/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: Canonical.com Patch Queue Manager
  • Date: 2007-11-22 20:17:22 UTC
  • mfrom: (3004.1.10 84043)
  • Revision ID: pqm@pqm.ubuntu.com-20071122201722-4l5bgljqn7qdux7f
commit now can invoke an external editor in a non-ascii directory.  (Watkins)

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
98
151
        self._dirstate = None
99
152
        self._inventory = None
100
153
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        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
154
 
108
155
    @needs_tree_write_lock
109
156
    def _add(self, files, ids, kinds):
137
184
    @needs_tree_write_lock
138
185
    def add_reference(self, sub_tree):
139
186
        # use standard implementation, which calls back to self._add
140
 
        #
 
187
        # 
141
188
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
189
        # it's just recorded at the moment of commit. 
143
190
        self._add_reference(sub_tree)
144
191
 
145
192
    def break_lock(self):
184
231
            WorkingTree3._comparison_data(self, entry, path)
185
232
        # it looks like a plain directory, but it's really a reference -- see
186
233
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
189
237
            kind = 'tree-reference'
190
238
        return kind, executable, stat_value
191
239
 
217
265
            return self._dirstate
218
266
        local_path = self.bzrdir.get_workingtree_transport(None
219
267
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
222
269
        return self._dirstate
223
270
 
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
271
    def filter_unversioned_files(self, paths):
237
272
        """Filter out paths that are versioned.
238
273
 
270
305
 
271
306
    def _generate_inventory(self):
272
307
        """Create and set self.inventory from the dirstate object.
273
 
 
 
308
        
274
309
        This is relatively expensive: we have to walk the entire dirstate.
275
310
        Ideally we would not, and can deprecate this function.
276
311
        """
280
315
        state._read_dirblocks_if_needed()
281
316
        root_key, current_entry = self._get_entry(path='')
282
317
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
318
        assert current_entry[0][0] == 'd' # directory
285
319
        inv = Inventory(root_id=current_id)
286
320
        # Turn some things into local variables
287
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
354
                    # add this entry to the parent map.
321
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
356
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
327
362
                    inv_entry.reference_revision = link_or_sha1 or None
328
363
                elif kind != 'symlink':
329
364
                    raise AssertionError("unknown kind %r" % kind)
330
365
                # 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,))
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
337
369
                inv_byid[file_id] = inv_entry
338
370
                parent_ie.children[name_unicode] = inv_entry
339
371
        self._inventory = inv
344
376
        If either file_id or path is supplied, it is used as the key to lookup.
345
377
        If both are supplied, the fastest lookup is used, and an error is
346
378
        raised if they do not both point at the same row.
347
 
 
 
379
        
348
380
        :param file_id: An optional unicode file_id to be looked up.
349
381
        :param path: An optional unicode path to be looked up.
350
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
374
406
                    return None
375
407
                else:
376
408
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
379
411
        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
 
412
            return link_or_sha1
390
413
        return None
391
414
 
392
415
    def _get_inventory(self):
406
429
    @needs_read_lock
407
430
    def get_parent_ids(self):
408
431
        """See Tree.get_parent_ids.
409
 
 
 
432
        
410
433
        This implementation requests the ids list from the dirstate file.
411
434
        """
412
435
        return self.current_dirstate().get_parent_ids()
434
457
        return osutils.lexists(pathjoin(
435
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
459
 
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
460
    @needs_read_lock
443
461
    def id2path(self, file_id):
444
462
        "Convert a file-id to a path."
474
492
 
475
493
            Note: The caller is expected to take a read-lock before calling this.
476
494
            """
477
 
            self._must_be_locked()
478
495
            if not path:
479
496
                path = self.id2path(file_id)
480
497
            mode = os.lstat(self.abspath(path)).st_mode
481
498
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
499
 
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
500
    @needs_read_lock
494
501
    def __iter__(self):
495
502
        """Iterate through file_ids for this tree.
508
515
        return iter(result)
509
516
 
510
517
    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
518
        for key, tree_details in self.current_dirstate()._iter_entries():
516
519
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
520
                # not relevant to the working tree
519
522
            if not key[1]:
520
523
                # the root is not a reference.
521
524
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
526
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
527
                if self._kind(path) == 'tree-reference':
 
528
                    yield path, key[2]
526
529
            except errors.NoSuchFile:
527
530
                # path is missing on disk.
528
531
                continue
529
532
 
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
533
    def kind(self, file_id):
537
534
        """Return the kind of a file.
538
535
 
542
539
        Note: The caller is expected to take a read-lock before calling this.
543
540
        """
544
541
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
542
        assert relpath != None, \
 
543
            "path for id {%s} is None!" % file_id
548
544
        return self._kind(relpath)
549
545
 
550
546
    def _kind(self, relpath):
551
547
        abspath = self.abspath(relpath)
552
548
        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'
 
549
        if (self._repo_supports_tree_reference and
 
550
            kind == 'directory' and
 
551
            self._directory_is_tree_reference(relpath)):
 
552
            kind = 'tree-reference'
558
553
        return kind
559
554
 
560
555
    @needs_read_lock
567
562
            return _mod_revision.NULL_REVISION
568
563
 
569
564
    def lock_read(self):
570
 
        """See Branch.lock_read, and WorkingTree.unlock.
571
 
 
572
 
        :return: An object with an unlock method which will release the lock
573
 
            obtained.
574
 
        """
 
565
        """See Branch.lock_read, and WorkingTree.unlock."""
575
566
        self.branch.lock_read()
576
567
        try:
577
568
            self._control_files.lock_read()
590
581
        except:
591
582
            self.branch.unlock()
592
583
            raise
593
 
        return self
594
584
 
595
585
    def _lock_self_write(self):
596
586
        """This should be called after the branch is locked."""
611
601
        except:
612
602
            self.branch.unlock()
613
603
            raise
614
 
        return self
615
604
 
616
605
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: An object with an unlock method which will release the lock
620
 
            obtained.
621
 
        """
 
606
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
622
607
        self.branch.lock_read()
623
 
        return self._lock_self_write()
 
608
        self._lock_self_write()
624
609
 
625
610
    def lock_write(self):
626
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
627
 
 
628
 
        :return: An object with an unlock method which will release the lock
629
 
            obtained.
630
 
        """
 
611
        """See MutableTree.lock_write, and WorkingTree.unlock."""
631
612
        self.branch.lock_write()
632
 
        return self._lock_self_write()
 
613
        self._lock_self_write()
633
614
 
634
615
    @needs_tree_write_lock
635
616
    def move(self, from_paths, to_dir, after=False):
637
618
        result = []
638
619
        if not from_paths:
639
620
            return result
 
621
 
640
622
        state = self.current_dirstate()
641
 
        if isinstance(from_paths, basestring):
642
 
            raise ValueError()
 
623
 
 
624
        assert not isinstance(from_paths, basestring)
643
625
        to_dir_utf8 = to_dir.encode('utf8')
644
626
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
645
627
        id_index = state._get_id_index()
667
649
        if self._inventory is not None:
668
650
            update_inventory = True
669
651
            inv = self.inventory
 
652
            to_dir_ie = inv[to_dir_id]
670
653
            to_dir_id = to_entry[0][2]
671
 
            to_dir_ie = inv[to_dir_id]
672
654
        else:
673
655
            update_inventory = False
674
656
 
705
687
            from_entry = self._get_entry(path=from_rel)
706
688
            if from_entry == (None, None):
707
689
                raise errors.BzrMoveFailedError(from_rel,to_dir,
708
 
                    errors.NotVersionedError(path=from_rel))
 
690
                    errors.NotVersionedError(path=str(from_rel)))
709
691
 
710
692
            from_id = from_entry[0][2]
711
693
            to_rel = pathjoin(to_dir, from_tail)
800
782
                if minikind == 'd':
801
783
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
802
784
                        """Recursively update all entries in this dirblock."""
803
 
                        if from_dir == '':
804
 
                            raise AssertionError("renaming root not supported")
 
785
                        assert from_dir != '', "renaming root not supported"
805
786
                        from_key = (from_dir, '')
806
787
                        from_block_idx, present = \
807
788
                            state._find_block_index_from_key(from_key)
820
801
 
821
802
                        # Grab a copy since move_one may update the list.
822
803
                        for entry in from_block[1][:]:
823
 
                            if not (entry[0][0] == from_dir):
824
 
                                raise AssertionError()
 
804
                            assert entry[0][0] == from_dir
825
805
                            cur_details = entry[1][0]
826
806
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
827
807
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
886
866
        for tree in trees:
887
867
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
888
868
                parents):
889
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
890
 
                    trees, require_versioned)
 
869
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
891
870
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
892
871
        # -- make all paths utf8 --
893
872
        paths_utf8 = set()
954
933
                raise errors.PathsNotVersionedError(paths)
955
934
        # -- remove redundancy in supplied paths to prevent over-scanning --
956
935
        search_paths = osutils.minimum_path_selection(paths)
957
 
        # sketch:
 
936
        # sketch: 
958
937
        # for all search_indexs in each path at or under each element of
959
938
        # search_paths, if the detail is relocated: add the id, and add the
960
939
        # relocated path as one to search if its not searched already. If the
1016
995
 
1017
996
    def read_working_inventory(self):
1018
997
        """Read the working inventory.
1019
 
 
 
998
        
1020
999
        This is a meaningless operation for dirstate, but we obey it anyhow.
1021
1000
        """
1022
1001
        return self.inventory
1040
1019
    def set_last_revision(self, new_revision):
1041
1020
        """Change the last revision in the working tree."""
1042
1021
        parents = self.get_parent_ids()
1043
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1044
 
            if len(parents) >= 2:
1045
 
                raise AssertionError(
1046
 
                    "setting the last parent to none with a pending merge is "
1047
 
                    "unsupported.")
 
1022
        if new_revision in (NULL_REVISION, None):
 
1023
            assert len(parents) < 2, (
 
1024
                "setting the last parent to none with a pending merge is "
 
1025
                "unsupported.")
1048
1026
            self.set_parent_ids([])
1049
1027
        else:
1050
1028
            self.set_parent_ids([new_revision] + parents[1:],
1053
1031
    @needs_tree_write_lock
1054
1032
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
1033
        """Set the parent ids to revision_ids.
1056
 
 
 
1034
        
1057
1035
        See also set_parent_trees. This api will try to retrieve the tree data
1058
1036
        for each element of revision_ids from the trees repository. If you have
1059
1037
        tree data already available, it is more efficient to use
1091
1069
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1092
1070
        real_trees = []
1093
1071
        ghosts = []
1094
 
 
1095
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1096
 
        graph = self.branch.repository.get_graph()
1097
 
        heads = graph.heads(parent_ids)
1098
 
        accepted_revisions = set()
1099
 
 
1100
1072
        # convert absent trees to the null tree, which we convert back to
1101
1073
        # missing on access.
1102
1074
        for rev_id, tree in parents_list:
1103
 
            if len(accepted_revisions) > 0:
1104
 
                # we always accept the first tree
1105
 
                if rev_id in accepted_revisions or rev_id not in heads:
1106
 
                    # We have already included either this tree, or its
1107
 
                    # descendent, so we skip it.
1108
 
                    continue
1109
1075
            _mod_revision.check_not_reserved_id(rev_id)
1110
1076
            if tree is not None:
1111
1077
                real_trees.append((rev_id, tree))
1112
1078
            else:
1113
1079
                real_trees.append((rev_id,
1114
 
                    self.branch.repository.revision_tree(
1115
 
                        _mod_revision.NULL_REVISION)))
 
1080
                    self.branch.repository.revision_tree(None)))
1116
1081
                ghosts.append(rev_id)
1117
 
            accepted_revisions.add(rev_id)
1118
1082
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1119
1083
        self._make_dirty(reset_inventory=False)
1120
1084
 
1213
1177
                # just forget the whole block.
1214
1178
                entry_index = 0
1215
1179
                while entry_index < len(block[1]):
 
1180
                    # Mark this file id as having been removed
1216
1181
                    entry = block[1][entry_index]
1217
 
                    if entry[1][0][0] in 'ar':
1218
 
                        # don't remove absent or renamed entries
 
1182
                    ids_to_unversion.discard(entry[0][2])
 
1183
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1184
                                               # entries
 
1185
                        or not state._make_absent(entry)):
1219
1186
                        entry_index += 1
1220
 
                    else:
1221
 
                        # Mark this file id as having been removed
1222
 
                        ids_to_unversion.discard(entry[0][2])
1223
 
                        if not state._make_absent(entry):
1224
 
                            # The block has not shrunk.
1225
 
                            entry_index += 1
1226
1187
                # go to the next block. (At the moment we dont delete empty
1227
1188
                # dirblocks)
1228
1189
                block_index += 1
1251
1212
            for file_id in file_ids:
1252
1213
                self._inventory.remove_recursive_id(file_id)
1253
1214
 
1254
 
    @needs_tree_write_lock
1255
 
    def rename_one(self, from_rel, to_rel, after=False):
1256
 
        """See WorkingTree.rename_one"""
1257
 
        self.flush()
1258
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1259
 
 
1260
 
    @needs_tree_write_lock
1261
 
    def apply_inventory_delta(self, changes):
1262
 
        """See MutableTree.apply_inventory_delta"""
1263
 
        state = self.current_dirstate()
1264
 
        state.update_by_delta(changes)
1265
 
        self._make_dirty(reset_inventory=True)
1266
 
 
1267
1215
    def update_basis_by_delta(self, new_revid, delta):
1268
1216
        """See MutableTree.update_basis_by_delta."""
1269
 
        if self.last_revision() == new_revid:
1270
 
            raise AssertionError()
 
1217
        assert self.last_revision() != new_revid
1271
1218
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1272
1219
 
1273
1220
    @needs_read_lock
1277
1224
    @needs_tree_write_lock
1278
1225
    def _write_inventory(self, inv):
1279
1226
        """Write inventory as the current inventory."""
1280
 
        if self._dirty:
1281
 
            raise AssertionError("attempting to write an inventory when the "
1282
 
                "dirstate is dirty will lose pending changes")
1283
 
        had_inventory = self._inventory is not None
1284
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1285
 
        # working inventory. We do this because self.inventory may be inv, or
1286
 
        # may have been modified, and either case would prevent a clean delta
1287
 
        # being created.
1288
 
        self._inventory = None
1289
 
        # generate a delta,
1290
 
        delta = inv._make_delta(self.inventory)
1291
 
        # and apply it.
1292
 
        self.apply_inventory_delta(delta)
1293
 
        if had_inventory:
 
1227
        assert not self._dirty, ("attempting to write an inventory when the "
 
1228
            "dirstate is dirty will cause data loss")
 
1229
        self.current_dirstate().set_state_from_inventory(inv)
 
1230
        self._make_dirty(reset_inventory=False)
 
1231
        if self._inventory is not None:
1294
1232
            self._inventory = inv
1295
1233
        self.flush()
1296
1234
 
1297
1235
 
1298
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1299
 
 
1300
 
    def __init__(self, tree):
1301
 
        self.tree = tree
1302
 
 
1303
 
    def sha1(self, abspath):
1304
 
        """See dirstate.SHA1Provider.sha1()."""
1305
 
        filters = self.tree._content_filter_stack(
1306
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1307
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1308
 
 
1309
 
    def stat_and_sha1(self, abspath):
1310
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1311
 
        filters = self.tree._content_filter_stack(
1312
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1313
 
        file_obj = file(abspath, 'rb', 65000)
1314
 
        try:
1315
 
            statvalue = os.fstat(file_obj.fileno())
1316
 
            if filters:
1317
 
                file_obj = filtered_input_file(file_obj, filters)
1318
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1319
 
        finally:
1320
 
            file_obj.close()
1321
 
        return statvalue, sha1
1322
 
 
1323
 
 
1324
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1325
 
    """Dirstate working tree that supports content filtering.
1326
 
 
1327
 
    The dirstate holds the hash and size of the canonical form of the file, 
1328
 
    and most methods must return that.
1329
 
    """
1330
 
 
1331
 
    def _file_content_summary(self, path, stat_result):
1332
 
        # This is to support the somewhat obsolete path_content_summary method
1333
 
        # with content filtering: see
1334
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1335
 
        #
1336
 
        # If the dirstate cache is up to date and knows the hash and size,
1337
 
        # return that.
1338
 
        # Otherwise if there are no content filters, return the on-disk size
1339
 
        # and leave the hash blank.
1340
 
        # Otherwise, read and filter the on-disk file and use its size and
1341
 
        # hash.
1342
 
        #
1343
 
        # The dirstate doesn't store the size of the canonical form so we
1344
 
        # can't trust it for content-filtered trees.  We just return None.
1345
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1346
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1347
 
        return ('file', None, executable, dirstate_sha1)
1348
 
 
1349
 
 
1350
 
class WorkingTree4(DirStateWorkingTree):
1351
 
    """This is the Format 4 working tree.
1352
 
 
1353
 
    This differs from WorkingTree3 by:
1354
 
     - Having a consolidated internal dirstate, stored in a
1355
 
       randomly-accessible sorted file on disk.
1356
 
     - Not having a regular inventory attribute.  One can be synthesized
1357
 
       on demand but this is expensive and should be avoided.
1358
 
 
1359
 
    This is new in bzr 0.15.
1360
 
    """
1361
 
 
1362
 
 
1363
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1364
 
    """This is the Format 5 working tree.
1365
 
 
1366
 
    This differs from WorkingTree4 by:
1367
 
     - Supporting content filtering.
1368
 
 
1369
 
    This is new in bzr 1.11.
1370
 
    """
1371
 
 
1372
 
 
1373
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1374
 
    """This is the Format 6 working tree.
1375
 
 
1376
 
    This differs from WorkingTree5 by:
1377
 
     - Supporting a current view that may mask the set of files in a tree
1378
 
       impacted by most user operations.
1379
 
 
1380
 
    This is new in bzr 1.14.
1381
 
    """
1382
 
 
1383
 
    def _make_views(self):
1384
 
        return views.PathBasedViews(self)
1385
 
 
1386
 
 
1387
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1388
 
 
1389
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1390
 
                   accelerator_tree=None, hardlink=False):
 
1236
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1237
    """The first consolidated dirstate working tree format.
 
1238
 
 
1239
    This format:
 
1240
        - exists within a metadir controlling .bzr
 
1241
        - includes an explicit version marker for the workingtree control
 
1242
          files, separate from the BzrDir format
 
1243
        - modifies the hash cache format
 
1244
        - is new in bzr 0.15
 
1245
        - uses a LockDir to guard access to it.
 
1246
    """
 
1247
 
 
1248
    upgrade_recommended = False
 
1249
 
 
1250
    def get_format_string(self):
 
1251
        """See WorkingTreeFormat.get_format_string()."""
 
1252
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1253
 
 
1254
    def get_format_description(self):
 
1255
        """See WorkingTreeFormat.get_format_description()."""
 
1256
        return "Working tree format 4"
 
1257
 
 
1258
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None):
1391
1259
        """See WorkingTreeFormat.initialize().
1392
1260
 
1393
1261
        :param revision_id: allows creating a working tree at a different
1394
1262
        revision than the branch is at.
1395
 
        :param accelerator_tree: A tree which can be used for retrieving file
1396
 
            contents more quickly than the revision tree, i.e. a workingtree.
1397
 
            The revision tree will be used for cases where accelerator_tree's
1398
 
            content is different.
1399
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1400
 
            where possible.
1401
1263
 
1402
1264
        These trees get an initial random root id, if their repository supports
1403
1265
        rich root data, TREE_ROOT otherwise.
1408
1270
        control_files = self._open_control_files(a_bzrdir)
1409
1271
        control_files.create_lock()
1410
1272
        control_files.lock_write()
1411
 
        transport.put_bytes('format', self.get_format_string(),
1412
 
            mode=a_bzrdir._get_file_mode())
 
1273
        control_files.put_utf8('format', self.get_format_string())
1413
1274
        if from_branch is not None:
1414
1275
            branch = from_branch
1415
1276
        else:
1421
1282
        state = dirstate.DirState.initialize(local_path)
1422
1283
        state.unlock()
1423
1284
        del state
1424
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1285
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1425
1286
                         branch,
1426
1287
                         _format=self,
1427
1288
                         _bzrdir=a_bzrdir,
1429
1290
        wt._new_tree()
1430
1291
        wt.lock_tree_write()
1431
1292
        try:
1432
 
            self._init_custom_control_files(wt)
1433
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1293
            if revision_id in (None, NULL_REVISION):
1434
1294
                if branch.repository.supports_rich_root():
1435
1295
                    wt._set_root_id(generate_ids.gen_root_id())
1436
1296
                else:
1437
1297
                    wt._set_root_id(ROOT_ID)
1438
1298
                wt.flush()
1439
 
            basis = None
1440
 
            # frequently, we will get here due to branching.  The accelerator
1441
 
            # tree will be the tree from the branch, so the desired basis
1442
 
            # tree will often be a parent of the accelerator tree.
1443
 
            if accelerator_tree is not None:
1444
 
                try:
1445
 
                    basis = accelerator_tree.revision_tree(revision_id)
1446
 
                except errors.NoSuchRevision:
1447
 
                    pass
1448
 
            if basis is None:
1449
 
                basis = branch.repository.revision_tree(revision_id)
1450
 
            if revision_id == _mod_revision.NULL_REVISION:
1451
 
                parents_list = []
1452
 
            else:
1453
 
                parents_list = [(revision_id, basis)]
 
1299
            wt.set_last_revision(revision_id)
 
1300
            wt.flush()
 
1301
            basis = wt.basis_tree()
1454
1302
            basis.lock_read()
1455
 
            try:
1456
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1303
            # if the basis has a root id we have to use that; otherwise we use
 
1304
            # a new random one
 
1305
            basis_root_id = basis.get_root_id()
 
1306
            if basis_root_id is not None:
 
1307
                wt._set_root_id(basis_root_id)
1457
1308
                wt.flush()
1458
 
                # if the basis has a root id we have to use that; otherwise we
1459
 
                # use a new random one
1460
 
                basis_root_id = basis.get_root_id()
1461
 
                if basis_root_id is not None:
1462
 
                    wt._set_root_id(basis_root_id)
1463
 
                    wt.flush()
1464
 
                if wt.supports_content_filtering():
1465
 
                    # The original tree may not have the same content filters
1466
 
                    # applied so we can't safely build the inventory delta from
1467
 
                    # the source tree.
1468
 
                    delta_from_tree = False
1469
 
                else:
1470
 
                    delta_from_tree = True
1471
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1472
 
                # because wt4.apply_inventory_delta does not mutate the input
1473
 
                # inventory entries.
1474
 
                transform.build_tree(basis, wt, accelerator_tree,
1475
 
                                     hardlink=hardlink,
1476
 
                                     delta_from_tree=delta_from_tree)
1477
 
            finally:
1478
 
                basis.unlock()
 
1309
            transform.build_tree(basis, wt)
 
1310
            basis.unlock()
1479
1311
        finally:
1480
1312
            control_files.unlock()
1481
1313
            wt.unlock()
1482
1314
        return wt
1483
1315
 
1484
 
    def _init_custom_control_files(self, wt):
1485
 
        """Subclasses with custom control files should override this method.
1486
 
 
1487
 
        The working tree and control files are locked for writing when this
1488
 
        method is called.
1489
 
 
1490
 
        :param wt: the WorkingTree object
1491
 
        """
1492
 
 
1493
1316
    def _open(self, a_bzrdir, control_files):
1494
1317
        """Open the tree itself.
1495
1318
 
1496
1319
        :param a_bzrdir: the dir for the tree.
1497
1320
        :param control_files: the control files for the tree.
1498
1321
        """
1499
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1322
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1500
1323
                           branch=a_bzrdir.open_branch(),
1501
1324
                           _format=self,
1502
1325
                           _bzrdir=a_bzrdir,
1503
1326
                           _control_files=control_files)
1504
1327
 
1505
1328
    def __get_matchingbzrdir(self):
1506
 
        return self._get_matchingbzrdir()
1507
 
 
1508
 
    def _get_matchingbzrdir(self):
1509
 
        """Overrideable method to get a bzrdir for testing."""
1510
1329
        # please test against something that will let us do tree references
1511
1330
        return bzrdir.format_registry.make_bzrdir(
1512
1331
            'dirstate-with-subtree')
1514
1333
    _matchingbzrdir = property(__get_matchingbzrdir)
1515
1334
 
1516
1335
 
1517
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1518
 
    """The first consolidated dirstate working tree format.
1519
 
 
1520
 
    This format:
1521
 
        - exists within a metadir controlling .bzr
1522
 
        - includes an explicit version marker for the workingtree control
1523
 
          files, separate from the BzrDir format
1524
 
        - modifies the hash cache format
1525
 
        - is new in bzr 0.15
1526
 
        - uses a LockDir to guard access to it.
1527
 
    """
1528
 
 
1529
 
    upgrade_recommended = False
1530
 
 
1531
 
    _tree_class = WorkingTree4
1532
 
 
1533
 
    def get_format_string(self):
1534
 
        """See WorkingTreeFormat.get_format_string()."""
1535
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1536
 
 
1537
 
    def get_format_description(self):
1538
 
        """See WorkingTreeFormat.get_format_description()."""
1539
 
        return "Working tree format 4"
1540
 
 
1541
 
 
1542
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1543
 
    """WorkingTree format supporting content filtering.
1544
 
    """
1545
 
 
1546
 
    upgrade_recommended = False
1547
 
 
1548
 
    _tree_class = WorkingTree5
1549
 
 
1550
 
    def get_format_string(self):
1551
 
        """See WorkingTreeFormat.get_format_string()."""
1552
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1553
 
 
1554
 
    def get_format_description(self):
1555
 
        """See WorkingTreeFormat.get_format_description()."""
1556
 
        return "Working tree format 5"
1557
 
 
1558
 
    def supports_content_filtering(self):
1559
 
        return True
1560
 
 
1561
 
 
1562
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1563
 
    """WorkingTree format supporting views.
1564
 
    """
1565
 
 
1566
 
    upgrade_recommended = False
1567
 
 
1568
 
    _tree_class = WorkingTree6
1569
 
 
1570
 
    def get_format_string(self):
1571
 
        """See WorkingTreeFormat.get_format_string()."""
1572
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1573
 
 
1574
 
    def get_format_description(self):
1575
 
        """See WorkingTreeFormat.get_format_description()."""
1576
 
        return "Working tree format 6"
1577
 
 
1578
 
    def _init_custom_control_files(self, wt):
1579
 
        """Subclasses with custom control files should override this method."""
1580
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1581
 
 
1582
 
    def supports_content_filtering(self):
1583
 
        return True
1584
 
 
1585
 
    def supports_views(self):
1586
 
        return True
1587
 
 
1588
 
 
1589
1336
class DirStateRevisionTree(Tree):
1590
 
    """A revision tree pulling the inventory from a dirstate.
1591
 
    
1592
 
    Note that this is one of the historical (ie revision) trees cached in the
1593
 
    dirstate for easy access, not the workingtree.
1594
 
    """
 
1337
    """A revision tree pulling the inventory from a dirstate."""
1595
1338
 
1596
1339
    def __init__(self, dirstate, revision_id, repository):
1597
1340
        self._dirstate = dirstate
1600
1343
        self._inventory = None
1601
1344
        self._locked = 0
1602
1345
        self._dirstate_locked = False
1603
 
        self._repo_supports_tree_reference = getattr(
1604
 
            repository._format, "supports_tree_reference",
1605
 
            False)
1606
1346
 
1607
1347
    def __repr__(self):
1608
1348
        return "<%s of %s in %s>" % \
1611
1351
    def annotate_iter(self, file_id,
1612
1352
                      default_revision=_mod_revision.CURRENT_REVISION):
1613
1353
        """See Tree.annotate_iter"""
1614
 
        text_key = (file_id, self.inventory[file_id].revision)
1615
 
        annotations = self._repository.texts.annotate(text_key)
1616
 
        return [(key[-1], line) for (key, line) in annotations]
 
1354
        w = self._get_weave(file_id)
 
1355
        return w.annotate_iter(self.inventory[file_id].revision)
1617
1356
 
1618
1357
    def _get_ancestors(self, default_revision):
1619
1358
        return set(self._repository.get_ancestry(self._revision_id,
1640
1379
    def get_root_id(self):
1641
1380
        return self.path2id('')
1642
1381
 
1643
 
    def id2path(self, file_id):
1644
 
        "Convert a file-id to a path."
1645
 
        entry = self._get_entry(file_id=file_id)
1646
 
        if entry == (None, None):
1647
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1648
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1649
 
        return path_utf8.decode('utf8')
1650
 
 
1651
 
    def iter_references(self):
1652
 
        if not self._repo_supports_tree_reference:
1653
 
            # When the repo doesn't support references, we will have nothing to
1654
 
            # return
1655
 
            return iter([])
1656
 
        # Otherwise, fall back to the default implementation
1657
 
        return super(DirStateRevisionTree, self).iter_references()
1658
 
 
1659
1382
    def _get_parent_index(self):
1660
1383
        """Return the index in the dirstate referenced by this tree."""
1661
1384
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1666
1389
        If either file_id or path is supplied, it is used as the key to lookup.
1667
1390
        If both are supplied, the fastest lookup is used, and an error is
1668
1391
        raised if they do not both point at the same row.
1669
 
 
 
1392
        
1670
1393
        :param file_id: An optional unicode file_id to be looked up.
1671
1394
        :param path: An optional unicode path to be looked up.
1672
1395
        :return: The dirstate row tuple for path/file_id, or (None, None)
1686
1409
 
1687
1410
        This is relatively expensive: we have to walk the entire dirstate.
1688
1411
        """
1689
 
        if not self._locked:
1690
 
            raise AssertionError(
1691
 
                'cannot generate inventory of an unlocked '
1692
 
                'dirstate revision tree')
 
1412
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1413
            'dirstate revision tree'
1693
1414
        # separate call for profiling - makes it clear where the costs are.
1694
1415
        self._dirstate._read_dirblocks_if_needed()
1695
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1696
 
            raise AssertionError(
1697
 
                'parent %s has disappeared from %s' % (
1698
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1416
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1417
            'parent %s has disappeared from %s' % (
 
1418
            self._revision_id, self._dirstate.get_parent_ids())
1699
1419
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1700
1420
        # This is identical now to the WorkingTree _generate_inventory except
1701
1421
        # for the tree index use.
1702
1422
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1703
1423
        current_id = root_key[2]
1704
 
        if current_entry[parent_index][0] != 'd':
1705
 
            raise AssertionError()
 
1424
        assert current_entry[parent_index][0] == 'd'
1706
1425
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1707
1426
        inv.root.revision = current_entry[parent_index][4]
1708
1427
        # Turn some things into local variables
1748
1467
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1749
1468
                            % entry)
1750
1469
                # These checks cost us around 40ms on a 55k entry tree
1751
 
                if file_id in inv_byid:
1752
 
                    raise AssertionError('file_id %s already in'
1753
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1754
 
                if name_unicode in parent_ie.children:
1755
 
                    raise AssertionError('name %r already in parent'
1756
 
                        % (name_unicode,))
 
1470
                assert file_id not in inv_byid
 
1471
                assert name_unicode not in parent_ie.children
1757
1472
                inv_byid[file_id] = inv_entry
1758
1473
                parent_ie.children[name_unicode] = inv_entry
1759
1474
        self._inventory = inv
1769
1484
            return None
1770
1485
        parent_index = self._get_parent_index()
1771
1486
        last_changed_revision = entry[1][parent_index][4]
1772
 
        try:
1773
 
            rev = self._repository.get_revision(last_changed_revision)
1774
 
        except errors.NoSuchRevision:
1775
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1776
 
        return rev.timestamp
 
1487
        return self._repository.get_revision(last_changed_revision).timestamp
1777
1488
 
1778
1489
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1779
1490
        entry = self._get_entry(file_id=file_id, path=path)
1783
1494
            return parent_details[1]
1784
1495
        return None
1785
1496
 
 
1497
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1498
    def get_weave(self, file_id):
 
1499
        return self._get_weave(file_id)
 
1500
 
 
1501
    def _get_weave(self, file_id):
 
1502
        return self._repository.weave_store.get_weave(file_id,
 
1503
                self._repository.get_transaction())
 
1504
 
1786
1505
    def get_file(self, file_id, path=None):
1787
1506
        return StringIO(self.get_file_text(file_id))
1788
1507
 
 
1508
    def get_file_lines(self, file_id):
 
1509
        ie = self.inventory[file_id]
 
1510
        return self._get_weave(file_id).get_lines(ie.revision)
 
1511
 
1789
1512
    def get_file_size(self, file_id):
1790
 
        """See Tree.get_file_size"""
1791
1513
        return self.inventory[file_id].text_size
1792
1514
 
1793
 
    def get_file_text(self, file_id, path=None):
1794
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1795
 
        return ''.join(content)
 
1515
    def get_file_text(self, file_id):
 
1516
        return ''.join(self.get_file_lines(file_id))
1796
1517
 
1797
1518
    def get_reference_revision(self, file_id, path=None):
1798
1519
        return self.inventory[file_id].reference_revision
1817
1538
        if entry[1][parent_index][0] != 'l':
1818
1539
            return None
1819
1540
        else:
1820
 
            target = entry[1][parent_index][1]
1821
 
            target = target.decode('utf8')
1822
 
            return target
 
1541
            # At present, none of the tree implementations supports non-ascii
 
1542
            # symlink targets. So we will just assume that the dirstate path is
 
1543
            # correct.
 
1544
            return entry[1][parent_index][1]
1823
1545
 
1824
1546
    def get_revision_id(self):
1825
1547
        """Return the revision id for this tree."""
1843
1565
        return bool(self.path2id(filename))
1844
1566
 
1845
1567
    def kind(self, file_id):
1846
 
        entry = self._get_entry(file_id=file_id)[1]
1847
 
        if entry is None:
1848
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1849
 
        parent_index = self._get_parent_index()
1850
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1851
 
 
1852
 
    def stored_kind(self, file_id):
1853
 
        """See Tree.stored_kind"""
1854
 
        return self.kind(file_id)
 
1568
        return self.inventory[file_id].kind
1855
1569
 
1856
1570
    def path_content_summary(self, path):
1857
1571
        """See Tree.path_content_summary."""
1873
1587
            return None
1874
1588
        return ie.executable
1875
1589
 
1876
 
    def is_locked(self):
1877
 
        return self._locked
1878
 
 
1879
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1590
    def list_files(self, include_root=False):
1880
1591
        # We use a standard implementation, because DirStateRevisionTree is
1881
1592
        # dealing with one of the parents of the current state
1882
1593
        inv = self._get_inventory()
1883
 
        if from_dir is None:
1884
 
            from_dir_id = None
1885
 
        else:
1886
 
            from_dir_id = inv.path2id(from_dir)
1887
 
            if from_dir_id is None:
1888
 
                # Directory not versioned
1889
 
                return
1890
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1891
 
        if inv.root is not None and not include_root and from_dir is None:
 
1594
        entries = inv.iter_entries()
 
1595
        if self.inventory.root is not None and not include_root:
1892
1596
            entries.next()
1893
1597
        for path, entry in entries:
1894
1598
            yield path, 'V', entry.kind, entry.file_id, entry
1895
1599
 
1896
1600
    def lock_read(self):
1897
 
        """Lock the tree for a set of operations.
1898
 
 
1899
 
        :return: An object with an unlock method which will release the lock
1900
 
            obtained.
1901
 
        """
 
1601
        """Lock the tree for a set of operations."""
1902
1602
        if not self._locked:
1903
1603
            self._repository.lock_read()
1904
1604
            if self._dirstate._lock_token is None:
1905
1605
                self._dirstate.lock_read()
1906
1606
                self._dirstate_locked = True
1907
1607
        self._locked += 1
1908
 
        return self
1909
1608
 
1910
1609
    def _must_be_locked(self):
1911
1610
        if not self._locked:
1932
1631
                self._dirstate_locked = False
1933
1632
            self._repository.unlock()
1934
1633
 
1935
 
    @needs_read_lock
1936
 
    def supports_tree_reference(self):
1937
 
        return self._repo_supports_tree_reference
1938
 
 
1939
1634
    def walkdirs(self, prefix=""):
1940
1635
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1941
 
        # implementation based on an inventory.
 
1636
        # implementation based on an inventory.  
1942
1637
        # This should be cleaned up to use the much faster Dirstate code
1943
1638
        # So for now, we just build up the parent inventory, and extract
1944
1639
        # it the same way RevisionTree does.
1973
1668
 
1974
1669
class InterDirStateTree(InterTree):
1975
1670
    """Fast path optimiser for changes_from with dirstate trees.
1976
 
 
1977
 
    This is used only when both trees are in the dirstate working file, and
1978
 
    the source is any parent within the dirstate, and the destination is
 
1671
    
 
1672
    This is used only when both trees are in the dirstate working file, and 
 
1673
    the source is any parent within the dirstate, and the destination is 
1979
1674
    the current working tree of the same dirstate.
1980
1675
    """
1981
1676
    # this could be generalized to allow comparisons between any trees in the
1994
1689
        target.set_parent_ids([revid])
1995
1690
        return target.basis_tree(), target
1996
1691
 
1997
 
    @classmethod
1998
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1999
 
        result = klass.make_source_parent_tree(source, target)
2000
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2001
 
        return result
2002
 
 
2003
 
    @classmethod
2004
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2005
 
                                                  target):
2006
 
        from bzrlib.tests.test__dirstate_helpers import \
2007
 
            compiled_dirstate_helpers_feature
2008
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2009
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2010
 
        result = klass.make_source_parent_tree(source, target)
2011
 
        result[1]._iter_changes = ProcessEntryC
2012
 
        return result
2013
 
 
2014
1692
    _matching_from_tree_format = WorkingTreeFormat4()
2015
1693
    _matching_to_tree_format = WorkingTreeFormat4()
2016
 
 
2017
 
    @classmethod
2018
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2019
 
        # This method shouldn't be called, because we have python and C
2020
 
        # specific flavours.
2021
 
        raise NotImplementedError
2022
 
 
2023
 
    def iter_changes(self, include_unchanged=False,
 
1694
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1695
 
 
1696
    def _iter_changes(self, include_unchanged=False,
2024
1697
                      specific_files=None, pb=None, extra_trees=[],
2025
1698
                      require_versioned=True, want_unversioned=False):
2026
1699
        """Return the changes from source to target.
2027
1700
 
2028
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1701
        :return: An iterator that yields tuples. See InterTree._iter_changes
2029
1702
            for details.
2030
1703
        :param specific_files: An optional list of file paths to restrict the
2031
1704
            comparison to. When mapping filenames to ids, all matches in all
2042
1715
            output. An unversioned file is defined as one with (False, False)
2043
1716
            for the versioned pair.
2044
1717
        """
 
1718
        utf8_decode = cache_utf8._utf8_decode
 
1719
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1720
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1721
        # NB: show_status depends on being able to pass in non-versioned files
 
1722
        # and report them as unknown
2045
1723
        # TODO: handle extra trees in the dirstate.
2046
 
        if (extra_trees or specific_files == []):
 
1724
        # TODO: handle comparisons as an empty tree as a different special
 
1725
        # case? mbp 20070226
 
1726
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1727
            or specific_files == []):
2047
1728
            # we can't fast-path these cases (yet)
2048
 
            return super(InterDirStateTree, self).iter_changes(
 
1729
            for f in super(InterDirStateTree, self)._iter_changes(
2049
1730
                include_unchanged, specific_files, pb, extra_trees,
2050
 
                require_versioned, want_unversioned=want_unversioned)
 
1731
                require_versioned, want_unversioned=want_unversioned):
 
1732
                yield f
 
1733
            return
2051
1734
        parent_ids = self.target.get_parent_ids()
2052
 
        if not (self.source._revision_id in parent_ids
2053
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2054
 
            raise AssertionError(
2055
 
                "revision {%s} is not stored in {%s}, but %s "
2056
 
                "can only be used for trees stored in the dirstate"
2057
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1735
        assert (self.source._revision_id in parent_ids), \
 
1736
                "revision {%s} is not stored in {%s}, but %s " \
 
1737
                "can only be used for trees stored in the dirstate" \
 
1738
                % (self.source._revision_id, self.target, self._iter_changes)
2058
1739
        target_index = 0
2059
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1740
        if self.source._revision_id == NULL_REVISION:
2060
1741
            source_index = None
2061
1742
            indices = (target_index,)
2062
1743
        else:
2063
 
            if not (self.source._revision_id in parent_ids):
2064
 
                raise AssertionError(
2065
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2066
 
                    self.source._revision_id, parent_ids))
 
1744
            assert (self.source._revision_id in parent_ids), \
 
1745
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1746
                self.source._revision_id, parent_ids)
2067
1747
            source_index = 1 + parent_ids.index(self.source._revision_id)
2068
 
            indices = (source_index, target_index)
 
1748
            indices = (source_index,target_index)
2069
1749
        # -- make all specific_files utf8 --
2070
1750
        if specific_files:
2071
1751
            specific_files_utf8 = set()
2072
1752
            for path in specific_files:
2073
 
                # Note, if there are many specific files, using cache_utf8
2074
 
                # would be good here.
2075
1753
                specific_files_utf8.add(path.encode('utf8'))
2076
1754
            specific_files = specific_files_utf8
2077
1755
        else:
2078
1756
            specific_files = set([''])
2079
1757
        # -- specific_files is now a utf8 path set --
2080
 
 
2081
1758
        # -- get the state object and prepare it.
2082
1759
        state = self.target.current_dirstate()
2083
1760
        state._read_dirblocks_if_needed()
 
1761
        def _entries_for_path(path):
 
1762
            """Return a list with all the entries that match path for all ids.
 
1763
            """
 
1764
            dirname, basename = os.path.split(path)
 
1765
            key = (dirname, basename, '')
 
1766
            block_index, present = state._find_block_index_from_key(key)
 
1767
            if not present:
 
1768
                # the block which should contain path is absent.
 
1769
                return []
 
1770
            result = []
 
1771
            block = state._dirblocks[block_index][1]
 
1772
            entry_index, _ = state._find_entry_index(key, block)
 
1773
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1774
            while (entry_index < len(block) and
 
1775
                block[entry_index][0][0:2] == key[0:2]):
 
1776
                result.append(block[entry_index])
 
1777
                entry_index += 1
 
1778
            return result
2084
1779
        if require_versioned:
2085
1780
            # -- check all supplied paths are versioned in a search tree. --
2086
 
            not_versioned = []
 
1781
            all_versioned = True
2087
1782
            for path in specific_files:
2088
 
                path_entries = state._entries_for_path(path)
 
1783
                path_entries = _entries_for_path(path)
2089
1784
                if not path_entries:
2090
1785
                    # this specified path is not present at all: error
2091
 
                    not_versioned.append(path)
2092
 
                    continue
 
1786
                    all_versioned = False
 
1787
                    break
2093
1788
                found_versioned = False
2094
1789
                # for each id at this path
2095
1790
                for entry in path_entries:
2102
1797
                if not found_versioned:
2103
1798
                    # none of the indexes was not 'absent' at all ids for this
2104
1799
                    # path.
2105
 
                    not_versioned.append(path)
2106
 
            if len(not_versioned) > 0:
2107
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1800
                    all_versioned = False
 
1801
                    break
 
1802
            if not all_versioned:
 
1803
                raise errors.PathsNotVersionedError(specific_files)
2108
1804
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2109
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1805
        search_specific_files = set()
 
1806
        for path in specific_files:
 
1807
            other_specific_files = specific_files.difference(set([path]))
 
1808
            if not osutils.is_inside_any(other_specific_files, path):
 
1809
                # this is a top level path, we must check it.
 
1810
                search_specific_files.add(path)
 
1811
        # sketch: 
 
1812
        # compare source_index and target_index at or under each element of search_specific_files.
 
1813
        # follow the following comparison table. Note that we only want to do diff operations when
 
1814
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1815
        # for the target.
 
1816
        # cases:
 
1817
        # 
 
1818
        # Source | Target | disk | action
 
1819
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1820
        #        |        |      | diff check on source-target
 
1821
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1822
        #        |        |      | ???
 
1823
        #   r    |  a     |      | add source to search
 
1824
        #   r    |  a     |  a   | 
 
1825
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1826
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1827
        #   a    | fdlt   |      | add new id
 
1828
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1829
        #   a    |  a     |      | not present in either tree, skip
 
1830
        #   a    |  a     |  a   | not present in any tree, skip
 
1831
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1832
        #        |        |      | may not be selected by the users list of paths.
 
1833
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1834
        #        |        |      | may not be selected by the users list of paths.
 
1835
        #  fdlt  | fdlt   |      | content in both: diff them
 
1836
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1837
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1838
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1839
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1840
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1841
        #        |        |      | this id at the other path.
 
1842
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1843
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1844
        #        |        |      | this id at the other path.
 
1845
 
 
1846
        # for all search_indexs in each path at or under each element of
 
1847
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1848
        # relocated path as one to search if its not searched already. If the
 
1849
        # detail is not relocated, add the id.
 
1850
        searched_specific_files = set()
 
1851
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1852
        # Using a list so that we can access the values and change them in
 
1853
        # nested scope. Each one is [path, file_id, entry]
 
1854
        last_source_parent = [None, None]
 
1855
        last_target_parent = [None, None]
2110
1856
 
2111
1857
        use_filesystem_for_exec = (sys.platform != 'win32')
2112
 
        iter_changes = self.target._iter_changes(include_unchanged,
2113
 
            use_filesystem_for_exec, search_specific_files, state,
2114
 
            source_index, target_index, want_unversioned, self.target)
2115
 
        return iter_changes.iter_changes()
 
1858
 
 
1859
        # Just a sentry, so that _process_entry can say that this
 
1860
        # record is handled, but isn't interesting to process (unchanged)
 
1861
        uninteresting = object()
 
1862
 
 
1863
 
 
1864
        old_dirname_to_file_id = {}
 
1865
        new_dirname_to_file_id = {}
 
1866
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1867
        #       keeping a cache of directories that we have seen.
 
1868
 
 
1869
        def _process_entry(entry, path_info):
 
1870
            """Compare an entry and real disk to generate delta information.
 
1871
 
 
1872
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1873
                the path of entry. If None, then the path is considered absent.
 
1874
                (Perhaps we should pass in a concrete entry for this ?)
 
1875
                Basename is returned as a utf8 string because we expect this
 
1876
                tuple will be ignored, and don't want to take the time to
 
1877
                decode.
 
1878
            :return: None if these don't match
 
1879
                     A tuple of information about the change, or
 
1880
                     the object 'uninteresting' if these match, but are
 
1881
                     basically identical.
 
1882
            """
 
1883
            if source_index is None:
 
1884
                source_details = NULL_PARENT_DETAILS
 
1885
            else:
 
1886
                source_details = entry[1][source_index]
 
1887
            target_details = entry[1][target_index]
 
1888
            target_minikind = target_details[0]
 
1889
            if path_info is not None and target_minikind in 'fdlt':
 
1890
                assert target_index == 0
 
1891
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1892
                                                  stat_value=path_info[3])
 
1893
                # The entry may have been modified by update_entry
 
1894
                target_details = entry[1][target_index]
 
1895
                target_minikind = target_details[0]
 
1896
            else:
 
1897
                link_or_sha1 = None
 
1898
            file_id = entry[0][2]
 
1899
            source_minikind = source_details[0]
 
1900
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1901
                # claimed content in both: diff
 
1902
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1903
                #        |        |      | diff check on source-target
 
1904
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1905
                #        |        |      | ???
 
1906
                if source_minikind in 'r':
 
1907
                    # add the source to the search path to find any children it
 
1908
                    # has.  TODO ? : only add if it is a container ?
 
1909
                    if not osutils.is_inside_any(searched_specific_files,
 
1910
                                                 source_details[1]):
 
1911
                        search_specific_files.add(source_details[1])
 
1912
                    # generate the old path; this is needed for stating later
 
1913
                    # as well.
 
1914
                    old_path = source_details[1]
 
1915
                    old_dirname, old_basename = os.path.split(old_path)
 
1916
                    path = pathjoin(entry[0][0], entry[0][1])
 
1917
                    old_entry = state._get_entry(source_index,
 
1918
                                                 path_utf8=old_path)
 
1919
                    # update the source details variable to be the real
 
1920
                    # location.
 
1921
                    source_details = old_entry[1][source_index]
 
1922
                    source_minikind = source_details[0]
 
1923
                else:
 
1924
                    old_dirname = entry[0][0]
 
1925
                    old_basename = entry[0][1]
 
1926
                    old_path = path = None
 
1927
                if path_info is None:
 
1928
                    # the file is missing on disk, show as removed.
 
1929
                    content_change = True
 
1930
                    target_kind = None
 
1931
                    target_exec = False
 
1932
                else:
 
1933
                    # source and target are both versioned and disk file is present.
 
1934
                    target_kind = path_info[2]
 
1935
                    if target_kind == 'directory':
 
1936
                        if path is None:
 
1937
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1938
                        new_dirname_to_file_id[path] = file_id
 
1939
                        if source_minikind != 'd':
 
1940
                            content_change = True
 
1941
                        else:
 
1942
                            # directories have no fingerprint
 
1943
                            content_change = False
 
1944
                        target_exec = False
 
1945
                    elif target_kind == 'file':
 
1946
                        if source_minikind != 'f':
 
1947
                            content_change = True
 
1948
                        else:
 
1949
                            # We could check the size, but we already have the
 
1950
                            # sha1 hash.
 
1951
                            content_change = (link_or_sha1 != source_details[1])
 
1952
                        # Target details is updated at update_entry time
 
1953
                        if use_filesystem_for_exec:
 
1954
                            # We don't need S_ISREG here, because we are sure
 
1955
                            # we are dealing with a file.
 
1956
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1957
                        else:
 
1958
                            target_exec = target_details[3]
 
1959
                    elif target_kind == 'symlink':
 
1960
                        if source_minikind != 'l':
 
1961
                            content_change = True
 
1962
                        else:
 
1963
                            content_change = (link_or_sha1 != source_details[1])
 
1964
                        target_exec = False
 
1965
                    elif target_kind == 'tree-reference':
 
1966
                        if source_minikind != 't':
 
1967
                            content_change = True
 
1968
                        else:
 
1969
                            content_change = False
 
1970
                        target_exec = False
 
1971
                    else:
 
1972
                        raise Exception, "unknown kind %s" % path_info[2]
 
1973
                if source_minikind == 'd':
 
1974
                    if path is None:
 
1975
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1976
                    old_dirname_to_file_id[old_path] = file_id
 
1977
                # parent id is the entry for the path in the target tree
 
1978
                if old_dirname == last_source_parent[0]:
 
1979
                    source_parent_id = last_source_parent[1]
 
1980
                else:
 
1981
                    try:
 
1982
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1983
                    except KeyError:
 
1984
                        source_parent_entry = state._get_entry(source_index,
 
1985
                                                               path_utf8=old_dirname)
 
1986
                        source_parent_id = source_parent_entry[0][2]
 
1987
                    if source_parent_id == entry[0][2]:
 
1988
                        # This is the root, so the parent is None
 
1989
                        source_parent_id = None
 
1990
                    else:
 
1991
                        last_source_parent[0] = old_dirname
 
1992
                        last_source_parent[1] = source_parent_id
 
1993
                new_dirname = entry[0][0]
 
1994
                if new_dirname == last_target_parent[0]:
 
1995
                    target_parent_id = last_target_parent[1]
 
1996
                else:
 
1997
                    try:
 
1998
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1999
                    except KeyError:
 
2000
                        # TODO: We don't always need to do the lookup, because the
 
2001
                        #       parent entry will be the same as the source entry.
 
2002
                        target_parent_entry = state._get_entry(target_index,
 
2003
                                                               path_utf8=new_dirname)
 
2004
                        assert target_parent_entry != (None, None), (
 
2005
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2006
                            % (new_dirname, entry))
 
2007
                        target_parent_id = target_parent_entry[0][2]
 
2008
                    if target_parent_id == entry[0][2]:
 
2009
                        # This is the root, so the parent is None
 
2010
                        target_parent_id = None
 
2011
                    else:
 
2012
                        last_target_parent[0] = new_dirname
 
2013
                        last_target_parent[1] = target_parent_id
 
2014
 
 
2015
                source_exec = source_details[3]
 
2016
                if (include_unchanged
 
2017
                    or content_change
 
2018
                    or source_parent_id != target_parent_id
 
2019
                    or old_basename != entry[0][1]
 
2020
                    or source_exec != target_exec
 
2021
                    ):
 
2022
                    if old_path is None:
 
2023
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2024
                        old_path_u = utf8_decode(old_path)[0]
 
2025
                        path_u = old_path_u
 
2026
                    else:
 
2027
                        old_path_u = utf8_decode(old_path)[0]
 
2028
                        if old_path == path:
 
2029
                            path_u = old_path_u
 
2030
                        else:
 
2031
                            path_u = utf8_decode(path)[0]
 
2032
                    source_kind = _minikind_to_kind[source_minikind]
 
2033
                    return (entry[0][2],
 
2034
                           (old_path_u, path_u),
 
2035
                           content_change,
 
2036
                           (True, True),
 
2037
                           (source_parent_id, target_parent_id),
 
2038
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2039
                           (source_kind, target_kind),
 
2040
                           (source_exec, target_exec))
 
2041
                else:
 
2042
                    return uninteresting
 
2043
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2044
                # looks like a new file
 
2045
                if path_info is not None:
 
2046
                    path = pathjoin(entry[0][0], entry[0][1])
 
2047
                    # parent id is the entry for the path in the target tree
 
2048
                    # TODO: these are the same for an entire directory: cache em.
 
2049
                    parent_id = state._get_entry(target_index,
 
2050
                                                 path_utf8=entry[0][0])[0][2]
 
2051
                    if parent_id == entry[0][2]:
 
2052
                        parent_id = None
 
2053
                    if use_filesystem_for_exec:
 
2054
                        # We need S_ISREG here, because we aren't sure if this
 
2055
                        # is a file or not.
 
2056
                        target_exec = bool(
 
2057
                            stat.S_ISREG(path_info[3].st_mode)
 
2058
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2059
                    else:
 
2060
                        target_exec = target_details[3]
 
2061
                    return (entry[0][2],
 
2062
                           (None, utf8_decode(path)[0]),
 
2063
                           True,
 
2064
                           (False, True),
 
2065
                           (None, parent_id),
 
2066
                           (None, utf8_decode(entry[0][1])[0]),
 
2067
                           (None, path_info[2]),
 
2068
                           (None, target_exec))
 
2069
                else:
 
2070
                    # but its not on disk: we deliberately treat this as just
 
2071
                    # never-present. (Why ?! - RBC 20070224)
 
2072
                    pass
 
2073
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2074
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2075
                # if its still on disk, *and* theres no other entry at this
 
2076
                # path [we dont know this in this routine at the moment -
 
2077
                # perhaps we should change this - then it would be an unknown.
 
2078
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2079
                # parent id is the entry for the path in the target tree
 
2080
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2081
                if parent_id == entry[0][2]:
 
2082
                    parent_id = None
 
2083
                return (entry[0][2],
 
2084
                       (utf8_decode(old_path)[0], None),
 
2085
                       True,
 
2086
                       (True, False),
 
2087
                       (parent_id, None),
 
2088
                       (utf8_decode(entry[0][1])[0], None),
 
2089
                       (_minikind_to_kind[source_minikind], None),
 
2090
                       (source_details[3], None))
 
2091
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2092
                # a rename; could be a true rename, or a rename inherited from
 
2093
                # a renamed parent. TODO: handle this efficiently. Its not
 
2094
                # common case to rename dirs though, so a correct but slow
 
2095
                # implementation will do.
 
2096
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2097
                    search_specific_files.add(target_details[1])
 
2098
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2099
                # neither of the selected trees contain this file,
 
2100
                # so skip over it. This is not currently directly tested, but
 
2101
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2102
                pass
 
2103
            else:
 
2104
                raise AssertionError("don't know how to compare "
 
2105
                    "source_minikind=%r, target_minikind=%r"
 
2106
                    % (source_minikind, target_minikind))
 
2107
                ## import pdb;pdb.set_trace()
 
2108
            return None
 
2109
 
 
2110
        while search_specific_files:
 
2111
            # TODO: the pending list should be lexically sorted?  the
 
2112
            # interface doesn't require it.
 
2113
            current_root = search_specific_files.pop()
 
2114
            current_root_unicode = current_root.decode('utf8')
 
2115
            searched_specific_files.add(current_root)
 
2116
            # process the entries for this containing directory: the rest will be
 
2117
            # found by their parents recursively.
 
2118
            root_entries = _entries_for_path(current_root)
 
2119
            root_abspath = self.target.abspath(current_root_unicode)
 
2120
            try:
 
2121
                root_stat = os.lstat(root_abspath)
 
2122
            except OSError, e:
 
2123
                if e.errno == errno.ENOENT:
 
2124
                    # the path does not exist: let _process_entry know that.
 
2125
                    root_dir_info = None
 
2126
                else:
 
2127
                    # some other random error: hand it up.
 
2128
                    raise
 
2129
            else:
 
2130
                root_dir_info = ('', current_root,
 
2131
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2132
                    root_abspath)
 
2133
                if root_dir_info[2] == 'directory':
 
2134
                    if self.target._directory_is_tree_reference(
 
2135
                        current_root.decode('utf8')):
 
2136
                        root_dir_info = root_dir_info[:2] + \
 
2137
                            ('tree-reference',) + root_dir_info[3:]
 
2138
 
 
2139
            if not root_entries and not root_dir_info:
 
2140
                # this specified path is not present at all, skip it.
 
2141
                continue
 
2142
            path_handled = False
 
2143
            for entry in root_entries:
 
2144
                result = _process_entry(entry, root_dir_info)
 
2145
                if result is not None:
 
2146
                    path_handled = True
 
2147
                    if result is not uninteresting:
 
2148
                        yield result
 
2149
            if want_unversioned and not path_handled and root_dir_info:
 
2150
                new_executable = bool(
 
2151
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2152
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2153
                yield (None,
 
2154
                       (None, current_root_unicode),
 
2155
                       True,
 
2156
                       (False, False),
 
2157
                       (None, None),
 
2158
                       (None, splitpath(current_root_unicode)[-1]),
 
2159
                       (None, root_dir_info[2]),
 
2160
                       (None, new_executable)
 
2161
                      )
 
2162
            initial_key = (current_root, '', '')
 
2163
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2164
            if block_index == 0:
 
2165
                # we have processed the total root already, but because the
 
2166
                # initial key matched it we should skip it here.
 
2167
                block_index +=1
 
2168
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2169
                current_dir_info = None
 
2170
            else:
 
2171
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2172
                try:
 
2173
                    current_dir_info = dir_iterator.next()
 
2174
                except OSError, e:
 
2175
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2176
                    # python 2.5 has e.errno == EINVAL,
 
2177
                    #            and e.winerror == ERROR_DIRECTORY
 
2178
                    e_winerror = getattr(e, 'winerror', None)
 
2179
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2180
                    # there may be directories in the inventory even though
 
2181
                    # this path is not a file on disk: so mark it as end of
 
2182
                    # iterator
 
2183
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2184
                        current_dir_info = None
 
2185
                    elif (sys.platform == 'win32'
 
2186
                          and (e.errno in win_errors
 
2187
                               or e_winerror in win_errors)):
 
2188
                        current_dir_info = None
 
2189
                    else:
 
2190
                        raise
 
2191
                else:
 
2192
                    if current_dir_info[0][0] == '':
 
2193
                        # remove .bzr from iteration
 
2194
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2195
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2196
                        del current_dir_info[1][bzr_index]
 
2197
            # walk until both the directory listing and the versioned metadata
 
2198
            # are exhausted. 
 
2199
            if (block_index < len(state._dirblocks) and
 
2200
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2201
                current_block = state._dirblocks[block_index]
 
2202
            else:
 
2203
                current_block = None
 
2204
            while (current_dir_info is not None or
 
2205
                   current_block is not None):
 
2206
                if (current_dir_info and current_block
 
2207
                    and current_dir_info[0][0] != current_block[0]):
 
2208
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2209
                        # filesystem data refers to paths not covered by the dirblock.
 
2210
                        # this has two possibilities:
 
2211
                        # A) it is versioned but empty, so there is no block for it
 
2212
                        # B) it is not versioned.
 
2213
 
 
2214
                        # if (A) then we need to recurse into it to check for
 
2215
                        # new unknown files or directories.
 
2216
                        # if (B) then we should ignore it, because we don't
 
2217
                        # recurse into unknown directories.
 
2218
                        path_index = 0
 
2219
                        while path_index < len(current_dir_info[1]):
 
2220
                                current_path_info = current_dir_info[1][path_index]
 
2221
                                if want_unversioned:
 
2222
                                    if current_path_info[2] == 'directory':
 
2223
                                        if self.target._directory_is_tree_reference(
 
2224
                                            current_path_info[0].decode('utf8')):
 
2225
                                            current_path_info = current_path_info[:2] + \
 
2226
                                                ('tree-reference',) + current_path_info[3:]
 
2227
                                    new_executable = bool(
 
2228
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2229
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2230
                                    yield (None,
 
2231
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2232
                                        True,
 
2233
                                        (False, False),
 
2234
                                        (None, None),
 
2235
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2236
                                        (None, current_path_info[2]),
 
2237
                                        (None, new_executable))
 
2238
                                # dont descend into this unversioned path if it is
 
2239
                                # a dir
 
2240
                                if current_path_info[2] in ('directory',
 
2241
                                                            'tree-reference'):
 
2242
                                    del current_dir_info[1][path_index]
 
2243
                                    path_index -= 1
 
2244
                                path_index += 1
 
2245
 
 
2246
                        # This dir info has been handled, go to the next
 
2247
                        try:
 
2248
                            current_dir_info = dir_iterator.next()
 
2249
                        except StopIteration:
 
2250
                            current_dir_info = None
 
2251
                    else:
 
2252
                        # We have a dirblock entry for this location, but there
 
2253
                        # is no filesystem path for this. This is most likely
 
2254
                        # because a directory was removed from the disk.
 
2255
                        # We don't have to report the missing directory,
 
2256
                        # because that should have already been handled, but we
 
2257
                        # need to handle all of the files that are contained
 
2258
                        # within.
 
2259
                        for current_entry in current_block[1]:
 
2260
                            # entry referring to file not present on disk.
 
2261
                            # advance the entry only, after processing.
 
2262
                            result = _process_entry(current_entry, None)
 
2263
                            if result is not None:
 
2264
                                if result is not uninteresting:
 
2265
                                    yield result
 
2266
                        block_index +=1
 
2267
                        if (block_index < len(state._dirblocks) and
 
2268
                            osutils.is_inside(current_root,
 
2269
                                              state._dirblocks[block_index][0])):
 
2270
                            current_block = state._dirblocks[block_index]
 
2271
                        else:
 
2272
                            current_block = None
 
2273
                    continue
 
2274
                entry_index = 0
 
2275
                if current_block and entry_index < len(current_block[1]):
 
2276
                    current_entry = current_block[1][entry_index]
 
2277
                else:
 
2278
                    current_entry = None
 
2279
                advance_entry = True
 
2280
                path_index = 0
 
2281
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2282
                    current_path_info = current_dir_info[1][path_index]
 
2283
                    if current_path_info[2] == 'directory':
 
2284
                        if self.target._directory_is_tree_reference(
 
2285
                            current_path_info[0].decode('utf8')):
 
2286
                            current_path_info = current_path_info[:2] + \
 
2287
                                ('tree-reference',) + current_path_info[3:]
 
2288
                else:
 
2289
                    current_path_info = None
 
2290
                advance_path = True
 
2291
                path_handled = False
 
2292
                while (current_entry is not None or
 
2293
                    current_path_info is not None):
 
2294
                    if current_entry is None:
 
2295
                        # the check for path_handled when the path is adnvaced
 
2296
                        # will yield this path if needed.
 
2297
                        pass
 
2298
                    elif current_path_info is None:
 
2299
                        # no path is fine: the per entry code will handle it.
 
2300
                        result = _process_entry(current_entry, current_path_info)
 
2301
                        if result is not None:
 
2302
                            if result is not uninteresting:
 
2303
                                yield result
 
2304
                    elif (current_entry[0][1] != current_path_info[1]
 
2305
                          or current_entry[1][target_index][0] in 'ar'):
 
2306
                        # The current path on disk doesn't match the dirblock
 
2307
                        # record. Either the dirblock is marked as absent, or
 
2308
                        # the file on disk is not present at all in the
 
2309
                        # dirblock. Either way, report about the dirblock
 
2310
                        # entry, and let other code handle the filesystem one.
 
2311
 
 
2312
                        # Compare the basename for these files to determine
 
2313
                        # which comes first
 
2314
                        if current_path_info[1] < current_entry[0][1]:
 
2315
                            # extra file on disk: pass for now, but only
 
2316
                            # increment the path, not the entry
 
2317
                            advance_entry = False
 
2318
                        else:
 
2319
                            # entry referring to file not present on disk.
 
2320
                            # advance the entry only, after processing.
 
2321
                            result = _process_entry(current_entry, None)
 
2322
                            if result is not None:
 
2323
                                if result is not uninteresting:
 
2324
                                    yield result
 
2325
                            advance_path = False
 
2326
                    else:
 
2327
                        result = _process_entry(current_entry, current_path_info)
 
2328
                        if result is not None:
 
2329
                            path_handled = True
 
2330
                            if result is not uninteresting:
 
2331
                                yield result
 
2332
                    if advance_entry and current_entry is not None:
 
2333
                        entry_index += 1
 
2334
                        if entry_index < len(current_block[1]):
 
2335
                            current_entry = current_block[1][entry_index]
 
2336
                        else:
 
2337
                            current_entry = None
 
2338
                    else:
 
2339
                        advance_entry = True # reset the advance flaga
 
2340
                    if advance_path and current_path_info is not None:
 
2341
                        if not path_handled:
 
2342
                            # unversioned in all regards
 
2343
                            if want_unversioned:
 
2344
                                new_executable = bool(
 
2345
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2346
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2347
                                yield (None,
 
2348
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2349
                                    True,
 
2350
                                    (False, False),
 
2351
                                    (None, None),
 
2352
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2353
                                    (None, current_path_info[2]),
 
2354
                                    (None, new_executable))
 
2355
                            # dont descend into this unversioned path if it is
 
2356
                            # a dir
 
2357
                            if current_path_info[2] in ('directory'):
 
2358
                                del current_dir_info[1][path_index]
 
2359
                                path_index -= 1
 
2360
                        # dont descend the disk iterator into any tree 
 
2361
                        # paths.
 
2362
                        if current_path_info[2] == 'tree-reference':
 
2363
                            del current_dir_info[1][path_index]
 
2364
                            path_index -= 1
 
2365
                        path_index += 1
 
2366
                        if path_index < len(current_dir_info[1]):
 
2367
                            current_path_info = current_dir_info[1][path_index]
 
2368
                            if current_path_info[2] == 'directory':
 
2369
                                if self.target._directory_is_tree_reference(
 
2370
                                    current_path_info[0].decode('utf8')):
 
2371
                                    current_path_info = current_path_info[:2] + \
 
2372
                                        ('tree-reference',) + current_path_info[3:]
 
2373
                        else:
 
2374
                            current_path_info = None
 
2375
                        path_handled = False
 
2376
                    else:
 
2377
                        advance_path = True # reset the advance flagg.
 
2378
                if current_block is not None:
 
2379
                    block_index += 1
 
2380
                    if (block_index < len(state._dirblocks) and
 
2381
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2382
                        current_block = state._dirblocks[block_index]
 
2383
                    else:
 
2384
                        current_block = None
 
2385
                if current_dir_info is not None:
 
2386
                    try:
 
2387
                        current_dir_info = dir_iterator.next()
 
2388
                    except StopIteration:
 
2389
                        current_dir_info = None
 
2390
 
2116
2391
 
2117
2392
    @staticmethod
2118
2393
    def is_compatible(source, target):
2119
2394
        # the target must be a dirstate working tree
2120
 
        if not isinstance(target, DirStateWorkingTree):
 
2395
        if not isinstance(target, WorkingTree4):
2121
2396
            return False
2122
 
        # the source must be a revtree or dirstate rev tree.
 
2397
        # the source must be a revtreee or dirstate rev tree.
2123
2398
        if not isinstance(source,
2124
2399
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2125
2400
            return False
2126
2401
        # the source revid must be in the target dirstate
2127
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2402
        if not (source._revision_id == NULL_REVISION or
2128
2403
            source._revision_id in target.get_parent_ids()):
2129
 
            # TODO: what about ghosts? it may well need to
 
2404
            # TODO: what about ghosts? it may well need to 
2130
2405
            # check for them explicitly.
2131
2406
            return False
2132
2407
        return True
2142
2417
 
2143
2418
    def convert(self, tree):
2144
2419
        # lock the control files not the tree, so that we dont get tree
2145
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2420
        # on-unlock behaviours, and so that noone else diddles with the 
2146
2421
        # tree during upgrade.
2147
2422
        tree._control_files.lock_write()
2148
2423
        try:
2174
2449
 
2175
2450
    def update_format(self, tree):
2176
2451
        """Change the format marker."""
2177
 
        tree._transport.put_bytes('format',
2178
 
            self.target_format.get_format_string(),
2179
 
            mode=tree.bzrdir._get_file_mode())
2180
 
 
2181
 
 
2182
 
class Converter4to5(object):
2183
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2184
 
 
2185
 
    def __init__(self):
2186
 
        self.target_format = WorkingTreeFormat5()
2187
 
 
2188
 
    def convert(self, tree):
2189
 
        # lock the control files not the tree, so that we don't get tree
2190
 
        # on-unlock behaviours, and so that no-one else diddles with the
2191
 
        # tree during upgrade.
2192
 
        tree._control_files.lock_write()
2193
 
        try:
2194
 
            self.update_format(tree)
2195
 
        finally:
2196
 
            tree._control_files.unlock()
2197
 
 
2198
 
    def update_format(self, tree):
2199
 
        """Change the format marker."""
2200
 
        tree._transport.put_bytes('format',
2201
 
            self.target_format.get_format_string(),
2202
 
            mode=tree.bzrdir._get_file_mode())
2203
 
 
2204
 
 
2205
 
class Converter4or5to6(object):
2206
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2207
 
 
2208
 
    def __init__(self):
2209
 
        self.target_format = WorkingTreeFormat6()
2210
 
 
2211
 
    def convert(self, tree):
2212
 
        # lock the control files not the tree, so that we don't get tree
2213
 
        # on-unlock behaviours, and so that no-one else diddles with the
2214
 
        # tree during upgrade.
2215
 
        tree._control_files.lock_write()
2216
 
        try:
2217
 
            self.init_custom_control_files(tree)
2218
 
            self.update_format(tree)
2219
 
        finally:
2220
 
            tree._control_files.unlock()
2221
 
 
2222
 
    def init_custom_control_files(self, tree):
2223
 
        """Initialize custom control files."""
2224
 
        tree._transport.put_bytes('views', '',
2225
 
            mode=tree.bzrdir._get_file_mode())
2226
 
 
2227
 
    def update_format(self, tree):
2228
 
        """Change the format marker."""
2229
 
        tree._transport.put_bytes('format',
2230
 
            self.target_format.get_format_string(),
2231
 
            mode=tree.bzrdir._get_file_mode())
 
2452
        tree._control_files.put_utf8('format',
 
2453
            self.target_format.get_format_string())