merge crew and main

mercurial/dirstate.py

@@ -1,800 +1,816 b''
 # dirstate.py - working directory tracking for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 import errno
 
 from node import nullid
 from i18n import _
 import scmutil, util, ignore, osutil, parsers, encoding
-import os, stat, errno
+import os, stat, errno, gc
 
 propertycache = util.propertycache
 filecache = scmutil.filecache
 _rangemask = 0x7fffffff
 
 class repocache(filecache):
     """filecache for files in .hg/"""
     def join(self, obj, fname):
         return obj._opener.join(fname)
 
 class rootcache(filecache):
     """filecache for files in the repository root"""
     def join(self, obj, fname):
         return obj._join(fname)
 
 def _finddirs(path):
     pos = path.rfind('/')
     while pos != -1:
         yield path[:pos]
         pos = path.rfind('/', 0, pos)
 
 def _incdirs(dirs, path):
     for base in _finddirs(path):
         if base in dirs:
             dirs[base] += 1
             return
         dirs[base] = 1
 
 def _decdirs(dirs, path):
     for base in _finddirs(path):
         if dirs[base] > 1:
             dirs[base] -= 1
             return
         del dirs[base]
 
 class dirstate(object):
 
     def __init__(self, opener, ui, root, validate):
         '''Create a new dirstate object.
 
         opener is an open()-like callable that can be used to open the
         dirstate file; root is the root of the directory tracked by
         the dirstate.
         '''
         self._opener = opener
         self._validate = validate
         self._root = root
         self._rootdir = os.path.join(root, '')
         self._dirty = False
         self._dirtypl = False
         self._lastnormaltime = 0
         self._ui = ui
         self._filecache = {}
 
     @propertycache
     def _map(self):
         '''Return the dirstate contents as a map from filename to
         (state, mode, size, time).'''
         self._read()
         return self._map
 
     @propertycache
     def _copymap(self):
         self._read()
         return self._copymap
 
     @propertycache
     def _foldmap(self):
         f = {}
         for name in self._map:
             f[util.normcase(name)] = name
         for name in self._dirs:
             f[util.normcase(name)] = name
         f['.'] = '.' # prevents useless util.fspath() invocation
         return f
 
     @repocache('branch')
     def _branch(self):
         try:
             return self._opener.read("branch").strip() or "default"
         except IOError, inst:
             if inst.errno != errno.ENOENT:
                 raise
             return "default"
 
     @propertycache
     def _pl(self):
         try:
             fp = self._opener("dirstate")
             st = fp.read(40)
             fp.close()
             l = len(st)
             if l == 40:
                 return st[:20], st[20:40]
             elif l > 0 and l < 40:
                 raise util.Abort(_('working directory state appears damaged!'))
         except IOError, err:
             if err.errno != errno.ENOENT:
                 raise
         return [nullid, nullid]
 
     @propertycache
     def _dirs(self):
         dirs = {}
         for f, s in self._map.iteritems():
             if s[0] != 'r':
                 _incdirs(dirs, f)
         return dirs
 
     def dirs(self):
         return self._dirs
 
     @rootcache('.hgignore')
     def _ignore(self):
         files = [self._join('.hgignore')]
         for name, path in self._ui.configitems("ui"):
             if name == 'ignore' or name.startswith('ignore.'):
                 files.append(util.expandpath(path))
         return ignore.ignore(self._root, files, self._ui.warn)
 
     @propertycache
     def _slash(self):
         return self._ui.configbool('ui', 'slash') and os.sep != '/'
 
     @propertycache
     def _checklink(self):
         return util.checklink(self._root)
 
     @propertycache
     def _checkexec(self):
         return util.checkexec(self._root)
 
     @propertycache
     def _checkcase(self):
         return not util.checkcase(self._join('.hg'))
 
     def _join(self, f):
         # much faster than os.path.join()
         # it's safe because f is always a relative path
         return self._rootdir + f
 
     def flagfunc(self, buildfallback):
         if self._checklink and self._checkexec:
             def f(x):
                 p = self._join(x)
                 if os.path.islink(p):
                     return 'l'
                 if util.isexec(p):
                     return 'x'
                 return ''
             return f
 
         fallback = buildfallback()
         if self._checklink:
             def f(x):
                 if os.path.islink(self._join(x)):
                     return 'l'
                 if 'x' in fallback(x):
                     return 'x'
                 return ''
             return f
         if self._checkexec:
             def f(x):
                 if 'l' in fallback(x):
                     return 'l'
                 if util.isexec(self._join(x)):
                     return 'x'
                 return ''
             return f
         else:
             return fallback
 
     def getcwd(self):
         cwd = os.getcwd()
         if cwd == self._root:
             return ''
         # self._root ends with a path separator if self._root is '/' or 'C:\'
         rootsep = self._root
         if not util.endswithsep(rootsep):
             rootsep += os.sep
         if cwd.startswith(rootsep):
             return cwd[len(rootsep):]
         else:
             # we're outside the repo. return an absolute path.
             return cwd
 
     def pathto(self, f, cwd=None):
         if cwd is None:
             cwd = self.getcwd()
         path = util.pathto(self._root, cwd, f)
         if self._slash:
             return util.normpath(path)
         return path
 
     def __getitem__(self, key):
         '''Return the current state of key (a filename) in the dirstate.
 
         States are:
           n  normal
           m  needs merging
           r  marked for removal
           a  marked for addition
           ?  not tracked
         '''
         return self._map.get(key, ("?",))[0]
 
     def __contains__(self, key):
         return key in self._map
 
     def __iter__(self):
         for x in sorted(self._map):
             yield x
 
     def parents(self):
         return [self._validate(p) for p in self._pl]
 
     def p1(self):
         return self._validate(self._pl[0])
 
     def p2(self):
         return self._validate(self._pl[1])
 
     def branch(self):
         return encoding.tolocal(self._branch)
 
     def setparents(self, p1, p2=nullid):
         """Set dirstate parents to p1 and p2.
 
         When moving from two parents to one, 'm' merged entries a
         adjusted to normal and previous copy records discarded and
         returned by the call.
 
         See localrepo.setparents()
         """
         self._dirty = self._dirtypl = True
         oldp2 = self._pl[1]
         self._pl = p1, p2
         copies = {}
         if oldp2 != nullid and p2 == nullid:
             # Discard 'm' markers when moving away from a merge state
             for f, s in self._map.iteritems():
                 if s[0] == 'm':
                     if f in self._copymap:
                         copies[f] = self._copymap[f]
                     self.normallookup(f)
         return copies
 
     def setbranch(self, branch):
         self._branch = encoding.fromlocal(branch)
         f = self._opener('branch', 'w', atomictemp=True)
         try:
             f.write(self._branch + '\n')
             f.close()
 
             # make sure filecache has the correct stat info for _branch after
             # replacing the underlying file
             ce = self._filecache['_branch']
             if ce:
                 ce.refresh()
         except: # re-raises
             f.discard()
             raise
 
     def _read(self):
         self._map = {}
         self._copymap = {}
         try:
             st = self._opener.read("dirstate")
         except IOError, err:
             if err.errno != errno.ENOENT:
                 raise
             return
         if not st:
             return
 
+        # 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.
+        gcenabled = gc.isenabled()
+        gc.disable()
+        try:
         p = parsers.parse_dirstate(self._map, self._copymap, st)
+        finally:
+            if gcenabled:
+                gc.enable()
         if not self._dirtypl:
             self._pl = p
 
     def invalidate(self):
         for a in ("_map", "_copymap", "_foldmap", "_branch", "_pl", "_dirs",
                 "_ignore"):
             if a in self.__dict__:
                 delattr(self, a)
         self._lastnormaltime = 0
         self._dirty = False
 
     def copy(self, source, dest):
         """Mark dest as a copy of source. Unmark dest if source is None."""
         if source == dest:
             return
         self._dirty = True
         if source is not None:
             self._copymap[dest] = source
         elif dest in self._copymap:
             del self._copymap[dest]
 
     def copied(self, file):
         return self._copymap.get(file, None)
 
     def copies(self):
         return self._copymap
 
     def _droppath(self, f):
         if self[f] not in "?r" and "_dirs" in self.__dict__:
             _decdirs(self._dirs, f)
 
     def _addpath(self, f, state, mode, size, mtime):
         oldstate = self[f]
         if state == 'a' or oldstate == 'r':
             scmutil.checkfilename(f)
             if f in self._dirs:
                 raise util.Abort(_('directory %r already in dirstate') % f)
             # shadows
             for d in _finddirs(f):
                 if d in self._dirs:
                     break
                 if d in self._map and self[d] != 'r':
                     raise util.Abort(
                         _('file %r in dirstate clashes with %r') % (d, f))
         if oldstate in "?r" and "_dirs" in self.__dict__:
             _incdirs(self._dirs, f)
         self._dirty = True
         self._map[f] = (state, mode, size, mtime)
 
     def normal(self, f):
         '''Mark a file normal and clean.'''
         s = os.lstat(self._join(f))
         mtime = int(s.st_mtime)
         self._addpath(f, 'n', s.st_mode,
                       s.st_size & _rangemask, mtime & _rangemask)
         if f in self._copymap:
             del self._copymap[f]
         if mtime > self._lastnormaltime:
             # Remember the most recent modification timeslot for status(),
             # to make sure we won't miss future size-preserving file content
             # modifications that happen within the same timeslot.
             self._lastnormaltime = mtime
 
     def normallookup(self, f):
         '''Mark a file normal, but possibly dirty.'''
         if self._pl[1] != nullid and f in self._map:
             # if there is a merge going on and the file was either
             # in state 'm' (-1) or coming from other parent (-2) before
             # being removed, restore that state.
             entry = self._map[f]
             if entry[0] == 'r' and entry[2] in (-1, -2):
                 source = self._copymap.get(f)
                 if entry[2] == -1:
                     self.merge(f)
                 elif entry[2] == -2:
                     self.otherparent(f)
                 if source:
                     self.copy(source, f)
                 return
             if entry[0] == 'm' or entry[0] == 'n' and entry[2] == -2:
                 return
         self._addpath(f, 'n', 0, -1, -1)
         if f in self._copymap:
             del self._copymap[f]
 
     def otherparent(self, f):
         '''Mark as coming from the other parent, always dirty.'''
         if self._pl[1] == nullid:
             raise util.Abort(_("setting %r to other parent "
                                "only allowed in merges") % f)
         self._addpath(f, 'n', 0, -2, -1)
         if f in self._copymap:
             del self._copymap[f]
 
     def add(self, f):
         '''Mark a file added.'''
         self._addpath(f, 'a', 0, -1, -1)
         if f in self._copymap:
             del self._copymap[f]
 
     def remove(self, f):
         '''Mark a file removed.'''
         self._dirty = True
         self._droppath(f)
         size = 0
         if self._pl[1] != nullid and f in self._map:
             # backup the previous state
             entry = self._map[f]
             if entry[0] == 'm': # merge
                 size = -1
             elif entry[0] == 'n' and entry[2] == -2: # other parent
                 size = -2
         self._map[f] = ('r', 0, size, 0)
         if size == 0 and f in self._copymap:
             del self._copymap[f]
 
     def merge(self, f):
         '''Mark a file merged.'''
         if self._pl[1] == nullid:
             return self.normallookup(f)
         s = os.lstat(self._join(f))
         self._addpath(f, 'm', s.st_mode,
                       s.st_size & _rangemask, int(s.st_mtime) & _rangemask)
         if f in self._copymap:
             del self._copymap[f]
 
     def drop(self, f):
         '''Drop a file from the dirstate'''
         if f in self._map:
             self._dirty = True
             self._droppath(f)
             del self._map[f]
 
     def _normalize(self, path, isknown, ignoremissing=False, exists=None):
         normed = util.normcase(path)
         folded = self._foldmap.get(normed, None)
         if folded is None:
             if isknown:
                 folded = path
             else:
                 if exists is None:
                     exists = os.path.lexists(os.path.join(self._root, path))
                 if not exists:
                     # Maybe a path component exists
                     if not ignoremissing and '/' in path:
                         d, f = path.rsplit('/', 1)
                         d = self._normalize(d, isknown, ignoremissing, None)
                         folded = d + "/" + f
                     else:
                         # No path components, preserve original case
                         folded = path
                 else:
                     # recursively normalize leading directory components
                     # against dirstate
                     if '/' in normed:
                         d, f = normed.rsplit('/', 1)
                         d = self._normalize(d, isknown, ignoremissing, True)
                         r = self._root + "/" + d
                         folded = d + "/" + util.fspath(f, r)
                     else:
                         folded = util.fspath(normed, self._root)
                     self._foldmap[normed] = folded
 
         return folded
 
     def normalize(self, path, isknown=False, ignoremissing=False):
         '''
         normalize the case of a pathname when on a casefolding filesystem
 
         isknown specifies whether the filename came from walking the
         disk, to avoid extra filesystem access.
 
         If ignoremissing is True, missing path are returned
         unchanged. Otherwise, we try harder to normalize possibly
         existing path components.
 
         The normalized case is determined based on the following precedence:
 
         - version of name already stored in the dirstate
         - version of name stored on disk
         - version provided via command arguments
         '''
 
         if self._checkcase:
             return self._normalize(path, isknown, ignoremissing)
         return path
 
     def clear(self):
         self._map = {}
         if "_dirs" in self.__dict__:
             delattr(self, "_dirs")
         self._copymap = {}
         self._pl = [nullid, nullid]
         self._lastnormaltime = 0
         self._dirty = True
 
     def rebuild(self, parent, files):
         self.clear()
         for f in files:
             if 'x' in files.flags(f):
                 self._map[f] = ('n', 0777, -1, 0)
             else:
                 self._map[f] = ('n', 0666, -1, 0)
         self._pl = (parent, nullid)
         self._dirty = True
 
     def write(self):
         if not self._dirty:
             return
         st = self._opener("dirstate", "w", atomictemp=True)
 
         def finish(s):
             st.write(s)
             st.close()
             self._lastnormaltime = 0
             self._dirty = self._dirtypl = False
 
         # use the modification time of the newly created temporary file as the
         # filesystem's notion of 'now'
         now = util.fstat(st).st_mtime
         finish(parsers.pack_dirstate(self._map, self._copymap, self._pl, now))
 
     def _dirignore(self, f):
         if f == '.':
             return False
         if self._ignore(f):
             return True
         for p in _finddirs(f):
             if self._ignore(p):
                 return True
         return False
 
     def walk(self, match, subrepos, unknown, ignored):
         '''
         Walk recursively through the directory tree, finding all files
         matched by match.
 
         Return a dict mapping filename to stat-like object (either
         mercurial.osutil.stat instance or return value of os.stat()).
         '''
 
         def fwarn(f, msg):
             self._ui.warn('%s: %s\n' % (self.pathto(f), msg))
             return False
 
         def badtype(mode):
             kind = _('unknown')
             if stat.S_ISCHR(mode):
                 kind = _('character device')
             elif stat.S_ISBLK(mode):
                 kind = _('block device')
             elif stat.S_ISFIFO(mode):
                 kind = _('fifo')
             elif stat.S_ISSOCK(mode):
                 kind = _('socket')
             elif stat.S_ISDIR(mode):
                 kind = _('directory')
             return _('unsupported file type (type is %s)') % kind
 
         ignore = self._ignore
         dirignore = self._dirignore
         if ignored:
             ignore = util.never
             dirignore = util.never
         elif not unknown:
             # if unknown and ignored are False, skip step 2
             ignore = util.always
             dirignore = util.always
 
         matchfn = match.matchfn
         badfn = match.bad
         dmap = self._map
         normpath = util.normpath
         listdir = osutil.listdir
         lstat = os.lstat
         getkind = stat.S_IFMT
         dirkind = stat.S_IFDIR
         regkind = stat.S_IFREG
         lnkkind = stat.S_IFLNK
         join = self._join
         work = []
         wadd = work.append
 
         exact = skipstep3 = False
         if matchfn == match.exact: # match.exact
             exact = True
             dirignore = util.always # skip step 2
         elif match.files() and not match.anypats(): # match.match, no patterns
             skipstep3 = True
 
         if not exact and self._checkcase:
             normalize = self._normalize
             skipstep3 = False
         else:
             normalize = None
 
         files = sorted(match.files())
         subrepos.sort()
         i, j = 0, 0
         while i < len(files) and j < len(subrepos):
             subpath = subrepos[j] + "/"
             if files[i] < subpath:
                 i += 1
                 continue
             while i < len(files) and files[i].startswith(subpath):
                 del files[i]
             j += 1
 
         if not files or '.' in files:
             files = ['']
         results = dict.fromkeys(subrepos)
         results['.hg'] = None
 
         # step 1: find all explicit files
         for ff in files:
             if normalize:
                 nf = normalize(normpath(ff), False, True)
             else:
                 nf = normpath(ff)
             if nf in results:
                 continue
 
             try:
                 st = lstat(join(nf))
                 kind = getkind(st.st_mode)
                 if kind == dirkind:
                     skipstep3 = False
                     if nf in dmap:
                         #file deleted on disk but still in dirstate
                         results[nf] = None
                     match.dir(nf)
                     if not dirignore(nf):
                         wadd(nf)
                 elif kind == regkind or kind == lnkkind:
                     results[nf] = st
                 else:
                     badfn(ff, badtype(kind))
                     if nf in dmap:
                         results[nf] = None
             except OSError, inst:
                 if nf in dmap: # does it exactly match a file?
                     results[nf] = None
                 else: # does it match a directory?
                     prefix = nf + "/"
                     for fn in dmap:
                         if fn.startswith(prefix):
                             match.dir(nf)
                             skipstep3 = False
                             break
                     else:
                         badfn(ff, inst.strerror)
 
         # step 2: visit subdirectories
         while work:
             nd = work.pop()
             skip = None
             if nd == '.':
                 nd = ''
             else:
                 skip = '.hg'
             try:
                 entries = listdir(join(nd), stat=True, skip=skip)
             except OSError, inst:
                 if inst.errno in (errno.EACCES, errno.ENOENT):
                     fwarn(nd, inst.strerror)
                     continue
                 raise
             for f, kind, st in entries:
                 if normalize:
                     nf = normalize(nd and (nd + "/" + f) or f, True, True)
                 else:
                     nf = nd and (nd + "/" + f) or f
                 if nf not in results:
                     if kind == dirkind:
                         if not ignore(nf):
                             match.dir(nf)
                             wadd(nf)
                         if nf in dmap and matchfn(nf):
                             results[nf] = None
                     elif kind == regkind or kind == lnkkind:
                         if nf in dmap:
                             if matchfn(nf):
                                 results[nf] = st
                         elif matchfn(nf) and not ignore(nf):
                             results[nf] = st
                     elif nf in dmap and matchfn(nf):
                         results[nf] = None
 
         # step 3: report unseen items in the dmap hash
         if not skipstep3 and not exact:
             visit = sorted([f for f in dmap if f not in results and matchfn(f)])
             if unknown:
                 # unknown == True means we walked the full directory tree above.
                 # So if a file is not seen it was either a) not matching matchfn
                 # b) ignored, c) missing, or d) under a symlink directory.
                 audit_path = scmutil.pathauditor(self._root)
 
                 for nf in iter(visit):
                     # Report ignored items in the dmap as long as they are not
                     # under a symlink directory.
                     if ignore(nf) and audit_path.check(nf):
                         results[nf] = util.statfiles([join(nf)])[0]
                     else:
                         # It's either missing or under a symlink directory
                         results[nf] = None
             else:
                 # We may not have walked the full directory tree above,
                 # so stat everything we missed.
                 nf = iter(visit).next
                 for st in util.statfiles([join(i) for i in visit]):
                     results[nf()] = st
         for s in subrepos:
             del results[s]
         del results['.hg']
         return results
 
     def status(self, match, subrepos, ignored, clean, unknown):
         '''Determine the status of the working copy relative to the
         dirstate and return a tuple of lists (unsure, modified, added,
         removed, deleted, unknown, ignored, clean), where:
 
           unsure:
             files that might have been modified since the dirstate was
             written, but need to be read to be sure (size is the same
             but mtime differs)
           modified:
             files that have definitely been modified since the dirstate
             was written (different size or mode)
           added:
             files that have been explicitly added with hg add
           removed:
             files that have been explicitly removed with hg remove
           deleted:
             files that have been deleted through other means ("missing")
           unknown:
             files not in the dirstate that are not ignored
           ignored:
             files not in the dirstate that are ignored
             (by _dirignore())
           clean:
             files that have definitely not been modified since the
             dirstate was written
         '''
         listignored, listclean, listunknown = ignored, clean, unknown
         lookup, modified, added, unknown, ignored = [], [], [], [], []
         removed, deleted, clean = [], [], []
 
         dmap = self._map
         ladd = lookup.append            # aka "unsure"
         madd = modified.append
         aadd = added.append
         uadd = unknown.append
         iadd = ignored.append
         radd = removed.append
         dadd = deleted.append
         cadd = clean.append
         mexact = match.exact
         dirignore = self._dirignore
         checkexec = self._checkexec
         checklink = self._checklink
         copymap = self._copymap
         lastnormaltime = self._lastnormaltime
 
         lnkkind = stat.S_IFLNK
 
         for fn, st in self.walk(match, subrepos, listunknown,
                                 listignored).iteritems():
             if fn not in dmap:
                 if (listignored or mexact(fn)) and dirignore(fn):
                     if listignored:
                         iadd(fn)
                 elif listunknown:
                     uadd(fn)
                 continue
 
             state, mode, size, time = dmap[fn]
 
             if not st and state in "nma":
                 dadd(fn)
             elif state == 'n':
                 # The "mode & lnkkind != lnkkind or self._checklink"
                 # lines are an expansion of "islink => checklink"
                 # where islink means "is this a link?" and checklink
                 # means "can we check links?".
                 mtime = int(st.st_mtime)
                 if (size >= 0 and
                     ((size != st.st_size and size != st.st_size & _rangemask)
                      or ((mode ^ st.st_mode) & 0100 and checkexec))
                     and (mode & lnkkind != lnkkind or checklink)
                     or size == -2 # other parent
                     or fn in copymap):
                     madd(fn)
                 elif ((time != mtime and time != mtime & _rangemask)
                       and (mode & lnkkind != lnkkind or checklink)):
                     ladd(fn)
                 elif mtime == lastnormaltime:
                     # fn may have been changed in the same timeslot without
                     # changing its size. This can happen if we quickly do
                     # multiple commits in a single transaction.
                     # Force lookup, so we don't miss such a racy file change.
                     ladd(fn)
                 elif listclean:
                     cadd(fn)
             elif state == 'm':
                 madd(fn)
             elif state == 'a':
                 aadd(fn)
             elif state == 'r':
                 radd(fn)
 
         return (lookup, modified, added, removed, deleted, unknown, ignored,
                 clean)

mercurial/merge.py

@@ -1,708 +1,716 b''
 # merge.py - directory-level update/merge handling for Mercurial
 #
 # Copyright 2006, 2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from node import nullid, nullrev, hex, bin
 from i18n import _
 import error, util, filemerge, copies, subrepo, worker
 import errno, os, shutil
 
 class mergestate(object):
     '''track 3-way merge state of individual files'''
     def __init__(self, repo):
         self._repo = repo
         self._dirty = False
         self._read()
     def reset(self, node=None):
         self._state = {}
         if node:
             self._local = node
         shutil.rmtree(self._repo.join("merge"), True)
         self._dirty = False
     def _read(self):
         self._state = {}
         try:
             f = self._repo.opener("merge/state")
             for i, l in enumerate(f):
                 if i == 0:
                     self._local = bin(l[:-1])
                 else:
                     bits = l[:-1].split("\0")
                     self._state[bits[0]] = bits[1:]
             f.close()
         except IOError, err:
             if err.errno != errno.ENOENT:
                 raise
         self._dirty = False
     def commit(self):
         if self._dirty:
             f = self._repo.opener("merge/state", "w")
             f.write(hex(self._local) + "\n")
             for d, v in self._state.iteritems():
                 f.write("\0".join([d] + v) + "\n")
             f.close()
             self._dirty = False
     def add(self, fcl, fco, fca, fd):
         hash = util.sha1(fcl.path()).hexdigest()
         self._repo.opener.write("merge/" + hash, fcl.data())
         self._state[fd] = ['u', hash, fcl.path(), fca.path(),
                            hex(fca.filenode()), fco.path(), fcl.flags()]
         self._dirty = True
     def __contains__(self, dfile):
         return dfile in self._state
     def __getitem__(self, dfile):
         return self._state[dfile][0]
     def __iter__(self):
         l = self._state.keys()
         l.sort()
         for f in l:
             yield f
     def mark(self, dfile, state):
         self._state[dfile][0] = state
         self._dirty = True
     def resolve(self, dfile, wctx, octx):
         if self[dfile] == 'r':
             return 0
         state, hash, lfile, afile, anode, ofile, flags = self._state[dfile]
         fcd = wctx[dfile]
         fco = octx[ofile]
         fca = self._repo.filectx(afile, fileid=anode)
         # "premerge" x flags
         flo = fco.flags()
         fla = fca.flags()
         if 'x' in flags + flo + fla and 'l' not in flags + flo + fla:
             if fca.node() == nullid:
                 self._repo.ui.warn(_('warning: cannot merge flags for %s\n') %
                                    afile)
             elif flags == fla:
                 flags = flo
         # restore local
         f = self._repo.opener("merge/" + hash)
         self._repo.wwrite(dfile, f.read(), flags)
         f.close()
         r = filemerge.filemerge(self._repo, self._local, lfile, fcd, fco, fca)
         if r is None:
             # no real conflict
             del self._state[dfile]
         elif not r:
             self.mark(dfile, 'r')
         return r
 
 def _checkunknownfile(repo, wctx, mctx, f):
     return (not repo.dirstate._ignore(f)
         and os.path.isfile(repo.wjoin(f))
         and repo.dirstate.normalize(f) not in repo.dirstate
         and mctx[f].cmp(wctx[f]))
 
 def _checkunknown(repo, wctx, mctx):
     "check for collisions between unknown files and files in mctx"
 
     error = False
     for f in mctx:
         if f not in wctx and _checkunknownfile(repo, wctx, mctx, f):
             error = True
             wctx._repo.ui.warn(_("%s: untracked file differs\n") % f)
     if error:
         raise util.Abort(_("untracked files in working directory differ "
                            "from files in requested revision"))
 
 def _remains(f, m, ma, workingctx=False):
     """check whether specified file remains after merge.
 
     It is assumed that specified file is not contained in the manifest
     of the other context.
     """
     if f in ma:
         n = m[f]
         if n != ma[f]:
             return True # because it is changed locally
             # even though it doesn't remain, if "remote deleted" is
             # chosen in manifestmerge()
         elif workingctx and n[20:] == "a":
             return True # because it is added locally (linear merge specific)
         else:
             return False # because it is removed remotely
     else:
         return True # because it is added locally
 
 def _checkcollision(mctx, extractxs):
     "check for case folding collisions in the destination context"
     folded = {}
     for fn in mctx:
         fold = util.normcase(fn)
         if fold in folded:
             raise util.Abort(_("case-folding collision between %s and %s")
                              % (fn, folded[fold]))
         folded[fold] = fn
 
     if extractxs:
         wctx, actx = extractxs
         # class to delay looking up copy mapping
         class pathcopies(object):
             @util.propertycache
             def map(self):
                 # {dst@mctx: src@wctx} copy mapping
                 return copies.pathcopies(wctx, mctx)
         pc = pathcopies()
 
         for fn in wctx:
             fold = util.normcase(fn)
             mfn = folded.get(fold, None)
             if (mfn and mfn != fn and pc.map.get(mfn) != fn and
                 _remains(fn, wctx.manifest(), actx.manifest(), True) and
                 _remains(mfn, mctx.manifest(), actx.manifest())):
                 raise util.Abort(_("case-folding collision between %s and %s")
                                  % (mfn, fn))
 
 def _forgetremoved(wctx, mctx, branchmerge):
     """
     Forget removed files
 
     If we're jumping between revisions (as opposed to merging), and if
     neither the working directory nor the target rev has the file,
     then we need to remove it from the dirstate, to prevent the
     dirstate from listing the file when it is no longer in the
     manifest.
 
     If we're merging, and the other revision has removed a file
     that is not present in the working directory, we need to mark it
     as removed.
     """
 
     actions = []
     state = branchmerge and 'r' or 'f'
     for f in wctx.deleted():
         if f not in mctx:
             actions.append((f, state, None, "forget deleted"))
 
     if not branchmerge:
         for f in wctx.removed():
             if f not in mctx:
                 actions.append((f, "f", None, "forget removed"))
 
     return actions
 
 def manifestmerge(repo, wctx, p2, pa, branchmerge, force, partial):
     """
     Merge p1 and p2 with ancestor pa and generate merge action list
 
     branchmerge and force are as passed in to update
     partial = function to filter file lists
     """
 
     overwrite = force and not branchmerge
     actions, copy, movewithdir = [], {}, {}
 
+    followcopies = False
     if overwrite:
         pa = wctx
     elif pa == p2: # backwards
         pa = wctx.p1()
     elif not branchmerge and not wctx.dirty(missing=True):
         pass
     elif pa and repo.ui.configbool("merge", "followcopies", True):
+        followcopies = True
+
+    # manifests fetched in order are going to be faster, so prime the caches
+    [x.manifest() for x in
+     sorted(wctx.parents() + [p2, pa], key=lambda x: x.rev())]
+
+    if followcopies:
         ret = copies.mergecopies(repo, wctx, p2, pa)
         copy, movewithdir, diverge, renamedelete = ret
         for of, fl in diverge.iteritems():
             actions.append((of, "dr", (fl,), "divergent renames"))
         for of, fl in renamedelete.iteritems():
             actions.append((of, "rd", (fl,), "rename and delete"))
 
     repo.ui.note(_("resolving manifests\n"))
     repo.ui.debug(" branchmerge: %s, force: %s, partial: %s\n"
                   % (bool(branchmerge), bool(force), bool(partial)))
     repo.ui.debug(" ancestor: %s, local: %s, remote: %s\n" % (pa, wctx, p2))
 
     m1, m2, ma = wctx.manifest(), p2.manifest(), pa.manifest()
     copied = set(copy.values())
     copied.update(movewithdir.values())
 
     if '.hgsubstate' in m1:
         # check whether sub state is modified
         for s in sorted(wctx.substate):
             if wctx.sub(s).dirty():
                 m1['.hgsubstate'] += "+"
                 break
 
     aborts, prompts = [], []
     # Compare manifests
     for f, n in m1.iteritems():
         if partial and not partial(f):
             continue
         if f in m2:
             n2 = m2[f]
             fl1, fl2, fla = m1.flags(f), m2.flags(f), ma.flags(f)
             nol = 'l' not in fl1 + fl2 + fla
             a = ma.get(f, nullid)
             if n == n2 and fl1 == fl2:
                 pass # same - keep local
             elif n2 == a and fl2 == fla:
                 pass # remote unchanged - keep local
             elif n == a and fl1 == fla: # local unchanged - use remote
                 if n == n2: # optimization: keep local content
                     actions.append((f, "e", (fl2,), "update permissions"))
                 else:
                     actions.append((f, "g", (fl2,), "remote is newer"))
             elif nol and n2 == a: # remote only changed 'x'
                 actions.append((f, "e", (fl2,), "update permissions"))
             elif nol and n == a: # local only changed 'x'
                 actions.append((f, "g", (fl1,), "remote is newer"))
             else: # both changed something
                 actions.append((f, "m", (f, f, False), "versions differ"))
         elif f in copied: # files we'll deal with on m2 side
             pass
         elif f in movewithdir: # directory rename
             f2 = movewithdir[f]
             actions.append((f, "d", (None, f2, m1.flags(f)),
                             "remote renamed directory to " + f2))
         elif f in copy:
             f2 = copy[f]
             actions.append((f, "m", (f2, f, False),
                             "local copied/moved to " + f2))
         elif f in ma: # clean, a different, no remote
             if n != ma[f]:
                 prompts.append((f, "cd")) # prompt changed/deleted
             elif n[20:] == "a": # added, no remote
                 actions.append((f, "f", None, "remote deleted"))
             else:
                 actions.append((f, "r", None, "other deleted"))
 
     for f, n in m2.iteritems():
         if partial and not partial(f):
             continue
         if f in m1 or f in copied: # files already visited
             continue
         if f in movewithdir:
             f2 = movewithdir[f]
             actions.append((None, "d", (f, f2, m2.flags(f)),
                             "local renamed directory to " + f2))
         elif f in copy:
             f2 = copy[f]
             if f2 in m2:
                 actions.append((f2, "m", (f, f, False),
                                 "remote copied to " + f))
             else:
                 actions.append((f2, "m", (f, f, True),
                                 "remote moved to " + f))
         elif f not in ma:
             # local unknown, remote created: the logic is described by the
             # following table:
             #
             # force  branchmerge  different  |  action
             #   n         *           n      |    get
             #   n         *           y      |   abort
             #   y         n           *      |    get
             #   y         y           n      |    get
             #   y         y           y      |   merge
             #
             # Checking whether the files are different is expensive, so we
             # don't do that when we can avoid it.
             if force and not branchmerge:
                 actions.append((f, "g", (m2.flags(f),), "remote created"))
             else:
                 different = _checkunknownfile(repo, wctx, p2, f)
                 if force and branchmerge and different:
                     actions.append((f, "m", (f, f, False),
                                     "remote differs from untracked local"))
                 elif not force and different:
                     aborts.append((f, "ud"))
                 else:
                     actions.append((f, "g", (m2.flags(f),), "remote created"))
         elif n != ma[f]:
             prompts.append((f, "dc")) # prompt deleted/changed
 
     for f, m in sorted(aborts):
         if m == "ud":
             repo.ui.warn(_("%s: untracked file differs\n") % f)
         else: assert False, m
     if aborts:
         raise util.Abort(_("untracked files in working directory differ "
                            "from files in requested revision"))
 
     for f, m in sorted(prompts):
         if m == "cd":
             if repo.ui.promptchoice(
                 _("local changed %s which remote deleted\n"
                   "use (c)hanged version or (d)elete?") % f,
                 (_("&Changed"), _("&Delete")), 0):
                 actions.append((f, "r", None, "prompt delete"))
             else:
                 actions.append((f, "a", None, "prompt keep"))
         elif m == "dc":
             if repo.ui.promptchoice(
                 _("remote changed %s which local deleted\n"
                   "use (c)hanged version or leave (d)eleted?") % f,
                 (_("&Changed"), _("&Deleted")), 0) == 0:
                 actions.append((f, "g", (m2.flags(f),), "prompt recreating"))
         else: assert False, m
     return actions
 
 def actionkey(a):
     return a[1] == "r" and -1 or 0, a
 
 def getremove(repo, mctx, overwrite, args):
     """apply usually-non-interactive updates to the working directory
 
     mctx is the context to be merged into the working copy
 
     yields tuples for progress updates
     """
     verbose = repo.ui.verbose
     unlink = util.unlinkpath
     wjoin = repo.wjoin
     fctx = mctx.filectx
     wwrite = repo.wwrite
     audit = repo.wopener.audit
     i = 0
     for arg in args:
         f = arg[0]
         if arg[1] == 'r':
             if verbose:
                 repo.ui.note(_("removing %s\n") % f)
             audit(f)
             try:
                 unlink(wjoin(f), ignoremissing=True)
             except OSError, inst:
                 repo.ui.warn(_("update failed to remove %s: %s!\n") %
                              (f, inst.strerror))
         else:
             if verbose:
                 repo.ui.note(_("getting %s\n") % f)
             wwrite(f, fctx(f).data(), arg[2][0])
         if i == 100:
             yield i, f
             i = 0
         i += 1
     if i > 0:
         yield i, f
 
 def applyupdates(repo, actions, wctx, mctx, actx, overwrite):
     """apply the merge action list to the working directory
 
     wctx is the working copy context
     mctx is the context to be merged into the working copy
     actx is the context of the common ancestor
 
     Return a tuple of counts (updated, merged, removed, unresolved) that
     describes how many files were affected by the update.
     """
 
     updated, merged, removed, unresolved = 0, 0, 0, 0
     ms = mergestate(repo)
     ms.reset(wctx.p1().node())
     moves = []
     actions.sort(key=actionkey)
 
     # prescan for merges
     for a in actions:
         f, m, args, msg = a
         repo.ui.debug(" %s: %s -> %s\n" % (f, msg, m))
         if m == "m": # merge
             f2, fd, move = args
             if fd == '.hgsubstate': # merged internally
                 continue
             repo.ui.debug("  preserving %s for resolve of %s\n" % (f, fd))
             fcl = wctx[f]
             fco = mctx[f2]
             if mctx == actx: # backwards, use working dir parent as ancestor
                 if fcl.parents():
                     fca = fcl.p1()
                 else:
                     fca = repo.filectx(f, fileid=nullrev)
             else:
                 fca = fcl.ancestor(fco, actx)
             if not fca:
                 fca = repo.filectx(f, fileid=nullrev)
             ms.add(fcl, fco, fca, fd)
             if f != fd and move:
                 moves.append(f)
 
     audit = repo.wopener.audit
 
     # remove renamed files after safely stored
     for f in moves:
         if os.path.lexists(repo.wjoin(f)):
             repo.ui.debug("removing %s\n" % f)
             audit(f)
             util.unlinkpath(repo.wjoin(f))
 
     numupdates = len(actions)
     workeractions = [a for a in actions if a[1] in 'gr']
     updated = len([a for a in workeractions if a[1] == 'g'])
     removed = len([a for a in workeractions if a[1] == 'r'])
     actions = [a for a in actions if a[1] not in 'gr']
 
     hgsub = [a[1] for a in workeractions if a[0] == '.hgsubstate']
     if hgsub and hgsub[0] == 'r':
         subrepo.submerge(repo, wctx, mctx, wctx, overwrite)
 
     z = 0
     prog = worker.worker(repo.ui, 0.001, getremove, (repo, mctx, overwrite),
                          workeractions)
     for i, item in prog:
         z += i
         repo.ui.progress(_('updating'), z, item=item, total=numupdates,
                          unit=_('files'))
 
     if hgsub and hgsub[0] == 'g':
         subrepo.submerge(repo, wctx, mctx, wctx, overwrite)
 
     _updating = _('updating')
     _files = _('files')
     progress = repo.ui.progress
 
     for i, a in enumerate(actions):
         f, m, args, msg = a
         progress(_updating, z + i + 1, item=f, total=numupdates, unit=_files)
         if m == "m": # merge
             if fd == '.hgsubstate': # subrepo states need updating
                 subrepo.submerge(repo, wctx, mctx, wctx.ancestor(mctx),
                                  overwrite)
                 continue
             f2, fd, move = args
             audit(fd)
             r = ms.resolve(fd, wctx, mctx)
             if r is not None and r > 0:
                 unresolved += 1
             else:
                 if r is None:
                     updated += 1
                 else:
                     merged += 1
         elif m == "d": # directory rename
             f2, fd, flags = args
             if f:
                 repo.ui.note(_("moving %s to %s\n") % (f, fd))
                 audit(f)
                 repo.wwrite(fd, wctx.filectx(f).data(), flags)
                 util.unlinkpath(repo.wjoin(f))
             if f2:
                 repo.ui.note(_("getting %s to %s\n") % (f2, fd))
                 repo.wwrite(fd, mctx.filectx(f2).data(), flags)
             updated += 1
         elif m == "dr": # divergent renames
             fl, = args
             repo.ui.warn(_("note: possible conflict - %s was renamed "
                            "multiple times to:\n") % f)
             for nf in fl:
                 repo.ui.warn(" %s\n" % nf)
         elif m == "rd": # rename and delete
             fl, = args
             repo.ui.warn(_("note: possible conflict - %s was deleted "
                            "and renamed to:\n") % f)
             for nf in fl:
                 repo.ui.warn(" %s\n" % nf)
         elif m == "e": # exec
             flags, = args
             audit(f)
             util.setflags(repo.wjoin(f), 'l' in flags, 'x' in flags)
             updated += 1
     ms.commit()
     progress(_updating, None, total=numupdates, unit=_files)
 
     return updated, merged, removed, unresolved
 
 def calculateupdates(repo, tctx, mctx, ancestor, branchmerge, force, partial):
     "Calculate the actions needed to merge mctx into tctx"
     actions = []
     folding = not util.checkcase(repo.path)
     if folding:
         # collision check is not needed for clean update
         if (not branchmerge and
             (force or not tctx.dirty(missing=True, branch=False))):
             _checkcollision(mctx, None)
         else:
             _checkcollision(mctx, (tctx, ancestor))
-    if tctx.rev() is None:
-        actions += _forgetremoved(tctx, mctx, branchmerge)
     actions += manifestmerge(repo, tctx, mctx,
                              ancestor,
                              branchmerge, force,
                              partial)
+    if tctx.rev() is None:
+        actions += _forgetremoved(tctx, mctx, branchmerge)
     return actions
 
 def recordupdates(repo, actions, branchmerge):
     "record merge actions to the dirstate"
 
     for a in actions:
         f, m, args, msg = a
         if m == "r": # remove
             if branchmerge:
                 repo.dirstate.remove(f)
             else:
                 repo.dirstate.drop(f)
         elif m == "a": # re-add
             if not branchmerge:
                 repo.dirstate.add(f)
         elif m == "f": # forget
             repo.dirstate.drop(f)
         elif m == "e": # exec change
             repo.dirstate.normallookup(f)
         elif m == "g": # get
             if branchmerge:
                 repo.dirstate.otherparent(f)
             else:
                 repo.dirstate.normal(f)
         elif m == "m": # merge
             f2, fd, move = args
             if branchmerge:
                 # We've done a branch merge, mark this file as merged
                 # so that we properly record the merger later
                 repo.dirstate.merge(fd)
                 if f != f2: # copy/rename
                     if move:
                         repo.dirstate.remove(f)
                     if f != fd:
                         repo.dirstate.copy(f, fd)
                     else:
                         repo.dirstate.copy(f2, fd)
             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 == fd: # file not locally copied/moved
                     repo.dirstate.normallookup(fd)
                 if move:
                     repo.dirstate.drop(f)
         elif m == "d": # directory rename
             f2, fd, flag = args
             if not f2 and f not in repo.dirstate:
                 # untracked file moved
                 continue
             if branchmerge:
                 repo.dirstate.add(fd)
                 if f:
                     repo.dirstate.remove(f)
                     repo.dirstate.copy(f, fd)
                 if f2:
                     repo.dirstate.copy(f2, fd)
             else:
                 repo.dirstate.normal(fd)
                 if f:
                     repo.dirstate.drop(f)
 
 def update(repo, node, branchmerge, force, partial, ancestor=None,
            mergeancestor=False):
     """
     Perform a merge between the working directory and the given node
 
     node = the node to update to, or None if unspecified
     branchmerge = whether to merge between branches
     force = whether to force branch merging or file overwriting
     partial = a function to filter file lists (dirstate not updated)
     mergeancestor = if false, merging with an ancestor (fast-forward)
       is only allowed between different named branches. This flag
       is used by rebase extension as a temporary fix and should be
       avoided in general.
 
     The table below shows all the behaviors of the update command
     given the -c and -C or no options, whether the working directory
     is dirty, whether a revision is specified, and the relationship of
     the parent rev to the target rev (linear, on the same named
     branch, or on another named branch).
 
     This logic is tested by test-update-branches.t.
 
     -c  -C  dirty  rev  |  linear   same  cross
      n   n    n     n   |    ok     (1)     x
      n   n    n     y   |    ok     ok     ok
      n   n    y     *   |   merge   (2)    (2)
      n   y    *     *   |    ---  discard  ---
      y   n    y     *   |    ---    (3)    ---
      y   n    n     *   |    ---    ok     ---
      y   y    *     *   |    ---    (4)    ---
 
     x = can't happen
     * = don't-care
     1 = abort: crosses branches (use 'hg merge' or 'hg update -c')
     2 = abort: crosses branches (use 'hg merge' to merge or
                  use 'hg update -C' to discard changes)
     3 = abort: uncommitted local changes
     4 = incompatible options (checked in commands.py)
 
     Return the same tuple as applyupdates().
     """
 
     onode = node
     wlock = repo.wlock()
     try:
         wc = repo[None]
         if node is None:
             # tip of current branch
             try:
                 node = repo.branchtip(wc.branch())
             except error.RepoLookupError:
                 if wc.branch() == "default": # no default branch!
                     node = repo.lookup("tip") # update to tip
                 else:
                     raise util.Abort(_("branch %s not found") % wc.branch())
         overwrite = force and not branchmerge
         pl = wc.parents()
         p1, p2 = pl[0], repo[node]
         if ancestor:
             pa = repo[ancestor]
         else:
             pa = p1.ancestor(p2)
 
         fp1, fp2, xp1, xp2 = p1.node(), p2.node(), str(p1), str(p2)
 
         ### check phase
         if not overwrite and len(pl) > 1:
             raise util.Abort(_("outstanding uncommitted merges"))
         if branchmerge:
             if pa == p2:
                 raise util.Abort(_("merging with a working directory ancestor"
                                    " has no effect"))
             elif pa == p1:
                 if not mergeancestor and p1.branch() == p2.branch():
                     raise util.Abort(_("nothing to merge"),
                                      hint=_("use 'hg update' "
                                             "or check 'hg heads'"))
             if not force and (wc.files() or wc.deleted()):
                 raise util.Abort(_("outstanding uncommitted changes"),
                                  hint=_("use 'hg status' to list changes"))
             for s in sorted(wc.substate):
                 if wc.sub(s).dirty():
                     raise util.Abort(_("outstanding uncommitted changes in "
                                        "subrepository '%s'") % s)
 
         elif not overwrite:
             if pa == p1 or pa == p2: # linear
                 pass # all good
             elif wc.dirty(missing=True):
                 raise util.Abort(_("crosses branches (merge branches or use"
                                    " --clean to discard changes)"))
             elif onode is None:
                 raise util.Abort(_("crosses branches (merge branches or update"
                                    " --check to force update)"))
             else:
                 # Allow jumping branches if clean and specific rev given
                 pa = p1
 
         ### calculate phase
         actions = calculateupdates(repo, wc, p2, pa,
                                    branchmerge, force, partial)
 
         ### apply phase
         if not branchmerge: # just jump to the new rev
             fp1, fp2, xp1, xp2 = fp2, nullid, xp2, ''
         if not partial:
             repo.hook('preupdate', throw=True, parent1=xp1, parent2=xp2)
 
         stats = applyupdates(repo, actions, wc, p2, pa, overwrite)
 
         if not partial:
             repo.setparents(fp1, fp2)
             recordupdates(repo, actions, branchmerge)
             if not branchmerge:
                 repo.dirstate.setbranch(p2.branch())
     finally:
         wlock.release()
 
     if not partial:
         repo.hook('update', parent1=xp1, parent2=xp2, error=stats[3])
     return stats