_GET: [],
         }
         for f, (r, action) in pycompat.iteritems(self._results):
             if action is not None:
                 actions[action].append((f, None, b"merge result"))
         return actions
 
 
 class mergestate(_mergestate_base):
 
     statepathv1 = b'merge/state'
     statepathv2 = b'merge/state2'
 
     @staticmethod
     def clean(repo):
         """Initialize a brand new merge state, removing any existing state on
         disk."""
         ms = mergestate(repo)
         ms.reset()
         return ms
 
     @staticmethod
     def read(repo):
         """Initialize the merge state, reading it from disk."""
         ms = mergestate(repo)
         ms._read()
         return ms
 
     def _read(self):
         """Analyse each record content to restore a serialized state from disk
 
         This function process "record" entry produced by the de-serialization
         of on disk file.
         """
         unsupported = set()
         records = self._readrecords()
         for rtype, record in records:
             if rtype == RECORD_LOCAL:
                 self._local = bin(record)
             elif rtype == RECORD_OTHER:
                 self._other = bin(record)
             elif rtype == LEGACY_MERGE_DRIVER_STATE:
                 pass
             elif rtype in (
                 RECORD_MERGED,
                 RECORD_CHANGEDELETE_CONFLICT,
                 RECORD_PATH_CONFLICT,
                 LEGACY_MERGE_DRIVER_MERGE,
                 LEGACY_RECORD_RESOLVED_OTHER,
             ):
                 bits = record.split(b'\0')
                 # merge entry type MERGE_RECORD_MERGED_OTHER is deprecated
                 # and we now store related information in _stateextras, so
                 # lets write to _stateextras directly
                 if bits[1] == MERGE_RECORD_MERGED_OTHER:
                     self._stateextras[bits[0]][b'filenode-source'] = b'other'
                 else:
                     self._state[bits[0]] = bits[1:]
             elif rtype == RECORD_FILE_VALUES:
                 filename, rawextras = record.split(b'\0', 1)
                 extraparts = rawextras.split(b'\0')
                 extras = {}
                 i = 0
                 while i < len(extraparts):
                     extras[extraparts[i]] = extraparts[i + 1]
                     i += 2
 
                 self._stateextras[filename] = extras
             elif rtype == RECORD_LABELS:
                 labels = record.split(b'\0', 2)
                 self._labels = [l for l in labels if len(l) > 0]
             elif not rtype.islower():
                 unsupported.add(rtype)
 
         if unsupported:
             raise error.UnsupportedMergeRecords(unsupported)
 
     def _readrecords(self):
         """Read merge state from disk and return a list of record (TYPE, data)
 
         We read data from both v1 and v2 files and decide which one to use.
 
         V1 has been used by version prior to 2.9.1 and contains less data than
         v2. We read both versions and check if no data in v2 contradicts
         v1. If there is not contradiction we can safely assume that both v1
         and v2 were written at the same time and use the extract data in v2. If
         there is contradiction we ignore v2 content as we assume an old version
         of Mercurial has overwritten the mergestate file and left an old v2
         file around.
 
         returns list of record [(TYPE, data), ...]"""
         v1records = self._readrecordsv1()
         v2records = self._readrecordsv2()
         if self._v1v2match(v1records, v2records):
             return v2records
         else:
             # v1 file is newer than v2 file, use it
             # we have to infer the "other" changeset of the merge
             # we cannot do better than that with v1 of the format
             mctx = self._repo[None].parents()[-1]
             v1records.append((RECORD_OTHER, mctx.hex()))
             # add place holder "other" file node information
             # nobody is using it yet so we do no need to fetch the data
             # if mctx was wrong `mctx[bits[-2]]` may fails.
             for idx, r in enumerate(v1records):
                 if r[0] == RECORD_MERGED:
                     bits = r[1].split(b'\0')
                     bits.insert(-2, b'')
                     v1records[idx] = (r[0], b'\0'.join(bits))
             return v1records
 
     def _v1v2match(self, v1records, v2records):
         oldv2 = set()  # old format version of v2 record
         for rec in v2records:
             if rec[0] == RECORD_LOCAL:
                 oldv2.add(rec)
             elif rec[0] == RECORD_MERGED:
                 # drop the onode data (not contained in v1)
                 oldv2.add((RECORD_MERGED, _droponode(rec[1])))
         for rec in v1records:
             if rec not in oldv2:
                 return False
         else:
             return True
 
     def _readrecordsv1(self):
         """read on disk merge state for version 1 file
 
         returns list of record [(TYPE, data), ...]
 
         Note: the "F" data from this file are one entry short
               (no "other file node" entry)
         """
         records = []
         try:
             f = self._repo.vfs(self.statepathv1)
             for i, l in enumerate(f):
                 if i == 0:
                     records.append((RECORD_LOCAL, l[:-1]))
                 else:
                     records.append((RECORD_MERGED, l[:-1]))
             f.close()
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
         return records
 
     def _readrecordsv2(self):
         """read on disk merge state for version 2 file
 
         This format is a list of arbitrary records of the form:
 
           [type][length][content]
 
         `type` is a single character, `length` is a 4 byte integer, and
         `content` is an arbitrary byte sequence of length `length`.
 
         Mercurial versions prior to 3.7 have a bug where if there are
         unsupported mandatory merge records, attempting to clear out the merge
         state with hg update --clean or similar aborts. The 't' record type
         works around that by writing out what those versions treat as an
         advisory record, but later versions interpret as special: the first
         character is the 'real' record type and everything onwards is the data.
 
         Returns list of records [(TYPE, data), ...]."""
         records = []
         try:
             f = self._repo.vfs(self.statepathv2)
             data = f.read()
             off = 0
             end = len(data)
             while off < end:
                 rtype = data[off : off + 1]
                 off += 1
                 length = _unpack(b'>I', data[off : (off + 4)])[0]
                 off += 4
                 record = data[off : (off + length)]
                 off += length
                 if rtype == RECORD_OVERRIDE:
                     rtype, record = record[0:1], record[1:]
                 records.append((rtype, record))
             f.close()
         except IOError as err:
             if err.errno != errno.ENOENT:
                 raise
         return records
 
     def commit(self):
         if self._dirty:
             records = self._makerecords()
             self._writerecords(records)
             self._dirty = False
 
     def _makerecords(self):
         records = []
         records.append((RECORD_LOCAL, hex(self._local)))
         records.append((RECORD_OTHER, hex(self._other)))
         # Write out state items. In all cases, the value of the state map entry
         # is written as the contents of the record. The record type depends on
         # the type of state that is stored, and capital-letter records are used
         # to prevent older versions of Mercurial that do not support the feature
         # from loading them.
         for filename, v in pycompat.iteritems(self._state):
             if v[0] in (
                 MERGE_RECORD_UNRESOLVED_PATH,
                 MERGE_RECORD_RESOLVED_PATH,
             ):
                 # Path conflicts. These are stored in 'P' records.  The current
                 # resolution state ('pu' or 'pr') is stored within the record.
                 records.append(
                     (RECORD_PATH_CONFLICT, b'\0'.join([filename] + v))
                 )
             elif (
                 v[1] == self._repo.nodeconstants.nullhex
                 or v[6] == self._repo.nodeconstants.nullhex
             ):
                 # Change/Delete or Delete/Change conflicts. These are stored in
                 # 'C' records. v[1] is the local file, and is nullhex when the
                 # file is deleted locally ('dc'). v[6] is the remote file, and
                 # is nullhex when the file is deleted remotely ('cd').
                 records.append(
                     (RECORD_CHANGEDELETE_CONFLICT, b'\0'.join([filename] + v))
                 )
             else:
                 # Normal files.  These are stored in 'F' records.
                 records.append((RECORD_MERGED, b'\0'.join([filename] + v)))
         for filename, extras in sorted(pycompat.iteritems(self._stateextras)):
             rawextras = b'\0'.join(
                 b'%s\0%s' % (k, v) for k, v in pycompat.iteritems(extras)
             )
             records.append(
                 (RECORD_FILE_VALUES, b'%s\0%s' % (filename, rawextras))
             )
         if self._labels is not None:
             labels = b'\0'.join(self._labels)
             records.append((RECORD_LABELS, labels))
         return records
 
     def _writerecords(self, records):
         """Write current state on disk (both v1 and v2)"""
         self._writerecordsv1(records)
         self._writerecordsv2(records)
 
     def _writerecordsv1(self, records):
         """Write current state on disk in a version 1 file"""
         f = self._repo.vfs(self.statepathv1, b'wb')
         irecords = iter(records)
         lrecords = next(irecords)
         assert lrecords[0] == RECORD_LOCAL
         f.write(hex(self._local) + b'\n')
         for rtype, data in irecords:
             if rtype == RECORD_MERGED:
                 f.write(b'%s\n' % _droponode(data))
         f.close()
 
     def _writerecordsv2(self, records):
         """Write current state on disk in a version 2 file
 
         See the docstring for _readrecordsv2 for why we use 't'."""
         # these are the records that all version 2 clients can read
         allowlist = (RECORD_LOCAL, RECORD_OTHER, RECORD_MERGED)
         f = self._repo.vfs(self.statepathv2, b'wb')
         for key, data in records:
             assert len(key) == 1
             if key not in allowlist:
                 key, data = RECORD_OVERRIDE, b'%s%s' % (key, data)
             format = b'>sI%is' % len(data)
             f.write(_pack(format, key, len(data), data))
         f.close()
 
     def _make_backup(self, fctx, localkey):
         self._repo.vfs.write(b'merge/' + localkey, fctx.data())
 
     def _restore_backup(self, fctx, localkey, flags):
         with self._repo.vfs(b'merge/' + localkey) as f:
             fctx.write(f.read(), flags)
 
     def reset(self):
         shutil.rmtree(self._repo.vfs.join(b'merge'), True)
 
 
 class memmergestate(_mergestate_base):
     def __init__(self, repo):
         super(memmergestate, self).__init__(repo)
         self._backups = {}
 
     def _make_backup(self, fctx, localkey):
         self._backups[localkey] = fctx.data()
 
     def _restore_backup(self, fctx, localkey, flags):
         fctx.write(self._backups[localkey], flags)
 
 
 def recordupdates(repo, actions, branchmerge, getfiledata):
     """record merge actions to the dirstate"""
     # remove (must come first)
     for f, args, msg in actions.get(ACTION_REMOVE, []):
         if branchmerge:
             repo.dirstate.update_file(f, p1_tracked=True, wc_tracked=False)
         else:
             repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=False)
 
     # forget (must come first)
     for f, args, msg in actions.get(ACTION_FORGET, []):
         repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=False)
 
     # resolve path conflicts
     for f, args, msg in actions.get(ACTION_PATH_CONFLICT_RESOLVE, []):
         (f0, origf0) = args
         repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=True)
         repo.dirstate.copy(origf0, f)
         if f0 == origf0:
             repo.dirstate.update_file(f0, p1_tracked=True, wc_tracked=False)
         else:
             repo.dirstate.update_file(f0, p1_tracked=False, wc_tracked=False)
 
     # re-add
     for f, args, msg in actions.get(ACTION_ADD, []):
         repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=True)
 
     # re-add/mark as modified
     for f, args, msg in actions.get(ACTION_ADD_MODIFIED, []):
         if branchmerge:
             repo.dirstate.update_file(
                 f, p1_tracked=True, wc_tracked=True, possibly_dirty=True
             )
         else:
             repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=True)
 
     # exec change
     for f, args, msg in actions.get(ACTION_EXEC, []):
         repo.dirstate.update_file(
             f, p1_tracked=True, wc_tracked=True, possibly_dirty=True
         )
 
     # keep
     for f, args, msg in actions.get(ACTION_KEEP, []):
         pass
 
     # keep deleted
     for f, args, msg in actions.get(ACTION_KEEP_ABSENT, []):
         pass
 
     # keep new
     for f, args, msg in actions.get(ACTION_KEEP_NEW, []):
         pass
 
     # get
     for f, args, msg in actions.get(ACTION_GET, []):
         if branchmerge:
             # tracked in p1 can be True also but update_file should not care
+            old_entry = repo.dirstate.get_entry(f)
+            p1_tracked = old_entry.any_tracked and not old_entry.added
             repo.dirstate.update_file(
                 f,
-                p1_tracked=False,
+                p1_tracked=p1_tracked,
                 p2_tracked=True,
                 wc_tracked=True,
-                clean_p2=True,
+                clean_p2=not p1_tracked,
+                merged=p1_tracked,
             )
         else:
             parentfiledata = getfiledata[f] if getfiledata else None
             repo.dirstate.update_file(
                 f,
                 p1_tracked=True,
                 wc_tracked=True,
                 parentfiledata=parentfiledata,
             )
 
     # merge
     for f, args, msg in actions.get(ACTION_MERGE, []):
         f1, f2, fa, move, anc = args
         if branchmerge:
             # We've done a branch merge, mark this file as merged
             # so that we properly record the merger later
             repo.dirstate.update_file(
                 f, p1_tracked=True, wc_tracked=True, merged=True
             )
             if f1 != f2:  # copy/rename
                 if move:
                     repo.dirstate.update_file(
                         f1, p1_tracked=True, wc_tracked=False
                     )
                 if f1 != f:
                     repo.dirstate.copy(f1, f)
                 else:
                     repo.dirstate.copy(f2, f)
         else:
             # We've update-merged a locally modified file, so
             # we set the dirstate to emulate a normal checkout
             # of that file some time in the past. Thus our
             # merge will appear as a normal local file
             # modification.
             if f2 == f:  # file not locally copied/moved
                 repo.dirstate.update_file(
                     f, p1_tracked=True, wc_tracked=True, possibly_dirty=True
                 )
             if move:
                 repo.dirstate.update_file(
                     f1, p1_tracked=False, wc_tracked=False
                 )
 
     # directory rename, move local
     for f, args, msg in actions.get(ACTION_DIR_RENAME_MOVE_LOCAL, []):
         f0, flag = args
         if branchmerge:
             repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=True)
             repo.dirstate.update_file(f0, p1_tracked=True, wc_tracked=False)
             repo.dirstate.copy(f0, f)
         else:
             repo.dirstate.update_file(f, p1_tracked=True, wc_tracked=True)
             repo.dirstate.update_file(f0, p1_tracked=False, wc_tracked=False)
 
     # directory rename, get
     for f, args, msg in actions.get(ACTION_LOCAL_DIR_RENAME_GET, []):
         f0, flag = args
         if branchmerge:
             repo.dirstate.update_file(f, p1_tracked=False, wc_tracked=True)
             repo.dirstate.copy(f0, f)
         else:
             repo.dirstate.update_file(f, p1_tracked=True, wc_tracked=True)