/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

Merge bzr.dev.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
38
46
    debug,
 
47
    delta,
39
48
    dirstate,
40
49
    errors,
41
50
    generate_ids,
 
51
    globbing,
 
52
    hashcache,
 
53
    ignores,
 
54
    merge,
42
55
    osutils,
43
56
    revision as _mod_revision,
44
57
    revisiontree,
 
58
    textui,
45
59
    trace,
46
60
    transform,
47
 
    views,
 
61
    urlutils,
 
62
    xml5,
 
63
    xml6,
48
64
    )
49
65
import bzrlib.branch
 
66
from bzrlib.transport import get_transport
50
67
import bzrlib.ui
51
68
""")
52
69
 
 
70
from bzrlib import symbol_versioning
53
71
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
73
from bzrlib.lockable_files import LockableFiles, TransportLock
 
74
from bzrlib.lockdir import LockDir
 
75
import bzrlib.mutabletree
56
76
from bzrlib.mutabletree import needs_tree_write_lock
57
77
from bzrlib.osutils import (
58
78
    file_kind,
59
79
    isdir,
 
80
    normpath,
60
81
    pathjoin,
 
82
    rand_chars,
61
83
    realpath,
62
84
    safe_unicode,
 
85
    splitpath,
63
86
    )
64
 
from bzrlib.trace import mutter
 
87
from bzrlib.trace import mutter, note
65
88
from bzrlib.transport.local import LocalTransport
66
89
from bzrlib.tree import InterTree
 
90
from bzrlib.progress import DummyProgress, ProgressPhase
 
91
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
92
from bzrlib.rio import RioReader, rio_file, Stanza
 
93
from bzrlib.symbol_versioning import (deprecated_passed,
 
94
        deprecated_method,
 
95
        deprecated_function,
 
96
        DEPRECATED_PARAMETER,
 
97
        )
67
98
from bzrlib.tree import Tree
68
99
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
100
 
70
101
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
102
# This is the Windows equivalent of ENOTDIR
 
103
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
104
# just an error code.
 
105
ERROR_PATH_NOT_FOUND = 3
 
106
ERROR_DIRECTORY = 267
 
107
 
 
108
 
 
109
class WorkingTree4(WorkingTree3):
 
110
    """This is the Format 4 working tree.
 
111
 
 
112
    This differs from WorkingTree3 by:
 
113
     - Having a consolidated internal dirstate, stored in a
 
114
       randomly-accessible sorted file on disk.
 
115
     - Not having a regular inventory attribute.  One can be synthesized 
 
116
       on demand but this is expensive and should be avoided.
 
117
 
 
118
    This is new in bzr 0.15.
 
119
    """
 
120
 
72
121
    def __init__(self, basedir,
73
122
                 branch,
74
123
                 _control_files=None,
83
132
        """
84
133
        self._format = _format
85
134
        self.bzrdir = _bzrdir
 
135
        from bzrlib.trace import note, mutter
 
136
        assert isinstance(basedir, basestring), \
 
137
            "base directory %r is not a string" % basedir
86
138
        basedir = safe_unicode(basedir)
87
139
        mutter("opening working tree %r", basedir)
88
140
        self._branch = branch
 
141
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
142
            "branch %r is not a Branch" % self.branch
89
143
        self.basedir = realpath(basedir)
90
144
        # if branch is at our basedir and is a format 6 or less
91
145
        # assume all other formats have their own control files.
 
146
        assert isinstance(_control_files, LockableFiles), \
 
147
            "_control_files must be a LockableFiles, not %r" % _control_files
92
148
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
149
        self._dirty = None
95
150
        #-------------
96
151
        # during a read or write lock these objects are set, and are
98
153
        self._dirstate = None
99
154
        self._inventory = None
100
155
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
156
 
108
157
    @needs_tree_write_lock
109
158
    def _add(self, files, ids, kinds):
111
160
        state = self.current_dirstate()
112
161
        for f, file_id, kind in zip(files, ids, kinds):
113
162
            f = f.strip('/')
 
163
            assert '//' not in f
 
164
            assert '..' not in f
114
165
            if self.path2id(f):
115
166
                # special case tree root handling.
116
167
                if f == '' and self.path2id(f) == ROOT_ID:
137
188
    @needs_tree_write_lock
138
189
    def add_reference(self, sub_tree):
139
190
        # use standard implementation, which calls back to self._add
140
 
        #
 
191
        # 
141
192
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
193
        # it's just recorded at the moment of commit. 
143
194
        self._add_reference(sub_tree)
144
195
 
145
196
    def break_lock(self):
184
235
            WorkingTree3._comparison_data(self, entry, path)
185
236
        # it looks like a plain directory, but it's really a reference -- see
186
237
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
238
        if (self._repo_supports_tree_reference and
 
239
            kind == 'directory' and
 
240
            self._directory_is_tree_reference(path)):
189
241
            kind = 'tree-reference'
190
242
        return kind, executable, stat_value
191
243
 
217
269
            return self._dirstate
218
270
        local_path = self.bzrdir.get_workingtree_transport(None
219
271
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
272
        self._dirstate = dirstate.DirState.on_file(local_path)
222
273
        return self._dirstate
223
274
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
235
 
 
236
275
    def filter_unversioned_files(self, paths):
237
276
        """Filter out paths that are versioned.
238
277
 
270
309
 
271
310
    def _generate_inventory(self):
272
311
        """Create and set self.inventory from the dirstate object.
273
 
 
 
312
        
274
313
        This is relatively expensive: we have to walk the entire dirstate.
275
314
        Ideally we would not, and can deprecate this function.
276
315
        """
280
319
        state._read_dirblocks_if_needed()
281
320
        root_key, current_entry = self._get_entry(path='')
282
321
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
322
        assert current_entry[0][0] == 'd' # directory
285
323
        inv = Inventory(root_id=current_id)
286
324
        # Turn some things into local variables
287
325
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
358
                    # add this entry to the parent map.
321
359
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
360
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
361
                    assert self._repo_supports_tree_reference, \
 
362
                        "repository of %r " \
 
363
                        "doesn't support tree references " \
 
364
                        "required by entry %r" \
 
365
                        % (self, name)
327
366
                    inv_entry.reference_revision = link_or_sha1 or None
328
367
                elif kind != 'symlink':
329
368
                    raise AssertionError("unknown kind %r" % kind)
330
369
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
370
                assert file_id not in inv_byid, ('file_id %s already in'
 
371
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
372
                assert name_unicode not in parent_ie.children
337
373
                inv_byid[file_id] = inv_entry
338
374
                parent_ie.children[name_unicode] = inv_entry
339
375
        self._inventory = inv
344
380
        If either file_id or path is supplied, it is used as the key to lookup.
345
381
        If both are supplied, the fastest lookup is used, and an error is
346
382
        raised if they do not both point at the same row.
347
 
 
 
383
        
348
384
        :param file_id: An optional unicode file_id to be looked up.
349
385
        :param path: An optional unicode path to be looked up.
350
386
        :return: The dirstate row tuple for path/file_id, or (None, None)
374
410
                    return None
375
411
                else:
376
412
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
413
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
414
                                          stat_value=stat_value)
379
415
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
416
            return link_or_sha1
390
417
        return None
391
418
 
392
419
    def _get_inventory(self):
406
433
    @needs_read_lock
407
434
    def get_parent_ids(self):
408
435
        """See Tree.get_parent_ids.
409
 
 
 
436
        
410
437
        This implementation requests the ids list from the dirstate file.
411
438
        """
412
439
        return self.current_dirstate().get_parent_ids()
428
455
 
429
456
    def has_id(self, file_id):
430
457
        state = self.current_dirstate()
 
458
        file_id = osutils.safe_file_id(file_id)
431
459
        row, parents = self._get_entry(file_id=file_id)
432
460
        if row is None:
433
461
            return False
434
462
        return osutils.lexists(pathjoin(
435
463
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
464
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
465
    @needs_read_lock
443
466
    def id2path(self, file_id):
444
467
        "Convert a file-id to a path."
 
468
        file_id = osutils.safe_file_id(file_id)
445
469
        state = self.current_dirstate()
446
470
        entry = self._get_entry(file_id=file_id)
447
471
        if entry == (None, None):
449
473
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
474
        return path_utf8.decode('utf8')
451
475
 
452
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
453
 
        entry = self._get_entry(path=path)
454
 
        if entry == (None, None):
455
 
            return False # Missing entries are not executable
456
 
        return entry[1][0][3] # Executable?
457
 
 
458
476
    if not osutils.supports_executable():
459
477
        def is_executable(self, file_id, path=None):
460
478
            """Test if a file is executable or not.
461
479
 
462
480
            Note: The caller is expected to take a read-lock before calling this.
463
481
            """
 
482
            file_id = osutils.safe_file_id(file_id)
464
483
            entry = self._get_entry(file_id=file_id, path=path)
465
484
            if entry == (None, None):
466
485
                return False
467
486
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
487
    else:
472
488
        def is_executable(self, file_id, path=None):
473
489
            """Test if a file is executable or not.
474
490
 
475
491
            Note: The caller is expected to take a read-lock before calling this.
476
492
            """
477
 
            self._must_be_locked()
478
493
            if not path:
 
494
                file_id = osutils.safe_file_id(file_id)
479
495
                path = self.id2path(file_id)
480
496
            mode = os.lstat(self.abspath(path)).st_mode
481
497
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
498
 
483
 
    def all_file_ids(self):
484
 
        """See Tree.iter_all_file_ids"""
485
 
        self._must_be_locked()
486
 
        result = set()
487
 
        for key, tree_details in self.current_dirstate()._iter_entries():
488
 
            if tree_details[0][0] in ('a', 'r'): # relocated
489
 
                continue
490
 
            result.add(key[2])
491
 
        return result
492
 
 
493
499
    @needs_read_lock
494
500
    def __iter__(self):
495
501
        """Iterate through file_ids for this tree.
508
514
        return iter(result)
509
515
 
510
516
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
517
        for key, tree_details in self.current_dirstate()._iter_entries():
516
518
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
519
                # not relevant to the working tree
519
521
            if not key[1]:
520
522
                # the root is not a reference.
521
523
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
524
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
525
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
526
                if self._kind(path) == 'tree-reference':
 
527
                    yield path, key[2]
526
528
            except errors.NoSuchFile:
527
529
                # path is missing on disk.
528
530
                continue
529
531
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
536
532
    def kind(self, file_id):
537
533
        """Return the kind of a file.
538
534
 
542
538
        Note: The caller is expected to take a read-lock before calling this.
543
539
        """
544
540
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
541
        assert relpath != None, \
 
542
            "path for id {%s} is None!" % file_id
548
543
        return self._kind(relpath)
549
544
 
550
545
    def _kind(self, relpath):
551
546
        abspath = self.abspath(relpath)
552
547
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    kind = 'tree-reference'
 
548
        if (self._repo_supports_tree_reference and
 
549
            kind == 'directory' and
 
550
            self._directory_is_tree_reference(relpath)):
 
551
            kind = 'tree-reference'
558
552
        return kind
559
553
 
560
554
    @needs_read_lock
623
617
        result = []
624
618
        if not from_paths:
625
619
            return result
 
620
 
626
621
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
622
 
 
623
        assert not isinstance(from_paths, basestring)
629
624
        to_dir_utf8 = to_dir.encode('utf8')
630
625
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
626
        id_index = state._get_id_index()
653
648
        if self._inventory is not None:
654
649
            update_inventory = True
655
650
            inv = self.inventory
 
651
            to_dir_ie = inv[to_dir_id]
656
652
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
653
        else:
659
654
            update_inventory = False
660
655
 
691
686
            from_entry = self._get_entry(path=from_rel)
692
687
            if from_entry == (None, None):
693
688
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
689
                    errors.NotVersionedError(path=str(from_rel)))
695
690
 
696
691
            from_id = from_entry[0][2]
697
692
            to_rel = pathjoin(to_dir, from_tail)
724
719
                if from_missing: # implicitly just update our path mapping
725
720
                    move_file = False
726
721
                elif not after:
727
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
722
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
723
                        extra="(Use --after to update the Bazaar id)")
728
724
 
729
725
            rollbacks = []
730
726
            def rollback_rename():
786
782
                if minikind == 'd':
787
783
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
784
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
785
                        assert from_dir != '', "renaming root not supported"
791
786
                        from_key = (from_dir, '')
792
787
                        from_block_idx, present = \
793
788
                            state._find_block_index_from_key(from_key)
806
801
 
807
802
                        # Grab a copy since move_one may update the list.
808
803
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
804
                            assert entry[0][0] == from_dir
811
805
                            cur_details = entry[1][0]
812
806
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
807
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
872
866
        for tree in trees:
873
867
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
868
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
869
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
870
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
871
        # -- make all paths utf8 --
879
872
        paths_utf8 = set()
939
932
            if not all_versioned:
940
933
                raise errors.PathsNotVersionedError(paths)
941
934
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
 
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
935
        search_paths = set()
 
936
        for path in paths:
 
937
            other_paths = paths.difference(set([path]))
 
938
            if not osutils.is_inside_any(other_paths, path):
 
939
                # this is a top level path, we must check it.
 
940
                search_paths.add(path)
 
941
        # sketch: 
944
942
        # for all search_indexs in each path at or under each element of
945
943
        # search_paths, if the detail is relocated: add the id, and add the
946
944
        # relocated path as one to search if its not searched already. If the
1002
1000
 
1003
1001
    def read_working_inventory(self):
1004
1002
        """Read the working inventory.
1005
 
 
 
1003
        
1006
1004
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
1005
        """
1008
1006
        return self.inventory
1013
1011
 
1014
1012
        WorkingTree4 supplies revision_trees for any basis tree.
1015
1013
        """
 
1014
        revision_id = osutils.safe_revision_id(revision_id)
1016
1015
        dirstate = self.current_dirstate()
1017
1016
        parent_ids = dirstate.get_parent_ids()
1018
1017
        if revision_id not in parent_ids:
1025
1024
    @needs_tree_write_lock
1026
1025
    def set_last_revision(self, new_revision):
1027
1026
        """Change the last revision in the working tree."""
 
1027
        new_revision = osutils.safe_revision_id(new_revision)
1028
1028
        parents = self.get_parent_ids()
1029
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1030
 
            if len(parents) >= 2:
1031
 
                raise AssertionError(
1032
 
                    "setting the last parent to none with a pending merge is "
1033
 
                    "unsupported.")
 
1029
        if new_revision in (NULL_REVISION, None):
 
1030
            assert len(parents) < 2, (
 
1031
                "setting the last parent to none with a pending merge is "
 
1032
                "unsupported.")
1034
1033
            self.set_parent_ids([])
1035
1034
        else:
1036
1035
            self.set_parent_ids([new_revision] + parents[1:],
1039
1038
    @needs_tree_write_lock
1040
1039
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
1040
        """Set the parent ids to revision_ids.
1042
 
 
 
1041
        
1043
1042
        See also set_parent_trees. This api will try to retrieve the tree data
1044
1043
        for each element of revision_ids from the trees repository. If you have
1045
1044
        tree data already available, it is more efficient to use
1049
1048
        :param revision_ids: The revision_ids to set as the parent ids of this
1050
1049
            working tree. Any of these may be ghosts.
1051
1050
        """
 
1051
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1052
1052
        trees = []
1053
1053
        for revision_id in revision_ids:
1054
1054
            try:
1077
1077
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1078
        real_trees = []
1079
1079
        ghosts = []
1080
 
 
1081
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1082
 
        graph = self.branch.repository.get_graph()
1083
 
        heads = graph.heads(parent_ids)
1084
 
        accepted_revisions = set()
1085
 
 
1086
1080
        # convert absent trees to the null tree, which we convert back to
1087
1081
        # missing on access.
1088
1082
        for rev_id, tree in parents_list:
1089
 
            if len(accepted_revisions) > 0:
1090
 
                # we always accept the first tree
1091
 
                if rev_id in accepted_revisions or rev_id not in heads:
1092
 
                    # We have already included either this tree, or its
1093
 
                    # descendent, so we skip it.
1094
 
                    continue
 
1083
            rev_id = osutils.safe_revision_id(rev_id)
1095
1084
            _mod_revision.check_not_reserved_id(rev_id)
1096
1085
            if tree is not None:
1097
1086
                real_trees.append((rev_id, tree))
1098
1087
            else:
1099
1088
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1089
                    self.branch.repository.revision_tree(None)))
1102
1090
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
1104
1091
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1105
1092
        self._make_dirty(reset_inventory=False)
1106
1093
 
1111
1098
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1112
1099
            self._make_dirty(reset_inventory=True)
1113
1100
 
1114
 
    def _sha_from_stat(self, path, stat_result):
1115
 
        """Get a sha digest from the tree's stat cache.
1116
 
 
1117
 
        The default implementation assumes no stat cache is present.
1118
 
 
1119
 
        :param path: The path.
1120
 
        :param stat_result: The stat result being looked up.
1121
 
        """
1122
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1123
 
 
1124
1101
    @needs_read_lock
1125
1102
    def supports_tree_reference(self):
1126
1103
        return self._repo_supports_tree_reference
1166
1143
            return
1167
1144
        state = self.current_dirstate()
1168
1145
        state._read_dirblocks_if_needed()
1169
 
        ids_to_unversion = set(file_ids)
 
1146
        ids_to_unversion = set()
 
1147
        for file_id in file_ids:
 
1148
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1170
1149
        paths_to_unversion = set()
1171
1150
        # sketch:
1172
1151
        # check if the root is to be unversioned, if so, assert for now.
1199
1178
                # just forget the whole block.
1200
1179
                entry_index = 0
1201
1180
                while entry_index < len(block[1]):
 
1181
                    # Mark this file id as having been removed
1202
1182
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1183
                    ids_to_unversion.discard(entry[0][2])
 
1184
                    if (entry[1][0][0] == 'a'
 
1185
                        or not state._make_absent(entry)):
1205
1186
                        entry_index += 1
1206
 
                    else:
1207
 
                        # Mark this file id as having been removed
1208
 
                        ids_to_unversion.discard(entry[0][2])
1209
 
                        if not state._make_absent(entry):
1210
 
                            # The block has not shrunk.
1211
 
                            entry_index += 1
1212
1187
                # go to the next block. (At the moment we dont delete empty
1213
1188
                # dirblocks)
1214
1189
                block_index += 1
1237
1212
            for file_id in file_ids:
1238
1213
                self._inventory.remove_recursive_id(file_id)
1239
1214
 
1240
 
    @needs_tree_write_lock
1241
 
    def rename_one(self, from_rel, to_rel, after=False):
1242
 
        """See WorkingTree.rename_one"""
1243
 
        self.flush()
1244
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1245
 
 
1246
 
    @needs_tree_write_lock
1247
 
    def apply_inventory_delta(self, changes):
1248
 
        """See MutableTree.apply_inventory_delta"""
1249
 
        state = self.current_dirstate()
1250
 
        state.update_by_delta(changes)
1251
 
        self._make_dirty(reset_inventory=True)
1252
 
 
1253
 
    def update_basis_by_delta(self, new_revid, delta):
1254
 
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
1257
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
 
 
1259
1215
    @needs_read_lock
1260
1216
    def _validate(self):
1261
1217
        self._dirstate._validate()
1263
1219
    @needs_tree_write_lock
1264
1220
    def _write_inventory(self, inv):
1265
1221
        """Write inventory as the current inventory."""
1266
 
        if self._dirty:
1267
 
            raise AssertionError("attempting to write an inventory when the "
1268
 
                "dirstate is dirty will lose pending changes")
1269
 
        had_inventory = self._inventory is not None
1270
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1271
 
        # working inventory. We do this because self.inventory may be inv, or
1272
 
        # may have been modified, and either case would prevent a clean delta
1273
 
        # being created.
1274
 
        self._inventory = None
1275
 
        # generate a delta,
1276
 
        delta = inv._make_delta(self.inventory)
1277
 
        # and apply it.
1278
 
        self.apply_inventory_delta(delta)
1279
 
        if had_inventory:
 
1222
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1223
        self.current_dirstate().set_state_from_inventory(inv)
 
1224
        self._make_dirty(reset_inventory=False)
 
1225
        if self._inventory is not None:
1280
1226
            self._inventory = inv
1281
1227
        self.flush()
1282
1228
 
1283
1229
 
1284
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1285
 
 
1286
 
    def __init__(self, tree):
1287
 
        self.tree = tree
1288
 
 
1289
 
    def sha1(self, abspath):
1290
 
        """See dirstate.SHA1Provider.sha1()."""
1291
 
        filters = self.tree._content_filter_stack(
1292
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1293
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1294
 
 
1295
 
    def stat_and_sha1(self, abspath):
1296
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1297
 
        filters = self.tree._content_filter_stack(
1298
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1299
 
        file_obj = file(abspath, 'rb', 65000)
1300
 
        try:
1301
 
            statvalue = os.fstat(file_obj.fileno())
1302
 
            if filters:
1303
 
                file_obj = filtered_input_file(file_obj, filters)
1304
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1305
 
        finally:
1306
 
            file_obj.close()
1307
 
        return statvalue, sha1
1308
 
 
1309
 
 
1310
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1311
 
    """Dirstate working tree that supports content filtering.
1312
 
 
1313
 
    The dirstate holds the hash and size of the canonical form of the file, 
1314
 
    and most methods must return that.
1315
 
    """
1316
 
 
1317
 
    def _file_content_summary(self, path, stat_result):
1318
 
        # This is to support the somewhat obsolete path_content_summary method
1319
 
        # with content filtering: see
1320
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1321
 
        #
1322
 
        # If the dirstate cache is up to date and knows the hash and size,
1323
 
        # return that.
1324
 
        # Otherwise if there are no content filters, return the on-disk size
1325
 
        # and leave the hash blank.
1326
 
        # Otherwise, read and filter the on-disk file and use its size and
1327
 
        # hash.
1328
 
        #
1329
 
        # The dirstate doesn't store the size of the canonical form so we
1330
 
        # can't trust it for content-filtered trees.  We just return None.
1331
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1332
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1333
 
        return ('file', None, executable, dirstate_sha1)
1334
 
 
1335
 
 
1336
 
class WorkingTree4(DirStateWorkingTree):
1337
 
    """This is the Format 4 working tree.
1338
 
 
1339
 
    This differs from WorkingTree3 by:
1340
 
     - Having a consolidated internal dirstate, stored in a
1341
 
       randomly-accessible sorted file on disk.
1342
 
     - Not having a regular inventory attribute.  One can be synthesized
1343
 
       on demand but this is expensive and should be avoided.
1344
 
 
1345
 
    This is new in bzr 0.15.
1346
 
    """
1347
 
 
1348
 
 
1349
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1350
 
    """This is the Format 5 working tree.
1351
 
 
1352
 
    This differs from WorkingTree4 by:
1353
 
     - Supporting content filtering.
1354
 
 
1355
 
    This is new in bzr 1.11.
1356
 
    """
1357
 
 
1358
 
 
1359
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1360
 
    """This is the Format 6 working tree.
1361
 
 
1362
 
    This differs from WorkingTree5 by:
1363
 
     - Supporting a current view that may mask the set of files in a tree
1364
 
       impacted by most user operations.
1365
 
 
1366
 
    This is new in bzr 1.14.
1367
 
    """
1368
 
 
1369
 
    def _make_views(self):
1370
 
        return views.PathBasedViews(self)
1371
 
 
1372
 
 
1373
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1374
 
 
1375
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
 
                   accelerator_tree=None, hardlink=False):
 
1230
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1231
    """The first consolidated dirstate working tree format.
 
1232
 
 
1233
    This format:
 
1234
        - exists within a metadir controlling .bzr
 
1235
        - includes an explicit version marker for the workingtree control
 
1236
          files, separate from the BzrDir format
 
1237
        - modifies the hash cache format
 
1238
        - is new in bzr 0.15
 
1239
        - uses a LockDir to guard access to it.
 
1240
    """
 
1241
 
 
1242
    upgrade_recommended = False
 
1243
 
 
1244
    def get_format_string(self):
 
1245
        """See WorkingTreeFormat.get_format_string()."""
 
1246
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1247
 
 
1248
    def get_format_description(self):
 
1249
        """See WorkingTreeFormat.get_format_description()."""
 
1250
        return "Working tree format 4"
 
1251
 
 
1252
    def initialize(self, a_bzrdir, revision_id=None):
1377
1253
        """See WorkingTreeFormat.initialize().
1378
1254
 
1379
1255
        :param revision_id: allows creating a working tree at a different
1380
1256
        revision than the branch is at.
1381
 
        :param accelerator_tree: A tree which can be used for retrieving file
1382
 
            contents more quickly than the revision tree, i.e. a workingtree.
1383
 
            The revision tree will be used for cases where accelerator_tree's
1384
 
            content is different.
1385
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1386
 
            where possible.
1387
1257
 
1388
1258
        These trees get an initial random root id, if their repository supports
1389
1259
        rich root data, TREE_ROOT otherwise.
1390
1260
        """
 
1261
        revision_id = osutils.safe_revision_id(revision_id)
1391
1262
        if not isinstance(a_bzrdir.transport, LocalTransport):
1392
1263
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1393
1264
        transport = a_bzrdir.get_workingtree_transport(self)
1394
1265
        control_files = self._open_control_files(a_bzrdir)
1395
1266
        control_files.create_lock()
1396
1267
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
1399
 
        if from_branch is not None:
1400
 
            branch = from_branch
1401
 
        else:
1402
 
            branch = a_bzrdir.open_branch()
 
1268
        control_files.put_utf8('format', self.get_format_string())
 
1269
        branch = a_bzrdir.open_branch()
1403
1270
        if revision_id is None:
1404
1271
            revision_id = branch.last_revision()
1405
1272
        local_path = transport.local_abspath('dirstate')
1407
1274
        state = dirstate.DirState.initialize(local_path)
1408
1275
        state.unlock()
1409
1276
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1277
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1278
                         branch,
1412
1279
                         _format=self,
1413
1280
                         _bzrdir=a_bzrdir,
1415
1282
        wt._new_tree()
1416
1283
        wt.lock_tree_write()
1417
1284
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1285
            if revision_id in (None, NULL_REVISION):
1420
1286
                if branch.repository.supports_rich_root():
1421
1287
                    wt._set_root_id(generate_ids.gen_root_id())
1422
1288
                else:
1423
1289
                    wt._set_root_id(ROOT_ID)
1424
1290
                wt.flush()
1425
 
            basis = None
1426
 
            # frequently, we will get here due to branching.  The accelerator
1427
 
            # tree will be the tree from the branch, so the desired basis
1428
 
            # tree will often be a parent of the accelerator tree.
1429
 
            if accelerator_tree is not None:
1430
 
                try:
1431
 
                    basis = accelerator_tree.revision_tree(revision_id)
1432
 
                except errors.NoSuchRevision:
1433
 
                    pass
1434
 
            if basis is None:
1435
 
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
1437
 
                parents_list = []
1438
 
            else:
1439
 
                parents_list = [(revision_id, basis)]
 
1291
            wt.set_last_revision(revision_id)
 
1292
            wt.flush()
 
1293
            basis = wt.basis_tree()
1440
1294
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1295
            # if the basis has a root id we have to use that; otherwise we use
 
1296
            # a new random one
 
1297
            basis_root_id = basis.get_root_id()
 
1298
            if basis_root_id is not None:
 
1299
                wt._set_root_id(basis_root_id)
1443
1300
                wt.flush()
1444
 
                # if the basis has a root id we have to use that; otherwise we
1445
 
                # use a new random one
1446
 
                basis_root_id = basis.get_root_id()
1447
 
                if basis_root_id is not None:
1448
 
                    wt._set_root_id(basis_root_id)
1449
 
                    wt.flush()
1450
 
                if wt.supports_content_filtering():
1451
 
                    # The original tree may not have the same content filters
1452
 
                    # applied so we can't safely build the inventory delta from
1453
 
                    # the source tree.
1454
 
                    delta_from_tree = False
1455
 
                else:
1456
 
                    delta_from_tree = True
1457
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1458
 
                # because wt4.apply_inventory_delta does not mutate the input
1459
 
                # inventory entries.
1460
 
                transform.build_tree(basis, wt, accelerator_tree,
1461
 
                                     hardlink=hardlink,
1462
 
                                     delta_from_tree=delta_from_tree)
1463
 
            finally:
1464
 
                basis.unlock()
 
1301
            transform.build_tree(basis, wt)
 
1302
            basis.unlock()
1465
1303
        finally:
1466
1304
            control_files.unlock()
1467
1305
            wt.unlock()
1468
1306
        return wt
1469
1307
 
1470
 
    def _init_custom_control_files(self, wt):
1471
 
        """Subclasses with custom control files should override this method.
1472
 
 
1473
 
        The working tree and control files are locked for writing when this
1474
 
        method is called.
1475
 
 
1476
 
        :param wt: the WorkingTree object
1477
 
        """
1478
 
 
1479
1308
    def _open(self, a_bzrdir, control_files):
1480
1309
        """Open the tree itself.
1481
1310
 
1482
1311
        :param a_bzrdir: the dir for the tree.
1483
1312
        :param control_files: the control files for the tree.
1484
1313
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1314
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1315
                           branch=a_bzrdir.open_branch(),
1487
1316
                           _format=self,
1488
1317
                           _bzrdir=a_bzrdir,
1489
1318
                           _control_files=control_files)
1490
1319
 
1491
1320
    def __get_matchingbzrdir(self):
1492
 
        return self._get_matchingbzrdir()
1493
 
 
1494
 
    def _get_matchingbzrdir(self):
1495
 
        """Overrideable method to get a bzrdir for testing."""
1496
1321
        # please test against something that will let us do tree references
1497
1322
        return bzrdir.format_registry.make_bzrdir(
1498
1323
            'dirstate-with-subtree')
1500
1325
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1326
 
1502
1327
 
1503
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1504
 
    """The first consolidated dirstate working tree format.
1505
 
 
1506
 
    This format:
1507
 
        - exists within a metadir controlling .bzr
1508
 
        - includes an explicit version marker for the workingtree control
1509
 
          files, separate from the BzrDir format
1510
 
        - modifies the hash cache format
1511
 
        - is new in bzr 0.15
1512
 
        - uses a LockDir to guard access to it.
1513
 
    """
1514
 
 
1515
 
    upgrade_recommended = False
1516
 
 
1517
 
    _tree_class = WorkingTree4
1518
 
 
1519
 
    def get_format_string(self):
1520
 
        """See WorkingTreeFormat.get_format_string()."""
1521
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1522
 
 
1523
 
    def get_format_description(self):
1524
 
        """See WorkingTreeFormat.get_format_description()."""
1525
 
        return "Working tree format 4"
1526
 
 
1527
 
 
1528
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1529
 
    """WorkingTree format supporting content filtering.
1530
 
    """
1531
 
 
1532
 
    upgrade_recommended = False
1533
 
 
1534
 
    _tree_class = WorkingTree5
1535
 
 
1536
 
    def get_format_string(self):
1537
 
        """See WorkingTreeFormat.get_format_string()."""
1538
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1539
 
 
1540
 
    def get_format_description(self):
1541
 
        """See WorkingTreeFormat.get_format_description()."""
1542
 
        return "Working tree format 5"
1543
 
 
1544
 
    def supports_content_filtering(self):
1545
 
        return True
1546
 
 
1547
 
 
1548
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1549
 
    """WorkingTree format supporting views.
1550
 
    """
1551
 
 
1552
 
    upgrade_recommended = False
1553
 
 
1554
 
    _tree_class = WorkingTree6
1555
 
 
1556
 
    def get_format_string(self):
1557
 
        """See WorkingTreeFormat.get_format_string()."""
1558
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1559
 
 
1560
 
    def get_format_description(self):
1561
 
        """See WorkingTreeFormat.get_format_description()."""
1562
 
        return "Working tree format 6"
1563
 
 
1564
 
    def _init_custom_control_files(self, wt):
1565
 
        """Subclasses with custom control files should override this method."""
1566
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1567
 
 
1568
 
    def supports_content_filtering(self):
1569
 
        return True
1570
 
 
1571
 
    def supports_views(self):
1572
 
        return True
1573
 
 
1574
 
 
1575
1328
class DirStateRevisionTree(Tree):
1576
 
    """A revision tree pulling the inventory from a dirstate.
1577
 
    
1578
 
    Note that this is one of the historical (ie revision) trees cached in the
1579
 
    dirstate for easy access, not the workingtree.
1580
 
    """
 
1329
    """A revision tree pulling the inventory from a dirstate."""
1581
1330
 
1582
1331
    def __init__(self, dirstate, revision_id, repository):
1583
1332
        self._dirstate = dirstate
1584
 
        self._revision_id = revision_id
 
1333
        self._revision_id = osutils.safe_revision_id(revision_id)
1585
1334
        self._repository = repository
1586
1335
        self._inventory = None
1587
1336
        self._locked = 0
1588
1337
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1338
 
1593
1339
    def __repr__(self):
1594
1340
        return "<%s of %s in %s>" % \
1597
1343
    def annotate_iter(self, file_id,
1598
1344
                      default_revision=_mod_revision.CURRENT_REVISION):
1599
1345
        """See Tree.annotate_iter"""
1600
 
        text_key = (file_id, self.inventory[file_id].revision)
1601
 
        annotations = self._repository.texts.annotate(text_key)
1602
 
        return [(key[-1], line) for (key, line) in annotations]
 
1346
        w = self._get_weave(file_id)
 
1347
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1348
 
1604
1349
    def _get_ancestors(self, default_revision):
1605
1350
        return set(self._repository.get_ancestry(self._revision_id,
1626
1371
    def get_root_id(self):
1627
1372
        return self.path2id('')
1628
1373
 
1629
 
    def id2path(self, file_id):
1630
 
        "Convert a file-id to a path."
1631
 
        entry = self._get_entry(file_id=file_id)
1632
 
        if entry == (None, None):
1633
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1634
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
 
        return path_utf8.decode('utf8')
1636
 
 
1637
 
    def iter_references(self):
1638
 
        if not self._repo_supports_tree_reference:
1639
 
            # When the repo doesn't support references, we will have nothing to
1640
 
            # return
1641
 
            return iter([])
1642
 
        # Otherwise, fall back to the default implementation
1643
 
        return super(DirStateRevisionTree, self).iter_references()
1644
 
 
1645
1374
    def _get_parent_index(self):
1646
1375
        """Return the index in the dirstate referenced by this tree."""
1647
1376
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1381
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1382
        If both are supplied, the fastest lookup is used, and an error is
1654
1383
        raised if they do not both point at the same row.
1655
 
 
 
1384
        
1656
1385
        :param file_id: An optional unicode file_id to be looked up.
1657
1386
        :param path: An optional unicode path to be looked up.
1658
1387
        :return: The dirstate row tuple for path/file_id, or (None, None)
1659
1388
        """
1660
1389
        if file_id is None and path is None:
1661
1390
            raise errors.BzrError('must supply file_id or path')
 
1391
        file_id = osutils.safe_file_id(file_id)
1662
1392
        if path is not None:
1663
1393
            path = path.encode('utf8')
1664
1394
        parent_index = self._get_parent_index()
1672
1402
 
1673
1403
        This is relatively expensive: we have to walk the entire dirstate.
1674
1404
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1405
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1406
            'dirstate revision tree'
1679
1407
        # separate call for profiling - makes it clear where the costs are.
1680
1408
        self._dirstate._read_dirblocks_if_needed()
1681
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1682
 
            raise AssertionError(
1683
 
                'parent %s has disappeared from %s' % (
1684
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1409
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1410
            'parent %s has disappeared from %s' % (
 
1411
            self._revision_id, self._dirstate.get_parent_ids())
1685
1412
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1413
        # This is identical now to the WorkingTree _generate_inventory except
1687
1414
        # for the tree index use.
1688
1415
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1416
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1417
        assert current_entry[parent_index][0] == 'd'
1692
1418
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1419
        inv.root.revision = current_entry[parent_index][4]
1694
1420
        # Turn some things into local variables
1734
1460
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1461
                            % entry)
1736
1462
                # These checks cost us around 40ms on a 55k entry tree
1737
 
                if file_id in inv_byid:
1738
 
                    raise AssertionError('file_id %s already in'
1739
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1740
 
                if name_unicode in parent_ie.children:
1741
 
                    raise AssertionError('name %r already in parent'
1742
 
                        % (name_unicode,))
 
1463
                assert file_id not in inv_byid
 
1464
                assert name_unicode not in parent_ie.children
1743
1465
                inv_byid[file_id] = inv_entry
1744
1466
                parent_ie.children[name_unicode] = inv_entry
1745
1467
        self._inventory = inv
1755
1477
            return None
1756
1478
        parent_index = self._get_parent_index()
1757
1479
        last_changed_revision = entry[1][parent_index][4]
1758
 
        try:
1759
 
            rev = self._repository.get_revision(last_changed_revision)
1760
 
        except errors.NoSuchRevision:
1761
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1762
 
        return rev.timestamp
 
1480
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1481
 
1764
1482
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1483
        entry = self._get_entry(file_id=file_id, path=path)
1769
1487
            return parent_details[1]
1770
1488
        return None
1771
1489
 
 
1490
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1491
    def get_weave(self, file_id):
 
1492
        return self._get_weave(file_id)
 
1493
 
 
1494
    def _get_weave(self, file_id):
 
1495
        return self._repository.weave_store.get_weave(file_id,
 
1496
                self._repository.get_transaction())
 
1497
 
1772
1498
    def get_file(self, file_id, path=None):
1773
1499
        return StringIO(self.get_file_text(file_id))
1774
1500
 
 
1501
    def get_file_lines(self, file_id):
 
1502
        ie = self.inventory[file_id]
 
1503
        return self._get_weave(file_id).get_lines(ie.revision)
 
1504
 
1775
1505
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1506
        return self.inventory[file_id].text_size
1778
1507
 
1779
 
    def get_file_text(self, file_id, path=None):
1780
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1781
 
        return ''.join(content)
 
1508
    def get_file_text(self, file_id):
 
1509
        return ''.join(self.get_file_lines(file_id))
1782
1510
 
1783
1511
    def get_reference_revision(self, file_id, path=None):
1784
1512
        return self.inventory[file_id].reference_revision
1803
1531
        if entry[1][parent_index][0] != 'l':
1804
1532
            return None
1805
1533
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1534
            # At present, none of the tree implementations supports non-ascii
 
1535
            # symlink targets. So we will just assume that the dirstate path is
 
1536
            # correct.
 
1537
            return entry[1][parent_index][1]
1809
1538
 
1810
1539
    def get_revision_id(self):
1811
1540
        """Return the revision id for this tree."""
1829
1558
        return bool(self.path2id(filename))
1830
1559
 
1831
1560
    def kind(self, file_id):
1832
 
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
1834
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1835
 
        parent_index = self._get_parent_index()
1836
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1837
 
 
1838
 
    def stored_kind(self, file_id):
1839
 
        """See Tree.stored_kind"""
1840
 
        return self.kind(file_id)
 
1561
        return self.inventory[file_id].kind
1841
1562
 
1842
1563
    def path_content_summary(self, path):
1843
1564
        """See Tree.path_content_summary."""
1859
1580
            return None
1860
1581
        return ie.executable
1861
1582
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1583
    def list_files(self, include_root=False):
1863
1584
        # We use a standard implementation, because DirStateRevisionTree is
1864
1585
        # dealing with one of the parents of the current state
1865
1586
        inv = self._get_inventory()
1866
 
        if from_dir is None:
1867
 
            from_dir_id = None
1868
 
        else:
1869
 
            from_dir_id = inv.path2id(from_dir)
1870
 
            if from_dir_id is None:
1871
 
                # Directory not versioned
1872
 
                return
1873
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1874
 
        if inv.root is not None and not include_root and from_dir is None:
 
1587
        entries = inv.iter_entries()
 
1588
        if self.inventory.root is not None and not include_root:
1875
1589
            entries.next()
1876
1590
        for path, entry in entries:
1877
1591
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1624
                self._dirstate_locked = False
1911
1625
            self._repository.unlock()
1912
1626
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1627
    def walkdirs(self, prefix=""):
1918
1628
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
 
1629
        # implementation based on an inventory.  
1920
1630
        # This should be cleaned up to use the much faster Dirstate code
1921
1631
        # So for now, we just build up the parent inventory, and extract
1922
1632
        # it the same way RevisionTree does.
1951
1661
 
1952
1662
class InterDirStateTree(InterTree):
1953
1663
    """Fast path optimiser for changes_from with dirstate trees.
1954
 
 
1955
 
    This is used only when both trees are in the dirstate working file, and
1956
 
    the source is any parent within the dirstate, and the destination is
 
1664
    
 
1665
    This is used only when both trees are in the dirstate working file, and 
 
1666
    the source is any parent within the dirstate, and the destination is 
1957
1667
    the current working tree of the same dirstate.
1958
1668
    """
1959
1669
    # this could be generalized to allow comparisons between any trees in the
1972
1682
        target.set_parent_ids([revid])
1973
1683
        return target.basis_tree(), target
1974
1684
 
1975
 
    @classmethod
1976
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1977
 
        result = klass.make_source_parent_tree(source, target)
1978
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1979
 
        return result
1980
 
 
1981
 
    @classmethod
1982
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
1983
 
                                                  target):
1984
 
        from bzrlib.tests.test__dirstate_helpers import \
1985
 
            compiled_dirstate_helpers_feature
1986
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
1987
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
1988
 
        result = klass.make_source_parent_tree(source, target)
1989
 
        result[1]._iter_changes = ProcessEntryC
1990
 
        return result
1991
 
 
1992
1685
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1686
    _matching_to_tree_format = WorkingTreeFormat4()
1994
 
 
1995
 
    @classmethod
1996
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
1997
 
        # This method shouldn't be called, because we have python and C
1998
 
        # specific flavours.
1999
 
        raise NotImplementedError
2000
 
 
2001
 
    def iter_changes(self, include_unchanged=False,
 
1687
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1688
 
 
1689
    def _iter_changes(self, include_unchanged=False,
2002
1690
                      specific_files=None, pb=None, extra_trees=[],
2003
1691
                      require_versioned=True, want_unversioned=False):
2004
1692
        """Return the changes from source to target.
2005
1693
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1694
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1695
            for details.
2008
1696
        :param specific_files: An optional list of file paths to restrict the
2009
1697
            comparison to. When mapping filenames to ids, all matches in all
2020
1708
            output. An unversioned file is defined as one with (False, False)
2021
1709
            for the versioned pair.
2022
1710
        """
 
1711
        utf8_decode = cache_utf8._utf8_decode
 
1712
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1713
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1714
        # NB: show_status depends on being able to pass in non-versioned files
 
1715
        # and report them as unknown
2023
1716
        # TODO: handle extra trees in the dirstate.
2024
 
        if (extra_trees or specific_files == []):
 
1717
        # TODO: handle comparisons as an empty tree as a different special
 
1718
        # case? mbp 20070226
 
1719
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1720
            or specific_files == []):
2025
1721
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1722
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1723
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1724
                require_versioned, want_unversioned=want_unversioned):
 
1725
                yield f
 
1726
            return
2029
1727
        parent_ids = self.target.get_parent_ids()
2030
 
        if not (self.source._revision_id in parent_ids
2031
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2032
 
            raise AssertionError(
2033
 
                "revision {%s} is not stored in {%s}, but %s "
2034
 
                "can only be used for trees stored in the dirstate"
2035
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1728
        assert (self.source._revision_id in parent_ids), \
 
1729
                "revision {%s} is not stored in {%s}, but %s " \
 
1730
                "can only be used for trees stored in the dirstate" \
 
1731
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1732
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1733
        if self.source._revision_id == NULL_REVISION:
2038
1734
            source_index = None
2039
1735
            indices = (target_index,)
2040
1736
        else:
2041
 
            if not (self.source._revision_id in parent_ids):
2042
 
                raise AssertionError(
2043
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2044
 
                    self.source._revision_id, parent_ids))
 
1737
            assert (self.source._revision_id in parent_ids), \
 
1738
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1739
                self.source._revision_id, parent_ids)
2045
1740
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
 
            indices = (source_index, target_index)
 
1741
            indices = (source_index,target_index)
2047
1742
        # -- make all specific_files utf8 --
2048
1743
        if specific_files:
2049
1744
            specific_files_utf8 = set()
2050
1745
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1746
                specific_files_utf8.add(path.encode('utf8'))
2054
1747
            specific_files = specific_files_utf8
2055
1748
        else:
2056
1749
            specific_files = set([''])
2057
1750
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1751
        # -- get the state object and prepare it.
2060
1752
        state = self.target.current_dirstate()
2061
1753
        state._read_dirblocks_if_needed()
 
1754
        def _entries_for_path(path):
 
1755
            """Return a list with all the entries that match path for all ids.
 
1756
            """
 
1757
            dirname, basename = os.path.split(path)
 
1758
            key = (dirname, basename, '')
 
1759
            block_index, present = state._find_block_index_from_key(key)
 
1760
            if not present:
 
1761
                # the block which should contain path is absent.
 
1762
                return []
 
1763
            result = []
 
1764
            block = state._dirblocks[block_index][1]
 
1765
            entry_index, _ = state._find_entry_index(key, block)
 
1766
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1767
            while (entry_index < len(block) and
 
1768
                block[entry_index][0][0:2] == key[0:2]):
 
1769
                result.append(block[entry_index])
 
1770
                entry_index += 1
 
1771
            return result
2062
1772
        if require_versioned:
2063
1773
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1774
            all_versioned = True
2065
1775
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1776
                path_entries = _entries_for_path(path)
2067
1777
                if not path_entries:
2068
1778
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1779
                    all_versioned = False
 
1780
                    break
2071
1781
                found_versioned = False
2072
1782
                # for each id at this path
2073
1783
                for entry in path_entries:
2080
1790
                if not found_versioned:
2081
1791
                    # none of the indexes was not 'absent' at all ids for this
2082
1792
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1793
                    all_versioned = False
 
1794
                    break
 
1795
            if not all_versioned:
 
1796
                raise errors.PathsNotVersionedError(specific_files)
2086
1797
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1798
        search_specific_files = set()
 
1799
        for path in specific_files:
 
1800
            other_specific_files = specific_files.difference(set([path]))
 
1801
            if not osutils.is_inside_any(other_specific_files, path):
 
1802
                # this is a top level path, we must check it.
 
1803
                search_specific_files.add(path)
 
1804
        # sketch: 
 
1805
        # compare source_index and target_index at or under each element of search_specific_files.
 
1806
        # follow the following comparison table. Note that we only want to do diff operations when
 
1807
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1808
        # for the target.
 
1809
        # cases:
 
1810
        # 
 
1811
        # Source | Target | disk | action
 
1812
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1813
        #        |        |      | diff check on source-target
 
1814
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1815
        #        |        |      | ???
 
1816
        #   r    |  a     |      | add source to search
 
1817
        #   r    |  a     |  a   | 
 
1818
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1819
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1820
        #   a    | fdlt   |      | add new id
 
1821
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1822
        #   a    |  a     |      | not present in either tree, skip
 
1823
        #   a    |  a     |  a   | not present in any tree, skip
 
1824
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1825
        #        |        |      | may not be selected by the users list of paths.
 
1826
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1827
        #        |        |      | may not be selected by the users list of paths.
 
1828
        #  fdlt  | fdlt   |      | content in both: diff them
 
1829
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1830
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1831
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1832
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1833
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1834
        #        |        |      | this id at the other path.
 
1835
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1836
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1837
        #        |        |      | this id at the other path.
 
1838
 
 
1839
        # for all search_indexs in each path at or under each element of
 
1840
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1841
        # relocated path as one to search if its not searched already. If the
 
1842
        # detail is not relocated, add the id.
 
1843
        searched_specific_files = set()
 
1844
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1845
        # Using a list so that we can access the values and change them in
 
1846
        # nested scope. Each one is [path, file_id, entry]
 
1847
        last_source_parent = [None, None]
 
1848
        last_target_parent = [None, None]
2088
1849
 
2089
1850
        use_filesystem_for_exec = (sys.platform != 'win32')
2090
 
        iter_changes = self.target._iter_changes(include_unchanged,
2091
 
            use_filesystem_for_exec, search_specific_files, state,
2092
 
            source_index, target_index, want_unversioned, self.target)
2093
 
        return iter_changes.iter_changes()
 
1851
 
 
1852
        # Just a sentry, so that _process_entry can say that this
 
1853
        # record is handled, but isn't interesting to process (unchanged)
 
1854
        uninteresting = object()
 
1855
 
 
1856
 
 
1857
        old_dirname_to_file_id = {}
 
1858
        new_dirname_to_file_id = {}
 
1859
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1860
        #       keeping a cache of directories that we have seen.
 
1861
 
 
1862
        def _process_entry(entry, path_info):
 
1863
            """Compare an entry and real disk to generate delta information.
 
1864
 
 
1865
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1866
                the path of entry. If None, then the path is considered absent.
 
1867
                (Perhaps we should pass in a concrete entry for this ?)
 
1868
                Basename is returned as a utf8 string because we expect this
 
1869
                tuple will be ignored, and don't want to take the time to
 
1870
                decode.
 
1871
            :return: None if these don't match
 
1872
                     A tuple of information about the change, or
 
1873
                     the object 'uninteresting' if these match, but are
 
1874
                     basically identical.
 
1875
            """
 
1876
            if source_index is None:
 
1877
                source_details = NULL_PARENT_DETAILS
 
1878
            else:
 
1879
                source_details = entry[1][source_index]
 
1880
            target_details = entry[1][target_index]
 
1881
            target_minikind = target_details[0]
 
1882
            if path_info is not None and target_minikind in 'fdlt':
 
1883
                assert target_index == 0
 
1884
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1885
                                                  stat_value=path_info[3])
 
1886
                # The entry may have been modified by update_entry
 
1887
                target_details = entry[1][target_index]
 
1888
                target_minikind = target_details[0]
 
1889
            else:
 
1890
                link_or_sha1 = None
 
1891
            file_id = entry[0][2]
 
1892
            source_minikind = source_details[0]
 
1893
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1894
                # claimed content in both: diff
 
1895
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1896
                #        |        |      | diff check on source-target
 
1897
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1898
                #        |        |      | ???
 
1899
                if source_minikind in 'r':
 
1900
                    # add the source to the search path to find any children it
 
1901
                    # has.  TODO ? : only add if it is a container ?
 
1902
                    if not osutils.is_inside_any(searched_specific_files,
 
1903
                                                 source_details[1]):
 
1904
                        search_specific_files.add(source_details[1])
 
1905
                    # generate the old path; this is needed for stating later
 
1906
                    # as well.
 
1907
                    old_path = source_details[1]
 
1908
                    old_dirname, old_basename = os.path.split(old_path)
 
1909
                    path = pathjoin(entry[0][0], entry[0][1])
 
1910
                    old_entry = state._get_entry(source_index,
 
1911
                                                 path_utf8=old_path)
 
1912
                    # update the source details variable to be the real
 
1913
                    # location.
 
1914
                    source_details = old_entry[1][source_index]
 
1915
                    source_minikind = source_details[0]
 
1916
                else:
 
1917
                    old_dirname = entry[0][0]
 
1918
                    old_basename = entry[0][1]
 
1919
                    old_path = path = None
 
1920
                if path_info is None:
 
1921
                    # the file is missing on disk, show as removed.
 
1922
                    content_change = True
 
1923
                    target_kind = None
 
1924
                    target_exec = False
 
1925
                else:
 
1926
                    # source and target are both versioned and disk file is present.
 
1927
                    target_kind = path_info[2]
 
1928
                    if target_kind == 'directory':
 
1929
                        if path is None:
 
1930
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1931
                        new_dirname_to_file_id[path] = file_id
 
1932
                        if source_minikind != 'd':
 
1933
                            content_change = True
 
1934
                        else:
 
1935
                            # directories have no fingerprint
 
1936
                            content_change = False
 
1937
                        target_exec = False
 
1938
                    elif target_kind == 'file':
 
1939
                        if source_minikind != 'f':
 
1940
                            content_change = True
 
1941
                        else:
 
1942
                            # We could check the size, but we already have the
 
1943
                            # sha1 hash.
 
1944
                            content_change = (link_or_sha1 != source_details[1])
 
1945
                        # Target details is updated at update_entry time
 
1946
                        if use_filesystem_for_exec:
 
1947
                            # We don't need S_ISREG here, because we are sure
 
1948
                            # we are dealing with a file.
 
1949
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1950
                        else:
 
1951
                            target_exec = target_details[3]
 
1952
                    elif target_kind == 'symlink':
 
1953
                        if source_minikind != 'l':
 
1954
                            content_change = True
 
1955
                        else:
 
1956
                            content_change = (link_or_sha1 != source_details[1])
 
1957
                        target_exec = False
 
1958
                    elif target_kind == 'tree-reference':
 
1959
                        if source_minikind != 't':
 
1960
                            content_change = True
 
1961
                        else:
 
1962
                            content_change = False
 
1963
                        target_exec = False
 
1964
                    else:
 
1965
                        raise Exception, "unknown kind %s" % path_info[2]
 
1966
                if source_minikind == 'd':
 
1967
                    if path is None:
 
1968
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1969
                    old_dirname_to_file_id[old_path] = file_id
 
1970
                # parent id is the entry for the path in the target tree
 
1971
                if old_dirname == last_source_parent[0]:
 
1972
                    source_parent_id = last_source_parent[1]
 
1973
                else:
 
1974
                    try:
 
1975
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1976
                    except KeyError:
 
1977
                        source_parent_entry = state._get_entry(source_index,
 
1978
                                                               path_utf8=old_dirname)
 
1979
                        source_parent_id = source_parent_entry[0][2]
 
1980
                    if source_parent_id == entry[0][2]:
 
1981
                        # This is the root, so the parent is None
 
1982
                        source_parent_id = None
 
1983
                    else:
 
1984
                        last_source_parent[0] = old_dirname
 
1985
                        last_source_parent[1] = source_parent_id
 
1986
                new_dirname = entry[0][0]
 
1987
                if new_dirname == last_target_parent[0]:
 
1988
                    target_parent_id = last_target_parent[1]
 
1989
                else:
 
1990
                    try:
 
1991
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1992
                    except KeyError:
 
1993
                        # TODO: We don't always need to do the lookup, because the
 
1994
                        #       parent entry will be the same as the source entry.
 
1995
                        target_parent_entry = state._get_entry(target_index,
 
1996
                                                               path_utf8=new_dirname)
 
1997
                        assert target_parent_entry != (None, None), (
 
1998
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1999
                            % (new_dirname, entry))
 
2000
                        target_parent_id = target_parent_entry[0][2]
 
2001
                    if target_parent_id == entry[0][2]:
 
2002
                        # This is the root, so the parent is None
 
2003
                        target_parent_id = None
 
2004
                    else:
 
2005
                        last_target_parent[0] = new_dirname
 
2006
                        last_target_parent[1] = target_parent_id
 
2007
 
 
2008
                source_exec = source_details[3]
 
2009
                if (include_unchanged
 
2010
                    or content_change
 
2011
                    or source_parent_id != target_parent_id
 
2012
                    or old_basename != entry[0][1]
 
2013
                    or source_exec != target_exec
 
2014
                    ):
 
2015
                    if old_path is None:
 
2016
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2017
                        old_path_u = utf8_decode(old_path)[0]
 
2018
                        path_u = old_path_u
 
2019
                    else:
 
2020
                        old_path_u = utf8_decode(old_path)[0]
 
2021
                        if old_path == path:
 
2022
                            path_u = old_path_u
 
2023
                        else:
 
2024
                            path_u = utf8_decode(path)[0]
 
2025
                    source_kind = _minikind_to_kind[source_minikind]
 
2026
                    return (entry[0][2],
 
2027
                           (old_path_u, path_u),
 
2028
                           content_change,
 
2029
                           (True, True),
 
2030
                           (source_parent_id, target_parent_id),
 
2031
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2032
                           (source_kind, target_kind),
 
2033
                           (source_exec, target_exec))
 
2034
                else:
 
2035
                    return uninteresting
 
2036
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2037
                # looks like a new file
 
2038
                if path_info is not None:
 
2039
                    path = pathjoin(entry[0][0], entry[0][1])
 
2040
                    # parent id is the entry for the path in the target tree
 
2041
                    # TODO: these are the same for an entire directory: cache em.
 
2042
                    parent_id = state._get_entry(target_index,
 
2043
                                                 path_utf8=entry[0][0])[0][2]
 
2044
                    if parent_id == entry[0][2]:
 
2045
                        parent_id = None
 
2046
                    if use_filesystem_for_exec:
 
2047
                        # We need S_ISREG here, because we aren't sure if this
 
2048
                        # is a file or not.
 
2049
                        target_exec = bool(
 
2050
                            stat.S_ISREG(path_info[3].st_mode)
 
2051
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2052
                    else:
 
2053
                        target_exec = target_details[3]
 
2054
                    return (entry[0][2],
 
2055
                           (None, utf8_decode(path)[0]),
 
2056
                           True,
 
2057
                           (False, True),
 
2058
                           (None, parent_id),
 
2059
                           (None, utf8_decode(entry[0][1])[0]),
 
2060
                           (None, path_info[2]),
 
2061
                           (None, target_exec))
 
2062
                else:
 
2063
                    # but its not on disk: we deliberately treat this as just
 
2064
                    # never-present. (Why ?! - RBC 20070224)
 
2065
                    pass
 
2066
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2067
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2068
                # if its still on disk, *and* theres no other entry at this
 
2069
                # path [we dont know this in this routine at the moment -
 
2070
                # perhaps we should change this - then it would be an unknown.
 
2071
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2072
                # parent id is the entry for the path in the target tree
 
2073
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2074
                if parent_id == entry[0][2]:
 
2075
                    parent_id = None
 
2076
                return (entry[0][2],
 
2077
                       (utf8_decode(old_path)[0], None),
 
2078
                       True,
 
2079
                       (True, False),
 
2080
                       (parent_id, None),
 
2081
                       (utf8_decode(entry[0][1])[0], None),
 
2082
                       (_minikind_to_kind[source_minikind], None),
 
2083
                       (source_details[3], None))
 
2084
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2085
                # a rename; could be a true rename, or a rename inherited from
 
2086
                # a renamed parent. TODO: handle this efficiently. Its not
 
2087
                # common case to rename dirs though, so a correct but slow
 
2088
                # implementation will do.
 
2089
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2090
                    search_specific_files.add(target_details[1])
 
2091
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2092
                # neither of the selected trees contain this file,
 
2093
                # so skip over it. This is not currently directly tested, but
 
2094
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2095
                pass
 
2096
            else:
 
2097
                raise AssertionError("don't know how to compare "
 
2098
                    "source_minikind=%r, target_minikind=%r"
 
2099
                    % (source_minikind, target_minikind))
 
2100
                ## import pdb;pdb.set_trace()
 
2101
            return None
 
2102
 
 
2103
        while search_specific_files:
 
2104
            # TODO: the pending list should be lexically sorted?  the
 
2105
            # interface doesn't require it.
 
2106
            current_root = search_specific_files.pop()
 
2107
            current_root_unicode = current_root.decode('utf8')
 
2108
            searched_specific_files.add(current_root)
 
2109
            # process the entries for this containing directory: the rest will be
 
2110
            # found by their parents recursively.
 
2111
            root_entries = _entries_for_path(current_root)
 
2112
            root_abspath = self.target.abspath(current_root_unicode)
 
2113
            try:
 
2114
                root_stat = os.lstat(root_abspath)
 
2115
            except OSError, e:
 
2116
                if e.errno == errno.ENOENT:
 
2117
                    # the path does not exist: let _process_entry know that.
 
2118
                    root_dir_info = None
 
2119
                else:
 
2120
                    # some other random error: hand it up.
 
2121
                    raise
 
2122
            else:
 
2123
                root_dir_info = ('', current_root,
 
2124
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2125
                    root_abspath)
 
2126
                if root_dir_info[2] == 'directory':
 
2127
                    if self.target._directory_is_tree_reference(
 
2128
                        current_root.decode('utf8')):
 
2129
                        root_dir_info = root_dir_info[:2] + \
 
2130
                            ('tree-reference',) + root_dir_info[3:]
 
2131
 
 
2132
            if not root_entries and not root_dir_info:
 
2133
                # this specified path is not present at all, skip it.
 
2134
                continue
 
2135
            path_handled = False
 
2136
            for entry in root_entries:
 
2137
                result = _process_entry(entry, root_dir_info)
 
2138
                if result is not None:
 
2139
                    path_handled = True
 
2140
                    if result is not uninteresting:
 
2141
                        yield result
 
2142
            if want_unversioned and not path_handled and root_dir_info:
 
2143
                new_executable = bool(
 
2144
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2145
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2146
                yield (None,
 
2147
                       (None, current_root_unicode),
 
2148
                       True,
 
2149
                       (False, False),
 
2150
                       (None, None),
 
2151
                       (None, splitpath(current_root_unicode)[-1]),
 
2152
                       (None, root_dir_info[2]),
 
2153
                       (None, new_executable)
 
2154
                      )
 
2155
            initial_key = (current_root, '', '')
 
2156
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2157
            if block_index == 0:
 
2158
                # we have processed the total root already, but because the
 
2159
                # initial key matched it we should skip it here.
 
2160
                block_index +=1
 
2161
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2162
                current_dir_info = None
 
2163
            else:
 
2164
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2165
                try:
 
2166
                    current_dir_info = dir_iterator.next()
 
2167
                except OSError, e:
 
2168
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2169
                    # python 2.5 has e.errno == EINVAL,
 
2170
                    #            and e.winerror == ERROR_DIRECTORY
 
2171
                    e_winerror = getattr(e, 'winerror', None)
 
2172
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2173
                    # there may be directories in the inventory even though
 
2174
                    # this path is not a file on disk: so mark it as end of
 
2175
                    # iterator
 
2176
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2177
                        current_dir_info = None
 
2178
                    elif (sys.platform == 'win32'
 
2179
                          and (e.errno in win_errors
 
2180
                               or e_winerror in win_errors)):
 
2181
                        current_dir_info = None
 
2182
                    else:
 
2183
                        raise
 
2184
                else:
 
2185
                    if current_dir_info[0][0] == '':
 
2186
                        # remove .bzr from iteration
 
2187
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2188
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2189
                        del current_dir_info[1][bzr_index]
 
2190
            # walk until both the directory listing and the versioned metadata
 
2191
            # are exhausted. 
 
2192
            if (block_index < len(state._dirblocks) and
 
2193
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2194
                current_block = state._dirblocks[block_index]
 
2195
            else:
 
2196
                current_block = None
 
2197
            while (current_dir_info is not None or
 
2198
                   current_block is not None):
 
2199
                if (current_dir_info and current_block
 
2200
                    and current_dir_info[0][0] != current_block[0]):
 
2201
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2202
                        # filesystem data refers to paths not covered by the dirblock.
 
2203
                        # this has two possibilities:
 
2204
                        # A) it is versioned but empty, so there is no block for it
 
2205
                        # B) it is not versioned.
 
2206
 
 
2207
                        # if (A) then we need to recurse into it to check for
 
2208
                        # new unknown files or directories.
 
2209
                        # if (B) then we should ignore it, because we don't
 
2210
                        # recurse into unknown directories.
 
2211
                        path_index = 0
 
2212
                        while path_index < len(current_dir_info[1]):
 
2213
                                current_path_info = current_dir_info[1][path_index]
 
2214
                                if want_unversioned:
 
2215
                                    if current_path_info[2] == 'directory':
 
2216
                                        if self.target._directory_is_tree_reference(
 
2217
                                            current_path_info[0].decode('utf8')):
 
2218
                                            current_path_info = current_path_info[:2] + \
 
2219
                                                ('tree-reference',) + current_path_info[3:]
 
2220
                                    new_executable = bool(
 
2221
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2222
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2223
                                    yield (None,
 
2224
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2225
                                        True,
 
2226
                                        (False, False),
 
2227
                                        (None, None),
 
2228
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2229
                                        (None, current_path_info[2]),
 
2230
                                        (None, new_executable))
 
2231
                                # dont descend into this unversioned path if it is
 
2232
                                # a dir
 
2233
                                if current_path_info[2] in ('directory',
 
2234
                                                            'tree-reference'):
 
2235
                                    del current_dir_info[1][path_index]
 
2236
                                    path_index -= 1
 
2237
                                path_index += 1
 
2238
 
 
2239
                        # This dir info has been handled, go to the next
 
2240
                        try:
 
2241
                            current_dir_info = dir_iterator.next()
 
2242
                        except StopIteration:
 
2243
                            current_dir_info = None
 
2244
                    else:
 
2245
                        # We have a dirblock entry for this location, but there
 
2246
                        # is no filesystem path for this. This is most likely
 
2247
                        # because a directory was removed from the disk.
 
2248
                        # We don't have to report the missing directory,
 
2249
                        # because that should have already been handled, but we
 
2250
                        # need to handle all of the files that are contained
 
2251
                        # within.
 
2252
                        for current_entry in current_block[1]:
 
2253
                            # entry referring to file not present on disk.
 
2254
                            # advance the entry only, after processing.
 
2255
                            result = _process_entry(current_entry, None)
 
2256
                            if result is not None:
 
2257
                                if result is not uninteresting:
 
2258
                                    yield result
 
2259
                        block_index +=1
 
2260
                        if (block_index < len(state._dirblocks) and
 
2261
                            osutils.is_inside(current_root,
 
2262
                                              state._dirblocks[block_index][0])):
 
2263
                            current_block = state._dirblocks[block_index]
 
2264
                        else:
 
2265
                            current_block = None
 
2266
                    continue
 
2267
                entry_index = 0
 
2268
                if current_block and entry_index < len(current_block[1]):
 
2269
                    current_entry = current_block[1][entry_index]
 
2270
                else:
 
2271
                    current_entry = None
 
2272
                advance_entry = True
 
2273
                path_index = 0
 
2274
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2275
                    current_path_info = current_dir_info[1][path_index]
 
2276
                    if current_path_info[2] == 'directory':
 
2277
                        if self.target._directory_is_tree_reference(
 
2278
                            current_path_info[0].decode('utf8')):
 
2279
                            current_path_info = current_path_info[:2] + \
 
2280
                                ('tree-reference',) + current_path_info[3:]
 
2281
                else:
 
2282
                    current_path_info = None
 
2283
                advance_path = True
 
2284
                path_handled = False
 
2285
                while (current_entry is not None or
 
2286
                    current_path_info is not None):
 
2287
                    if current_entry is None:
 
2288
                        # the check for path_handled when the path is adnvaced
 
2289
                        # will yield this path if needed.
 
2290
                        pass
 
2291
                    elif current_path_info is None:
 
2292
                        # no path is fine: the per entry code will handle it.
 
2293
                        result = _process_entry(current_entry, current_path_info)
 
2294
                        if result is not None:
 
2295
                            if result is not uninteresting:
 
2296
                                yield result
 
2297
                    elif (current_entry[0][1] != current_path_info[1]
 
2298
                          or current_entry[1][target_index][0] in 'ar'):
 
2299
                        # The current path on disk doesn't match the dirblock
 
2300
                        # record. Either the dirblock is marked as absent, or
 
2301
                        # the file on disk is not present at all in the
 
2302
                        # dirblock. Either way, report about the dirblock
 
2303
                        # entry, and let other code handle the filesystem one.
 
2304
 
 
2305
                        # Compare the basename for these files to determine
 
2306
                        # which comes first
 
2307
                        if current_path_info[1] < current_entry[0][1]:
 
2308
                            # extra file on disk: pass for now, but only
 
2309
                            # increment the path, not the entry
 
2310
                            advance_entry = False
 
2311
                        else:
 
2312
                            # entry referring to file not present on disk.
 
2313
                            # advance the entry only, after processing.
 
2314
                            result = _process_entry(current_entry, None)
 
2315
                            if result is not None:
 
2316
                                if result is not uninteresting:
 
2317
                                    yield result
 
2318
                            advance_path = False
 
2319
                    else:
 
2320
                        result = _process_entry(current_entry, current_path_info)
 
2321
                        if result is not None:
 
2322
                            path_handled = True
 
2323
                            if result is not uninteresting:
 
2324
                                yield result
 
2325
                    if advance_entry and current_entry is not None:
 
2326
                        entry_index += 1
 
2327
                        if entry_index < len(current_block[1]):
 
2328
                            current_entry = current_block[1][entry_index]
 
2329
                        else:
 
2330
                            current_entry = None
 
2331
                    else:
 
2332
                        advance_entry = True # reset the advance flaga
 
2333
                    if advance_path and current_path_info is not None:
 
2334
                        if not path_handled:
 
2335
                            # unversioned in all regards
 
2336
                            if want_unversioned:
 
2337
                                new_executable = bool(
 
2338
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2339
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2340
                                yield (None,
 
2341
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2342
                                    True,
 
2343
                                    (False, False),
 
2344
                                    (None, None),
 
2345
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2346
                                    (None, current_path_info[2]),
 
2347
                                    (None, new_executable))
 
2348
                            # dont descend into this unversioned path if it is
 
2349
                            # a dir
 
2350
                            if current_path_info[2] in ('directory'):
 
2351
                                del current_dir_info[1][path_index]
 
2352
                                path_index -= 1
 
2353
                        # dont descend the disk iterator into any tree 
 
2354
                        # paths.
 
2355
                        if current_path_info[2] == 'tree-reference':
 
2356
                            del current_dir_info[1][path_index]
 
2357
                            path_index -= 1
 
2358
                        path_index += 1
 
2359
                        if path_index < len(current_dir_info[1]):
 
2360
                            current_path_info = current_dir_info[1][path_index]
 
2361
                            if current_path_info[2] == 'directory':
 
2362
                                if self.target._directory_is_tree_reference(
 
2363
                                    current_path_info[0].decode('utf8')):
 
2364
                                    current_path_info = current_path_info[:2] + \
 
2365
                                        ('tree-reference',) + current_path_info[3:]
 
2366
                        else:
 
2367
                            current_path_info = None
 
2368
                        path_handled = False
 
2369
                    else:
 
2370
                        advance_path = True # reset the advance flagg.
 
2371
                if current_block is not None:
 
2372
                    block_index += 1
 
2373
                    if (block_index < len(state._dirblocks) and
 
2374
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2375
                        current_block = state._dirblocks[block_index]
 
2376
                    else:
 
2377
                        current_block = None
 
2378
                if current_dir_info is not None:
 
2379
                    try:
 
2380
                        current_dir_info = dir_iterator.next()
 
2381
                    except StopIteration:
 
2382
                        current_dir_info = None
 
2383
 
2094
2384
 
2095
2385
    @staticmethod
2096
2386
    def is_compatible(source, target):
2097
2387
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2388
        if not isinstance(target, WorkingTree4):
2099
2389
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2390
        # the source must be a revtreee or dirstate rev tree.
2101
2391
        if not isinstance(source,
2102
2392
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2393
            return False
2104
2394
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2395
        if not (source._revision_id == NULL_REVISION or
2106
2396
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2397
            # TODO: what about ghosts? it may well need to 
2108
2398
            # check for them explicitly.
2109
2399
            return False
2110
2400
        return True
2120
2410
 
2121
2411
    def convert(self, tree):
2122
2412
        # lock the control files not the tree, so that we dont get tree
2123
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2413
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2414
        # tree during upgrade.
2125
2415
        tree._control_files.lock_write()
2126
2416
        try:
2152
2442
 
2153
2443
    def update_format(self, tree):
2154
2444
        """Change the format marker."""
2155
 
        tree._transport.put_bytes('format',
2156
 
            self.target_format.get_format_string(),
2157
 
            mode=tree.bzrdir._get_file_mode())
2158
 
 
2159
 
 
2160
 
class Converter4to5(object):
2161
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2162
 
 
2163
 
    def __init__(self):
2164
 
        self.target_format = WorkingTreeFormat5()
2165
 
 
2166
 
    def convert(self, tree):
2167
 
        # lock the control files not the tree, so that we don't get tree
2168
 
        # on-unlock behaviours, and so that no-one else diddles with the
2169
 
        # tree during upgrade.
2170
 
        tree._control_files.lock_write()
2171
 
        try:
2172
 
            self.update_format(tree)
2173
 
        finally:
2174
 
            tree._control_files.unlock()
2175
 
 
2176
 
    def update_format(self, tree):
2177
 
        """Change the format marker."""
2178
 
        tree._transport.put_bytes('format',
2179
 
            self.target_format.get_format_string(),
2180
 
            mode=tree.bzrdir._get_file_mode())
2181
 
 
2182
 
 
2183
 
class Converter4or5to6(object):
2184
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2185
 
 
2186
 
    def __init__(self):
2187
 
        self.target_format = WorkingTreeFormat6()
2188
 
 
2189
 
    def convert(self, tree):
2190
 
        # lock the control files not the tree, so that we don't get tree
2191
 
        # on-unlock behaviours, and so that no-one else diddles with the
2192
 
        # tree during upgrade.
2193
 
        tree._control_files.lock_write()
2194
 
        try:
2195
 
            self.init_custom_control_files(tree)
2196
 
            self.update_format(tree)
2197
 
        finally:
2198
 
            tree._control_files.unlock()
2199
 
 
2200
 
    def init_custom_control_files(self, tree):
2201
 
        """Initialize custom control files."""
2202
 
        tree._transport.put_bytes('views', '',
2203
 
            mode=tree.bzrdir._get_file_mode())
2204
 
 
2205
 
    def update_format(self, tree):
2206
 
        """Change the format marker."""
2207
 
        tree._transport.put_bytes('format',
2208
 
            self.target_format.get_format_string(),
2209
 
            mode=tree.bzrdir._get_file_mode())
 
2445
        tree._control_files.put_utf8('format',
 
2446
            self.target_format.get_format_string())