/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: Aaron Bentley
  • Date: 2007-12-20 19:16:02 UTC
  • mto: This revision was merged to the branch mainline in revision 3235.
  • Revision ID: abentley@panoramicfeedback.com-20071220191602-q1p5o3svshujfodd
Add support for hardlinks to TreeTransform

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."
474
494
 
475
495
            Note: The caller is expected to take a read-lock before calling this.
476
496
            """
477
 
            self._must_be_locked()
478
497
            if not path:
479
498
                path = self.id2path(file_id)
480
499
            mode = os.lstat(self.abspath(path)).st_mode
481
500
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
501
 
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
502
    @needs_read_lock
494
503
    def __iter__(self):
495
504
        """Iterate through file_ids for this tree.
508
517
        return iter(result)
509
518
 
510
519
    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
520
        for key, tree_details in self.current_dirstate()._iter_entries():
516
521
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
522
                # not relevant to the working tree
519
524
            if not key[1]:
520
525
                # the root is not a reference.
521
526
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
527
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
528
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
529
                if self._kind(path) == 'tree-reference':
 
530
                    yield path, key[2]
526
531
            except errors.NoSuchFile:
527
532
                # path is missing on disk.
528
533
                continue
529
534
 
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
535
    def kind(self, file_id):
537
536
        """Return the kind of a file.
538
537
 
542
541
        Note: The caller is expected to take a read-lock before calling this.
543
542
        """
544
543
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
544
        assert relpath != None, \
 
545
            "path for id {%s} is None!" % file_id
548
546
        return self._kind(relpath)
549
547
 
550
548
    def _kind(self, relpath):
551
549
        abspath = self.abspath(relpath)
552
550
        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'
 
551
        if (self._repo_supports_tree_reference and
 
552
            kind == 'directory' and
 
553
            self._directory_is_tree_reference(relpath)):
 
554
            kind = 'tree-reference'
558
555
        return kind
559
556
 
560
557
    @needs_read_lock
623
620
        result = []
624
621
        if not from_paths:
625
622
            return result
 
623
 
626
624
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
625
 
 
626
        assert not isinstance(from_paths, basestring)
629
627
        to_dir_utf8 = to_dir.encode('utf8')
630
628
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
629
        id_index = state._get_id_index()
653
651
        if self._inventory is not None:
654
652
            update_inventory = True
655
653
            inv = self.inventory
 
654
            to_dir_ie = inv[to_dir_id]
656
655
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
656
        else:
659
657
            update_inventory = False
660
658
 
691
689
            from_entry = self._get_entry(path=from_rel)
692
690
            if from_entry == (None, None):
693
691
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
692
                    errors.NotVersionedError(path=str(from_rel)))
695
693
 
696
694
            from_id = from_entry[0][2]
697
695
            to_rel = pathjoin(to_dir, from_tail)
786
784
                if minikind == 'd':
787
785
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
786
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
787
                        assert from_dir != '', "renaming root not supported"
791
788
                        from_key = (from_dir, '')
792
789
                        from_block_idx, present = \
793
790
                            state._find_block_index_from_key(from_key)
806
803
 
807
804
                        # Grab a copy since move_one may update the list.
808
805
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
806
                            assert entry[0][0] == from_dir
811
807
                            cur_details = entry[1][0]
812
808
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
809
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
872
868
        for tree in trees:
873
869
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
870
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
871
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
872
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
873
        # -- make all paths utf8 --
879
874
        paths_utf8 = set()
940
935
                raise errors.PathsNotVersionedError(paths)
941
936
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
937
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
938
        # sketch: 
944
939
        # for all search_indexs in each path at or under each element of
945
940
        # search_paths, if the detail is relocated: add the id, and add the
946
941
        # relocated path as one to search if its not searched already. If the
1002
997
 
1003
998
    def read_working_inventory(self):
1004
999
        """Read the working inventory.
1005
 
 
 
1000
        
1006
1001
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
1002
        """
1008
1003
        return self.inventory
1026
1021
    def set_last_revision(self, new_revision):
1027
1022
        """Change the last revision in the working tree."""
1028
1023
        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.")
 
1024
        if new_revision in (NULL_REVISION, None):
 
1025
            assert len(parents) < 2, (
 
1026
                "setting the last parent to none with a pending merge is "
 
1027
                "unsupported.")
1034
1028
            self.set_parent_ids([])
1035
1029
        else:
1036
1030
            self.set_parent_ids([new_revision] + parents[1:],
1039
1033
    @needs_tree_write_lock
1040
1034
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
1035
        """Set the parent ids to revision_ids.
1042
 
 
 
1036
        
1043
1037
        See also set_parent_trees. This api will try to retrieve the tree data
1044
1038
        for each element of revision_ids from the trees repository. If you have
1045
1039
        tree data already available, it is more efficient to use
1077
1071
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1072
        real_trees = []
1079
1073
        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
1074
        # convert absent trees to the null tree, which we convert back to
1087
1075
        # missing on access.
1088
1076
        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
1077
            _mod_revision.check_not_reserved_id(rev_id)
1096
1078
            if tree is not None:
1097
1079
                real_trees.append((rev_id, tree))
1098
1080
            else:
1099
1081
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1082
                    self.branch.repository.revision_tree(None)))
1102
1083
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
1104
1084
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1105
1085
        self._make_dirty(reset_inventory=False)
1106
1086
 
1199
1179
                # just forget the whole block.
1200
1180
                entry_index = 0
1201
1181
                while entry_index < len(block[1]):
 
1182
                    # Mark this file id as having been removed
1202
1183
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1184
                    ids_to_unversion.discard(entry[0][2])
 
1185
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1186
                                               # entries
 
1187
                        or not state._make_absent(entry)):
1205
1188
                        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
1189
                # go to the next block. (At the moment we dont delete empty
1213
1190
                # dirblocks)
1214
1191
                block_index += 1
1237
1214
            for file_id in file_ids:
1238
1215
                self._inventory.remove_recursive_id(file_id)
1239
1216
 
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
1217
    def update_basis_by_delta(self, new_revid, delta):
1254
1218
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
 
1219
        assert self.last_revision() != new_revid
1257
1220
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
1221
 
1259
1222
    @needs_read_lock
1263
1226
    @needs_tree_write_lock
1264
1227
    def _write_inventory(self, inv):
1265
1228
        """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:
 
1229
        assert not self._dirty, ("attempting to write an inventory when the "
 
1230
            "dirstate is dirty will cause data loss")
 
1231
        self.current_dirstate().set_state_from_inventory(inv)
 
1232
        self._make_dirty(reset_inventory=False)
 
1233
        if self._inventory is not None:
1280
1234
            self._inventory = inv
1281
1235
        self.flush()
1282
1236
 
1283
1237
 
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):
 
1238
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1239
    """The first consolidated dirstate working tree format.
 
1240
 
 
1241
    This format:
 
1242
        - exists within a metadir controlling .bzr
 
1243
        - includes an explicit version marker for the workingtree control
 
1244
          files, separate from the BzrDir format
 
1245
        - modifies the hash cache format
 
1246
        - is new in bzr 0.15
 
1247
        - uses a LockDir to guard access to it.
 
1248
    """
 
1249
 
 
1250
    upgrade_recommended = False
 
1251
 
 
1252
    def get_format_string(self):
 
1253
        """See WorkingTreeFormat.get_format_string()."""
 
1254
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1255
 
 
1256
    def get_format_description(self):
 
1257
        """See WorkingTreeFormat.get_format_description()."""
 
1258
        return "Working tree format 4"
1374
1259
 
1375
1260
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
 
                   accelerator_tree=None, hardlink=False):
 
1261
                   accelerator_tree=None):
1377
1262
        """See WorkingTreeFormat.initialize().
1378
1263
 
1379
1264
        :param revision_id: allows creating a working tree at a different
1382
1267
            contents more quickly than the revision tree, i.e. a workingtree.
1383
1268
            The revision tree will be used for cases where accelerator_tree's
1384
1269
            content is different.
1385
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1386
 
            where possible.
1387
1270
 
1388
1271
        These trees get an initial random root id, if their repository supports
1389
1272
        rich root data, TREE_ROOT otherwise.
1394
1277
        control_files = self._open_control_files(a_bzrdir)
1395
1278
        control_files.create_lock()
1396
1279
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
 
1280
        control_files.put_utf8('format', self.get_format_string())
1399
1281
        if from_branch is not None:
1400
1282
            branch = from_branch
1401
1283
        else:
1407
1289
        state = dirstate.DirState.initialize(local_path)
1408
1290
        state.unlock()
1409
1291
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1292
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1293
                         branch,
1412
1294
                         _format=self,
1413
1295
                         _bzrdir=a_bzrdir,
1415
1297
        wt._new_tree()
1416
1298
        wt.lock_tree_write()
1417
1299
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1300
            if revision_id in (None, NULL_REVISION):
1420
1301
                if branch.repository.supports_rich_root():
1421
1302
                    wt._set_root_id(generate_ids.gen_root_id())
1422
1303
                else:
1433
1314
                    pass
1434
1315
            if basis is None:
1435
1316
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1317
            if revision_id == NULL_REVISION:
1437
1318
                parents_list = []
1438
1319
            else:
1439
1320
                parents_list = [(revision_id, basis)]
1440
1321
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1322
            wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1323
            wt.flush()
 
1324
            # if the basis has a root id we have to use that; otherwise we use
 
1325
            # a new random one
 
1326
            basis_root_id = basis.get_root_id()
 
1327
            if basis_root_id is not None:
 
1328
                wt._set_root_id(basis_root_id)
1443
1329
                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()
 
1330
            transform.build_tree(basis, wt, accelerator_tree)
 
1331
            basis.unlock()
1465
1332
        finally:
1466
1333
            control_files.unlock()
1467
1334
            wt.unlock()
1468
1335
        return wt
1469
1336
 
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
1337
    def _open(self, a_bzrdir, control_files):
1480
1338
        """Open the tree itself.
1481
1339
 
1482
1340
        :param a_bzrdir: the dir for the tree.
1483
1341
        :param control_files: the control files for the tree.
1484
1342
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1343
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1344
                           branch=a_bzrdir.open_branch(),
1487
1345
                           _format=self,
1488
1346
                           _bzrdir=a_bzrdir,
1489
1347
                           _control_files=control_files)
1490
1348
 
1491
1349
    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
1350
        # please test against something that will let us do tree references
1497
1351
        return bzrdir.format_registry.make_bzrdir(
1498
1352
            'dirstate-with-subtree')
1500
1354
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1355
 
1502
1356
 
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
1357
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
 
    """
 
1358
    """A revision tree pulling the inventory from a dirstate."""
1581
1359
 
1582
1360
    def __init__(self, dirstate, revision_id, repository):
1583
1361
        self._dirstate = dirstate
1586
1364
        self._inventory = None
1587
1365
        self._locked = 0
1588
1366
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1367
 
1593
1368
    def __repr__(self):
1594
1369
        return "<%s of %s in %s>" % \
1597
1372
    def annotate_iter(self, file_id,
1598
1373
                      default_revision=_mod_revision.CURRENT_REVISION):
1599
1374
        """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]
 
1375
        w = self._get_weave(file_id)
 
1376
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1377
 
1604
1378
    def _get_ancestors(self, default_revision):
1605
1379
        return set(self._repository.get_ancestry(self._revision_id,
1634
1408
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
1409
        return path_utf8.decode('utf8')
1636
1410
 
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
1411
    def _get_parent_index(self):
1646
1412
        """Return the index in the dirstate referenced by this tree."""
1647
1413
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1418
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1419
        If both are supplied, the fastest lookup is used, and an error is
1654
1420
        raised if they do not both point at the same row.
1655
 
 
 
1421
        
1656
1422
        :param file_id: An optional unicode file_id to be looked up.
1657
1423
        :param path: An optional unicode path to be looked up.
1658
1424
        :return: The dirstate row tuple for path/file_id, or (None, None)
1672
1438
 
1673
1439
        This is relatively expensive: we have to walk the entire dirstate.
1674
1440
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1441
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1442
            'dirstate revision tree'
1679
1443
        # separate call for profiling - makes it clear where the costs are.
1680
1444
        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()))
 
1445
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1446
            'parent %s has disappeared from %s' % (
 
1447
            self._revision_id, self._dirstate.get_parent_ids())
1685
1448
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1449
        # This is identical now to the WorkingTree _generate_inventory except
1687
1450
        # for the tree index use.
1688
1451
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1452
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1453
        assert current_entry[parent_index][0] == 'd'
1692
1454
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1455
        inv.root.revision = current_entry[parent_index][4]
1694
1456
        # Turn some things into local variables
1734
1496
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1497
                            % entry)
1736
1498
                # 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,))
 
1499
                assert file_id not in inv_byid
 
1500
                assert name_unicode not in parent_ie.children
1743
1501
                inv_byid[file_id] = inv_entry
1744
1502
                parent_ie.children[name_unicode] = inv_entry
1745
1503
        self._inventory = inv
1755
1513
            return None
1756
1514
        parent_index = self._get_parent_index()
1757
1515
        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
 
1516
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1517
 
1764
1518
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1519
        entry = self._get_entry(file_id=file_id, path=path)
1769
1523
            return parent_details[1]
1770
1524
        return None
1771
1525
 
 
1526
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1527
    def get_weave(self, file_id):
 
1528
        return self._get_weave(file_id)
 
1529
 
 
1530
    def _get_weave(self, file_id):
 
1531
        return self._repository.weave_store.get_weave(file_id,
 
1532
                self._repository.get_transaction())
 
1533
 
1772
1534
    def get_file(self, file_id, path=None):
1773
1535
        return StringIO(self.get_file_text(file_id))
1774
1536
 
 
1537
    def get_file_lines(self, file_id):
 
1538
        entry = self._get_entry(file_id=file_id)[1]
 
1539
        if entry == None:
 
1540
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1541
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1542
 
1775
1543
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1544
        return self.inventory[file_id].text_size
1778
1545
 
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)
 
1546
    def get_file_text(self, file_id):
 
1547
        return ''.join(self.get_file_lines(file_id))
1782
1548
 
1783
1549
    def get_reference_revision(self, file_id, path=None):
1784
1550
        return self.inventory[file_id].reference_revision
1803
1569
        if entry[1][parent_index][0] != 'l':
1804
1570
            return None
1805
1571
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1572
            # At present, none of the tree implementations supports non-ascii
 
1573
            # symlink targets. So we will just assume that the dirstate path is
 
1574
            # correct.
 
1575
            return entry[1][parent_index][1]
1809
1576
 
1810
1577
    def get_revision_id(self):
1811
1578
        """Return the revision id for this tree."""
1830
1597
 
1831
1598
    def kind(self, file_id):
1832
1599
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
 
1600
        if entry == None:
1834
1601
            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)
 
1602
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1841
1603
 
1842
1604
    def path_content_summary(self, path):
1843
1605
        """See Tree.path_content_summary."""
1859
1621
            return None
1860
1622
        return ie.executable
1861
1623
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1624
    def list_files(self, include_root=False):
1863
1625
        # We use a standard implementation, because DirStateRevisionTree is
1864
1626
        # dealing with one of the parents of the current state
1865
1627
        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:
 
1628
        entries = inv.iter_entries()
 
1629
        if self.inventory.root is not None and not include_root:
1875
1630
            entries.next()
1876
1631
        for path, entry in entries:
1877
1632
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1665
                self._dirstate_locked = False
1911
1666
            self._repository.unlock()
1912
1667
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1668
    def walkdirs(self, prefix=""):
1918
1669
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
 
1670
        # implementation based on an inventory.  
1920
1671
        # This should be cleaned up to use the much faster Dirstate code
1921
1672
        # So for now, we just build up the parent inventory, and extract
1922
1673
        # it the same way RevisionTree does.
1951
1702
 
1952
1703
class InterDirStateTree(InterTree):
1953
1704
    """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
 
1705
    
 
1706
    This is used only when both trees are in the dirstate working file, and 
 
1707
    the source is any parent within the dirstate, and the destination is 
1957
1708
    the current working tree of the same dirstate.
1958
1709
    """
1959
1710
    # this could be generalized to allow comparisons between any trees in the
1972
1723
        target.set_parent_ids([revid])
1973
1724
        return target.basis_tree(), target
1974
1725
 
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
1726
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1727
    _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,
 
1728
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1729
 
 
1730
    def _iter_changes(self, include_unchanged=False,
2002
1731
                      specific_files=None, pb=None, extra_trees=[],
2003
1732
                      require_versioned=True, want_unversioned=False):
2004
1733
        """Return the changes from source to target.
2005
1734
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1735
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1736
            for details.
2008
1737
        :param specific_files: An optional list of file paths to restrict the
2009
1738
            comparison to. When mapping filenames to ids, all matches in all
2020
1749
            output. An unversioned file is defined as one with (False, False)
2021
1750
            for the versioned pair.
2022
1751
        """
 
1752
        utf8_decode = cache_utf8._utf8_decode
 
1753
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1754
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1755
        # NB: show_status depends on being able to pass in non-versioned files
 
1756
        # and report them as unknown
2023
1757
        # TODO: handle extra trees in the dirstate.
2024
1758
        if (extra_trees or specific_files == []):
2025
1759
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1760
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1761
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1762
                require_versioned, want_unversioned=want_unversioned):
 
1763
                yield f
 
1764
            return
2029
1765
        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))
 
1766
        assert (self.source._revision_id in parent_ids
 
1767
                or self.source._revision_id == NULL_REVISION), \
 
1768
                "revision {%s} is not stored in {%s}, but %s " \
 
1769
                "can only be used for trees stored in the dirstate" \
 
1770
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1771
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1772
        if self.source._revision_id == NULL_REVISION:
2038
1773
            source_index = None
2039
1774
            indices = (target_index,)
2040
1775
        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))
 
1776
            assert (self.source._revision_id in parent_ids), \
 
1777
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1778
                self.source._revision_id, parent_ids)
2045
1779
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
1780
            indices = (source_index, target_index)
2047
1781
        # -- make all specific_files utf8 --
2048
1782
        if specific_files:
2049
1783
            specific_files_utf8 = set()
2050
1784
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1785
                specific_files_utf8.add(path.encode('utf8'))
2054
1786
            specific_files = specific_files_utf8
2055
1787
        else:
2056
1788
            specific_files = set([''])
2057
1789
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1790
        # -- get the state object and prepare it.
2060
1791
        state = self.target.current_dirstate()
2061
1792
        state._read_dirblocks_if_needed()
 
1793
        def _entries_for_path(path):
 
1794
            """Return a list with all the entries that match path for all ids.
 
1795
            """
 
1796
            dirname, basename = os.path.split(path)
 
1797
            key = (dirname, basename, '')
 
1798
            block_index, present = state._find_block_index_from_key(key)
 
1799
            if not present:
 
1800
                # the block which should contain path is absent.
 
1801
                return []
 
1802
            result = []
 
1803
            block = state._dirblocks[block_index][1]
 
1804
            entry_index, _ = state._find_entry_index(key, block)
 
1805
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1806
            while (entry_index < len(block) and
 
1807
                block[entry_index][0][0:2] == key[0:2]):
 
1808
                result.append(block[entry_index])
 
1809
                entry_index += 1
 
1810
            return result
2062
1811
        if require_versioned:
2063
1812
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1813
            all_versioned = True
2065
1814
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1815
                path_entries = _entries_for_path(path)
2067
1816
                if not path_entries:
2068
1817
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1818
                    all_versioned = False
 
1819
                    break
2071
1820
                found_versioned = False
2072
1821
                # for each id at this path
2073
1822
                for entry in path_entries:
2080
1829
                if not found_versioned:
2081
1830
                    # none of the indexes was not 'absent' at all ids for this
2082
1831
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1832
                    all_versioned = False
 
1833
                    break
 
1834
            if not all_versioned:
 
1835
                raise errors.PathsNotVersionedError(specific_files)
2086
1836
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1837
        search_specific_files = set()
 
1838
        for path in specific_files:
 
1839
            other_specific_files = specific_files.difference(set([path]))
 
1840
            if not osutils.is_inside_any(other_specific_files, path):
 
1841
                # this is a top level path, we must check it.
 
1842
                search_specific_files.add(path)
 
1843
        # sketch: 
 
1844
        # compare source_index and target_index at or under each element of search_specific_files.
 
1845
        # follow the following comparison table. Note that we only want to do diff operations when
 
1846
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1847
        # for the target.
 
1848
        # cases:
 
1849
        # 
 
1850
        # Source | Target | disk | action
 
1851
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1852
        #        |        |      | diff check on source-target
 
1853
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1854
        #        |        |      | ???
 
1855
        #   r    |  a     |      | add source to search
 
1856
        #   r    |  a     |  a   | 
 
1857
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1858
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1859
        #   a    | fdlt   |      | add new id
 
1860
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1861
        #   a    |  a     |      | not present in either tree, skip
 
1862
        #   a    |  a     |  a   | not present in any tree, skip
 
1863
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1864
        #        |        |      | may not be selected by the users list of paths.
 
1865
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1866
        #        |        |      | may not be selected by the users list of paths.
 
1867
        #  fdlt  | fdlt   |      | content in both: diff them
 
1868
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1869
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1870
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1871
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1872
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1873
        #        |        |      | this id at the other path.
 
1874
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1875
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1876
        #        |        |      | this id at the other path.
 
1877
 
 
1878
        # for all search_indexs in each path at or under each element of
 
1879
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1880
        # relocated path as one to search if its not searched already. If the
 
1881
        # detail is not relocated, add the id.
 
1882
        searched_specific_files = set()
 
1883
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1884
        # Using a list so that we can access the values and change them in
 
1885
        # nested scope. Each one is [path, file_id, entry]
 
1886
        last_source_parent = [None, None]
 
1887
        last_target_parent = [None, None]
2088
1888
 
2089
1889
        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()
 
1890
 
 
1891
        # Just a sentry, so that _process_entry can say that this
 
1892
        # record is handled, but isn't interesting to process (unchanged)
 
1893
        uninteresting = object()
 
1894
 
 
1895
 
 
1896
        old_dirname_to_file_id = {}
 
1897
        new_dirname_to_file_id = {}
 
1898
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1899
        #       keeping a cache of directories that we have seen.
 
1900
 
 
1901
        def _process_entry(entry, path_info):
 
1902
            """Compare an entry and real disk to generate delta information.
 
1903
 
 
1904
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1905
                the path of entry. If None, then the path is considered absent.
 
1906
                (Perhaps we should pass in a concrete entry for this ?)
 
1907
                Basename is returned as a utf8 string because we expect this
 
1908
                tuple will be ignored, and don't want to take the time to
 
1909
                decode.
 
1910
            :return: None if these don't match
 
1911
                     A tuple of information about the change, or
 
1912
                     the object 'uninteresting' if these match, but are
 
1913
                     basically identical.
 
1914
            """
 
1915
            if source_index is None:
 
1916
                source_details = NULL_PARENT_DETAILS
 
1917
            else:
 
1918
                source_details = entry[1][source_index]
 
1919
            target_details = entry[1][target_index]
 
1920
            target_minikind = target_details[0]
 
1921
            if path_info is not None and target_minikind in 'fdlt':
 
1922
                assert target_index == 0
 
1923
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1924
                                                  stat_value=path_info[3])
 
1925
                # The entry may have been modified by update_entry
 
1926
                target_details = entry[1][target_index]
 
1927
                target_minikind = target_details[0]
 
1928
            else:
 
1929
                link_or_sha1 = None
 
1930
            file_id = entry[0][2]
 
1931
            source_minikind = source_details[0]
 
1932
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1933
                # claimed content in both: diff
 
1934
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1935
                #        |        |      | diff check on source-target
 
1936
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1937
                #        |        |      | ???
 
1938
                if source_minikind in 'r':
 
1939
                    # add the source to the search path to find any children it
 
1940
                    # has.  TODO ? : only add if it is a container ?
 
1941
                    if not osutils.is_inside_any(searched_specific_files,
 
1942
                                                 source_details[1]):
 
1943
                        search_specific_files.add(source_details[1])
 
1944
                    # generate the old path; this is needed for stating later
 
1945
                    # as well.
 
1946
                    old_path = source_details[1]
 
1947
                    old_dirname, old_basename = os.path.split(old_path)
 
1948
                    path = pathjoin(entry[0][0], entry[0][1])
 
1949
                    old_entry = state._get_entry(source_index,
 
1950
                                                 path_utf8=old_path)
 
1951
                    # update the source details variable to be the real
 
1952
                    # location.
 
1953
                    source_details = old_entry[1][source_index]
 
1954
                    source_minikind = source_details[0]
 
1955
                else:
 
1956
                    old_dirname = entry[0][0]
 
1957
                    old_basename = entry[0][1]
 
1958
                    old_path = path = None
 
1959
                if path_info is None:
 
1960
                    # the file is missing on disk, show as removed.
 
1961
                    content_change = True
 
1962
                    target_kind = None
 
1963
                    target_exec = False
 
1964
                else:
 
1965
                    # source and target are both versioned and disk file is present.
 
1966
                    target_kind = path_info[2]
 
1967
                    if target_kind == 'directory':
 
1968
                        if path is None:
 
1969
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1970
                        new_dirname_to_file_id[path] = file_id
 
1971
                        if source_minikind != 'd':
 
1972
                            content_change = True
 
1973
                        else:
 
1974
                            # directories have no fingerprint
 
1975
                            content_change = False
 
1976
                        target_exec = False
 
1977
                    elif target_kind == 'file':
 
1978
                        if source_minikind != 'f':
 
1979
                            content_change = True
 
1980
                        else:
 
1981
                            # We could check the size, but we already have the
 
1982
                            # sha1 hash.
 
1983
                            content_change = (link_or_sha1 != source_details[1])
 
1984
                        # Target details is updated at update_entry time
 
1985
                        if use_filesystem_for_exec:
 
1986
                            # We don't need S_ISREG here, because we are sure
 
1987
                            # we are dealing with a file.
 
1988
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1989
                        else:
 
1990
                            target_exec = target_details[3]
 
1991
                    elif target_kind == 'symlink':
 
1992
                        if source_minikind != 'l':
 
1993
                            content_change = True
 
1994
                        else:
 
1995
                            content_change = (link_or_sha1 != source_details[1])
 
1996
                        target_exec = False
 
1997
                    elif target_kind == 'tree-reference':
 
1998
                        if source_minikind != 't':
 
1999
                            content_change = True
 
2000
                        else:
 
2001
                            content_change = False
 
2002
                        target_exec = False
 
2003
                    else:
 
2004
                        raise Exception, "unknown kind %s" % path_info[2]
 
2005
                if source_minikind == 'd':
 
2006
                    if path is None:
 
2007
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2008
                    old_dirname_to_file_id[old_path] = file_id
 
2009
                # parent id is the entry for the path in the target tree
 
2010
                if old_dirname == last_source_parent[0]:
 
2011
                    source_parent_id = last_source_parent[1]
 
2012
                else:
 
2013
                    try:
 
2014
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2015
                    except KeyError:
 
2016
                        source_parent_entry = state._get_entry(source_index,
 
2017
                                                               path_utf8=old_dirname)
 
2018
                        source_parent_id = source_parent_entry[0][2]
 
2019
                    if source_parent_id == entry[0][2]:
 
2020
                        # This is the root, so the parent is None
 
2021
                        source_parent_id = None
 
2022
                    else:
 
2023
                        last_source_parent[0] = old_dirname
 
2024
                        last_source_parent[1] = source_parent_id
 
2025
                new_dirname = entry[0][0]
 
2026
                if new_dirname == last_target_parent[0]:
 
2027
                    target_parent_id = last_target_parent[1]
 
2028
                else:
 
2029
                    try:
 
2030
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2031
                    except KeyError:
 
2032
                        # TODO: We don't always need to do the lookup, because the
 
2033
                        #       parent entry will be the same as the source entry.
 
2034
                        target_parent_entry = state._get_entry(target_index,
 
2035
                                                               path_utf8=new_dirname)
 
2036
                        assert target_parent_entry != (None, None), (
 
2037
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2038
                            % (new_dirname, entry))
 
2039
                        target_parent_id = target_parent_entry[0][2]
 
2040
                    if target_parent_id == entry[0][2]:
 
2041
                        # This is the root, so the parent is None
 
2042
                        target_parent_id = None
 
2043
                    else:
 
2044
                        last_target_parent[0] = new_dirname
 
2045
                        last_target_parent[1] = target_parent_id
 
2046
 
 
2047
                source_exec = source_details[3]
 
2048
                if (include_unchanged
 
2049
                    or content_change
 
2050
                    or source_parent_id != target_parent_id
 
2051
                    or old_basename != entry[0][1]
 
2052
                    or source_exec != target_exec
 
2053
                    ):
 
2054
                    if old_path is None:
 
2055
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2056
                        old_path_u = utf8_decode(old_path)[0]
 
2057
                        path_u = old_path_u
 
2058
                    else:
 
2059
                        old_path_u = utf8_decode(old_path)[0]
 
2060
                        if old_path == path:
 
2061
                            path_u = old_path_u
 
2062
                        else:
 
2063
                            path_u = utf8_decode(path)[0]
 
2064
                    source_kind = _minikind_to_kind[source_minikind]
 
2065
                    return (entry[0][2],
 
2066
                           (old_path_u, path_u),
 
2067
                           content_change,
 
2068
                           (True, True),
 
2069
                           (source_parent_id, target_parent_id),
 
2070
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2071
                           (source_kind, target_kind),
 
2072
                           (source_exec, target_exec))
 
2073
                else:
 
2074
                    return uninteresting
 
2075
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2076
                # looks like a new file
 
2077
                if path_info is not None:
 
2078
                    path = pathjoin(entry[0][0], entry[0][1])
 
2079
                    # parent id is the entry for the path in the target tree
 
2080
                    # TODO: these are the same for an entire directory: cache em.
 
2081
                    parent_id = state._get_entry(target_index,
 
2082
                                                 path_utf8=entry[0][0])[0][2]
 
2083
                    if parent_id == entry[0][2]:
 
2084
                        parent_id = None
 
2085
                    if use_filesystem_for_exec:
 
2086
                        # We need S_ISREG here, because we aren't sure if this
 
2087
                        # is a file or not.
 
2088
                        target_exec = bool(
 
2089
                            stat.S_ISREG(path_info[3].st_mode)
 
2090
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2091
                    else:
 
2092
                        target_exec = target_details[3]
 
2093
                    return (entry[0][2],
 
2094
                           (None, utf8_decode(path)[0]),
 
2095
                           True,
 
2096
                           (False, True),
 
2097
                           (None, parent_id),
 
2098
                           (None, utf8_decode(entry[0][1])[0]),
 
2099
                           (None, path_info[2]),
 
2100
                           (None, target_exec))
 
2101
                else:
 
2102
                    # but its not on disk: we deliberately treat this as just
 
2103
                    # never-present. (Why ?! - RBC 20070224)
 
2104
                    pass
 
2105
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2106
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2107
                # if its still on disk, *and* theres no other entry at this
 
2108
                # path [we dont know this in this routine at the moment -
 
2109
                # perhaps we should change this - then it would be an unknown.
 
2110
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2111
                # parent id is the entry for the path in the target tree
 
2112
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2113
                if parent_id == entry[0][2]:
 
2114
                    parent_id = None
 
2115
                return (entry[0][2],
 
2116
                       (utf8_decode(old_path)[0], None),
 
2117
                       True,
 
2118
                       (True, False),
 
2119
                       (parent_id, None),
 
2120
                       (utf8_decode(entry[0][1])[0], None),
 
2121
                       (_minikind_to_kind[source_minikind], None),
 
2122
                       (source_details[3], None))
 
2123
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2124
                # a rename; could be a true rename, or a rename inherited from
 
2125
                # a renamed parent. TODO: handle this efficiently. Its not
 
2126
                # common case to rename dirs though, so a correct but slow
 
2127
                # implementation will do.
 
2128
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2129
                    search_specific_files.add(target_details[1])
 
2130
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2131
                # neither of the selected trees contain this file,
 
2132
                # so skip over it. This is not currently directly tested, but
 
2133
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2134
                pass
 
2135
            else:
 
2136
                raise AssertionError("don't know how to compare "
 
2137
                    "source_minikind=%r, target_minikind=%r"
 
2138
                    % (source_minikind, target_minikind))
 
2139
                ## import pdb;pdb.set_trace()
 
2140
            return None
 
2141
 
 
2142
        while search_specific_files:
 
2143
            # TODO: the pending list should be lexically sorted?  the
 
2144
            # interface doesn't require it.
 
2145
            current_root = search_specific_files.pop()
 
2146
            current_root_unicode = current_root.decode('utf8')
 
2147
            searched_specific_files.add(current_root)
 
2148
            # process the entries for this containing directory: the rest will be
 
2149
            # found by their parents recursively.
 
2150
            root_entries = _entries_for_path(current_root)
 
2151
            root_abspath = self.target.abspath(current_root_unicode)
 
2152
            try:
 
2153
                root_stat = os.lstat(root_abspath)
 
2154
            except OSError, e:
 
2155
                if e.errno == errno.ENOENT:
 
2156
                    # the path does not exist: let _process_entry know that.
 
2157
                    root_dir_info = None
 
2158
                else:
 
2159
                    # some other random error: hand it up.
 
2160
                    raise
 
2161
            else:
 
2162
                root_dir_info = ('', current_root,
 
2163
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2164
                    root_abspath)
 
2165
                if root_dir_info[2] == 'directory':
 
2166
                    if self.target._directory_is_tree_reference(
 
2167
                        current_root.decode('utf8')):
 
2168
                        root_dir_info = root_dir_info[:2] + \
 
2169
                            ('tree-reference',) + root_dir_info[3:]
 
2170
 
 
2171
            if not root_entries and not root_dir_info:
 
2172
                # this specified path is not present at all, skip it.
 
2173
                continue
 
2174
            path_handled = False
 
2175
            for entry in root_entries:
 
2176
                result = _process_entry(entry, root_dir_info)
 
2177
                if result is not None:
 
2178
                    path_handled = True
 
2179
                    if result is not uninteresting:
 
2180
                        yield result
 
2181
            if want_unversioned and not path_handled and root_dir_info:
 
2182
                new_executable = bool(
 
2183
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2184
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2185
                yield (None,
 
2186
                       (None, current_root_unicode),
 
2187
                       True,
 
2188
                       (False, False),
 
2189
                       (None, None),
 
2190
                       (None, splitpath(current_root_unicode)[-1]),
 
2191
                       (None, root_dir_info[2]),
 
2192
                       (None, new_executable)
 
2193
                      )
 
2194
            initial_key = (current_root, '', '')
 
2195
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2196
            if block_index == 0:
 
2197
                # we have processed the total root already, but because the
 
2198
                # initial key matched it we should skip it here.
 
2199
                block_index +=1
 
2200
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2201
                current_dir_info = None
 
2202
            else:
 
2203
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2204
                try:
 
2205
                    current_dir_info = dir_iterator.next()
 
2206
                except OSError, e:
 
2207
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2208
                    # python 2.5 has e.errno == EINVAL,
 
2209
                    #            and e.winerror == ERROR_DIRECTORY
 
2210
                    e_winerror = getattr(e, 'winerror', None)
 
2211
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2212
                    # there may be directories in the inventory even though
 
2213
                    # this path is not a file on disk: so mark it as end of
 
2214
                    # iterator
 
2215
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2216
                        current_dir_info = None
 
2217
                    elif (sys.platform == 'win32'
 
2218
                          and (e.errno in win_errors
 
2219
                               or e_winerror in win_errors)):
 
2220
                        current_dir_info = None
 
2221
                    else:
 
2222
                        raise
 
2223
                else:
 
2224
                    if current_dir_info[0][0] == '':
 
2225
                        # remove .bzr from iteration
 
2226
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2227
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2228
                        del current_dir_info[1][bzr_index]
 
2229
            # walk until both the directory listing and the versioned metadata
 
2230
            # are exhausted. 
 
2231
            if (block_index < len(state._dirblocks) and
 
2232
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2233
                current_block = state._dirblocks[block_index]
 
2234
            else:
 
2235
                current_block = None
 
2236
            while (current_dir_info is not None or
 
2237
                   current_block is not None):
 
2238
                if (current_dir_info and current_block
 
2239
                    and current_dir_info[0][0] != current_block[0]):
 
2240
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2241
                        # filesystem data refers to paths not covered by the dirblock.
 
2242
                        # this has two possibilities:
 
2243
                        # A) it is versioned but empty, so there is no block for it
 
2244
                        # B) it is not versioned.
 
2245
 
 
2246
                        # if (A) then we need to recurse into it to check for
 
2247
                        # new unknown files or directories.
 
2248
                        # if (B) then we should ignore it, because we don't
 
2249
                        # recurse into unknown directories.
 
2250
                        path_index = 0
 
2251
                        while path_index < len(current_dir_info[1]):
 
2252
                                current_path_info = current_dir_info[1][path_index]
 
2253
                                if want_unversioned:
 
2254
                                    if current_path_info[2] == 'directory':
 
2255
                                        if self.target._directory_is_tree_reference(
 
2256
                                            current_path_info[0].decode('utf8')):
 
2257
                                            current_path_info = current_path_info[:2] + \
 
2258
                                                ('tree-reference',) + current_path_info[3:]
 
2259
                                    new_executable = bool(
 
2260
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2261
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2262
                                    yield (None,
 
2263
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2264
                                        True,
 
2265
                                        (False, False),
 
2266
                                        (None, None),
 
2267
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2268
                                        (None, current_path_info[2]),
 
2269
                                        (None, new_executable))
 
2270
                                # dont descend into this unversioned path if it is
 
2271
                                # a dir
 
2272
                                if current_path_info[2] in ('directory',
 
2273
                                                            'tree-reference'):
 
2274
                                    del current_dir_info[1][path_index]
 
2275
                                    path_index -= 1
 
2276
                                path_index += 1
 
2277
 
 
2278
                        # This dir info has been handled, go to the next
 
2279
                        try:
 
2280
                            current_dir_info = dir_iterator.next()
 
2281
                        except StopIteration:
 
2282
                            current_dir_info = None
 
2283
                    else:
 
2284
                        # We have a dirblock entry for this location, but there
 
2285
                        # is no filesystem path for this. This is most likely
 
2286
                        # because a directory was removed from the disk.
 
2287
                        # We don't have to report the missing directory,
 
2288
                        # because that should have already been handled, but we
 
2289
                        # need to handle all of the files that are contained
 
2290
                        # within.
 
2291
                        for current_entry in current_block[1]:
 
2292
                            # entry referring to file not present on disk.
 
2293
                            # advance the entry only, after processing.
 
2294
                            result = _process_entry(current_entry, None)
 
2295
                            if result is not None:
 
2296
                                if result is not uninteresting:
 
2297
                                    yield result
 
2298
                        block_index +=1
 
2299
                        if (block_index < len(state._dirblocks) and
 
2300
                            osutils.is_inside(current_root,
 
2301
                                              state._dirblocks[block_index][0])):
 
2302
                            current_block = state._dirblocks[block_index]
 
2303
                        else:
 
2304
                            current_block = None
 
2305
                    continue
 
2306
                entry_index = 0
 
2307
                if current_block and entry_index < len(current_block[1]):
 
2308
                    current_entry = current_block[1][entry_index]
 
2309
                else:
 
2310
                    current_entry = None
 
2311
                advance_entry = True
 
2312
                path_index = 0
 
2313
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2314
                    current_path_info = current_dir_info[1][path_index]
 
2315
                    if current_path_info[2] == 'directory':
 
2316
                        if self.target._directory_is_tree_reference(
 
2317
                            current_path_info[0].decode('utf8')):
 
2318
                            current_path_info = current_path_info[:2] + \
 
2319
                                ('tree-reference',) + current_path_info[3:]
 
2320
                else:
 
2321
                    current_path_info = None
 
2322
                advance_path = True
 
2323
                path_handled = False
 
2324
                while (current_entry is not None or
 
2325
                    current_path_info is not None):
 
2326
                    if current_entry is None:
 
2327
                        # the check for path_handled when the path is adnvaced
 
2328
                        # will yield this path if needed.
 
2329
                        pass
 
2330
                    elif current_path_info is None:
 
2331
                        # no path is fine: the per entry code will handle it.
 
2332
                        result = _process_entry(current_entry, current_path_info)
 
2333
                        if result is not None:
 
2334
                            if result is not uninteresting:
 
2335
                                yield result
 
2336
                    elif (current_entry[0][1] != current_path_info[1]
 
2337
                          or current_entry[1][target_index][0] in 'ar'):
 
2338
                        # The current path on disk doesn't match the dirblock
 
2339
                        # record. Either the dirblock is marked as absent, or
 
2340
                        # the file on disk is not present at all in the
 
2341
                        # dirblock. Either way, report about the dirblock
 
2342
                        # entry, and let other code handle the filesystem one.
 
2343
 
 
2344
                        # Compare the basename for these files to determine
 
2345
                        # which comes first
 
2346
                        if current_path_info[1] < current_entry[0][1]:
 
2347
                            # extra file on disk: pass for now, but only
 
2348
                            # increment the path, not the entry
 
2349
                            advance_entry = False
 
2350
                        else:
 
2351
                            # entry referring to file not present on disk.
 
2352
                            # advance the entry only, after processing.
 
2353
                            result = _process_entry(current_entry, None)
 
2354
                            if result is not None:
 
2355
                                if result is not uninteresting:
 
2356
                                    yield result
 
2357
                            advance_path = False
 
2358
                    else:
 
2359
                        result = _process_entry(current_entry, current_path_info)
 
2360
                        if result is not None:
 
2361
                            path_handled = True
 
2362
                            if result is not uninteresting:
 
2363
                                yield result
 
2364
                    if advance_entry and current_entry is not None:
 
2365
                        entry_index += 1
 
2366
                        if entry_index < len(current_block[1]):
 
2367
                            current_entry = current_block[1][entry_index]
 
2368
                        else:
 
2369
                            current_entry = None
 
2370
                    else:
 
2371
                        advance_entry = True # reset the advance flaga
 
2372
                    if advance_path and current_path_info is not None:
 
2373
                        if not path_handled:
 
2374
                            # unversioned in all regards
 
2375
                            if want_unversioned:
 
2376
                                new_executable = bool(
 
2377
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2378
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2379
                                yield (None,
 
2380
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2381
                                    True,
 
2382
                                    (False, False),
 
2383
                                    (None, None),
 
2384
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2385
                                    (None, current_path_info[2]),
 
2386
                                    (None, new_executable))
 
2387
                            # dont descend into this unversioned path if it is
 
2388
                            # a dir
 
2389
                            if current_path_info[2] in ('directory'):
 
2390
                                del current_dir_info[1][path_index]
 
2391
                                path_index -= 1
 
2392
                        # dont descend the disk iterator into any tree 
 
2393
                        # paths.
 
2394
                        if current_path_info[2] == 'tree-reference':
 
2395
                            del current_dir_info[1][path_index]
 
2396
                            path_index -= 1
 
2397
                        path_index += 1
 
2398
                        if path_index < len(current_dir_info[1]):
 
2399
                            current_path_info = current_dir_info[1][path_index]
 
2400
                            if current_path_info[2] == 'directory':
 
2401
                                if self.target._directory_is_tree_reference(
 
2402
                                    current_path_info[0].decode('utf8')):
 
2403
                                    current_path_info = current_path_info[:2] + \
 
2404
                                        ('tree-reference',) + current_path_info[3:]
 
2405
                        else:
 
2406
                            current_path_info = None
 
2407
                        path_handled = False
 
2408
                    else:
 
2409
                        advance_path = True # reset the advance flagg.
 
2410
                if current_block is not None:
 
2411
                    block_index += 1
 
2412
                    if (block_index < len(state._dirblocks) and
 
2413
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2414
                        current_block = state._dirblocks[block_index]
 
2415
                    else:
 
2416
                        current_block = None
 
2417
                if current_dir_info is not None:
 
2418
                    try:
 
2419
                        current_dir_info = dir_iterator.next()
 
2420
                    except StopIteration:
 
2421
                        current_dir_info = None
 
2422
 
2094
2423
 
2095
2424
    @staticmethod
2096
2425
    def is_compatible(source, target):
2097
2426
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2427
        if not isinstance(target, WorkingTree4):
2099
2428
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2429
        # the source must be a revtreee or dirstate rev tree.
2101
2430
        if not isinstance(source,
2102
2431
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2432
            return False
2104
2433
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2434
        if not (source._revision_id == NULL_REVISION or
2106
2435
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2436
            # TODO: what about ghosts? it may well need to 
2108
2437
            # check for them explicitly.
2109
2438
            return False
2110
2439
        return True
2120
2449
 
2121
2450
    def convert(self, tree):
2122
2451
        # 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
 
2452
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2453
        # tree during upgrade.
2125
2454
        tree._control_files.lock_write()
2126
2455
        try:
2152
2481
 
2153
2482
    def update_format(self, tree):
2154
2483
        """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())
 
2484
        tree._control_files.put_utf8('format',
 
2485
            self.target_format.get_format_string())