~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2008-07-09 21:42:24 UTC
  • mto: This revision was merged to the branch mainline in revision 3543.
  • Revision ID: john@arbash-meinel.com-20080709214224-r75k87r6a01pfc3h
Restore a real weave merge to 'bzr merge --weave'.

To do so efficiently, we only add the simple LCAs to the final weave
object, unless we run into complexities with the merge graph.
This gives the same effective result as adding all the texts,
with the advantage of not having to extract all of them.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2012 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007, 2008 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
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
25
 
from __future__ import absolute_import
26
 
 
27
25
from cStringIO import StringIO
28
26
import os
29
27
import sys
30
28
 
31
29
from bzrlib.lazy_import import lazy_import
32
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
33
34
import errno
 
35
import itertools
 
36
import operator
34
37
import stat
 
38
from time import time
 
39
import warnings
35
40
 
 
41
import bzrlib
36
42
from bzrlib import (
37
43
    bzrdir,
38
44
    cache_utf8,
39
 
    config,
40
45
    conflicts as _mod_conflicts,
41
 
    controldir,
42
46
    debug,
 
47
    delta,
43
48
    dirstate,
44
49
    errors,
45
 
    filters as _mod_filters,
46
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
47
54
    osutils,
48
55
    revision as _mod_revision,
49
56
    revisiontree,
 
57
    textui,
50
58
    trace,
51
59
    transform,
52
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
53
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
54
67
""")
55
68
 
 
69
from bzrlib import symbol_versioning
56
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
58
 
from bzrlib.lock import LogicalLockResult
59
 
from bzrlib.lockable_files import LockableFiles
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
60
73
from bzrlib.lockdir import LockDir
61
 
from bzrlib.mutabletree import (
62
 
    MutableTree,
63
 
    needs_tree_write_lock,
64
 
    )
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
65
76
from bzrlib.osutils import (
66
77
    file_kind,
67
78
    isdir,
 
79
    normpath,
68
80
    pathjoin,
 
81
    rand_chars,
69
82
    realpath,
70
83
    safe_unicode,
71
 
    )
72
 
from bzrlib.symbol_versioning import (
73
 
    deprecated_in,
74
 
    deprecated_method,
75
 
    )
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
76
87
from bzrlib.transport.local import LocalTransport
77
 
from bzrlib.tree import (
78
 
    InterTree,
79
 
    InventoryTree,
80
 
    )
81
 
from bzrlib.workingtree import (
82
 
    InventoryWorkingTree,
83
 
    WorkingTree,
84
 
    WorkingTreeFormatMetaDir,
85
 
    )
86
 
 
87
 
 
88
 
class DirStateWorkingTree(InventoryWorkingTree):
 
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
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
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
    """
89
119
 
90
120
    def __init__(self, basedir,
91
121
                 branch,
102
132
        self._format = _format
103
133
        self.bzrdir = _bzrdir
104
134
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
135
        mutter("opening working tree %r", basedir)
106
136
        self._branch = branch
107
137
        self.basedir = realpath(basedir)
108
138
        # if branch is at our basedir and is a format 6 or less
118
148
        #-------------
119
149
        self._setup_directory_is_tree_reference()
120
150
        self._detect_case_handling()
121
 
        self._rules_searcher = None
122
 
        self.views = self._make_views()
123
 
        #--- allow tests to select the dirstate iter_changes implementation
124
 
        self._iter_changes = dirstate._process_entry
125
151
 
126
152
    @needs_tree_write_lock
127
153
    def _add(self, files, ids, kinds):
142
168
            state.add(f, file_id, kind, None, '')
143
169
        self._make_dirty(reset_inventory=True)
144
170
 
145
 
    def _get_check_refs(self):
146
 
        """Return the references needed to perform a check of this tree."""
147
 
        return [('trees', self.last_revision())]
148
 
 
149
171
    def _make_dirty(self, reset_inventory):
150
172
        """Make the tree state dirty.
151
173
 
159
181
    @needs_tree_write_lock
160
182
    def add_reference(self, sub_tree):
161
183
        # use standard implementation, which calls back to self._add
162
 
        #
 
184
        # 
163
185
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
186
        # it's just recorded at the moment of commit. 
165
187
        self._add_reference(sub_tree)
166
188
 
167
189
    def break_lock(self):
203
225
 
204
226
    def _comparison_data(self, entry, path):
205
227
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
228
            WorkingTree3._comparison_data(self, entry, path)
207
229
        # it looks like a plain directory, but it's really a reference -- see
208
230
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
231
        if (self._repo_supports_tree_reference and
 
232
            kind == 'directory' and
 
233
            self._directory_is_tree_reference(path)):
211
234
            kind = 'tree-reference'
212
235
        return kind, executable, stat_value
213
236
 
215
238
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
239
        # mark the tree as dirty post commit - commit
217
240
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
241
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
242
        self._make_dirty(reset_inventory=True)
220
243
        return result
221
244
 
239
262
            return self._dirstate
240
263
        local_path = self.bzrdir.get_workingtree_transport(None
241
264
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
265
        self._dirstate = dirstate.DirState.on_file(local_path)
244
266
        return self._dirstate
245
267
 
246
 
    def _sha1_provider(self):
247
 
        """A function that returns a SHA1Provider suitable for this tree.
248
 
 
249
 
        :return: None if content filtering is not supported by this tree.
250
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
251
 
          form of files, i.e. after read filters are applied.
252
 
        """
253
 
        if self.supports_content_filtering():
254
 
            return ContentFilterAwareSHA1Provider(self)
255
 
        else:
256
 
            return None
257
 
 
258
 
    def _worth_saving_limit(self):
259
 
        """How many hash changes are ok before we must save the dirstate.
260
 
 
261
 
        :return: an integer. -1 means never save.
262
 
        """
263
 
        conf = self.get_config_stack()
264
 
        return conf.get('bzr.workingtree.worth_saving_limit')
265
 
 
266
268
    def filter_unversioned_files(self, paths):
267
269
        """Filter out paths that are versioned.
268
270
 
300
302
 
301
303
    def _generate_inventory(self):
302
304
        """Create and set self.inventory from the dirstate object.
303
 
 
 
305
        
304
306
        This is relatively expensive: we have to walk the entire dirstate.
305
307
        Ideally we would not, and can deprecate this function.
306
308
        """
351
353
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
354
                elif kind == 'tree-reference':
353
355
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
356
                        raise AssertionError(
 
357
                            "repository of %r "
 
358
                            "doesn't support tree references "
 
359
                            "required by entry %r"
 
360
                            % (self, name))
357
361
                    inv_entry.reference_revision = link_or_sha1 or None
358
362
                elif kind != 'symlink':
359
363
                    raise AssertionError("unknown kind %r" % kind)
374
378
        If either file_id or path is supplied, it is used as the key to lookup.
375
379
        If both are supplied, the fastest lookup is used, and an error is
376
380
        raised if they do not both point at the same row.
377
 
 
 
381
        
378
382
        :param file_id: An optional unicode file_id to be looked up.
379
383
        :param path: An optional unicode path to be looked up.
380
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
398
402
        state = self.current_dirstate()
399
403
        if stat_value is None:
400
404
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
405
                stat_value = os.lstat(file_abspath)
402
406
            except OSError, e:
403
407
                if e.errno == errno.ENOENT:
404
408
                    return None
405
409
                else:
406
410
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
409
413
        if entry[1][0][0] == 'f':
410
 
            if link_or_sha1 is None:
411
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
412
 
                try:
413
 
                    sha1 = osutils.sha_file(file_obj)
414
 
                finally:
415
 
                    file_obj.close()
416
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
417
 
                return sha1
418
 
            else:
419
 
                return link_or_sha1
 
414
            return link_or_sha1
420
415
        return None
421
416
 
422
 
    def _get_root_inventory(self):
 
417
    def _get_inventory(self):
423
418
        """Get the inventory for the tree. This is only valid within a lock."""
424
419
        if 'evil' in debug.debug_flags:
425
420
            trace.mutter_callsite(2,
430
425
        self._generate_inventory()
431
426
        return self._inventory
432
427
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
428
    inventory = property(_get_inventory,
438
429
                         doc="Inventory of this Tree")
439
430
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
431
    @needs_read_lock
444
432
    def get_parent_ids(self):
445
433
        """See Tree.get_parent_ids.
446
 
 
 
434
        
447
435
        This implementation requests the ids list from the dirstate file.
448
436
        """
449
437
        return self.current_dirstate().get_parent_ids()
471
459
        return osutils.lexists(pathjoin(
472
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
461
 
474
 
    def has_or_had_id(self, file_id):
475
 
        state = self.current_dirstate()
476
 
        row, parents = self._get_entry(file_id=file_id)
477
 
        return row is not None
478
 
 
479
462
    @needs_read_lock
480
463
    def id2path(self, file_id):
481
464
        "Convert a file-id to a path."
492
475
            return False # Missing entries are not executable
493
476
        return entry[1][0][3] # Executable?
494
477
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
497
481
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
501
484
            entry = self._get_entry(file_id=file_id, path=path)
502
485
            if entry == (None, None):
503
486
                return False
504
487
            return entry[1][0][3]
505
 
        else:
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
506
497
            self._must_be_locked()
507
498
            if not path:
508
499
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
500
            mode = os.lstat(self.abspath(path)).st_mode
510
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
502
 
512
503
    def all_file_ids(self):
556
547
                # path is missing on disk.
557
548
                continue
558
549
 
559
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
560
 
        """See MutableTree._observed_sha1."""
561
 
        state = self.current_dirstate()
562
 
        entry = self._get_entry(file_id=file_id, path=path)
563
 
        state._observed_sha1(entry, sha1, statvalue)
564
 
 
565
550
    def kind(self, file_id):
566
551
        """Return the kind of a file.
567
552
 
579
564
    def _kind(self, relpath):
580
565
        abspath = self.abspath(relpath)
581
566
        kind = file_kind(abspath)
582
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
583
 
            entry = self._get_entry(path=relpath)
584
 
            if entry[1] is not None:
585
 
                if entry[1][0][0] == 't':
586
 
                    kind = 'tree-reference'
 
567
        if (self._repo_supports_tree_reference and
 
568
            kind == 'directory' and
 
569
            self._directory_is_tree_reference(relpath)):
 
570
            kind = 'tree-reference'
587
571
        return kind
588
572
 
589
573
    @needs_read_lock
596
580
            return _mod_revision.NULL_REVISION
597
581
 
598
582
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
583
        """See Branch.lock_read, and WorkingTree.unlock."""
603
584
        self.branch.lock_read()
604
585
        try:
605
586
            self._control_files.lock_read()
618
599
        except:
619
600
            self.branch.unlock()
620
601
            raise
621
 
        return LogicalLockResult(self.unlock)
622
602
 
623
603
    def _lock_self_write(self):
624
604
        """This should be called after the branch is locked."""
639
619
        except:
640
620
            self.branch.unlock()
641
621
            raise
642
 
        return LogicalLockResult(self.unlock)
643
622
 
644
623
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
624
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
625
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
626
        self._lock_self_write()
651
627
 
652
628
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
629
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
630
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
631
        self._lock_self_write()
659
632
 
660
633
    @needs_tree_write_lock
661
634
    def move(self, from_paths, to_dir, after=False):
692
665
 
693
666
        if self._inventory is not None:
694
667
            update_inventory = True
695
 
            inv = self.root_inventory
 
668
            inv = self.inventory
696
669
            to_dir_id = to_entry[0][2]
697
670
            to_dir_ie = inv[to_dir_id]
698
671
        else:
731
704
            from_entry = self._get_entry(path=from_rel)
732
705
            if from_entry == (None, None):
733
706
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
707
                    errors.NotVersionedError(path=str(from_rel)))
735
708
 
736
709
            from_id = from_entry[0][2]
737
710
            to_rel = pathjoin(to_dir, from_tail)
878
851
                rollback_rename()
879
852
                raise
880
853
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
854
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
855
            self._make_dirty(reset_inventory=False)
883
856
 
884
857
        return result
894
867
    @needs_read_lock
895
868
    def path2id(self, path):
896
869
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
870
        path = path.strip('/')
902
871
        entry = self._get_entry(path=path)
903
872
        if entry == (None, None):
916
885
        for tree in trees:
917
886
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
887
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
888
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
889
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
890
        # -- make all paths utf8 --
923
891
        paths_utf8 = set()
981
949
                    all_versioned = False
982
950
                    break
983
951
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
952
                raise errors.PathsNotVersionedError(paths)
986
953
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
954
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
955
        # sketch: 
989
956
        # for all search_indexs in each path at or under each element of
990
957
        # search_paths, if the detail is relocated: add the id, and add the
991
958
        # relocated path as one to search if its not searched already. If the
1037
1004
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
1005
            for dir_name in split_paths:
1039
1006
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
1007
                    raise errors.PathsNotVersionedError(paths)
1042
1008
 
1043
1009
        for dir_name_id, trees_info in found.iteritems():
1044
1010
            for index in search_indexes:
1048
1014
 
1049
1015
    def read_working_inventory(self):
1050
1016
        """Read the working inventory.
1051
 
 
 
1017
        
1052
1018
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
1019
        """
1054
 
        return self.root_inventory
 
1020
        return self.inventory
1055
1021
 
1056
1022
    @needs_read_lock
1057
1023
    def revision_tree(self, revision_id):
1072
1038
    def set_last_revision(self, new_revision):
1073
1039
        """Change the last revision in the working tree."""
1074
1040
        parents = self.get_parent_ids()
1075
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
 
1041
        if new_revision in (NULL_REVISION, None):
1076
1042
            if len(parents) >= 2:
1077
1043
                raise AssertionError(
1078
1044
                    "setting the last parent to none with a pending merge is "
1085
1051
    @needs_tree_write_lock
1086
1052
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1053
        """Set the parent ids to revision_ids.
1088
 
 
 
1054
        
1089
1055
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1056
        for each element of revision_ids from the trees repository. If you have
1091
1057
        tree data already available, it is more efficient to use
1143
1109
                real_trees.append((rev_id, tree))
1144
1110
            else:
1145
1111
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1112
                    self.branch.repository.revision_tree(None)))
1148
1113
                ghosts.append(rev_id)
1149
1114
            accepted_revisions.add(rev_id)
1150
 
        updated = False
1151
 
        if (len(real_trees) == 1
1152
 
            and not ghosts
1153
 
            and self.branch.repository._format.fast_deltas
1154
 
            and isinstance(real_trees[0][1],
1155
 
                revisiontree.InventoryRevisionTree)
1156
 
            and self.get_parent_ids()):
1157
 
            rev_id, rev_tree = real_trees[0]
1158
 
            basis_id = self.get_parent_ids()[0]
1159
 
            # There are times when basis_tree won't be in
1160
 
            # self.branch.repository, (switch, for example)
1161
 
            try:
1162
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1163
 
            except errors.NoSuchRevision:
1164
 
                # Fall back to the set_parent_trees(), since we can't use
1165
 
                # _make_delta if we can't get the RevisionTree
1166
 
                pass
1167
 
            else:
1168
 
                delta = rev_tree.root_inventory._make_delta(
1169
 
                    basis_tree.root_inventory)
1170
 
                dirstate.update_basis_by_delta(delta, rev_id)
1171
 
                updated = True
1172
 
        if not updated:
1173
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1115
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1116
        self._make_dirty(reset_inventory=False)
1175
1117
 
1176
1118
    def _set_root_id(self, file_id):
1196
1138
 
1197
1139
    def unlock(self):
1198
1140
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1141
        # do non-implementation specific cleanup
 
1142
        self._cleanup()
 
1143
 
1199
1144
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1145
            # eventually we should do signature checking during read locks for
1204
1146
            # dirstate updates.
1205
1147
            if self._control_files._lock_mode == 'w':
1268
1210
                # just forget the whole block.
1269
1211
                entry_index = 0
1270
1212
                while entry_index < len(block[1]):
 
1213
                    # Mark this file id as having been removed
1271
1214
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1215
                    ids_to_unversion.discard(entry[0][2])
 
1216
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1217
                                               # entries
 
1218
                        or not state._make_absent(entry)):
1274
1219
                        entry_index += 1
1275
 
                    else:
1276
 
                        # Mark this file id as having been removed
1277
 
                        ids_to_unversion.discard(entry[0][2])
1278
 
                        if not state._make_absent(entry):
1279
 
                            # The block has not shrunk.
1280
 
                            entry_index += 1
1281
1220
                # go to the next block. (At the moment we dont delete empty
1282
1221
                # dirblocks)
1283
1222
                block_index += 1
1304
1243
        # have to change the legacy inventory too.
1305
1244
        if self._inventory is not None:
1306
1245
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
 
1246
                self._inventory.remove_recursive_id(file_id)
1309
1247
 
1310
1248
    @needs_tree_write_lock
1311
1249
    def rename_one(self, from_rel, to_rel, after=False):
1312
1250
        """See WorkingTree.rename_one"""
1313
1251
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1252
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1315
1253
 
1316
1254
    @needs_tree_write_lock
1317
1255
    def apply_inventory_delta(self, changes):
1336
1274
        if self._dirty:
1337
1275
            raise AssertionError("attempting to write an inventory when the "
1338
1276
                "dirstate is dirty will lose pending changes")
1339
 
        had_inventory = self._inventory is not None
1340
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1341
 
        # working inventory. We do this because self.inventory may be inv, or
1342
 
        # may have been modified, and either case would prevent a clean delta
1343
 
        # being created.
1344
 
        self._inventory = None
1345
 
        # generate a delta,
1346
 
        delta = inv._make_delta(self.root_inventory)
1347
 
        # and apply it.
1348
 
        self.apply_inventory_delta(delta)
1349
 
        if had_inventory:
 
1277
        self.current_dirstate().set_state_from_inventory(inv)
 
1278
        self._make_dirty(reset_inventory=False)
 
1279
        if self._inventory is not None:
1350
1280
            self._inventory = inv
1351
1281
        self.flush()
1352
1282
 
1353
 
    @needs_tree_write_lock
1354
 
    def reset_state(self, revision_ids=None):
1355
 
        """Reset the state of the working tree.
1356
 
 
1357
 
        This does a hard-reset to a last-known-good state. This is a way to
1358
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1359
 
        """
1360
 
        if revision_ids is None:
1361
 
            revision_ids = self.get_parent_ids()
1362
 
        if not revision_ids:
1363
 
            base_tree = self.branch.repository.revision_tree(
1364
 
                _mod_revision.NULL_REVISION)
1365
 
            trees = []
1366
 
        else:
1367
 
            trees = zip(revision_ids,
1368
 
                        self.branch.repository.revision_trees(revision_ids))
1369
 
            base_tree = trees[0][1]
1370
 
        state = self.current_dirstate()
1371
 
        # We don't support ghosts yet
1372
 
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
1373
 
 
1374
 
 
1375
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1376
 
 
1377
 
    def __init__(self, tree):
1378
 
        self.tree = tree
1379
 
 
1380
 
    def sha1(self, abspath):
1381
 
        """See dirstate.SHA1Provider.sha1()."""
1382
 
        filters = self.tree._content_filter_stack(
1383
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1384
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1385
 
 
1386
 
    def stat_and_sha1(self, abspath):
1387
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1388
 
        filters = self.tree._content_filter_stack(
1389
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1390
 
        file_obj = file(abspath, 'rb', 65000)
1391
 
        try:
1392
 
            statvalue = os.fstat(file_obj.fileno())
1393
 
            if filters:
1394
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1395
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1396
 
        finally:
1397
 
            file_obj.close()
1398
 
        return statvalue, sha1
1399
 
 
1400
 
 
1401
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1402
 
    """Dirstate working tree that supports content filtering.
1403
 
 
1404
 
    The dirstate holds the hash and size of the canonical form of the file, 
1405
 
    and most methods must return that.
1406
 
    """
1407
 
 
1408
 
    def _file_content_summary(self, path, stat_result):
1409
 
        # This is to support the somewhat obsolete path_content_summary method
1410
 
        # with content filtering: see
1411
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1412
 
        #
1413
 
        # If the dirstate cache is up to date and knows the hash and size,
1414
 
        # return that.
1415
 
        # Otherwise if there are no content filters, return the on-disk size
1416
 
        # and leave the hash blank.
1417
 
        # Otherwise, read and filter the on-disk file and use its size and
1418
 
        # hash.
1419
 
        #
1420
 
        # The dirstate doesn't store the size of the canonical form so we
1421
 
        # can't trust it for content-filtered trees.  We just return None.
1422
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1423
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1424
 
        return ('file', None, executable, dirstate_sha1)
1425
 
 
1426
 
 
1427
 
class WorkingTree4(DirStateWorkingTree):
1428
 
    """This is the Format 4 working tree.
1429
 
 
1430
 
    This differs from WorkingTree by:
1431
 
     - Having a consolidated internal dirstate, stored in a
1432
 
       randomly-accessible sorted file on disk.
1433
 
     - Not having a regular inventory attribute.  One can be synthesized
1434
 
       on demand but this is expensive and should be avoided.
1435
 
 
1436
 
    This is new in bzr 0.15.
1437
 
    """
1438
 
 
1439
 
 
1440
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1441
 
    """This is the Format 5 working tree.
1442
 
 
1443
 
    This differs from WorkingTree4 by:
1444
 
     - Supporting content filtering.
1445
 
 
1446
 
    This is new in bzr 1.11.
1447
 
    """
1448
 
 
1449
 
 
1450
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1451
 
    """This is the Format 6 working tree.
1452
 
 
1453
 
    This differs from WorkingTree5 by:
1454
 
     - Supporting a current view that may mask the set of files in a tree
1455
 
       impacted by most user operations.
1456
 
 
1457
 
    This is new in bzr 1.14.
1458
 
    """
1459
 
 
1460
 
    def _make_views(self):
1461
 
        return views.PathBasedViews(self)
1462
 
 
1463
 
 
1464
 
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
1465
 
 
1466
 
    missing_parent_conflicts = True
1467
 
 
1468
 
    supports_versioned_directories = True
1469
 
 
1470
 
    _lock_class = LockDir
1471
 
    _lock_file_name = 'lock'
1472
 
 
1473
 
    def _open_control_files(self, a_bzrdir):
1474
 
        transport = a_bzrdir.get_workingtree_transport(None)
1475
 
        return LockableFiles(transport, self._lock_file_name,
1476
 
                             self._lock_class)
 
1283
 
 
1284
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1285
    """The first consolidated dirstate working tree format.
 
1286
 
 
1287
    This format:
 
1288
        - exists within a metadir controlling .bzr
 
1289
        - includes an explicit version marker for the workingtree control
 
1290
          files, separate from the BzrDir format
 
1291
        - modifies the hash cache format
 
1292
        - is new in bzr 0.15
 
1293
        - uses a LockDir to guard access to it.
 
1294
    """
 
1295
 
 
1296
    upgrade_recommended = False
 
1297
 
 
1298
    def get_format_string(self):
 
1299
        """See WorkingTreeFormat.get_format_string()."""
 
1300
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1301
 
 
1302
    def get_format_description(self):
 
1303
        """See WorkingTreeFormat.get_format_description()."""
 
1304
        return "Working tree format 4"
1477
1305
 
1478
1306
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
1307
                   accelerator_tree=None, hardlink=False):
1480
1308
        """See WorkingTreeFormat.initialize().
1481
1309
 
1482
1310
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
 
1311
        revision than the branch is at.
1484
1312
        :param accelerator_tree: A tree which can be used for retrieving file
1485
1313
            contents more quickly than the revision tree, i.e. a workingtree.
1486
1314
            The revision tree will be used for cases where accelerator_tree's
1497
1325
        control_files = self._open_control_files(a_bzrdir)
1498
1326
        control_files.create_lock()
1499
1327
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
 
1328
        transport.put_bytes('format', self.get_format_string(),
1501
1329
            mode=a_bzrdir._get_file_mode())
1502
1330
        if from_branch is not None:
1503
1331
            branch = from_branch
1510
1338
        state = dirstate.DirState.initialize(local_path)
1511
1339
        state.unlock()
1512
1340
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1341
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1342
                         branch,
1515
1343
                         _format=self,
1516
1344
                         _bzrdir=a_bzrdir,
1518
1346
        wt._new_tree()
1519
1347
        wt.lock_tree_write()
1520
1348
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1349
            if revision_id in (None, NULL_REVISION):
1523
1350
                if branch.repository.supports_rich_root():
1524
1351
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1352
                else:
1536
1363
                    pass
1537
1364
            if basis is None:
1538
1365
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1366
            if revision_id == NULL_REVISION:
1540
1367
                parents_list = []
1541
1368
            else:
1542
1369
                parents_list = [(revision_id, basis)]
1550
1377
                if basis_root_id is not None:
1551
1378
                    wt._set_root_id(basis_root_id)
1552
1379
                    wt.flush()
1553
 
                if wt.supports_content_filtering():
1554
 
                    # The original tree may not have the same content filters
1555
 
                    # applied so we can't safely build the inventory delta from
1556
 
                    # the source tree.
1557
 
                    delta_from_tree = False
1558
 
                else:
1559
 
                    delta_from_tree = True
1560
1380
                # delta_from_tree is safe even for DirStateRevisionTrees,
1561
1381
                # because wt4.apply_inventory_delta does not mutate the input
1562
1382
                # inventory entries.
1563
1383
                transform.build_tree(basis, wt, accelerator_tree,
1564
 
                                     hardlink=hardlink,
1565
 
                                     delta_from_tree=delta_from_tree)
1566
 
                for hook in MutableTree.hooks['post_build_tree']:
1567
 
                    hook(wt)
 
1384
                                     hardlink=hardlink, delta_from_tree=True)
1568
1385
            finally:
1569
1386
                basis.unlock()
1570
1387
        finally:
1572
1389
            wt.unlock()
1573
1390
        return wt
1574
1391
 
1575
 
    def _init_custom_control_files(self, wt):
1576
 
        """Subclasses with custom control files should override this method.
1577
 
 
1578
 
        The working tree and control files are locked for writing when this
1579
 
        method is called.
1580
 
 
1581
 
        :param wt: the WorkingTree object
1582
 
        """
1583
 
 
1584
 
    def open(self, a_bzrdir, _found=False):
1585
 
        """Return the WorkingTree object for a_bzrdir
1586
 
 
1587
 
        _found is a private parameter, do not use it. It is used to indicate
1588
 
               if format probing has already been done.
1589
 
        """
1590
 
        if not _found:
1591
 
            # we are being called directly and must probe.
1592
 
            raise NotImplementedError
1593
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1594
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1595
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1596
 
        return wt
1597
 
 
1598
1392
    def _open(self, a_bzrdir, control_files):
1599
1393
        """Open the tree itself.
1600
1394
 
1601
1395
        :param a_bzrdir: the dir for the tree.
1602
1396
        :param control_files: the control files for the tree.
1603
1397
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1398
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1399
                           branch=a_bzrdir.open_branch(),
1606
1400
                           _format=self,
1607
1401
                           _bzrdir=a_bzrdir,
1608
1402
                           _control_files=control_files)
1609
1403
 
1610
1404
    def __get_matchingbzrdir(self):
1611
 
        return self._get_matchingbzrdir()
1612
 
 
1613
 
    def _get_matchingbzrdir(self):
1614
 
        """Overrideable method to get a bzrdir for testing."""
1615
1405
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1406
        return bzrdir.format_registry.make_bzrdir(
 
1407
            'dirstate-with-subtree')
1618
1408
 
1619
1409
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1410
 
1621
1411
 
1622
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1623
 
    """The first consolidated dirstate working tree format.
1624
 
 
1625
 
    This format:
1626
 
        - exists within a metadir controlling .bzr
1627
 
        - includes an explicit version marker for the workingtree control
1628
 
          files, separate from the ControlDir format
1629
 
        - modifies the hash cache format
1630
 
        - is new in bzr 0.15
1631
 
        - uses a LockDir to guard access to it.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree4
1637
 
 
1638
 
    @classmethod
1639
 
    def get_format_string(cls):
1640
 
        """See WorkingTreeFormat.get_format_string()."""
1641
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1642
 
 
1643
 
    def get_format_description(self):
1644
 
        """See WorkingTreeFormat.get_format_description()."""
1645
 
        return "Working tree format 4"
1646
 
 
1647
 
 
1648
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1649
 
    """WorkingTree format supporting content filtering.
1650
 
    """
1651
 
 
1652
 
    upgrade_recommended = False
1653
 
 
1654
 
    _tree_class = WorkingTree5
1655
 
 
1656
 
    @classmethod
1657
 
    def get_format_string(cls):
1658
 
        """See WorkingTreeFormat.get_format_string()."""
1659
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1660
 
 
1661
 
    def get_format_description(self):
1662
 
        """See WorkingTreeFormat.get_format_description()."""
1663
 
        return "Working tree format 5"
1664
 
 
1665
 
    def supports_content_filtering(self):
1666
 
        return True
1667
 
 
1668
 
 
1669
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1670
 
    """WorkingTree format supporting views.
1671
 
    """
1672
 
 
1673
 
    upgrade_recommended = False
1674
 
 
1675
 
    _tree_class = WorkingTree6
1676
 
 
1677
 
    @classmethod
1678
 
    def get_format_string(cls):
1679
 
        """See WorkingTreeFormat.get_format_string()."""
1680
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1681
 
 
1682
 
    def get_format_description(self):
1683
 
        """See WorkingTreeFormat.get_format_description()."""
1684
 
        return "Working tree format 6"
1685
 
 
1686
 
    def _init_custom_control_files(self, wt):
1687
 
        """Subclasses with custom control files should override this method."""
1688
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1689
 
 
1690
 
    def supports_content_filtering(self):
1691
 
        return True
1692
 
 
1693
 
    def supports_views(self):
1694
 
        return True
1695
 
 
1696
 
    def _get_matchingbzrdir(self):
1697
 
        """Overrideable method to get a bzrdir for testing."""
1698
 
        # We use 'development-subtree' instead of '2a', because we have a
1699
 
        # few tests that want to test tree references
1700
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
1701
 
 
1702
 
 
1703
 
class DirStateRevisionTree(InventoryTree):
1704
 
    """A revision tree pulling the inventory from a dirstate.
1705
 
    
1706
 
    Note that this is one of the historical (ie revision) trees cached in the
1707
 
    dirstate for easy access, not the workingtree.
1708
 
    """
 
1412
class DirStateRevisionTree(Tree):
 
1413
    """A revision tree pulling the inventory from a dirstate."""
1709
1414
 
1710
1415
    def __init__(self, dirstate, revision_id, repository):
1711
1416
        self._dirstate = dirstate
1725
1430
    def annotate_iter(self, file_id,
1726
1431
                      default_revision=_mod_revision.CURRENT_REVISION):
1727
1432
        """See Tree.annotate_iter"""
1728
 
        text_key = (file_id, self.get_file_revision(file_id))
 
1433
        text_key = (file_id, self.inventory[file_id].revision)
1729
1434
        annotations = self._repository.texts.annotate(text_key)
1730
1435
        return [(key[-1], line) for (key, line) in annotations]
1731
1436
 
 
1437
    def _get_ancestors(self, default_revision):
 
1438
        return set(self._repository.get_ancestry(self._revision_id,
 
1439
                                                 topo_sorted=False))
1732
1440
    def _comparison_data(self, entry, path):
1733
1441
        """See Tree._comparison_data."""
1734
1442
        if entry is None:
1777
1485
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1486
        If both are supplied, the fastest lookup is used, and an error is
1779
1487
        raised if they do not both point at the same row.
1780
 
 
 
1488
        
1781
1489
        :param file_id: An optional unicode file_id to be looked up.
1782
1490
        :param path: An optional unicode path to be looked up.
1783
1491
        :return: The dirstate row tuple for path/file_id, or (None, None)
1787
1495
        if path is not None:
1788
1496
            path = path.encode('utf8')
1789
1497
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1498
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1499
 
1793
1500
    def _generate_inventory(self):
1794
1501
        """Create and set self.inventory from the dirstate object.
1851
1558
                elif kind == 'directory':
1852
1559
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1560
                elif kind == 'symlink':
 
1561
                    inv_entry.executable = False
 
1562
                    inv_entry.text_size = None
1854
1563
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1564
                elif kind == 'tree-reference':
1856
1565
                    inv_entry.reference_revision = fingerprint or None
1876
1585
        # Make sure the file exists
1877
1586
        entry = self._get_entry(file_id, path=path)
1878
1587
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1588
            return None
1880
1589
        parent_index = self._get_parent_index()
1881
1590
        last_changed_revision = entry[1][parent_index][4]
1882
 
        try:
1883
 
            rev = self._repository.get_revision(last_changed_revision)
1884
 
        except errors.NoSuchRevision:
1885
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1886
 
        return rev.timestamp
 
1591
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1592
 
1888
1593
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1594
        entry = self._get_entry(file_id=file_id, path=path)
1893
1598
            return parent_details[1]
1894
1599
        return None
1895
1600
 
1896
 
    @needs_read_lock
1897
 
    def get_file_revision(self, file_id):
1898
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1899
 
        return inv[inv_file_id].revision
1900
 
 
1901
1601
    def get_file(self, file_id, path=None):
1902
1602
        return StringIO(self.get_file_text(file_id))
1903
1603
 
 
1604
    def get_file_lines(self, file_id):
 
1605
        return osutils.split_lines(self.get_file_text(file_id))
 
1606
 
1904
1607
    def get_file_size(self, file_id):
1905
1608
        """See Tree.get_file_size"""
1906
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1907
 
        return inv[inv_file_id].text_size
 
1609
        return self.inventory[file_id].text_size
1908
1610
 
1909
 
    def get_file_text(self, file_id, path=None):
1910
 
        content = None
1911
 
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
1912
 
            if content is not None:
1913
 
                raise AssertionError('iter_files_bytes returned'
1914
 
                    ' too many entries')
1915
 
            # For each entry returned by iter_files_bytes, we must consume the
1916
 
            # content_iter before we step the files iterator.
1917
 
            content = ''.join(content_iter)
1918
 
        if content is None:
1919
 
            raise AssertionError('iter_files_bytes did not return'
1920
 
                ' the requested data')
1921
 
        return content
 
1611
    def get_file_text(self, file_id):
 
1612
        return list(self.iter_files_bytes([(file_id, None)]))[0][1]
1922
1613
 
1923
1614
    def get_reference_revision(self, file_id, path=None):
1924
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1925
 
        return inv[inv_file_id].reference_revision
 
1615
        return self.inventory[file_id].reference_revision
1926
1616
 
1927
1617
    def iter_files_bytes(self, desired_files):
1928
1618
        """See Tree.iter_files_bytes.
1938
1628
                                       identifier))
1939
1629
        return self._repository.iter_files_bytes(repo_desired_files)
1940
1630
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1631
    def get_symlink_target(self, file_id):
1942
1632
        entry = self._get_entry(file_id=file_id)
1943
1633
        parent_index = self._get_parent_index()
1944
1634
        if entry[1][parent_index][0] != 'l':
1945
1635
            return None
1946
1636
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1637
            # At present, none of the tree implementations supports non-ascii
 
1638
            # symlink targets. So we will just assume that the dirstate path is
 
1639
            # correct.
 
1640
            return entry[1][parent_index][1]
1950
1641
 
1951
1642
    def get_revision_id(self):
1952
1643
        """Return the revision id for this tree."""
1953
1644
        return self._revision_id
1954
1645
 
1955
 
    def _get_root_inventory(self):
 
1646
    def _get_inventory(self):
1956
1647
        if self._inventory is not None:
1957
1648
            return self._inventory
1958
1649
        self._must_be_locked()
1959
1650
        self._generate_inventory()
1960
1651
        return self._inventory
1961
1652
 
1962
 
    root_inventory = property(_get_root_inventory,
1963
 
                         doc="Inventory of this Tree")
1964
 
 
1965
 
    @deprecated_method(deprecated_in((2, 5, 0)))
1966
 
    def _get_inventory(self):
1967
 
        return self.root_inventory
1968
 
 
1969
1653
    inventory = property(_get_inventory,
1970
1654
                         doc="Inventory of this Tree")
1971
1655
 
1980
1664
        entry = self._get_entry(file_id=file_id)[1]
1981
1665
        if entry is None:
1982
1666
            raise errors.NoSuchId(tree=self, file_id=file_id)
1983
 
        parent_index = self._get_parent_index()
1984
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1667
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1985
1668
 
1986
1669
    def stored_kind(self, file_id):
1987
1670
        """See Tree.stored_kind"""
1989
1672
 
1990
1673
    def path_content_summary(self, path):
1991
1674
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
 
1675
        id = self.inventory.path2id(path)
 
1676
        if id is None:
1994
1677
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
 
1678
        entry = self._inventory[id]
1996
1679
        kind = entry.kind
1997
1680
        if kind == 'file':
1998
1681
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
2002
1685
            return (kind, None, None, None)
2003
1686
 
2004
1687
    def is_executable(self, file_id, path=None):
2005
 
        inv, inv_file_id = self._unpack_file_id(file_id)
2006
 
        ie = inv[inv_file_id]
 
1688
        ie = self.inventory[file_id]
2007
1689
        if ie.kind != "file":
2008
 
            return False
 
1690
            return None
2009
1691
        return ie.executable
2010
1692
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1693
    def list_files(self, include_root=False):
2015
1694
        # We use a standard implementation, because DirStateRevisionTree is
2016
1695
        # dealing with one of the parents of the current state
2017
 
        if from_dir is None:
2018
 
            inv = self.root_inventory
2019
 
            from_dir_id = None
2020
 
        else:
2021
 
            inv, from_dir_id = self._path2inv_file_id(from_dir)
2022
 
            if from_dir_id is None:
2023
 
                # Directory not versioned
2024
 
                return
2025
 
        # FIXME: Support nested trees
2026
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
2027
 
        if inv.root is not None and not include_root and from_dir is None:
 
1696
        inv = self._get_inventory()
 
1697
        entries = inv.iter_entries()
 
1698
        if self.inventory.root is not None and not include_root:
2028
1699
            entries.next()
2029
1700
        for path, entry in entries:
2030
1701
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1702
 
2032
1703
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1704
        """Lock the tree for a set of operations."""
2037
1705
        if not self._locked:
2038
1706
            self._repository.lock_read()
2039
1707
            if self._dirstate._lock_token is None:
2040
1708
                self._dirstate.lock_read()
2041
1709
                self._dirstate_locked = True
2042
1710
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1711
 
2045
1712
    def _must_be_locked(self):
2046
1713
        if not self._locked:
2050
1717
    def path2id(self, path):
2051
1718
        """Return the id for path in this tree."""
2052
1719
        # lookup by path: faster than splitting and walking the ivnentory.
2053
 
        if isinstance(path, list):
2054
 
            if path == []:
2055
 
                path = [""]
2056
 
            path = osutils.pathjoin(*path)
2057
1720
        entry = self._get_entry(path=path)
2058
1721
        if entry == (None, None):
2059
1722
            return None
2077
1740
 
2078
1741
    def walkdirs(self, prefix=""):
2079
1742
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1743
        # implementation based on an inventory.  
2081
1744
        # This should be cleaned up to use the much faster Dirstate code
2082
1745
        # So for now, we just build up the parent inventory, and extract
2083
1746
        # it the same way RevisionTree does.
2084
1747
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1748
        inv = self._get_inventory()
2086
1749
        top_id = inv.path2id(prefix)
2087
1750
        if top_id is None:
2088
1751
            pending = []
2112
1775
 
2113
1776
class InterDirStateTree(InterTree):
2114
1777
    """Fast path optimiser for changes_from with dirstate trees.
2115
 
 
2116
 
    This is used only when both trees are in the dirstate working file, and
2117
 
    the source is any parent within the dirstate, and the destination is
 
1778
    
 
1779
    This is used only when both trees are in the dirstate working file, and 
 
1780
    the source is any parent within the dirstate, and the destination is 
2118
1781
    the current working tree of the same dirstate.
2119
1782
    """
2120
1783
    # this could be generalized to allow comparisons between any trees in the
2129
1792
    def make_source_parent_tree(source, target):
2130
1793
        """Change the source tree into a parent of the target."""
2131
1794
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1795
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1796
        target.set_parent_ids([revid])
2134
1797
        return target.basis_tree(), target
2135
1798
 
2136
 
    @classmethod
2137
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2138
 
        result = klass.make_source_parent_tree(source, target)
2139
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2140
 
        return result
2141
 
 
2142
 
    @classmethod
2143
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2144
 
                                                  target):
2145
 
        from bzrlib.tests.test__dirstate_helpers import \
2146
 
            compiled_dirstate_helpers_feature
2147
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2148
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2149
 
        result = klass.make_source_parent_tree(source, target)
2150
 
        result[1]._iter_changes = ProcessEntryC
2151
 
        return result
2152
 
 
2153
1799
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1800
    _matching_to_tree_format = WorkingTreeFormat4()
2155
 
 
2156
 
    @classmethod
2157
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2158
 
        # This method shouldn't be called, because we have python and C
2159
 
        # specific flavours.
2160
 
        raise NotImplementedError
 
1801
    _test_mutable_trees_to_test_trees = make_source_parent_tree
2161
1802
 
2162
1803
    def iter_changes(self, include_unchanged=False,
2163
1804
                      specific_files=None, pb=None, extra_trees=[],
2181
1822
            output. An unversioned file is defined as one with (False, False)
2182
1823
            for the versioned pair.
2183
1824
        """
 
1825
        utf8_decode = cache_utf8._utf8_decode
 
1826
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1827
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1828
        # NB: show_status depends on being able to pass in non-versioned files
 
1829
        # and report them as unknown
2184
1830
        # TODO: handle extra trees in the dirstate.
2185
1831
        if (extra_trees or specific_files == []):
2186
1832
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1833
            for f in super(InterDirStateTree, self).iter_changes(
2188
1834
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1835
                require_versioned, want_unversioned=want_unversioned):
 
1836
                yield f
 
1837
            return
2190
1838
        parent_ids = self.target.get_parent_ids()
2191
1839
        if not (self.source._revision_id in parent_ids
2192
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
 
1840
                or self.source._revision_id == NULL_REVISION):
2193
1841
            raise AssertionError(
2194
1842
                "revision {%s} is not stored in {%s}, but %s "
2195
1843
                "can only be used for trees stored in the dirstate"
2196
1844
                % (self.source._revision_id, self.target, self.iter_changes))
2197
1845
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1846
        if self.source._revision_id == NULL_REVISION:
2199
1847
            source_index = None
2200
1848
            indices = (target_index,)
2201
1849
        else:
2209
1857
        if specific_files:
2210
1858
            specific_files_utf8 = set()
2211
1859
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1860
                specific_files_utf8.add(path.encode('utf8'))
2215
1861
            specific_files = specific_files_utf8
2216
1862
        else:
2217
1863
            specific_files = set([''])
2218
1864
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1865
        # -- get the state object and prepare it.
2221
1866
        state = self.target.current_dirstate()
2222
1867
        state._read_dirblocks_if_needed()
 
1868
        def _entries_for_path(path):
 
1869
            """Return a list with all the entries that match path for all ids.
 
1870
            """
 
1871
            dirname, basename = os.path.split(path)
 
1872
            key = (dirname, basename, '')
 
1873
            block_index, present = state._find_block_index_from_key(key)
 
1874
            if not present:
 
1875
                # the block which should contain path is absent.
 
1876
                return []
 
1877
            result = []
 
1878
            block = state._dirblocks[block_index][1]
 
1879
            entry_index, _ = state._find_entry_index(key, block)
 
1880
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1881
            while (entry_index < len(block) and
 
1882
                block[entry_index][0][0:2] == key[0:2]):
 
1883
                result.append(block[entry_index])
 
1884
                entry_index += 1
 
1885
            return result
2223
1886
        if require_versioned:
2224
1887
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1888
            all_versioned = True
2226
1889
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1890
                path_entries = _entries_for_path(path)
2228
1891
                if not path_entries:
2229
1892
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1893
                    all_versioned = False
 
1894
                    break
2232
1895
                found_versioned = False
2233
1896
                # for each id at this path
2234
1897
                for entry in path_entries:
2241
1904
                if not found_versioned:
2242
1905
                    # none of the indexes was not 'absent' at all ids for this
2243
1906
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1907
                    all_versioned = False
 
1908
                    break
 
1909
            if not all_versioned:
 
1910
                raise errors.PathsNotVersionedError(specific_files)
2247
1911
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1912
        search_specific_files = set()
 
1913
        for path in specific_files:
 
1914
            other_specific_files = specific_files.difference(set([path]))
 
1915
            if not osutils.is_inside_any(other_specific_files, path):
 
1916
                # this is a top level path, we must check it.
 
1917
                search_specific_files.add(path)
 
1918
        # sketch: 
 
1919
        # compare source_index and target_index at or under each element of search_specific_files.
 
1920
        # follow the following comparison table. Note that we only want to do diff operations when
 
1921
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1922
        # for the target.
 
1923
        # cases:
 
1924
        # 
 
1925
        # Source | Target | disk | action
 
1926
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1927
        #        |        |      | diff check on source-target
 
1928
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1929
        #        |        |      | ???
 
1930
        #   r    |  a     |      | add source to search
 
1931
        #   r    |  a     |  a   | 
 
1932
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1933
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1934
        #   a    | fdlt   |      | add new id
 
1935
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1936
        #   a    |  a     |      | not present in either tree, skip
 
1937
        #   a    |  a     |  a   | not present in any tree, skip
 
1938
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1939
        #        |        |      | may not be selected by the users list of paths.
 
1940
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1941
        #        |        |      | may not be selected by the users list of paths.
 
1942
        #  fdlt  | fdlt   |      | content in both: diff them
 
1943
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1944
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1945
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1946
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1947
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1948
        #        |        |      | this id at the other path.
 
1949
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1950
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1951
        #        |        |      | this id at the other path.
 
1952
 
 
1953
        # for all search_indexs in each path at or under each element of
 
1954
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1955
        # relocated path as one to search if its not searched already. If the
 
1956
        # detail is not relocated, add the id.
 
1957
        searched_specific_files = set()
 
1958
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1959
        # Using a list so that we can access the values and change them in
 
1960
        # nested scope. Each one is [path, file_id, entry]
 
1961
        last_source_parent = [None, None]
 
1962
        last_target_parent = [None, None]
2249
1963
 
2250
1964
        use_filesystem_for_exec = (sys.platform != 'win32')
2251
 
        iter_changes = self.target._iter_changes(include_unchanged,
2252
 
            use_filesystem_for_exec, search_specific_files, state,
2253
 
            source_index, target_index, want_unversioned, self.target)
2254
 
        return iter_changes.iter_changes()
 
1965
 
 
1966
        # Just a sentry, so that _process_entry can say that this
 
1967
        # record is handled, but isn't interesting to process (unchanged)
 
1968
        uninteresting = object()
 
1969
 
 
1970
 
 
1971
        old_dirname_to_file_id = {}
 
1972
        new_dirname_to_file_id = {}
 
1973
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1974
        #       keeping a cache of directories that we have seen.
 
1975
 
 
1976
        def _process_entry(entry, path_info):
 
1977
            """Compare an entry and real disk to generate delta information.
 
1978
 
 
1979
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1980
                the path of entry. If None, then the path is considered absent.
 
1981
                (Perhaps we should pass in a concrete entry for this ?)
 
1982
                Basename is returned as a utf8 string because we expect this
 
1983
                tuple will be ignored, and don't want to take the time to
 
1984
                decode.
 
1985
            :return: None if these don't match
 
1986
                     A tuple of information about the change, or
 
1987
                     the object 'uninteresting' if these match, but are
 
1988
                     basically identical.
 
1989
            """
 
1990
            if source_index is None:
 
1991
                source_details = NULL_PARENT_DETAILS
 
1992
            else:
 
1993
                source_details = entry[1][source_index]
 
1994
            target_details = entry[1][target_index]
 
1995
            target_minikind = target_details[0]
 
1996
            if path_info is not None and target_minikind in 'fdlt':
 
1997
                if not (target_index == 0):
 
1998
                    raise AssertionError()
 
1999
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
2000
                                                  stat_value=path_info[3])
 
2001
                # The entry may have been modified by update_entry
 
2002
                target_details = entry[1][target_index]
 
2003
                target_minikind = target_details[0]
 
2004
            else:
 
2005
                link_or_sha1 = None
 
2006
            file_id = entry[0][2]
 
2007
            source_minikind = source_details[0]
 
2008
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
2009
                # claimed content in both: diff
 
2010
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
2011
                #        |        |      | diff check on source-target
 
2012
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
2013
                #        |        |      | ???
 
2014
                if source_minikind in 'r':
 
2015
                    # add the source to the search path to find any children it
 
2016
                    # has.  TODO ? : only add if it is a container ?
 
2017
                    if not osutils.is_inside_any(searched_specific_files,
 
2018
                                                 source_details[1]):
 
2019
                        search_specific_files.add(source_details[1])
 
2020
                    # generate the old path; this is needed for stating later
 
2021
                    # as well.
 
2022
                    old_path = source_details[1]
 
2023
                    old_dirname, old_basename = os.path.split(old_path)
 
2024
                    path = pathjoin(entry[0][0], entry[0][1])
 
2025
                    old_entry = state._get_entry(source_index,
 
2026
                                                 path_utf8=old_path)
 
2027
                    # update the source details variable to be the real
 
2028
                    # location.
 
2029
                    if old_entry == (None, None):
 
2030
                        raise errors.CorruptDirstate(state._filename,
 
2031
                            "entry '%s/%s' is considered renamed from %r"
 
2032
                            " but source does not exist\n"
 
2033
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
2034
                    source_details = old_entry[1][source_index]
 
2035
                    source_minikind = source_details[0]
 
2036
                else:
 
2037
                    old_dirname = entry[0][0]
 
2038
                    old_basename = entry[0][1]
 
2039
                    old_path = path = None
 
2040
                if path_info is None:
 
2041
                    # the file is missing on disk, show as removed.
 
2042
                    content_change = True
 
2043
                    target_kind = None
 
2044
                    target_exec = False
 
2045
                else:
 
2046
                    # source and target are both versioned and disk file is present.
 
2047
                    target_kind = path_info[2]
 
2048
                    if target_kind == 'directory':
 
2049
                        if path is None:
 
2050
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2051
                        new_dirname_to_file_id[path] = file_id
 
2052
                        if source_minikind != 'd':
 
2053
                            content_change = True
 
2054
                        else:
 
2055
                            # directories have no fingerprint
 
2056
                            content_change = False
 
2057
                        target_exec = False
 
2058
                    elif target_kind == 'file':
 
2059
                        if source_minikind != 'f':
 
2060
                            content_change = True
 
2061
                        else:
 
2062
                            # We could check the size, but we already have the
 
2063
                            # sha1 hash.
 
2064
                            content_change = (link_or_sha1 != source_details[1])
 
2065
                        # Target details is updated at update_entry time
 
2066
                        if use_filesystem_for_exec:
 
2067
                            # We don't need S_ISREG here, because we are sure
 
2068
                            # we are dealing with a file.
 
2069
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2070
                        else:
 
2071
                            target_exec = target_details[3]
 
2072
                    elif target_kind == 'symlink':
 
2073
                        if source_minikind != 'l':
 
2074
                            content_change = True
 
2075
                        else:
 
2076
                            content_change = (link_or_sha1 != source_details[1])
 
2077
                        target_exec = False
 
2078
                    elif target_kind == 'tree-reference':
 
2079
                        if source_minikind != 't':
 
2080
                            content_change = True
 
2081
                        else:
 
2082
                            content_change = False
 
2083
                        target_exec = False
 
2084
                    else:
 
2085
                        raise Exception, "unknown kind %s" % path_info[2]
 
2086
                if source_minikind == 'd':
 
2087
                    if path is None:
 
2088
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2089
                    old_dirname_to_file_id[old_path] = file_id
 
2090
                # parent id is the entry for the path in the target tree
 
2091
                if old_dirname == last_source_parent[0]:
 
2092
                    source_parent_id = last_source_parent[1]
 
2093
                else:
 
2094
                    try:
 
2095
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2096
                    except KeyError:
 
2097
                        source_parent_entry = state._get_entry(source_index,
 
2098
                                                               path_utf8=old_dirname)
 
2099
                        source_parent_id = source_parent_entry[0][2]
 
2100
                    if source_parent_id == entry[0][2]:
 
2101
                        # This is the root, so the parent is None
 
2102
                        source_parent_id = None
 
2103
                    else:
 
2104
                        last_source_parent[0] = old_dirname
 
2105
                        last_source_parent[1] = source_parent_id
 
2106
                new_dirname = entry[0][0]
 
2107
                if new_dirname == last_target_parent[0]:
 
2108
                    target_parent_id = last_target_parent[1]
 
2109
                else:
 
2110
                    try:
 
2111
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2112
                    except KeyError:
 
2113
                        # TODO: We don't always need to do the lookup, because the
 
2114
                        #       parent entry will be the same as the source entry.
 
2115
                        target_parent_entry = state._get_entry(target_index,
 
2116
                                                               path_utf8=new_dirname)
 
2117
                        if target_parent_entry == (None, None):
 
2118
                            raise AssertionError(
 
2119
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2120
                                % (new_dirname, entry))
 
2121
                        target_parent_id = target_parent_entry[0][2]
 
2122
                    if target_parent_id == entry[0][2]:
 
2123
                        # This is the root, so the parent is None
 
2124
                        target_parent_id = None
 
2125
                    else:
 
2126
                        last_target_parent[0] = new_dirname
 
2127
                        last_target_parent[1] = target_parent_id
 
2128
 
 
2129
                source_exec = source_details[3]
 
2130
                if (include_unchanged
 
2131
                    or content_change
 
2132
                    or source_parent_id != target_parent_id
 
2133
                    or old_basename != entry[0][1]
 
2134
                    or source_exec != target_exec
 
2135
                    ):
 
2136
                    if old_path is None:
 
2137
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2138
                        old_path_u = utf8_decode(old_path)[0]
 
2139
                        path_u = old_path_u
 
2140
                    else:
 
2141
                        old_path_u = utf8_decode(old_path)[0]
 
2142
                        if old_path == path:
 
2143
                            path_u = old_path_u
 
2144
                        else:
 
2145
                            path_u = utf8_decode(path)[0]
 
2146
                    source_kind = _minikind_to_kind[source_minikind]
 
2147
                    return (entry[0][2],
 
2148
                           (old_path_u, path_u),
 
2149
                           content_change,
 
2150
                           (True, True),
 
2151
                           (source_parent_id, target_parent_id),
 
2152
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2153
                           (source_kind, target_kind),
 
2154
                           (source_exec, target_exec))
 
2155
                else:
 
2156
                    return uninteresting
 
2157
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2158
                # looks like a new file
 
2159
                if path_info is not None:
 
2160
                    path = pathjoin(entry[0][0], entry[0][1])
 
2161
                    # parent id is the entry for the path in the target tree
 
2162
                    # TODO: these are the same for an entire directory: cache em.
 
2163
                    parent_id = state._get_entry(target_index,
 
2164
                                                 path_utf8=entry[0][0])[0][2]
 
2165
                    if parent_id == entry[0][2]:
 
2166
                        parent_id = None
 
2167
                    if use_filesystem_for_exec:
 
2168
                        # We need S_ISREG here, because we aren't sure if this
 
2169
                        # is a file or not.
 
2170
                        target_exec = bool(
 
2171
                            stat.S_ISREG(path_info[3].st_mode)
 
2172
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2173
                    else:
 
2174
                        target_exec = target_details[3]
 
2175
                    return (entry[0][2],
 
2176
                           (None, utf8_decode(path)[0]),
 
2177
                           True,
 
2178
                           (False, True),
 
2179
                           (None, parent_id),
 
2180
                           (None, utf8_decode(entry[0][1])[0]),
 
2181
                           (None, path_info[2]),
 
2182
                           (None, target_exec))
 
2183
                else:
 
2184
                    # but its not on disk: we deliberately treat this as just
 
2185
                    # never-present. (Why ?! - RBC 20070224)
 
2186
                    pass
 
2187
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2188
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2189
                # if its still on disk, *and* theres no other entry at this
 
2190
                # path [we dont know this in this routine at the moment -
 
2191
                # perhaps we should change this - then it would be an unknown.
 
2192
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2193
                # parent id is the entry for the path in the target tree
 
2194
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2195
                if parent_id == entry[0][2]:
 
2196
                    parent_id = None
 
2197
                return (entry[0][2],
 
2198
                       (utf8_decode(old_path)[0], None),
 
2199
                       True,
 
2200
                       (True, False),
 
2201
                       (parent_id, None),
 
2202
                       (utf8_decode(entry[0][1])[0], None),
 
2203
                       (_minikind_to_kind[source_minikind], None),
 
2204
                       (source_details[3], None))
 
2205
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2206
                # a rename; could be a true rename, or a rename inherited from
 
2207
                # a renamed parent. TODO: handle this efficiently. Its not
 
2208
                # common case to rename dirs though, so a correct but slow
 
2209
                # implementation will do.
 
2210
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2211
                    search_specific_files.add(target_details[1])
 
2212
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2213
                # neither of the selected trees contain this file,
 
2214
                # so skip over it. This is not currently directly tested, but
 
2215
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2216
                pass
 
2217
            else:
 
2218
                raise AssertionError("don't know how to compare "
 
2219
                    "source_minikind=%r, target_minikind=%r"
 
2220
                    % (source_minikind, target_minikind))
 
2221
                ## import pdb;pdb.set_trace()
 
2222
            return None
 
2223
 
 
2224
        while search_specific_files:
 
2225
            # TODO: the pending list should be lexically sorted?  the
 
2226
            # interface doesn't require it.
 
2227
            current_root = search_specific_files.pop()
 
2228
            current_root_unicode = current_root.decode('utf8')
 
2229
            searched_specific_files.add(current_root)
 
2230
            # process the entries for this containing directory: the rest will be
 
2231
            # found by their parents recursively.
 
2232
            root_entries = _entries_for_path(current_root)
 
2233
            root_abspath = self.target.abspath(current_root_unicode)
 
2234
            try:
 
2235
                root_stat = os.lstat(root_abspath)
 
2236
            except OSError, e:
 
2237
                if e.errno == errno.ENOENT:
 
2238
                    # the path does not exist: let _process_entry know that.
 
2239
                    root_dir_info = None
 
2240
                else:
 
2241
                    # some other random error: hand it up.
 
2242
                    raise
 
2243
            else:
 
2244
                root_dir_info = ('', current_root,
 
2245
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2246
                    root_abspath)
 
2247
                if root_dir_info[2] == 'directory':
 
2248
                    if self.target._directory_is_tree_reference(
 
2249
                        current_root.decode('utf8')):
 
2250
                        root_dir_info = root_dir_info[:2] + \
 
2251
                            ('tree-reference',) + root_dir_info[3:]
 
2252
 
 
2253
            if not root_entries and not root_dir_info:
 
2254
                # this specified path is not present at all, skip it.
 
2255
                continue
 
2256
            path_handled = False
 
2257
            for entry in root_entries:
 
2258
                result = _process_entry(entry, root_dir_info)
 
2259
                if result is not None:
 
2260
                    path_handled = True
 
2261
                    if result is not uninteresting:
 
2262
                        yield result
 
2263
            if want_unversioned and not path_handled and root_dir_info:
 
2264
                new_executable = bool(
 
2265
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2266
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2267
                yield (None,
 
2268
                       (None, current_root_unicode),
 
2269
                       True,
 
2270
                       (False, False),
 
2271
                       (None, None),
 
2272
                       (None, splitpath(current_root_unicode)[-1]),
 
2273
                       (None, root_dir_info[2]),
 
2274
                       (None, new_executable)
 
2275
                      )
 
2276
            initial_key = (current_root, '', '')
 
2277
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2278
            if block_index == 0:
 
2279
                # we have processed the total root already, but because the
 
2280
                # initial key matched it we should skip it here.
 
2281
                block_index +=1
 
2282
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2283
                current_dir_info = None
 
2284
            else:
 
2285
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2286
                try:
 
2287
                    current_dir_info = dir_iterator.next()
 
2288
                except OSError, e:
 
2289
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2290
                    # python 2.5 has e.errno == EINVAL,
 
2291
                    #            and e.winerror == ERROR_DIRECTORY
 
2292
                    e_winerror = getattr(e, 'winerror', None)
 
2293
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2294
                    # there may be directories in the inventory even though
 
2295
                    # this path is not a file on disk: so mark it as end of
 
2296
                    # iterator
 
2297
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2298
                        current_dir_info = None
 
2299
                    elif (sys.platform == 'win32'
 
2300
                          and (e.errno in win_errors
 
2301
                               or e_winerror in win_errors)):
 
2302
                        current_dir_info = None
 
2303
                    else:
 
2304
                        raise
 
2305
                else:
 
2306
                    if current_dir_info[0][0] == '':
 
2307
                        # remove .bzr from iteration
 
2308
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2309
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2310
                            raise AssertionError()
 
2311
                        del current_dir_info[1][bzr_index]
 
2312
            # walk until both the directory listing and the versioned metadata
 
2313
            # are exhausted. 
 
2314
            if (block_index < len(state._dirblocks) and
 
2315
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2316
                current_block = state._dirblocks[block_index]
 
2317
            else:
 
2318
                current_block = None
 
2319
            while (current_dir_info is not None or
 
2320
                   current_block is not None):
 
2321
                if (current_dir_info and current_block
 
2322
                    and current_dir_info[0][0] != current_block[0]):
 
2323
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2324
                        # filesystem data refers to paths not covered by the dirblock.
 
2325
                        # this has two possibilities:
 
2326
                        # A) it is versioned but empty, so there is no block for it
 
2327
                        # B) it is not versioned.
 
2328
 
 
2329
                        # if (A) then we need to recurse into it to check for
 
2330
                        # new unknown files or directories.
 
2331
                        # if (B) then we should ignore it, because we don't
 
2332
                        # recurse into unknown directories.
 
2333
                        path_index = 0
 
2334
                        while path_index < len(current_dir_info[1]):
 
2335
                                current_path_info = current_dir_info[1][path_index]
 
2336
                                if want_unversioned:
 
2337
                                    if current_path_info[2] == 'directory':
 
2338
                                        if self.target._directory_is_tree_reference(
 
2339
                                            current_path_info[0].decode('utf8')):
 
2340
                                            current_path_info = current_path_info[:2] + \
 
2341
                                                ('tree-reference',) + current_path_info[3:]
 
2342
                                    new_executable = bool(
 
2343
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2344
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2345
                                    yield (None,
 
2346
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2347
                                        True,
 
2348
                                        (False, False),
 
2349
                                        (None, None),
 
2350
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2351
                                        (None, current_path_info[2]),
 
2352
                                        (None, new_executable))
 
2353
                                # dont descend into this unversioned path if it is
 
2354
                                # a dir
 
2355
                                if current_path_info[2] in ('directory',
 
2356
                                                            'tree-reference'):
 
2357
                                    del current_dir_info[1][path_index]
 
2358
                                    path_index -= 1
 
2359
                                path_index += 1
 
2360
 
 
2361
                        # This dir info has been handled, go to the next
 
2362
                        try:
 
2363
                            current_dir_info = dir_iterator.next()
 
2364
                        except StopIteration:
 
2365
                            current_dir_info = None
 
2366
                    else:
 
2367
                        # We have a dirblock entry for this location, but there
 
2368
                        # is no filesystem path for this. This is most likely
 
2369
                        # because a directory was removed from the disk.
 
2370
                        # We don't have to report the missing directory,
 
2371
                        # because that should have already been handled, but we
 
2372
                        # need to handle all of the files that are contained
 
2373
                        # within.
 
2374
                        for current_entry in current_block[1]:
 
2375
                            # entry referring to file not present on disk.
 
2376
                            # advance the entry only, after processing.
 
2377
                            result = _process_entry(current_entry, None)
 
2378
                            if result is not None:
 
2379
                                if result is not uninteresting:
 
2380
                                    yield result
 
2381
                        block_index +=1
 
2382
                        if (block_index < len(state._dirblocks) and
 
2383
                            osutils.is_inside(current_root,
 
2384
                                              state._dirblocks[block_index][0])):
 
2385
                            current_block = state._dirblocks[block_index]
 
2386
                        else:
 
2387
                            current_block = None
 
2388
                    continue
 
2389
                entry_index = 0
 
2390
                if current_block and entry_index < len(current_block[1]):
 
2391
                    current_entry = current_block[1][entry_index]
 
2392
                else:
 
2393
                    current_entry = None
 
2394
                advance_entry = True
 
2395
                path_index = 0
 
2396
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2397
                    current_path_info = current_dir_info[1][path_index]
 
2398
                    if current_path_info[2] == 'directory':
 
2399
                        if self.target._directory_is_tree_reference(
 
2400
                            current_path_info[0].decode('utf8')):
 
2401
                            current_path_info = current_path_info[:2] + \
 
2402
                                ('tree-reference',) + current_path_info[3:]
 
2403
                else:
 
2404
                    current_path_info = None
 
2405
                advance_path = True
 
2406
                path_handled = False
 
2407
                while (current_entry is not None or
 
2408
                    current_path_info is not None):
 
2409
                    if current_entry is None:
 
2410
                        # the check for path_handled when the path is adnvaced
 
2411
                        # will yield this path if needed.
 
2412
                        pass
 
2413
                    elif current_path_info is None:
 
2414
                        # no path is fine: the per entry code will handle it.
 
2415
                        result = _process_entry(current_entry, current_path_info)
 
2416
                        if result is not None:
 
2417
                            if result is not uninteresting:
 
2418
                                yield result
 
2419
                    elif (current_entry[0][1] != current_path_info[1]
 
2420
                          or current_entry[1][target_index][0] in 'ar'):
 
2421
                        # The current path on disk doesn't match the dirblock
 
2422
                        # record. Either the dirblock is marked as absent, or
 
2423
                        # the file on disk is not present at all in the
 
2424
                        # dirblock. Either way, report about the dirblock
 
2425
                        # entry, and let other code handle the filesystem one.
 
2426
 
 
2427
                        # Compare the basename for these files to determine
 
2428
                        # which comes first
 
2429
                        if current_path_info[1] < current_entry[0][1]:
 
2430
                            # extra file on disk: pass for now, but only
 
2431
                            # increment the path, not the entry
 
2432
                            advance_entry = False
 
2433
                        else:
 
2434
                            # entry referring to file not present on disk.
 
2435
                            # advance the entry only, after processing.
 
2436
                            result = _process_entry(current_entry, None)
 
2437
                            if result is not None:
 
2438
                                if result is not uninteresting:
 
2439
                                    yield result
 
2440
                            advance_path = False
 
2441
                    else:
 
2442
                        result = _process_entry(current_entry, current_path_info)
 
2443
                        if result is not None:
 
2444
                            path_handled = True
 
2445
                            if result is not uninteresting:
 
2446
                                yield result
 
2447
                    if advance_entry and current_entry is not None:
 
2448
                        entry_index += 1
 
2449
                        if entry_index < len(current_block[1]):
 
2450
                            current_entry = current_block[1][entry_index]
 
2451
                        else:
 
2452
                            current_entry = None
 
2453
                    else:
 
2454
                        advance_entry = True # reset the advance flaga
 
2455
                    if advance_path and current_path_info is not None:
 
2456
                        if not path_handled:
 
2457
                            # unversioned in all regards
 
2458
                            if want_unversioned:
 
2459
                                new_executable = bool(
 
2460
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2461
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2462
                                yield (None,
 
2463
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2464
                                    True,
 
2465
                                    (False, False),
 
2466
                                    (None, None),
 
2467
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2468
                                    (None, current_path_info[2]),
 
2469
                                    (None, new_executable))
 
2470
                            # dont descend into this unversioned path if it is
 
2471
                            # a dir
 
2472
                            if current_path_info[2] in ('directory'):
 
2473
                                del current_dir_info[1][path_index]
 
2474
                                path_index -= 1
 
2475
                        # dont descend the disk iterator into any tree 
 
2476
                        # paths.
 
2477
                        if current_path_info[2] == 'tree-reference':
 
2478
                            del current_dir_info[1][path_index]
 
2479
                            path_index -= 1
 
2480
                        path_index += 1
 
2481
                        if path_index < len(current_dir_info[1]):
 
2482
                            current_path_info = current_dir_info[1][path_index]
 
2483
                            if current_path_info[2] == 'directory':
 
2484
                                if self.target._directory_is_tree_reference(
 
2485
                                    current_path_info[0].decode('utf8')):
 
2486
                                    current_path_info = current_path_info[:2] + \
 
2487
                                        ('tree-reference',) + current_path_info[3:]
 
2488
                        else:
 
2489
                            current_path_info = None
 
2490
                        path_handled = False
 
2491
                    else:
 
2492
                        advance_path = True # reset the advance flagg.
 
2493
                if current_block is not None:
 
2494
                    block_index += 1
 
2495
                    if (block_index < len(state._dirblocks) and
 
2496
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2497
                        current_block = state._dirblocks[block_index]
 
2498
                    else:
 
2499
                        current_block = None
 
2500
                if current_dir_info is not None:
 
2501
                    try:
 
2502
                        current_dir_info = dir_iterator.next()
 
2503
                    except StopIteration:
 
2504
                        current_dir_info = None
 
2505
 
2255
2506
 
2256
2507
    @staticmethod
2257
2508
    def is_compatible(source, target):
2258
2509
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2510
        if not isinstance(target, WorkingTree4):
2260
2511
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2512
        # the source must be a revtreee or dirstate rev tree.
2262
2513
        if not isinstance(source,
2263
2514
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2515
            return False
2265
2516
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2517
        if not (source._revision_id == NULL_REVISION or
2267
2518
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2519
            # TODO: what about ghosts? it may well need to 
2269
2520
            # check for them explicitly.
2270
2521
            return False
2271
2522
        return True
2281
2532
 
2282
2533
    def convert(self, tree):
2283
2534
        # lock the control files not the tree, so that we dont get tree
2284
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2535
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2536
        # tree during upgrade.
2286
2537
        tree._control_files.lock_write()
2287
2538
        try:
2314
2565
    def update_format(self, tree):
2315
2566
        """Change the format marker."""
2316
2567
        tree._transport.put_bytes('format',
2317
 
            self.target_format.as_string(),
2318
 
            mode=tree.bzrdir._get_file_mode())
2319
 
 
2320
 
 
2321
 
class Converter4to5(object):
2322
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2323
 
 
2324
 
    def __init__(self):
2325
 
        self.target_format = WorkingTreeFormat5()
2326
 
 
2327
 
    def convert(self, tree):
2328
 
        # lock the control files not the tree, so that we don't get tree
2329
 
        # on-unlock behaviours, and so that no-one else diddles with the
2330
 
        # tree during upgrade.
2331
 
        tree._control_files.lock_write()
2332
 
        try:
2333
 
            self.update_format(tree)
2334
 
        finally:
2335
 
            tree._control_files.unlock()
2336
 
 
2337
 
    def update_format(self, tree):
2338
 
        """Change the format marker."""
2339
 
        tree._transport.put_bytes('format',
2340
 
            self.target_format.as_string(),
2341
 
            mode=tree.bzrdir._get_file_mode())
2342
 
 
2343
 
 
2344
 
class Converter4or5to6(object):
2345
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2346
 
 
2347
 
    def __init__(self):
2348
 
        self.target_format = WorkingTreeFormat6()
2349
 
 
2350
 
    def convert(self, tree):
2351
 
        # lock the control files not the tree, so that we don't get tree
2352
 
        # on-unlock behaviours, and so that no-one else diddles with the
2353
 
        # tree during upgrade.
2354
 
        tree._control_files.lock_write()
2355
 
        try:
2356
 
            self.init_custom_control_files(tree)
2357
 
            self.update_format(tree)
2358
 
        finally:
2359
 
            tree._control_files.unlock()
2360
 
 
2361
 
    def init_custom_control_files(self, tree):
2362
 
        """Initialize custom control files."""
2363
 
        tree._transport.put_bytes('views', '',
2364
 
            mode=tree.bzrdir._get_file_mode())
2365
 
 
2366
 
    def update_format(self, tree):
2367
 
        """Change the format marker."""
2368
 
        tree._transport.put_bytes('format',
2369
 
            self.target_format.as_string(),
 
2568
            self.target_format.get_format_string(),
2370
2569
            mode=tree.bzrdir._get_file_mode())