/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2008-01-25 10:46:10 UTC
  • mto: This revision was merged to the branch mainline in revision 3211.
  • Revision ID: robertc@robertcollins.net-20080125104610-4qweeqqch2feyxdv
 * The ``register-branch`` command will now use the public url of the branch
   containing the current directory, if one has been set and no explicit branch is provided.
   (Robert Collins)

 * New error ``NoPublicBranch`` for commands that need a public branch to
   operate. (Robert Collins)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
38
46
    debug,
 
47
    delta,
39
48
    dirstate,
40
49
    errors,
41
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
55
    revision as _mod_revision,
44
56
    revisiontree,
 
57
    textui,
45
58
    trace,
46
59
    transform,
47
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
48
63
    )
49
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
50
66
import bzrlib.ui
51
67
""")
52
68
 
 
69
from bzrlib import symbol_versioning
53
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
56
75
from bzrlib.mutabletree import needs_tree_write_lock
57
76
from bzrlib.osutils import (
58
77
    file_kind,
59
78
    isdir,
 
79
    normpath,
60
80
    pathjoin,
 
81
    rand_chars,
61
82
    realpath,
62
83
    safe_unicode,
 
84
    splitpath,
63
85
    )
64
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
65
87
from bzrlib.transport.local import LocalTransport
66
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
67
97
from bzrlib.tree import Tree
68
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
99
 
70
100
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
72
120
    def __init__(self, basedir,
73
121
                 branch,
74
122
                 _control_files=None,
83
131
        """
84
132
        self._format = _format
85
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
86
136
        basedir = safe_unicode(basedir)
87
137
        mutter("opening working tree %r", basedir)
88
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
89
141
        self.basedir = realpath(basedir)
90
142
        # if branch is at our basedir and is a format 6 or less
91
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
92
146
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
147
        self._dirty = None
95
148
        #-------------
96
149
        # during a read or write lock these objects are set, and are
100
153
        #-------------
101
154
        self._setup_directory_is_tree_reference()
102
155
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
156
 
108
157
    @needs_tree_write_lock
109
158
    def _add(self, files, ids, kinds):
137
186
    @needs_tree_write_lock
138
187
    def add_reference(self, sub_tree):
139
188
        # use standard implementation, which calls back to self._add
140
 
        #
 
189
        # 
141
190
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
143
192
        self._add_reference(sub_tree)
144
193
 
145
194
    def break_lock(self):
184
233
            WorkingTree3._comparison_data(self, entry, path)
185
234
        # it looks like a plain directory, but it's really a reference -- see
186
235
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
189
239
            kind = 'tree-reference'
190
240
        return kind, executable, stat_value
191
241
 
217
267
            return self._dirstate
218
268
        local_path = self.bzrdir.get_workingtree_transport(None
219
269
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
222
271
        return self._dirstate
223
272
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
235
 
 
236
273
    def filter_unversioned_files(self, paths):
237
274
        """Filter out paths that are versioned.
238
275
 
270
307
 
271
308
    def _generate_inventory(self):
272
309
        """Create and set self.inventory from the dirstate object.
273
 
 
 
310
        
274
311
        This is relatively expensive: we have to walk the entire dirstate.
275
312
        Ideally we would not, and can deprecate this function.
276
313
        """
280
317
        state._read_dirblocks_if_needed()
281
318
        root_key, current_entry = self._get_entry(path='')
282
319
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
285
321
        inv = Inventory(root_id=current_id)
286
322
        # Turn some things into local variables
287
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
356
                    # add this entry to the parent map.
321
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
358
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
327
364
                    inv_entry.reference_revision = link_or_sha1 or None
328
365
                elif kind != 'symlink':
329
366
                    raise AssertionError("unknown kind %r" % kind)
330
367
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
337
371
                inv_byid[file_id] = inv_entry
338
372
                parent_ie.children[name_unicode] = inv_entry
339
373
        self._inventory = inv
344
378
        If either file_id or path is supplied, it is used as the key to lookup.
345
379
        If both are supplied, the fastest lookup is used, and an error is
346
380
        raised if they do not both point at the same row.
347
 
 
 
381
        
348
382
        :param file_id: An optional unicode file_id to be looked up.
349
383
        :param path: An optional unicode path to be looked up.
350
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
374
408
                    return None
375
409
                else:
376
410
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
379
413
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
414
            return link_or_sha1
390
415
        return None
391
416
 
392
417
    def _get_inventory(self):
406
431
    @needs_read_lock
407
432
    def get_parent_ids(self):
408
433
        """See Tree.get_parent_ids.
409
 
 
 
434
        
410
435
        This implementation requests the ids list from the dirstate file.
411
436
        """
412
437
        return self.current_dirstate().get_parent_ids()
434
459
        return osutils.lexists(pathjoin(
435
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
461
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
462
    @needs_read_lock
443
463
    def id2path(self, file_id):
444
464
        "Convert a file-id to a path."
480
500
            mode = os.lstat(self.abspath(path)).st_mode
481
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
502
 
483
 
    def all_file_ids(self):
484
 
        """See Tree.iter_all_file_ids"""
485
 
        self._must_be_locked()
486
 
        result = set()
487
 
        for key, tree_details in self.current_dirstate()._iter_entries():
488
 
            if tree_details[0][0] in ('a', 'r'): # relocated
489
 
                continue
490
 
            result.add(key[2])
491
 
        return result
492
 
 
493
503
    @needs_read_lock
494
504
    def __iter__(self):
495
505
        """Iterate through file_ids for this tree.
508
518
        return iter(result)
509
519
 
510
520
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
521
        for key, tree_details in self.current_dirstate()._iter_entries():
516
522
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
523
                # not relevant to the working tree
519
525
            if not key[1]:
520
526
                # the root is not a reference.
521
527
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
528
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
529
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
530
                if self._kind(path) == 'tree-reference':
 
531
                    yield path, key[2]
526
532
            except errors.NoSuchFile:
527
533
                # path is missing on disk.
528
534
                continue
529
535
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
536
536
    def kind(self, file_id):
537
537
        """Return the kind of a file.
538
538
 
542
542
        Note: The caller is expected to take a read-lock before calling this.
543
543
        """
544
544
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
545
        assert relpath != None, \
 
546
            "path for id {%s} is None!" % file_id
548
547
        return self._kind(relpath)
549
548
 
550
549
    def _kind(self, relpath):
551
550
        abspath = self.abspath(relpath)
552
551
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    kind = 'tree-reference'
 
552
        if (self._repo_supports_tree_reference and
 
553
            kind == 'directory' and
 
554
            self._directory_is_tree_reference(relpath)):
 
555
            kind = 'tree-reference'
558
556
        return kind
559
557
 
560
558
    @needs_read_lock
623
621
        result = []
624
622
        if not from_paths:
625
623
            return result
 
624
 
626
625
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
626
 
 
627
        assert not isinstance(from_paths, basestring)
629
628
        to_dir_utf8 = to_dir.encode('utf8')
630
629
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
630
        id_index = state._get_id_index()
653
652
        if self._inventory is not None:
654
653
            update_inventory = True
655
654
            inv = self.inventory
 
655
            to_dir_ie = inv[to_dir_id]
656
656
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
657
        else:
659
658
            update_inventory = False
660
659
 
691
690
            from_entry = self._get_entry(path=from_rel)
692
691
            if from_entry == (None, None):
693
692
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
693
                    errors.NotVersionedError(path=str(from_rel)))
695
694
 
696
695
            from_id = from_entry[0][2]
697
696
            to_rel = pathjoin(to_dir, from_tail)
786
785
                if minikind == 'd':
787
786
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
787
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
788
                        assert from_dir != '', "renaming root not supported"
791
789
                        from_key = (from_dir, '')
792
790
                        from_block_idx, present = \
793
791
                            state._find_block_index_from_key(from_key)
806
804
 
807
805
                        # Grab a copy since move_one may update the list.
808
806
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
807
                            assert entry[0][0] == from_dir
811
808
                            cur_details = entry[1][0]
812
809
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
810
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
872
869
        for tree in trees:
873
870
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
871
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
872
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
873
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
874
        # -- make all paths utf8 --
879
875
        paths_utf8 = set()
940
936
                raise errors.PathsNotVersionedError(paths)
941
937
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
938
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
939
        # sketch: 
944
940
        # for all search_indexs in each path at or under each element of
945
941
        # search_paths, if the detail is relocated: add the id, and add the
946
942
        # relocated path as one to search if its not searched already. If the
1002
998
 
1003
999
    def read_working_inventory(self):
1004
1000
        """Read the working inventory.
1005
 
 
 
1001
        
1006
1002
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
1003
        """
1008
1004
        return self.inventory
1026
1022
    def set_last_revision(self, new_revision):
1027
1023
        """Change the last revision in the working tree."""
1028
1024
        parents = self.get_parent_ids()
1029
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1030
 
            if len(parents) >= 2:
1031
 
                raise AssertionError(
1032
 
                    "setting the last parent to none with a pending merge is "
1033
 
                    "unsupported.")
 
1025
        if new_revision in (NULL_REVISION, None):
 
1026
            assert len(parents) < 2, (
 
1027
                "setting the last parent to none with a pending merge is "
 
1028
                "unsupported.")
1034
1029
            self.set_parent_ids([])
1035
1030
        else:
1036
1031
            self.set_parent_ids([new_revision] + parents[1:],
1039
1034
    @needs_tree_write_lock
1040
1035
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
1036
        """Set the parent ids to revision_ids.
1042
 
 
 
1037
        
1043
1038
        See also set_parent_trees. This api will try to retrieve the tree data
1044
1039
        for each element of revision_ids from the trees repository. If you have
1045
1040
        tree data already available, it is more efficient to use
1077
1072
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1073
        real_trees = []
1079
1074
        ghosts = []
1080
 
 
1081
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1082
 
        graph = self.branch.repository.get_graph()
1083
 
        heads = graph.heads(parent_ids)
1084
 
        accepted_revisions = set()
1085
 
 
1086
1075
        # convert absent trees to the null tree, which we convert back to
1087
1076
        # missing on access.
1088
1077
        for rev_id, tree in parents_list:
1089
 
            if len(accepted_revisions) > 0:
1090
 
                # we always accept the first tree
1091
 
                if rev_id in accepted_revisions or rev_id not in heads:
1092
 
                    # We have already included either this tree, or its
1093
 
                    # descendent, so we skip it.
1094
 
                    continue
1095
1078
            _mod_revision.check_not_reserved_id(rev_id)
1096
1079
            if tree is not None:
1097
1080
                real_trees.append((rev_id, tree))
1098
1081
            else:
1099
1082
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1083
                    self.branch.repository.revision_tree(None)))
1102
1084
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
1104
1085
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1105
1086
        self._make_dirty(reset_inventory=False)
1106
1087
 
1199
1180
                # just forget the whole block.
1200
1181
                entry_index = 0
1201
1182
                while entry_index < len(block[1]):
 
1183
                    # Mark this file id as having been removed
1202
1184
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1185
                    ids_to_unversion.discard(entry[0][2])
 
1186
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1187
                                               # entries
 
1188
                        or not state._make_absent(entry)):
1205
1189
                        entry_index += 1
1206
 
                    else:
1207
 
                        # Mark this file id as having been removed
1208
 
                        ids_to_unversion.discard(entry[0][2])
1209
 
                        if not state._make_absent(entry):
1210
 
                            # The block has not shrunk.
1211
 
                            entry_index += 1
1212
1190
                # go to the next block. (At the moment we dont delete empty
1213
1191
                # dirblocks)
1214
1192
                block_index += 1
1237
1215
            for file_id in file_ids:
1238
1216
                self._inventory.remove_recursive_id(file_id)
1239
1217
 
1240
 
    @needs_tree_write_lock
1241
 
    def rename_one(self, from_rel, to_rel, after=False):
1242
 
        """See WorkingTree.rename_one"""
1243
 
        self.flush()
1244
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1245
 
 
1246
 
    @needs_tree_write_lock
1247
 
    def apply_inventory_delta(self, changes):
1248
 
        """See MutableTree.apply_inventory_delta"""
1249
 
        state = self.current_dirstate()
1250
 
        state.update_by_delta(changes)
1251
 
        self._make_dirty(reset_inventory=True)
1252
 
 
1253
1218
    def update_basis_by_delta(self, new_revid, delta):
1254
1219
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
 
1220
        assert self.last_revision() != new_revid
1257
1221
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
1222
 
1259
1223
    @needs_read_lock
1263
1227
    @needs_tree_write_lock
1264
1228
    def _write_inventory(self, inv):
1265
1229
        """Write inventory as the current inventory."""
1266
 
        if self._dirty:
1267
 
            raise AssertionError("attempting to write an inventory when the "
1268
 
                "dirstate is dirty will lose pending changes")
1269
 
        had_inventory = self._inventory is not None
1270
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1271
 
        # working inventory. We do this because self.inventory may be inv, or
1272
 
        # may have been modified, and either case would prevent a clean delta
1273
 
        # being created.
1274
 
        self._inventory = None
1275
 
        # generate a delta,
1276
 
        delta = inv._make_delta(self.inventory)
1277
 
        # and apply it.
1278
 
        self.apply_inventory_delta(delta)
1279
 
        if had_inventory:
 
1230
        assert not self._dirty, ("attempting to write an inventory when the "
 
1231
            "dirstate is dirty will cause data loss")
 
1232
        self.current_dirstate().set_state_from_inventory(inv)
 
1233
        self._make_dirty(reset_inventory=False)
 
1234
        if self._inventory is not None:
1280
1235
            self._inventory = inv
1281
1236
        self.flush()
1282
1237
 
1283
1238
 
1284
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1285
 
 
1286
 
    def __init__(self, tree):
1287
 
        self.tree = tree
1288
 
 
1289
 
    def sha1(self, abspath):
1290
 
        """See dirstate.SHA1Provider.sha1()."""
1291
 
        filters = self.tree._content_filter_stack(
1292
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1293
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1294
 
 
1295
 
    def stat_and_sha1(self, abspath):
1296
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1297
 
        filters = self.tree._content_filter_stack(
1298
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1299
 
        file_obj = file(abspath, 'rb', 65000)
1300
 
        try:
1301
 
            statvalue = os.fstat(file_obj.fileno())
1302
 
            if filters:
1303
 
                file_obj = filtered_input_file(file_obj, filters)
1304
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1305
 
        finally:
1306
 
            file_obj.close()
1307
 
        return statvalue, sha1
1308
 
 
1309
 
 
1310
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1311
 
    """Dirstate working tree that supports content filtering.
1312
 
 
1313
 
    The dirstate holds the hash and size of the canonical form of the file, 
1314
 
    and most methods must return that.
1315
 
    """
1316
 
 
1317
 
    def _file_content_summary(self, path, stat_result):
1318
 
        # This is to support the somewhat obsolete path_content_summary method
1319
 
        # with content filtering: see
1320
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1321
 
        #
1322
 
        # If the dirstate cache is up to date and knows the hash and size,
1323
 
        # return that.
1324
 
        # Otherwise if there are no content filters, return the on-disk size
1325
 
        # and leave the hash blank.
1326
 
        # Otherwise, read and filter the on-disk file and use its size and
1327
 
        # hash.
1328
 
        #
1329
 
        # The dirstate doesn't store the size of the canonical form so we
1330
 
        # can't trust it for content-filtered trees.  We just return None.
1331
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1332
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1333
 
        return ('file', None, executable, dirstate_sha1)
1334
 
 
1335
 
 
1336
 
class WorkingTree4(DirStateWorkingTree):
1337
 
    """This is the Format 4 working tree.
1338
 
 
1339
 
    This differs from WorkingTree3 by:
1340
 
     - Having a consolidated internal dirstate, stored in a
1341
 
       randomly-accessible sorted file on disk.
1342
 
     - Not having a regular inventory attribute.  One can be synthesized
1343
 
       on demand but this is expensive and should be avoided.
1344
 
 
1345
 
    This is new in bzr 0.15.
1346
 
    """
1347
 
 
1348
 
 
1349
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1350
 
    """This is the Format 5 working tree.
1351
 
 
1352
 
    This differs from WorkingTree4 by:
1353
 
     - Supporting content filtering.
1354
 
 
1355
 
    This is new in bzr 1.11.
1356
 
    """
1357
 
 
1358
 
 
1359
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1360
 
    """This is the Format 6 working tree.
1361
 
 
1362
 
    This differs from WorkingTree5 by:
1363
 
     - Supporting a current view that may mask the set of files in a tree
1364
 
       impacted by most user operations.
1365
 
 
1366
 
    This is new in bzr 1.14.
1367
 
    """
1368
 
 
1369
 
    def _make_views(self):
1370
 
        return views.PathBasedViews(self)
1371
 
 
1372
 
 
1373
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
 
1239
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1240
    """The first consolidated dirstate working tree format.
 
1241
 
 
1242
    This format:
 
1243
        - exists within a metadir controlling .bzr
 
1244
        - includes an explicit version marker for the workingtree control
 
1245
          files, separate from the BzrDir format
 
1246
        - modifies the hash cache format
 
1247
        - is new in bzr 0.15
 
1248
        - uses a LockDir to guard access to it.
 
1249
    """
 
1250
 
 
1251
    upgrade_recommended = False
 
1252
 
 
1253
    def get_format_string(self):
 
1254
        """See WorkingTreeFormat.get_format_string()."""
 
1255
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1256
 
 
1257
    def get_format_description(self):
 
1258
        """See WorkingTreeFormat.get_format_description()."""
 
1259
        return "Working tree format 4"
1374
1260
 
1375
1261
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
 
                   accelerator_tree=None, hardlink=False):
 
1262
                   accelerator_tree=None):
1377
1263
        """See WorkingTreeFormat.initialize().
1378
1264
 
1379
1265
        :param revision_id: allows creating a working tree at a different
1382
1268
            contents more quickly than the revision tree, i.e. a workingtree.
1383
1269
            The revision tree will be used for cases where accelerator_tree's
1384
1270
            content is different.
1385
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1386
 
            where possible.
1387
1271
 
1388
1272
        These trees get an initial random root id, if their repository supports
1389
1273
        rich root data, TREE_ROOT otherwise.
1394
1278
        control_files = self._open_control_files(a_bzrdir)
1395
1279
        control_files.create_lock()
1396
1280
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
 
1281
        control_files.put_utf8('format', self.get_format_string())
1399
1282
        if from_branch is not None:
1400
1283
            branch = from_branch
1401
1284
        else:
1407
1290
        state = dirstate.DirState.initialize(local_path)
1408
1291
        state.unlock()
1409
1292
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1293
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1294
                         branch,
1412
1295
                         _format=self,
1413
1296
                         _bzrdir=a_bzrdir,
1415
1298
        wt._new_tree()
1416
1299
        wt.lock_tree_write()
1417
1300
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1301
            if revision_id in (None, NULL_REVISION):
1420
1302
                if branch.repository.supports_rich_root():
1421
1303
                    wt._set_root_id(generate_ids.gen_root_id())
1422
1304
                else:
1433
1315
                    pass
1434
1316
            if basis is None:
1435
1317
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1318
            if revision_id == NULL_REVISION:
1437
1319
                parents_list = []
1438
1320
            else:
1439
1321
                parents_list = [(revision_id, basis)]
1440
1322
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1323
            wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1324
            wt.flush()
 
1325
            # if the basis has a root id we have to use that; otherwise we use
 
1326
            # a new random one
 
1327
            basis_root_id = basis.get_root_id()
 
1328
            if basis_root_id is not None:
 
1329
                wt._set_root_id(basis_root_id)
1443
1330
                wt.flush()
1444
 
                # if the basis has a root id we have to use that; otherwise we
1445
 
                # use a new random one
1446
 
                basis_root_id = basis.get_root_id()
1447
 
                if basis_root_id is not None:
1448
 
                    wt._set_root_id(basis_root_id)
1449
 
                    wt.flush()
1450
 
                if wt.supports_content_filtering():
1451
 
                    # The original tree may not have the same content filters
1452
 
                    # applied so we can't safely build the inventory delta from
1453
 
                    # the source tree.
1454
 
                    delta_from_tree = False
1455
 
                else:
1456
 
                    delta_from_tree = True
1457
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1458
 
                # because wt4.apply_inventory_delta does not mutate the input
1459
 
                # inventory entries.
1460
 
                transform.build_tree(basis, wt, accelerator_tree,
1461
 
                                     hardlink=hardlink,
1462
 
                                     delta_from_tree=delta_from_tree)
1463
 
            finally:
1464
 
                basis.unlock()
 
1331
            transform.build_tree(basis, wt, accelerator_tree)
 
1332
            basis.unlock()
1465
1333
        finally:
1466
1334
            control_files.unlock()
1467
1335
            wt.unlock()
1468
1336
        return wt
1469
1337
 
1470
 
    def _init_custom_control_files(self, wt):
1471
 
        """Subclasses with custom control files should override this method.
1472
 
 
1473
 
        The working tree and control files are locked for writing when this
1474
 
        method is called.
1475
 
 
1476
 
        :param wt: the WorkingTree object
1477
 
        """
1478
 
 
1479
1338
    def _open(self, a_bzrdir, control_files):
1480
1339
        """Open the tree itself.
1481
1340
 
1482
1341
        :param a_bzrdir: the dir for the tree.
1483
1342
        :param control_files: the control files for the tree.
1484
1343
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1344
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1345
                           branch=a_bzrdir.open_branch(),
1487
1346
                           _format=self,
1488
1347
                           _bzrdir=a_bzrdir,
1489
1348
                           _control_files=control_files)
1490
1349
 
1491
1350
    def __get_matchingbzrdir(self):
1492
 
        return self._get_matchingbzrdir()
1493
 
 
1494
 
    def _get_matchingbzrdir(self):
1495
 
        """Overrideable method to get a bzrdir for testing."""
1496
1351
        # please test against something that will let us do tree references
1497
1352
        return bzrdir.format_registry.make_bzrdir(
1498
1353
            'dirstate-with-subtree')
1500
1355
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1356
 
1502
1357
 
1503
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1504
 
    """The first consolidated dirstate working tree format.
1505
 
 
1506
 
    This format:
1507
 
        - exists within a metadir controlling .bzr
1508
 
        - includes an explicit version marker for the workingtree control
1509
 
          files, separate from the BzrDir format
1510
 
        - modifies the hash cache format
1511
 
        - is new in bzr 0.15
1512
 
        - uses a LockDir to guard access to it.
1513
 
    """
1514
 
 
1515
 
    upgrade_recommended = False
1516
 
 
1517
 
    _tree_class = WorkingTree4
1518
 
 
1519
 
    def get_format_string(self):
1520
 
        """See WorkingTreeFormat.get_format_string()."""
1521
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1522
 
 
1523
 
    def get_format_description(self):
1524
 
        """See WorkingTreeFormat.get_format_description()."""
1525
 
        return "Working tree format 4"
1526
 
 
1527
 
 
1528
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1529
 
    """WorkingTree format supporting content filtering.
1530
 
    """
1531
 
 
1532
 
    upgrade_recommended = False
1533
 
 
1534
 
    _tree_class = WorkingTree5
1535
 
 
1536
 
    def get_format_string(self):
1537
 
        """See WorkingTreeFormat.get_format_string()."""
1538
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1539
 
 
1540
 
    def get_format_description(self):
1541
 
        """See WorkingTreeFormat.get_format_description()."""
1542
 
        return "Working tree format 5"
1543
 
 
1544
 
    def supports_content_filtering(self):
1545
 
        return True
1546
 
 
1547
 
 
1548
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1549
 
    """WorkingTree format supporting views.
1550
 
    """
1551
 
 
1552
 
    upgrade_recommended = False
1553
 
 
1554
 
    _tree_class = WorkingTree6
1555
 
 
1556
 
    def get_format_string(self):
1557
 
        """See WorkingTreeFormat.get_format_string()."""
1558
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1559
 
 
1560
 
    def get_format_description(self):
1561
 
        """See WorkingTreeFormat.get_format_description()."""
1562
 
        return "Working tree format 6"
1563
 
 
1564
 
    def _init_custom_control_files(self, wt):
1565
 
        """Subclasses with custom control files should override this method."""
1566
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1567
 
 
1568
 
    def supports_content_filtering(self):
1569
 
        return True
1570
 
 
1571
 
    def supports_views(self):
1572
 
        return True
1573
 
 
1574
 
 
1575
1358
class DirStateRevisionTree(Tree):
1576
 
    """A revision tree pulling the inventory from a dirstate.
1577
 
    
1578
 
    Note that this is one of the historical (ie revision) trees cached in the
1579
 
    dirstate for easy access, not the workingtree.
1580
 
    """
 
1359
    """A revision tree pulling the inventory from a dirstate."""
1581
1360
 
1582
1361
    def __init__(self, dirstate, revision_id, repository):
1583
1362
        self._dirstate = dirstate
1586
1365
        self._inventory = None
1587
1366
        self._locked = 0
1588
1367
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1368
 
1593
1369
    def __repr__(self):
1594
1370
        return "<%s of %s in %s>" % \
1597
1373
    def annotate_iter(self, file_id,
1598
1374
                      default_revision=_mod_revision.CURRENT_REVISION):
1599
1375
        """See Tree.annotate_iter"""
1600
 
        text_key = (file_id, self.inventory[file_id].revision)
1601
 
        annotations = self._repository.texts.annotate(text_key)
1602
 
        return [(key[-1], line) for (key, line) in annotations]
 
1376
        w = self._get_weave(file_id)
 
1377
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1378
 
1604
1379
    def _get_ancestors(self, default_revision):
1605
1380
        return set(self._repository.get_ancestry(self._revision_id,
1634
1409
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
1410
        return path_utf8.decode('utf8')
1636
1411
 
1637
 
    def iter_references(self):
1638
 
        if not self._repo_supports_tree_reference:
1639
 
            # When the repo doesn't support references, we will have nothing to
1640
 
            # return
1641
 
            return iter([])
1642
 
        # Otherwise, fall back to the default implementation
1643
 
        return super(DirStateRevisionTree, self).iter_references()
1644
 
 
1645
1412
    def _get_parent_index(self):
1646
1413
        """Return the index in the dirstate referenced by this tree."""
1647
1414
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1419
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1420
        If both are supplied, the fastest lookup is used, and an error is
1654
1421
        raised if they do not both point at the same row.
1655
 
 
 
1422
        
1656
1423
        :param file_id: An optional unicode file_id to be looked up.
1657
1424
        :param path: An optional unicode path to be looked up.
1658
1425
        :return: The dirstate row tuple for path/file_id, or (None, None)
1672
1439
 
1673
1440
        This is relatively expensive: we have to walk the entire dirstate.
1674
1441
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1442
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1443
            'dirstate revision tree'
1679
1444
        # separate call for profiling - makes it clear where the costs are.
1680
1445
        self._dirstate._read_dirblocks_if_needed()
1681
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1682
 
            raise AssertionError(
1683
 
                'parent %s has disappeared from %s' % (
1684
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1446
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1447
            'parent %s has disappeared from %s' % (
 
1448
            self._revision_id, self._dirstate.get_parent_ids())
1685
1449
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1450
        # This is identical now to the WorkingTree _generate_inventory except
1687
1451
        # for the tree index use.
1688
1452
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1453
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1454
        assert current_entry[parent_index][0] == 'd'
1692
1455
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1456
        inv.root.revision = current_entry[parent_index][4]
1694
1457
        # Turn some things into local variables
1734
1497
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1498
                            % entry)
1736
1499
                # These checks cost us around 40ms on a 55k entry tree
1737
 
                if file_id in inv_byid:
1738
 
                    raise AssertionError('file_id %s already in'
1739
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1740
 
                if name_unicode in parent_ie.children:
1741
 
                    raise AssertionError('name %r already in parent'
1742
 
                        % (name_unicode,))
 
1500
                assert file_id not in inv_byid
 
1501
                assert name_unicode not in parent_ie.children
1743
1502
                inv_byid[file_id] = inv_entry
1744
1503
                parent_ie.children[name_unicode] = inv_entry
1745
1504
        self._inventory = inv
1755
1514
            return None
1756
1515
        parent_index = self._get_parent_index()
1757
1516
        last_changed_revision = entry[1][parent_index][4]
1758
 
        try:
1759
 
            rev = self._repository.get_revision(last_changed_revision)
1760
 
        except errors.NoSuchRevision:
1761
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1762
 
        return rev.timestamp
 
1517
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1518
 
1764
1519
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1520
        entry = self._get_entry(file_id=file_id, path=path)
1769
1524
            return parent_details[1]
1770
1525
        return None
1771
1526
 
 
1527
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1528
    def get_weave(self, file_id):
 
1529
        return self._get_weave(file_id)
 
1530
 
 
1531
    def _get_weave(self, file_id):
 
1532
        return self._repository.weave_store.get_weave(file_id,
 
1533
                self._repository.get_transaction())
 
1534
 
1772
1535
    def get_file(self, file_id, path=None):
1773
1536
        return StringIO(self.get_file_text(file_id))
1774
1537
 
 
1538
    def get_file_lines(self, file_id):
 
1539
        entry = self._get_entry(file_id=file_id)[1]
 
1540
        if entry == None:
 
1541
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1542
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1543
 
1775
1544
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1545
        return self.inventory[file_id].text_size
1778
1546
 
1779
 
    def get_file_text(self, file_id, path=None):
1780
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1781
 
        return ''.join(content)
 
1547
    def get_file_text(self, file_id):
 
1548
        return ''.join(self.get_file_lines(file_id))
1782
1549
 
1783
1550
    def get_reference_revision(self, file_id, path=None):
1784
1551
        return self.inventory[file_id].reference_revision
1803
1570
        if entry[1][parent_index][0] != 'l':
1804
1571
            return None
1805
1572
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1573
            # At present, none of the tree implementations supports non-ascii
 
1574
            # symlink targets. So we will just assume that the dirstate path is
 
1575
            # correct.
 
1576
            return entry[1][parent_index][1]
1809
1577
 
1810
1578
    def get_revision_id(self):
1811
1579
        """Return the revision id for this tree."""
1830
1598
 
1831
1599
    def kind(self, file_id):
1832
1600
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
 
1601
        if entry == None:
1834
1602
            raise errors.NoSuchId(tree=self, file_id=file_id)
1835
 
        parent_index = self._get_parent_index()
1836
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1837
 
 
1838
 
    def stored_kind(self, file_id):
1839
 
        """See Tree.stored_kind"""
1840
 
        return self.kind(file_id)
 
1603
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1841
1604
 
1842
1605
    def path_content_summary(self, path):
1843
1606
        """See Tree.path_content_summary."""
1859
1622
            return None
1860
1623
        return ie.executable
1861
1624
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1625
    def list_files(self, include_root=False):
1863
1626
        # We use a standard implementation, because DirStateRevisionTree is
1864
1627
        # dealing with one of the parents of the current state
1865
1628
        inv = self._get_inventory()
1866
 
        if from_dir is None:
1867
 
            from_dir_id = None
1868
 
        else:
1869
 
            from_dir_id = inv.path2id(from_dir)
1870
 
            if from_dir_id is None:
1871
 
                # Directory not versioned
1872
 
                return
1873
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1874
 
        if inv.root is not None and not include_root and from_dir is None:
 
1629
        entries = inv.iter_entries()
 
1630
        if self.inventory.root is not None and not include_root:
1875
1631
            entries.next()
1876
1632
        for path, entry in entries:
1877
1633
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1666
                self._dirstate_locked = False
1911
1667
            self._repository.unlock()
1912
1668
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1669
    def walkdirs(self, prefix=""):
1918
1670
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
 
1671
        # implementation based on an inventory.  
1920
1672
        # This should be cleaned up to use the much faster Dirstate code
1921
1673
        # So for now, we just build up the parent inventory, and extract
1922
1674
        # it the same way RevisionTree does.
1951
1703
 
1952
1704
class InterDirStateTree(InterTree):
1953
1705
    """Fast path optimiser for changes_from with dirstate trees.
1954
 
 
1955
 
    This is used only when both trees are in the dirstate working file, and
1956
 
    the source is any parent within the dirstate, and the destination is
 
1706
    
 
1707
    This is used only when both trees are in the dirstate working file, and 
 
1708
    the source is any parent within the dirstate, and the destination is 
1957
1709
    the current working tree of the same dirstate.
1958
1710
    """
1959
1711
    # this could be generalized to allow comparisons between any trees in the
1972
1724
        target.set_parent_ids([revid])
1973
1725
        return target.basis_tree(), target
1974
1726
 
1975
 
    @classmethod
1976
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1977
 
        result = klass.make_source_parent_tree(source, target)
1978
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1979
 
        return result
1980
 
 
1981
 
    @classmethod
1982
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
1983
 
                                                  target):
1984
 
        from bzrlib.tests.test__dirstate_helpers import \
1985
 
            compiled_dirstate_helpers_feature
1986
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
1987
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
1988
 
        result = klass.make_source_parent_tree(source, target)
1989
 
        result[1]._iter_changes = ProcessEntryC
1990
 
        return result
1991
 
 
1992
1727
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1728
    _matching_to_tree_format = WorkingTreeFormat4()
1994
 
 
1995
 
    @classmethod
1996
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
1997
 
        # This method shouldn't be called, because we have python and C
1998
 
        # specific flavours.
1999
 
        raise NotImplementedError
2000
 
 
2001
 
    def iter_changes(self, include_unchanged=False,
 
1729
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1730
 
 
1731
    def _iter_changes(self, include_unchanged=False,
2002
1732
                      specific_files=None, pb=None, extra_trees=[],
2003
1733
                      require_versioned=True, want_unversioned=False):
2004
1734
        """Return the changes from source to target.
2005
1735
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1736
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1737
            for details.
2008
1738
        :param specific_files: An optional list of file paths to restrict the
2009
1739
            comparison to. When mapping filenames to ids, all matches in all
2020
1750
            output. An unversioned file is defined as one with (False, False)
2021
1751
            for the versioned pair.
2022
1752
        """
 
1753
        utf8_decode = cache_utf8._utf8_decode
 
1754
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1755
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1756
        # NB: show_status depends on being able to pass in non-versioned files
 
1757
        # and report them as unknown
2023
1758
        # TODO: handle extra trees in the dirstate.
2024
1759
        if (extra_trees or specific_files == []):
2025
1760
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1761
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1762
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1763
                require_versioned, want_unversioned=want_unversioned):
 
1764
                yield f
 
1765
            return
2029
1766
        parent_ids = self.target.get_parent_ids()
2030
 
        if not (self.source._revision_id in parent_ids
2031
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2032
 
            raise AssertionError(
2033
 
                "revision {%s} is not stored in {%s}, but %s "
2034
 
                "can only be used for trees stored in the dirstate"
2035
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1767
        assert (self.source._revision_id in parent_ids
 
1768
                or self.source._revision_id == NULL_REVISION), \
 
1769
                "revision {%s} is not stored in {%s}, but %s " \
 
1770
                "can only be used for trees stored in the dirstate" \
 
1771
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1772
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1773
        if self.source._revision_id == NULL_REVISION:
2038
1774
            source_index = None
2039
1775
            indices = (target_index,)
2040
1776
        else:
2041
 
            if not (self.source._revision_id in parent_ids):
2042
 
                raise AssertionError(
2043
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2044
 
                    self.source._revision_id, parent_ids))
 
1777
            assert (self.source._revision_id in parent_ids), \
 
1778
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1779
                self.source._revision_id, parent_ids)
2045
1780
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
1781
            indices = (source_index, target_index)
2047
1782
        # -- make all specific_files utf8 --
2048
1783
        if specific_files:
2049
1784
            specific_files_utf8 = set()
2050
1785
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1786
                specific_files_utf8.add(path.encode('utf8'))
2054
1787
            specific_files = specific_files_utf8
2055
1788
        else:
2056
1789
            specific_files = set([''])
2057
1790
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1791
        # -- get the state object and prepare it.
2060
1792
        state = self.target.current_dirstate()
2061
1793
        state._read_dirblocks_if_needed()
 
1794
        def _entries_for_path(path):
 
1795
            """Return a list with all the entries that match path for all ids.
 
1796
            """
 
1797
            dirname, basename = os.path.split(path)
 
1798
            key = (dirname, basename, '')
 
1799
            block_index, present = state._find_block_index_from_key(key)
 
1800
            if not present:
 
1801
                # the block which should contain path is absent.
 
1802
                return []
 
1803
            result = []
 
1804
            block = state._dirblocks[block_index][1]
 
1805
            entry_index, _ = state._find_entry_index(key, block)
 
1806
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1807
            while (entry_index < len(block) and
 
1808
                block[entry_index][0][0:2] == key[0:2]):
 
1809
                result.append(block[entry_index])
 
1810
                entry_index += 1
 
1811
            return result
2062
1812
        if require_versioned:
2063
1813
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1814
            all_versioned = True
2065
1815
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1816
                path_entries = _entries_for_path(path)
2067
1817
                if not path_entries:
2068
1818
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1819
                    all_versioned = False
 
1820
                    break
2071
1821
                found_versioned = False
2072
1822
                # for each id at this path
2073
1823
                for entry in path_entries:
2080
1830
                if not found_versioned:
2081
1831
                    # none of the indexes was not 'absent' at all ids for this
2082
1832
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1833
                    all_versioned = False
 
1834
                    break
 
1835
            if not all_versioned:
 
1836
                raise errors.PathsNotVersionedError(specific_files)
2086
1837
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1838
        search_specific_files = set()
 
1839
        for path in specific_files:
 
1840
            other_specific_files = specific_files.difference(set([path]))
 
1841
            if not osutils.is_inside_any(other_specific_files, path):
 
1842
                # this is a top level path, we must check it.
 
1843
                search_specific_files.add(path)
 
1844
        # sketch: 
 
1845
        # compare source_index and target_index at or under each element of search_specific_files.
 
1846
        # follow the following comparison table. Note that we only want to do diff operations when
 
1847
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1848
        # for the target.
 
1849
        # cases:
 
1850
        # 
 
1851
        # Source | Target | disk | action
 
1852
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1853
        #        |        |      | diff check on source-target
 
1854
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1855
        #        |        |      | ???
 
1856
        #   r    |  a     |      | add source to search
 
1857
        #   r    |  a     |  a   | 
 
1858
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1859
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1860
        #   a    | fdlt   |      | add new id
 
1861
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1862
        #   a    |  a     |      | not present in either tree, skip
 
1863
        #   a    |  a     |  a   | not present in any tree, skip
 
1864
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1865
        #        |        |      | may not be selected by the users list of paths.
 
1866
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1867
        #        |        |      | may not be selected by the users list of paths.
 
1868
        #  fdlt  | fdlt   |      | content in both: diff them
 
1869
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1870
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1871
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1872
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1873
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1874
        #        |        |      | this id at the other path.
 
1875
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1876
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1877
        #        |        |      | this id at the other path.
 
1878
 
 
1879
        # for all search_indexs in each path at or under each element of
 
1880
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1881
        # relocated path as one to search if its not searched already. If the
 
1882
        # detail is not relocated, add the id.
 
1883
        searched_specific_files = set()
 
1884
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1885
        # Using a list so that we can access the values and change them in
 
1886
        # nested scope. Each one is [path, file_id, entry]
 
1887
        last_source_parent = [None, None]
 
1888
        last_target_parent = [None, None]
2088
1889
 
2089
1890
        use_filesystem_for_exec = (sys.platform != 'win32')
2090
 
        iter_changes = self.target._iter_changes(include_unchanged,
2091
 
            use_filesystem_for_exec, search_specific_files, state,
2092
 
            source_index, target_index, want_unversioned, self.target)
2093
 
        return iter_changes.iter_changes()
 
1891
 
 
1892
        # Just a sentry, so that _process_entry can say that this
 
1893
        # record is handled, but isn't interesting to process (unchanged)
 
1894
        uninteresting = object()
 
1895
 
 
1896
 
 
1897
        old_dirname_to_file_id = {}
 
1898
        new_dirname_to_file_id = {}
 
1899
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1900
        #       keeping a cache of directories that we have seen.
 
1901
 
 
1902
        def _process_entry(entry, path_info):
 
1903
            """Compare an entry and real disk to generate delta information.
 
1904
 
 
1905
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1906
                the path of entry. If None, then the path is considered absent.
 
1907
                (Perhaps we should pass in a concrete entry for this ?)
 
1908
                Basename is returned as a utf8 string because we expect this
 
1909
                tuple will be ignored, and don't want to take the time to
 
1910
                decode.
 
1911
            :return: None if these don't match
 
1912
                     A tuple of information about the change, or
 
1913
                     the object 'uninteresting' if these match, but are
 
1914
                     basically identical.
 
1915
            """
 
1916
            if source_index is None:
 
1917
                source_details = NULL_PARENT_DETAILS
 
1918
            else:
 
1919
                source_details = entry[1][source_index]
 
1920
            target_details = entry[1][target_index]
 
1921
            target_minikind = target_details[0]
 
1922
            if path_info is not None and target_minikind in 'fdlt':
 
1923
                assert target_index == 0
 
1924
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1925
                                                  stat_value=path_info[3])
 
1926
                # The entry may have been modified by update_entry
 
1927
                target_details = entry[1][target_index]
 
1928
                target_minikind = target_details[0]
 
1929
            else:
 
1930
                link_or_sha1 = None
 
1931
            file_id = entry[0][2]
 
1932
            source_minikind = source_details[0]
 
1933
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1934
                # claimed content in both: diff
 
1935
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1936
                #        |        |      | diff check on source-target
 
1937
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1938
                #        |        |      | ???
 
1939
                if source_minikind in 'r':
 
1940
                    # add the source to the search path to find any children it
 
1941
                    # has.  TODO ? : only add if it is a container ?
 
1942
                    if not osutils.is_inside_any(searched_specific_files,
 
1943
                                                 source_details[1]):
 
1944
                        search_specific_files.add(source_details[1])
 
1945
                    # generate the old path; this is needed for stating later
 
1946
                    # as well.
 
1947
                    old_path = source_details[1]
 
1948
                    old_dirname, old_basename = os.path.split(old_path)
 
1949
                    path = pathjoin(entry[0][0], entry[0][1])
 
1950
                    old_entry = state._get_entry(source_index,
 
1951
                                                 path_utf8=old_path)
 
1952
                    # update the source details variable to be the real
 
1953
                    # location.
 
1954
                    source_details = old_entry[1][source_index]
 
1955
                    source_minikind = source_details[0]
 
1956
                else:
 
1957
                    old_dirname = entry[0][0]
 
1958
                    old_basename = entry[0][1]
 
1959
                    old_path = path = None
 
1960
                if path_info is None:
 
1961
                    # the file is missing on disk, show as removed.
 
1962
                    content_change = True
 
1963
                    target_kind = None
 
1964
                    target_exec = False
 
1965
                else:
 
1966
                    # source and target are both versioned and disk file is present.
 
1967
                    target_kind = path_info[2]
 
1968
                    if target_kind == 'directory':
 
1969
                        if path is None:
 
1970
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1971
                        new_dirname_to_file_id[path] = file_id
 
1972
                        if source_minikind != 'd':
 
1973
                            content_change = True
 
1974
                        else:
 
1975
                            # directories have no fingerprint
 
1976
                            content_change = False
 
1977
                        target_exec = False
 
1978
                    elif target_kind == 'file':
 
1979
                        if source_minikind != 'f':
 
1980
                            content_change = True
 
1981
                        else:
 
1982
                            # We could check the size, but we already have the
 
1983
                            # sha1 hash.
 
1984
                            content_change = (link_or_sha1 != source_details[1])
 
1985
                        # Target details is updated at update_entry time
 
1986
                        if use_filesystem_for_exec:
 
1987
                            # We don't need S_ISREG here, because we are sure
 
1988
                            # we are dealing with a file.
 
1989
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1990
                        else:
 
1991
                            target_exec = target_details[3]
 
1992
                    elif target_kind == 'symlink':
 
1993
                        if source_minikind != 'l':
 
1994
                            content_change = True
 
1995
                        else:
 
1996
                            content_change = (link_or_sha1 != source_details[1])
 
1997
                        target_exec = False
 
1998
                    elif target_kind == 'tree-reference':
 
1999
                        if source_minikind != 't':
 
2000
                            content_change = True
 
2001
                        else:
 
2002
                            content_change = False
 
2003
                        target_exec = False
 
2004
                    else:
 
2005
                        raise Exception, "unknown kind %s" % path_info[2]
 
2006
                if source_minikind == 'd':
 
2007
                    if path is None:
 
2008
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2009
                    old_dirname_to_file_id[old_path] = file_id
 
2010
                # parent id is the entry for the path in the target tree
 
2011
                if old_dirname == last_source_parent[0]:
 
2012
                    source_parent_id = last_source_parent[1]
 
2013
                else:
 
2014
                    try:
 
2015
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2016
                    except KeyError:
 
2017
                        source_parent_entry = state._get_entry(source_index,
 
2018
                                                               path_utf8=old_dirname)
 
2019
                        source_parent_id = source_parent_entry[0][2]
 
2020
                    if source_parent_id == entry[0][2]:
 
2021
                        # This is the root, so the parent is None
 
2022
                        source_parent_id = None
 
2023
                    else:
 
2024
                        last_source_parent[0] = old_dirname
 
2025
                        last_source_parent[1] = source_parent_id
 
2026
                new_dirname = entry[0][0]
 
2027
                if new_dirname == last_target_parent[0]:
 
2028
                    target_parent_id = last_target_parent[1]
 
2029
                else:
 
2030
                    try:
 
2031
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2032
                    except KeyError:
 
2033
                        # TODO: We don't always need to do the lookup, because the
 
2034
                        #       parent entry will be the same as the source entry.
 
2035
                        target_parent_entry = state._get_entry(target_index,
 
2036
                                                               path_utf8=new_dirname)
 
2037
                        assert target_parent_entry != (None, None), (
 
2038
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2039
                            % (new_dirname, entry))
 
2040
                        target_parent_id = target_parent_entry[0][2]
 
2041
                    if target_parent_id == entry[0][2]:
 
2042
                        # This is the root, so the parent is None
 
2043
                        target_parent_id = None
 
2044
                    else:
 
2045
                        last_target_parent[0] = new_dirname
 
2046
                        last_target_parent[1] = target_parent_id
 
2047
 
 
2048
                source_exec = source_details[3]
 
2049
                if (include_unchanged
 
2050
                    or content_change
 
2051
                    or source_parent_id != target_parent_id
 
2052
                    or old_basename != entry[0][1]
 
2053
                    or source_exec != target_exec
 
2054
                    ):
 
2055
                    if old_path is None:
 
2056
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2057
                        old_path_u = utf8_decode(old_path)[0]
 
2058
                        path_u = old_path_u
 
2059
                    else:
 
2060
                        old_path_u = utf8_decode(old_path)[0]
 
2061
                        if old_path == path:
 
2062
                            path_u = old_path_u
 
2063
                        else:
 
2064
                            path_u = utf8_decode(path)[0]
 
2065
                    source_kind = _minikind_to_kind[source_minikind]
 
2066
                    return (entry[0][2],
 
2067
                           (old_path_u, path_u),
 
2068
                           content_change,
 
2069
                           (True, True),
 
2070
                           (source_parent_id, target_parent_id),
 
2071
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2072
                           (source_kind, target_kind),
 
2073
                           (source_exec, target_exec))
 
2074
                else:
 
2075
                    return uninteresting
 
2076
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2077
                # looks like a new file
 
2078
                if path_info is not None:
 
2079
                    path = pathjoin(entry[0][0], entry[0][1])
 
2080
                    # parent id is the entry for the path in the target tree
 
2081
                    # TODO: these are the same for an entire directory: cache em.
 
2082
                    parent_id = state._get_entry(target_index,
 
2083
                                                 path_utf8=entry[0][0])[0][2]
 
2084
                    if parent_id == entry[0][2]:
 
2085
                        parent_id = None
 
2086
                    if use_filesystem_for_exec:
 
2087
                        # We need S_ISREG here, because we aren't sure if this
 
2088
                        # is a file or not.
 
2089
                        target_exec = bool(
 
2090
                            stat.S_ISREG(path_info[3].st_mode)
 
2091
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2092
                    else:
 
2093
                        target_exec = target_details[3]
 
2094
                    return (entry[0][2],
 
2095
                           (None, utf8_decode(path)[0]),
 
2096
                           True,
 
2097
                           (False, True),
 
2098
                           (None, parent_id),
 
2099
                           (None, utf8_decode(entry[0][1])[0]),
 
2100
                           (None, path_info[2]),
 
2101
                           (None, target_exec))
 
2102
                else:
 
2103
                    # but its not on disk: we deliberately treat this as just
 
2104
                    # never-present. (Why ?! - RBC 20070224)
 
2105
                    pass
 
2106
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2107
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2108
                # if its still on disk, *and* theres no other entry at this
 
2109
                # path [we dont know this in this routine at the moment -
 
2110
                # perhaps we should change this - then it would be an unknown.
 
2111
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2112
                # parent id is the entry for the path in the target tree
 
2113
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2114
                if parent_id == entry[0][2]:
 
2115
                    parent_id = None
 
2116
                return (entry[0][2],
 
2117
                       (utf8_decode(old_path)[0], None),
 
2118
                       True,
 
2119
                       (True, False),
 
2120
                       (parent_id, None),
 
2121
                       (utf8_decode(entry[0][1])[0], None),
 
2122
                       (_minikind_to_kind[source_minikind], None),
 
2123
                       (source_details[3], None))
 
2124
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2125
                # a rename; could be a true rename, or a rename inherited from
 
2126
                # a renamed parent. TODO: handle this efficiently. Its not
 
2127
                # common case to rename dirs though, so a correct but slow
 
2128
                # implementation will do.
 
2129
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2130
                    search_specific_files.add(target_details[1])
 
2131
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2132
                # neither of the selected trees contain this file,
 
2133
                # so skip over it. This is not currently directly tested, but
 
2134
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2135
                pass
 
2136
            else:
 
2137
                raise AssertionError("don't know how to compare "
 
2138
                    "source_minikind=%r, target_minikind=%r"
 
2139
                    % (source_minikind, target_minikind))
 
2140
                ## import pdb;pdb.set_trace()
 
2141
            return None
 
2142
 
 
2143
        while search_specific_files:
 
2144
            # TODO: the pending list should be lexically sorted?  the
 
2145
            # interface doesn't require it.
 
2146
            current_root = search_specific_files.pop()
 
2147
            current_root_unicode = current_root.decode('utf8')
 
2148
            searched_specific_files.add(current_root)
 
2149
            # process the entries for this containing directory: the rest will be
 
2150
            # found by their parents recursively.
 
2151
            root_entries = _entries_for_path(current_root)
 
2152
            root_abspath = self.target.abspath(current_root_unicode)
 
2153
            try:
 
2154
                root_stat = os.lstat(root_abspath)
 
2155
            except OSError, e:
 
2156
                if e.errno == errno.ENOENT:
 
2157
                    # the path does not exist: let _process_entry know that.
 
2158
                    root_dir_info = None
 
2159
                else:
 
2160
                    # some other random error: hand it up.
 
2161
                    raise
 
2162
            else:
 
2163
                root_dir_info = ('', current_root,
 
2164
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2165
                    root_abspath)
 
2166
                if root_dir_info[2] == 'directory':
 
2167
                    if self.target._directory_is_tree_reference(
 
2168
                        current_root.decode('utf8')):
 
2169
                        root_dir_info = root_dir_info[:2] + \
 
2170
                            ('tree-reference',) + root_dir_info[3:]
 
2171
 
 
2172
            if not root_entries and not root_dir_info:
 
2173
                # this specified path is not present at all, skip it.
 
2174
                continue
 
2175
            path_handled = False
 
2176
            for entry in root_entries:
 
2177
                result = _process_entry(entry, root_dir_info)
 
2178
                if result is not None:
 
2179
                    path_handled = True
 
2180
                    if result is not uninteresting:
 
2181
                        yield result
 
2182
            if want_unversioned and not path_handled and root_dir_info:
 
2183
                new_executable = bool(
 
2184
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2185
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2186
                yield (None,
 
2187
                       (None, current_root_unicode),
 
2188
                       True,
 
2189
                       (False, False),
 
2190
                       (None, None),
 
2191
                       (None, splitpath(current_root_unicode)[-1]),
 
2192
                       (None, root_dir_info[2]),
 
2193
                       (None, new_executable)
 
2194
                      )
 
2195
            initial_key = (current_root, '', '')
 
2196
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2197
            if block_index == 0:
 
2198
                # we have processed the total root already, but because the
 
2199
                # initial key matched it we should skip it here.
 
2200
                block_index +=1
 
2201
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2202
                current_dir_info = None
 
2203
            else:
 
2204
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2205
                try:
 
2206
                    current_dir_info = dir_iterator.next()
 
2207
                except OSError, e:
 
2208
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2209
                    # python 2.5 has e.errno == EINVAL,
 
2210
                    #            and e.winerror == ERROR_DIRECTORY
 
2211
                    e_winerror = getattr(e, 'winerror', None)
 
2212
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2213
                    # there may be directories in the inventory even though
 
2214
                    # this path is not a file on disk: so mark it as end of
 
2215
                    # iterator
 
2216
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2217
                        current_dir_info = None
 
2218
                    elif (sys.platform == 'win32'
 
2219
                          and (e.errno in win_errors
 
2220
                               or e_winerror in win_errors)):
 
2221
                        current_dir_info = None
 
2222
                    else:
 
2223
                        raise
 
2224
                else:
 
2225
                    if current_dir_info[0][0] == '':
 
2226
                        # remove .bzr from iteration
 
2227
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2228
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2229
                        del current_dir_info[1][bzr_index]
 
2230
            # walk until both the directory listing and the versioned metadata
 
2231
            # are exhausted. 
 
2232
            if (block_index < len(state._dirblocks) and
 
2233
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2234
                current_block = state._dirblocks[block_index]
 
2235
            else:
 
2236
                current_block = None
 
2237
            while (current_dir_info is not None or
 
2238
                   current_block is not None):
 
2239
                if (current_dir_info and current_block
 
2240
                    and current_dir_info[0][0] != current_block[0]):
 
2241
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2242
                        # filesystem data refers to paths not covered by the dirblock.
 
2243
                        # this has two possibilities:
 
2244
                        # A) it is versioned but empty, so there is no block for it
 
2245
                        # B) it is not versioned.
 
2246
 
 
2247
                        # if (A) then we need to recurse into it to check for
 
2248
                        # new unknown files or directories.
 
2249
                        # if (B) then we should ignore it, because we don't
 
2250
                        # recurse into unknown directories.
 
2251
                        path_index = 0
 
2252
                        while path_index < len(current_dir_info[1]):
 
2253
                                current_path_info = current_dir_info[1][path_index]
 
2254
                                if want_unversioned:
 
2255
                                    if current_path_info[2] == 'directory':
 
2256
                                        if self.target._directory_is_tree_reference(
 
2257
                                            current_path_info[0].decode('utf8')):
 
2258
                                            current_path_info = current_path_info[:2] + \
 
2259
                                                ('tree-reference',) + current_path_info[3:]
 
2260
                                    new_executable = bool(
 
2261
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2262
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2263
                                    yield (None,
 
2264
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2265
                                        True,
 
2266
                                        (False, False),
 
2267
                                        (None, None),
 
2268
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2269
                                        (None, current_path_info[2]),
 
2270
                                        (None, new_executable))
 
2271
                                # dont descend into this unversioned path if it is
 
2272
                                # a dir
 
2273
                                if current_path_info[2] in ('directory',
 
2274
                                                            'tree-reference'):
 
2275
                                    del current_dir_info[1][path_index]
 
2276
                                    path_index -= 1
 
2277
                                path_index += 1
 
2278
 
 
2279
                        # This dir info has been handled, go to the next
 
2280
                        try:
 
2281
                            current_dir_info = dir_iterator.next()
 
2282
                        except StopIteration:
 
2283
                            current_dir_info = None
 
2284
                    else:
 
2285
                        # We have a dirblock entry for this location, but there
 
2286
                        # is no filesystem path for this. This is most likely
 
2287
                        # because a directory was removed from the disk.
 
2288
                        # We don't have to report the missing directory,
 
2289
                        # because that should have already been handled, but we
 
2290
                        # need to handle all of the files that are contained
 
2291
                        # within.
 
2292
                        for current_entry in current_block[1]:
 
2293
                            # entry referring to file not present on disk.
 
2294
                            # advance the entry only, after processing.
 
2295
                            result = _process_entry(current_entry, None)
 
2296
                            if result is not None:
 
2297
                                if result is not uninteresting:
 
2298
                                    yield result
 
2299
                        block_index +=1
 
2300
                        if (block_index < len(state._dirblocks) and
 
2301
                            osutils.is_inside(current_root,
 
2302
                                              state._dirblocks[block_index][0])):
 
2303
                            current_block = state._dirblocks[block_index]
 
2304
                        else:
 
2305
                            current_block = None
 
2306
                    continue
 
2307
                entry_index = 0
 
2308
                if current_block and entry_index < len(current_block[1]):
 
2309
                    current_entry = current_block[1][entry_index]
 
2310
                else:
 
2311
                    current_entry = None
 
2312
                advance_entry = True
 
2313
                path_index = 0
 
2314
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2315
                    current_path_info = current_dir_info[1][path_index]
 
2316
                    if current_path_info[2] == 'directory':
 
2317
                        if self.target._directory_is_tree_reference(
 
2318
                            current_path_info[0].decode('utf8')):
 
2319
                            current_path_info = current_path_info[:2] + \
 
2320
                                ('tree-reference',) + current_path_info[3:]
 
2321
                else:
 
2322
                    current_path_info = None
 
2323
                advance_path = True
 
2324
                path_handled = False
 
2325
                while (current_entry is not None or
 
2326
                    current_path_info is not None):
 
2327
                    if current_entry is None:
 
2328
                        # the check for path_handled when the path is adnvaced
 
2329
                        # will yield this path if needed.
 
2330
                        pass
 
2331
                    elif current_path_info is None:
 
2332
                        # no path is fine: the per entry code will handle it.
 
2333
                        result = _process_entry(current_entry, current_path_info)
 
2334
                        if result is not None:
 
2335
                            if result is not uninteresting:
 
2336
                                yield result
 
2337
                    elif (current_entry[0][1] != current_path_info[1]
 
2338
                          or current_entry[1][target_index][0] in 'ar'):
 
2339
                        # The current path on disk doesn't match the dirblock
 
2340
                        # record. Either the dirblock is marked as absent, or
 
2341
                        # the file on disk is not present at all in the
 
2342
                        # dirblock. Either way, report about the dirblock
 
2343
                        # entry, and let other code handle the filesystem one.
 
2344
 
 
2345
                        # Compare the basename for these files to determine
 
2346
                        # which comes first
 
2347
                        if current_path_info[1] < current_entry[0][1]:
 
2348
                            # extra file on disk: pass for now, but only
 
2349
                            # increment the path, not the entry
 
2350
                            advance_entry = False
 
2351
                        else:
 
2352
                            # entry referring to file not present on disk.
 
2353
                            # advance the entry only, after processing.
 
2354
                            result = _process_entry(current_entry, None)
 
2355
                            if result is not None:
 
2356
                                if result is not uninteresting:
 
2357
                                    yield result
 
2358
                            advance_path = False
 
2359
                    else:
 
2360
                        result = _process_entry(current_entry, current_path_info)
 
2361
                        if result is not None:
 
2362
                            path_handled = True
 
2363
                            if result is not uninteresting:
 
2364
                                yield result
 
2365
                    if advance_entry and current_entry is not None:
 
2366
                        entry_index += 1
 
2367
                        if entry_index < len(current_block[1]):
 
2368
                            current_entry = current_block[1][entry_index]
 
2369
                        else:
 
2370
                            current_entry = None
 
2371
                    else:
 
2372
                        advance_entry = True # reset the advance flaga
 
2373
                    if advance_path and current_path_info is not None:
 
2374
                        if not path_handled:
 
2375
                            # unversioned in all regards
 
2376
                            if want_unversioned:
 
2377
                                new_executable = bool(
 
2378
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2379
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2380
                                yield (None,
 
2381
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2382
                                    True,
 
2383
                                    (False, False),
 
2384
                                    (None, None),
 
2385
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2386
                                    (None, current_path_info[2]),
 
2387
                                    (None, new_executable))
 
2388
                            # dont descend into this unversioned path if it is
 
2389
                            # a dir
 
2390
                            if current_path_info[2] in ('directory'):
 
2391
                                del current_dir_info[1][path_index]
 
2392
                                path_index -= 1
 
2393
                        # dont descend the disk iterator into any tree 
 
2394
                        # paths.
 
2395
                        if current_path_info[2] == 'tree-reference':
 
2396
                            del current_dir_info[1][path_index]
 
2397
                            path_index -= 1
 
2398
                        path_index += 1
 
2399
                        if path_index < len(current_dir_info[1]):
 
2400
                            current_path_info = current_dir_info[1][path_index]
 
2401
                            if current_path_info[2] == 'directory':
 
2402
                                if self.target._directory_is_tree_reference(
 
2403
                                    current_path_info[0].decode('utf8')):
 
2404
                                    current_path_info = current_path_info[:2] + \
 
2405
                                        ('tree-reference',) + current_path_info[3:]
 
2406
                        else:
 
2407
                            current_path_info = None
 
2408
                        path_handled = False
 
2409
                    else:
 
2410
                        advance_path = True # reset the advance flagg.
 
2411
                if current_block is not None:
 
2412
                    block_index += 1
 
2413
                    if (block_index < len(state._dirblocks) and
 
2414
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2415
                        current_block = state._dirblocks[block_index]
 
2416
                    else:
 
2417
                        current_block = None
 
2418
                if current_dir_info is not None:
 
2419
                    try:
 
2420
                        current_dir_info = dir_iterator.next()
 
2421
                    except StopIteration:
 
2422
                        current_dir_info = None
 
2423
 
2094
2424
 
2095
2425
    @staticmethod
2096
2426
    def is_compatible(source, target):
2097
2427
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2428
        if not isinstance(target, WorkingTree4):
2099
2429
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2430
        # the source must be a revtreee or dirstate rev tree.
2101
2431
        if not isinstance(source,
2102
2432
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2433
            return False
2104
2434
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2435
        if not (source._revision_id == NULL_REVISION or
2106
2436
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2437
            # TODO: what about ghosts? it may well need to 
2108
2438
            # check for them explicitly.
2109
2439
            return False
2110
2440
        return True
2120
2450
 
2121
2451
    def convert(self, tree):
2122
2452
        # lock the control files not the tree, so that we dont get tree
2123
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2453
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2454
        # tree during upgrade.
2125
2455
        tree._control_files.lock_write()
2126
2456
        try:
2152
2482
 
2153
2483
    def update_format(self, tree):
2154
2484
        """Change the format marker."""
2155
 
        tree._transport.put_bytes('format',
2156
 
            self.target_format.get_format_string(),
2157
 
            mode=tree.bzrdir._get_file_mode())
2158
 
 
2159
 
 
2160
 
class Converter4to5(object):
2161
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2162
 
 
2163
 
    def __init__(self):
2164
 
        self.target_format = WorkingTreeFormat5()
2165
 
 
2166
 
    def convert(self, tree):
2167
 
        # lock the control files not the tree, so that we don't get tree
2168
 
        # on-unlock behaviours, and so that no-one else diddles with the
2169
 
        # tree during upgrade.
2170
 
        tree._control_files.lock_write()
2171
 
        try:
2172
 
            self.update_format(tree)
2173
 
        finally:
2174
 
            tree._control_files.unlock()
2175
 
 
2176
 
    def update_format(self, tree):
2177
 
        """Change the format marker."""
2178
 
        tree._transport.put_bytes('format',
2179
 
            self.target_format.get_format_string(),
2180
 
            mode=tree.bzrdir._get_file_mode())
2181
 
 
2182
 
 
2183
 
class Converter4or5to6(object):
2184
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2185
 
 
2186
 
    def __init__(self):
2187
 
        self.target_format = WorkingTreeFormat6()
2188
 
 
2189
 
    def convert(self, tree):
2190
 
        # lock the control files not the tree, so that we don't get tree
2191
 
        # on-unlock behaviours, and so that no-one else diddles with the
2192
 
        # tree during upgrade.
2193
 
        tree._control_files.lock_write()
2194
 
        try:
2195
 
            self.init_custom_control_files(tree)
2196
 
            self.update_format(tree)
2197
 
        finally:
2198
 
            tree._control_files.unlock()
2199
 
 
2200
 
    def init_custom_control_files(self, tree):
2201
 
        """Initialize custom control files."""
2202
 
        tree._transport.put_bytes('views', '',
2203
 
            mode=tree.bzrdir._get_file_mode())
2204
 
 
2205
 
    def update_format(self, tree):
2206
 
        """Change the format marker."""
2207
 
        tree._transport.put_bytes('format',
2208
 
            self.target_format.get_format_string(),
2209
 
            mode=tree.bzrdir._get_file_mode())
 
2485
        tree._control_files.put_utf8('format',
 
2486
            self.target_format.get_format_string())