/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Martin Pool
  • Date: 2007-10-03 08:06:44 UTC
  • mto: This revision was merged to the branch mainline in revision 2901.
  • Revision ID: mbp@sourcefrog.net-20071003080644-oivy0gkg98sex0ed
Avoid internal error tracebacks on failure to lock on readonly transport (#129701).

Add new LockFailed, which doesn't imply that we failed to get it because of
contention.  Raise this if we fail to create the pending or lock directories
because of Transport errors.

UnlockableTransport is not an internal error.

ReadOnlyLockError has a message which didn't match its name or usage; it's now
deprecated and callers are updated to use LockFailed which is more appropriate.

Add zero_ninetytwo deprecation symbol.

Unify assertMatchesRe with TestCase.assertContainsRe.

When the constructor is deprecated, just say that the class is deprecated, not
the __init__ method - this works better with applyDeprecated in tests.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
38
46
    debug,
 
47
    delta,
39
48
    dirstate,
40
49
    errors,
41
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
55
    revision as _mod_revision,
44
56
    revisiontree,
 
57
    textui,
45
58
    trace,
46
59
    transform,
47
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
48
63
    )
49
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
50
66
import bzrlib.ui
51
67
""")
52
68
 
 
69
from bzrlib import symbol_versioning
53
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
56
75
from bzrlib.mutabletree import needs_tree_write_lock
57
76
from bzrlib.osutils import (
58
77
    file_kind,
59
78
    isdir,
 
79
    normpath,
60
80
    pathjoin,
 
81
    rand_chars,
61
82
    realpath,
62
83
    safe_unicode,
 
84
    splitpath,
63
85
    )
64
 
from bzrlib.trace import mutter
 
86
from bzrlib.trace import mutter, note
65
87
from bzrlib.transport.local import LocalTransport
66
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
67
97
from bzrlib.tree import Tree
68
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
99
 
70
100
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
72
120
    def __init__(self, basedir,
73
121
                 branch,
74
122
                 _control_files=None,
83
131
        """
84
132
        self._format = _format
85
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
86
136
        basedir = safe_unicode(basedir)
87
137
        mutter("opening working tree %r", basedir)
88
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
89
141
        self.basedir = realpath(basedir)
90
142
        # if branch is at our basedir and is a format 6 or less
91
143
        # assume all other formats have their own control files.
 
144
        assert isinstance(_control_files, LockableFiles), \
 
145
            "_control_files must be a LockableFiles, not %r" % _control_files
92
146
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
147
        self._dirty = None
95
148
        #-------------
96
149
        # during a read or write lock these objects are set, and are
98
151
        self._dirstate = None
99
152
        self._inventory = None
100
153
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
154
 
108
155
    @needs_tree_write_lock
109
156
    def _add(self, files, ids, kinds):
111
158
        state = self.current_dirstate()
112
159
        for f, file_id, kind in zip(files, ids, kinds):
113
160
            f = f.strip('/')
 
161
            assert '//' not in f
 
162
            assert '..' not in f
114
163
            if self.path2id(f):
115
164
                # special case tree root handling.
116
165
                if f == '' and self.path2id(f) == ROOT_ID:
137
186
    @needs_tree_write_lock
138
187
    def add_reference(self, sub_tree):
139
188
        # use standard implementation, which calls back to self._add
140
 
        #
 
189
        # 
141
190
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
143
192
        self._add_reference(sub_tree)
144
193
 
145
194
    def break_lock(self):
184
233
            WorkingTree3._comparison_data(self, entry, path)
185
234
        # it looks like a plain directory, but it's really a reference -- see
186
235
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
236
        if (self._repo_supports_tree_reference and
 
237
            kind == 'directory' and
 
238
            self._directory_is_tree_reference(path)):
189
239
            kind = 'tree-reference'
190
240
        return kind, executable, stat_value
191
241
 
217
267
            return self._dirstate
218
268
        local_path = self.bzrdir.get_workingtree_transport(None
219
269
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
222
271
        return self._dirstate
223
272
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
235
 
 
236
273
    def filter_unversioned_files(self, paths):
237
274
        """Filter out paths that are versioned.
238
275
 
270
307
 
271
308
    def _generate_inventory(self):
272
309
        """Create and set self.inventory from the dirstate object.
273
 
 
 
310
        
274
311
        This is relatively expensive: we have to walk the entire dirstate.
275
312
        Ideally we would not, and can deprecate this function.
276
313
        """
280
317
        state._read_dirblocks_if_needed()
281
318
        root_key, current_entry = self._get_entry(path='')
282
319
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
285
321
        inv = Inventory(root_id=current_id)
286
322
        # Turn some things into local variables
287
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
320
356
                    # add this entry to the parent map.
321
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
358
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
359
                    assert self._repo_supports_tree_reference, \
 
360
                        "repository of %r " \
 
361
                        "doesn't support tree references " \
 
362
                        "required by entry %r" \
 
363
                        % (self, name)
327
364
                    inv_entry.reference_revision = link_or_sha1 or None
328
365
                elif kind != 'symlink':
329
366
                    raise AssertionError("unknown kind %r" % kind)
330
367
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
368
                assert file_id not in inv_byid, ('file_id %s already in'
 
369
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
370
                assert name_unicode not in parent_ie.children
337
371
                inv_byid[file_id] = inv_entry
338
372
                parent_ie.children[name_unicode] = inv_entry
339
373
        self._inventory = inv
344
378
        If either file_id or path is supplied, it is used as the key to lookup.
345
379
        If both are supplied, the fastest lookup is used, and an error is
346
380
        raised if they do not both point at the same row.
347
 
 
 
381
        
348
382
        :param file_id: An optional unicode file_id to be looked up.
349
383
        :param path: An optional unicode path to be looked up.
350
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
374
408
                    return None
375
409
                else:
376
410
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
379
413
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
414
            return link_or_sha1
390
415
        return None
391
416
 
392
417
    def _get_inventory(self):
406
431
    @needs_read_lock
407
432
    def get_parent_ids(self):
408
433
        """See Tree.get_parent_ids.
409
 
 
 
434
        
410
435
        This implementation requests the ids list from the dirstate file.
411
436
        """
412
437
        return self.current_dirstate().get_parent_ids()
428
453
 
429
454
    def has_id(self, file_id):
430
455
        state = self.current_dirstate()
 
456
        file_id = osutils.safe_file_id(file_id)
431
457
        row, parents = self._get_entry(file_id=file_id)
432
458
        if row is None:
433
459
            return False
434
460
        return osutils.lexists(pathjoin(
435
461
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
462
 
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
463
    @needs_read_lock
443
464
    def id2path(self, file_id):
444
465
        "Convert a file-id to a path."
 
466
        file_id = osutils.safe_file_id(file_id)
445
467
        state = self.current_dirstate()
446
468
        entry = self._get_entry(file_id=file_id)
447
469
        if entry == (None, None):
449
471
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
472
        return path_utf8.decode('utf8')
451
473
 
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
474
    if not osutils.supports_executable():
459
475
        def is_executable(self, file_id, path=None):
460
476
            """Test if a file is executable or not.
461
477
 
462
478
            Note: The caller is expected to take a read-lock before calling this.
463
479
            """
 
480
            file_id = osutils.safe_file_id(file_id)
464
481
            entry = self._get_entry(file_id=file_id, path=path)
465
482
            if entry == (None, None):
466
483
                return False
467
484
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
485
    else:
472
486
        def is_executable(self, file_id, path=None):
473
487
            """Test if a file is executable or not.
474
488
 
475
489
            Note: The caller is expected to take a read-lock before calling this.
476
490
            """
477
 
            self._must_be_locked()
478
491
            if not path:
 
492
                file_id = osutils.safe_file_id(file_id)
479
493
                path = self.id2path(file_id)
480
494
            mode = os.lstat(self.abspath(path)).st_mode
481
495
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
496
 
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
497
    @needs_read_lock
494
498
    def __iter__(self):
495
499
        """Iterate through file_ids for this tree.
508
512
        return iter(result)
509
513
 
510
514
    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
515
        for key, tree_details in self.current_dirstate()._iter_entries():
516
516
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
517
                # not relevant to the working tree
519
519
            if not key[1]:
520
520
                # the root is not a reference.
521
521
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
522
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
523
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
524
                if self._kind(path) == 'tree-reference':
 
525
                    yield path, key[2]
526
526
            except errors.NoSuchFile:
527
527
                # path is missing on disk.
528
528
                continue
529
529
 
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
530
    def kind(self, file_id):
537
531
        """Return the kind of a file.
538
532
 
542
536
        Note: The caller is expected to take a read-lock before calling this.
543
537
        """
544
538
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
539
        assert relpath != None, \
 
540
            "path for id {%s} is None!" % file_id
548
541
        return self._kind(relpath)
549
542
 
550
543
    def _kind(self, relpath):
551
544
        abspath = self.abspath(relpath)
552
545
        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'
 
546
        if (self._repo_supports_tree_reference and
 
547
            kind == 'directory' and
 
548
            self._directory_is_tree_reference(relpath)):
 
549
            kind = 'tree-reference'
558
550
        return kind
559
551
 
560
552
    @needs_read_lock
623
615
        result = []
624
616
        if not from_paths:
625
617
            return result
 
618
 
626
619
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
620
 
 
621
        assert not isinstance(from_paths, basestring)
629
622
        to_dir_utf8 = to_dir.encode('utf8')
630
623
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
624
        id_index = state._get_id_index()
653
646
        if self._inventory is not None:
654
647
            update_inventory = True
655
648
            inv = self.inventory
 
649
            to_dir_ie = inv[to_dir_id]
656
650
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
651
        else:
659
652
            update_inventory = False
660
653
 
691
684
            from_entry = self._get_entry(path=from_rel)
692
685
            if from_entry == (None, None):
693
686
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
687
                    errors.NotVersionedError(path=str(from_rel)))
695
688
 
696
689
            from_id = from_entry[0][2]
697
690
            to_rel = pathjoin(to_dir, from_tail)
724
717
                if from_missing: # implicitly just update our path mapping
725
718
                    move_file = False
726
719
                elif not after:
727
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
720
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
721
                        extra="(Use --after to update the Bazaar id)")
728
722
 
729
723
            rollbacks = []
730
724
            def rollback_rename():
786
780
                if minikind == 'd':
787
781
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
782
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
783
                        assert from_dir != '', "renaming root not supported"
791
784
                        from_key = (from_dir, '')
792
785
                        from_block_idx, present = \
793
786
                            state._find_block_index_from_key(from_key)
806
799
 
807
800
                        # Grab a copy since move_one may update the list.
808
801
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
802
                            assert entry[0][0] == from_dir
811
803
                            cur_details = entry[1][0]
812
804
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
805
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
872
864
        for tree in trees:
873
865
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
866
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
867
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
868
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
869
        # -- make all paths utf8 --
879
870
        paths_utf8 = set()
940
931
                raise errors.PathsNotVersionedError(paths)
941
932
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
933
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
934
        # sketch: 
944
935
        # for all search_indexs in each path at or under each element of
945
936
        # search_paths, if the detail is relocated: add the id, and add the
946
937
        # relocated path as one to search if its not searched already. If the
1002
993
 
1003
994
    def read_working_inventory(self):
1004
995
        """Read the working inventory.
1005
 
 
 
996
        
1006
997
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
998
        """
1008
999
        return self.inventory
1013
1004
 
1014
1005
        WorkingTree4 supplies revision_trees for any basis tree.
1015
1006
        """
 
1007
        revision_id = osutils.safe_revision_id(revision_id)
1016
1008
        dirstate = self.current_dirstate()
1017
1009
        parent_ids = dirstate.get_parent_ids()
1018
1010
        if revision_id not in parent_ids:
1025
1017
    @needs_tree_write_lock
1026
1018
    def set_last_revision(self, new_revision):
1027
1019
        """Change the last revision in the working tree."""
 
1020
        new_revision = osutils.safe_revision_id(new_revision)
1028
1021
        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.")
 
1022
        if new_revision in (NULL_REVISION, None):
 
1023
            assert len(parents) < 2, (
 
1024
                "setting the last parent to none with a pending merge is "
 
1025
                "unsupported.")
1034
1026
            self.set_parent_ids([])
1035
1027
        else:
1036
1028
            self.set_parent_ids([new_revision] + parents[1:],
1039
1031
    @needs_tree_write_lock
1040
1032
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
1033
        """Set the parent ids to revision_ids.
1042
 
 
 
1034
        
1043
1035
        See also set_parent_trees. This api will try to retrieve the tree data
1044
1036
        for each element of revision_ids from the trees repository. If you have
1045
1037
        tree data already available, it is more efficient to use
1049
1041
        :param revision_ids: The revision_ids to set as the parent ids of this
1050
1042
            working tree. Any of these may be ghosts.
1051
1043
        """
 
1044
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1052
1045
        trees = []
1053
1046
        for revision_id in revision_ids:
1054
1047
            try:
1077
1070
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
1071
        real_trees = []
1079
1072
        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
1073
        # convert absent trees to the null tree, which we convert back to
1087
1074
        # missing on access.
1088
1075
        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
 
1076
            rev_id = osutils.safe_revision_id(rev_id)
1095
1077
            _mod_revision.check_not_reserved_id(rev_id)
1096
1078
            if tree is not None:
1097
1079
                real_trees.append((rev_id, tree))
1098
1080
            else:
1099
1081
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1082
                    self.branch.repository.revision_tree(None)))
1102
1083
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
1104
1084
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1105
1085
        self._make_dirty(reset_inventory=False)
1106
1086
 
1119
1099
        :param path: The path.
1120
1100
        :param stat_result: The stat result being looked up.
1121
1101
        """
1122
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1102
        state = self.current_dirstate()
 
1103
        # XXX: should we make the path be passed in as utf8 ?
 
1104
        entry = state._get_entry(0, path_utf8=cache_utf8.encode(path))
 
1105
        tree_details = entry[1][0]
 
1106
        packed_stat = dirstate.pack_stat(stat_result)
 
1107
        if tree_details[4] == packed_stat:
 
1108
            return tree_details[1]
 
1109
        else:
 
1110
            return None
1123
1111
 
1124
1112
    @needs_read_lock
1125
1113
    def supports_tree_reference(self):
1166
1154
            return
1167
1155
        state = self.current_dirstate()
1168
1156
        state._read_dirblocks_if_needed()
1169
 
        ids_to_unversion = set(file_ids)
 
1157
        ids_to_unversion = set()
 
1158
        for file_id in file_ids:
 
1159
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1170
1160
        paths_to_unversion = set()
1171
1161
        # sketch:
1172
1162
        # check if the root is to be unversioned, if so, assert for now.
1199
1189
                # just forget the whole block.
1200
1190
                entry_index = 0
1201
1191
                while entry_index < len(block[1]):
 
1192
                    # Mark this file id as having been removed
1202
1193
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1194
                    ids_to_unversion.discard(entry[0][2])
 
1195
                    if (entry[1][0][0] == 'a'
 
1196
                        or not state._make_absent(entry)):
1205
1197
                        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
1198
                # go to the next block. (At the moment we dont delete empty
1213
1199
                # dirblocks)
1214
1200
                block_index += 1
1237
1223
            for file_id in file_ids:
1238
1224
                self._inventory.remove_recursive_id(file_id)
1239
1225
 
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
1226
    @needs_read_lock
1260
1227
    def _validate(self):
1261
1228
        self._dirstate._validate()
1263
1230
    @needs_tree_write_lock
1264
1231
    def _write_inventory(self, inv):
1265
1232
        """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:
 
1233
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1234
        self.current_dirstate().set_state_from_inventory(inv)
 
1235
        self._make_dirty(reset_inventory=False)
 
1236
        if self._inventory is not None:
1280
1237
            self._inventory = inv
1281
1238
        self.flush()
1282
1239
 
1283
1240
 
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):
 
1241
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1242
    """The first consolidated dirstate working tree format.
 
1243
 
 
1244
    This format:
 
1245
        - exists within a metadir controlling .bzr
 
1246
        - includes an explicit version marker for the workingtree control
 
1247
          files, separate from the BzrDir format
 
1248
        - modifies the hash cache format
 
1249
        - is new in bzr 0.15
 
1250
        - uses a LockDir to guard access to it.
 
1251
    """
 
1252
 
 
1253
    upgrade_recommended = False
 
1254
 
 
1255
    def get_format_string(self):
 
1256
        """See WorkingTreeFormat.get_format_string()."""
 
1257
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1258
 
 
1259
    def get_format_description(self):
 
1260
        """See WorkingTreeFormat.get_format_description()."""
 
1261
        return "Working tree format 4"
 
1262
 
 
1263
    def initialize(self, a_bzrdir, revision_id=None):
1377
1264
        """See WorkingTreeFormat.initialize().
1378
1265
 
1379
1266
        :param revision_id: allows creating a working tree at a different
1380
1267
        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
1268
 
1388
1269
        These trees get an initial random root id, if their repository supports
1389
1270
        rich root data, TREE_ROOT otherwise.
1390
1271
        """
 
1272
        revision_id = osutils.safe_revision_id(revision_id)
1391
1273
        if not isinstance(a_bzrdir.transport, LocalTransport):
1392
1274
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1393
1275
        transport = a_bzrdir.get_workingtree_transport(self)
1394
1276
        control_files = self._open_control_files(a_bzrdir)
1395
1277
        control_files.create_lock()
1396
1278
        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()
 
1279
        control_files.put_utf8('format', self.get_format_string())
 
1280
        branch = a_bzrdir.open_branch()
1403
1281
        if revision_id is None:
1404
1282
            revision_id = branch.last_revision()
1405
1283
        local_path = transport.local_abspath('dirstate')
1407
1285
        state = dirstate.DirState.initialize(local_path)
1408
1286
        state.unlock()
1409
1287
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1288
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1289
                         branch,
1412
1290
                         _format=self,
1413
1291
                         _bzrdir=a_bzrdir,
1415
1293
        wt._new_tree()
1416
1294
        wt.lock_tree_write()
1417
1295
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1296
            if revision_id in (None, NULL_REVISION):
1420
1297
                if branch.repository.supports_rich_root():
1421
1298
                    wt._set_root_id(generate_ids.gen_root_id())
1422
1299
                else:
1423
1300
                    wt._set_root_id(ROOT_ID)
1424
1301
                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)]
 
1302
            wt.set_last_revision(revision_id)
 
1303
            wt.flush()
 
1304
            basis = wt.basis_tree()
1440
1305
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1306
            # if the basis has a root id we have to use that; otherwise we use
 
1307
            # a new random one
 
1308
            basis_root_id = basis.get_root_id()
 
1309
            if basis_root_id is not None:
 
1310
                wt._set_root_id(basis_root_id)
1443
1311
                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()
 
1312
            transform.build_tree(basis, wt)
 
1313
            basis.unlock()
1465
1314
        finally:
1466
1315
            control_files.unlock()
1467
1316
            wt.unlock()
1468
1317
        return wt
1469
1318
 
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
1319
    def _open(self, a_bzrdir, control_files):
1480
1320
        """Open the tree itself.
1481
1321
 
1482
1322
        :param a_bzrdir: the dir for the tree.
1483
1323
        :param control_files: the control files for the tree.
1484
1324
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1325
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1326
                           branch=a_bzrdir.open_branch(),
1487
1327
                           _format=self,
1488
1328
                           _bzrdir=a_bzrdir,
1489
1329
                           _control_files=control_files)
1490
1330
 
1491
1331
    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
1332
        # please test against something that will let us do tree references
1497
1333
        return bzrdir.format_registry.make_bzrdir(
1498
1334
            'dirstate-with-subtree')
1500
1336
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1337
 
1502
1338
 
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
1339
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
 
    """
 
1340
    """A revision tree pulling the inventory from a dirstate."""
1581
1341
 
1582
1342
    def __init__(self, dirstate, revision_id, repository):
1583
1343
        self._dirstate = dirstate
1584
 
        self._revision_id = revision_id
 
1344
        self._revision_id = osutils.safe_revision_id(revision_id)
1585
1345
        self._repository = repository
1586
1346
        self._inventory = None
1587
1347
        self._locked = 0
1588
1348
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1349
 
1593
1350
    def __repr__(self):
1594
1351
        return "<%s of %s in %s>" % \
1597
1354
    def annotate_iter(self, file_id,
1598
1355
                      default_revision=_mod_revision.CURRENT_REVISION):
1599
1356
        """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]
 
1357
        w = self._get_weave(file_id)
 
1358
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1359
 
1604
1360
    def _get_ancestors(self, default_revision):
1605
1361
        return set(self._repository.get_ancestry(self._revision_id,
1626
1382
    def get_root_id(self):
1627
1383
        return self.path2id('')
1628
1384
 
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
1385
    def _get_parent_index(self):
1646
1386
        """Return the index in the dirstate referenced by this tree."""
1647
1387
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1392
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1393
        If both are supplied, the fastest lookup is used, and an error is
1654
1394
        raised if they do not both point at the same row.
1655
 
 
 
1395
        
1656
1396
        :param file_id: An optional unicode file_id to be looked up.
1657
1397
        :param path: An optional unicode path to be looked up.
1658
1398
        :return: The dirstate row tuple for path/file_id, or (None, None)
1659
1399
        """
1660
1400
        if file_id is None and path is None:
1661
1401
            raise errors.BzrError('must supply file_id or path')
 
1402
        file_id = osutils.safe_file_id(file_id)
1662
1403
        if path is not None:
1663
1404
            path = path.encode('utf8')
1664
1405
        parent_index = self._get_parent_index()
1672
1413
 
1673
1414
        This is relatively expensive: we have to walk the entire dirstate.
1674
1415
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1416
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1417
            'dirstate revision tree'
1679
1418
        # separate call for profiling - makes it clear where the costs are.
1680
1419
        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()))
 
1420
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1421
            'parent %s has disappeared from %s' % (
 
1422
            self._revision_id, self._dirstate.get_parent_ids())
1685
1423
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1424
        # This is identical now to the WorkingTree _generate_inventory except
1687
1425
        # for the tree index use.
1688
1426
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1427
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1428
        assert current_entry[parent_index][0] == 'd'
1692
1429
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1430
        inv.root.revision = current_entry[parent_index][4]
1694
1431
        # Turn some things into local variables
1734
1471
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1472
                            % entry)
1736
1473
                # 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,))
 
1474
                assert file_id not in inv_byid
 
1475
                assert name_unicode not in parent_ie.children
1743
1476
                inv_byid[file_id] = inv_entry
1744
1477
                parent_ie.children[name_unicode] = inv_entry
1745
1478
        self._inventory = inv
1755
1488
            return None
1756
1489
        parent_index = self._get_parent_index()
1757
1490
        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
 
1491
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1492
 
1764
1493
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1494
        entry = self._get_entry(file_id=file_id, path=path)
1769
1498
            return parent_details[1]
1770
1499
        return None
1771
1500
 
 
1501
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1502
    def get_weave(self, file_id):
 
1503
        return self._get_weave(file_id)
 
1504
 
 
1505
    def _get_weave(self, file_id):
 
1506
        return self._repository.weave_store.get_weave(file_id,
 
1507
                self._repository.get_transaction())
 
1508
 
1772
1509
    def get_file(self, file_id, path=None):
1773
1510
        return StringIO(self.get_file_text(file_id))
1774
1511
 
 
1512
    def get_file_lines(self, file_id):
 
1513
        ie = self.inventory[file_id]
 
1514
        return self._get_weave(file_id).get_lines(ie.revision)
 
1515
 
1775
1516
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1517
        return self.inventory[file_id].text_size
1778
1518
 
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)
 
1519
    def get_file_text(self, file_id):
 
1520
        return ''.join(self.get_file_lines(file_id))
1782
1521
 
1783
1522
    def get_reference_revision(self, file_id, path=None):
1784
1523
        return self.inventory[file_id].reference_revision
1803
1542
        if entry[1][parent_index][0] != 'l':
1804
1543
            return None
1805
1544
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1545
            # At present, none of the tree implementations supports non-ascii
 
1546
            # symlink targets. So we will just assume that the dirstate path is
 
1547
            # correct.
 
1548
            return entry[1][parent_index][1]
1809
1549
 
1810
1550
    def get_revision_id(self):
1811
1551
        """Return the revision id for this tree."""
1829
1569
        return bool(self.path2id(filename))
1830
1570
 
1831
1571
    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)
 
1572
        return self.inventory[file_id].kind
1841
1573
 
1842
1574
    def path_content_summary(self, path):
1843
1575
        """See Tree.path_content_summary."""
1859
1591
            return None
1860
1592
        return ie.executable
1861
1593
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1594
    def list_files(self, include_root=False):
1863
1595
        # We use a standard implementation, because DirStateRevisionTree is
1864
1596
        # dealing with one of the parents of the current state
1865
1597
        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:
 
1598
        entries = inv.iter_entries()
 
1599
        if self.inventory.root is not None and not include_root:
1875
1600
            entries.next()
1876
1601
        for path, entry in entries:
1877
1602
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1635
                self._dirstate_locked = False
1911
1636
            self._repository.unlock()
1912
1637
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1638
    def walkdirs(self, prefix=""):
1918
1639
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
 
1640
        # implementation based on an inventory.  
1920
1641
        # This should be cleaned up to use the much faster Dirstate code
1921
1642
        # So for now, we just build up the parent inventory, and extract
1922
1643
        # it the same way RevisionTree does.
1951
1672
 
1952
1673
class InterDirStateTree(InterTree):
1953
1674
    """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
 
1675
    
 
1676
    This is used only when both trees are in the dirstate working file, and 
 
1677
    the source is any parent within the dirstate, and the destination is 
1957
1678
    the current working tree of the same dirstate.
1958
1679
    """
1959
1680
    # this could be generalized to allow comparisons between any trees in the
1972
1693
        target.set_parent_ids([revid])
1973
1694
        return target.basis_tree(), target
1974
1695
 
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
1696
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1697
    _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,
 
1698
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1699
 
 
1700
    def _iter_changes(self, include_unchanged=False,
2002
1701
                      specific_files=None, pb=None, extra_trees=[],
2003
1702
                      require_versioned=True, want_unversioned=False):
2004
1703
        """Return the changes from source to target.
2005
1704
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1705
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1706
            for details.
2008
1707
        :param specific_files: An optional list of file paths to restrict the
2009
1708
            comparison to. When mapping filenames to ids, all matches in all
2020
1719
            output. An unversioned file is defined as one with (False, False)
2021
1720
            for the versioned pair.
2022
1721
        """
 
1722
        utf8_decode = cache_utf8._utf8_decode
 
1723
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1724
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1725
        # NB: show_status depends on being able to pass in non-versioned files
 
1726
        # and report them as unknown
2023
1727
        # TODO: handle extra trees in the dirstate.
2024
 
        if (extra_trees or specific_files == []):
 
1728
        # TODO: handle comparisons as an empty tree as a different special
 
1729
        # case? mbp 20070226
 
1730
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1731
            or specific_files == []):
2025
1732
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1733
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1734
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1735
                require_versioned, want_unversioned=want_unversioned):
 
1736
                yield f
 
1737
            return
2029
1738
        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))
 
1739
        assert (self.source._revision_id in parent_ids), \
 
1740
                "revision {%s} is not stored in {%s}, but %s " \
 
1741
                "can only be used for trees stored in the dirstate" \
 
1742
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1743
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1744
        if self.source._revision_id == NULL_REVISION:
2038
1745
            source_index = None
2039
1746
            indices = (target_index,)
2040
1747
        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))
 
1748
            assert (self.source._revision_id in parent_ids), \
 
1749
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1750
                self.source._revision_id, parent_ids)
2045
1751
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
 
            indices = (source_index, target_index)
 
1752
            indices = (source_index,target_index)
2047
1753
        # -- make all specific_files utf8 --
2048
1754
        if specific_files:
2049
1755
            specific_files_utf8 = set()
2050
1756
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1757
                specific_files_utf8.add(path.encode('utf8'))
2054
1758
            specific_files = specific_files_utf8
2055
1759
        else:
2056
1760
            specific_files = set([''])
2057
1761
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1762
        # -- get the state object and prepare it.
2060
1763
        state = self.target.current_dirstate()
2061
1764
        state._read_dirblocks_if_needed()
 
1765
        def _entries_for_path(path):
 
1766
            """Return a list with all the entries that match path for all ids.
 
1767
            """
 
1768
            dirname, basename = os.path.split(path)
 
1769
            key = (dirname, basename, '')
 
1770
            block_index, present = state._find_block_index_from_key(key)
 
1771
            if not present:
 
1772
                # the block which should contain path is absent.
 
1773
                return []
 
1774
            result = []
 
1775
            block = state._dirblocks[block_index][1]
 
1776
            entry_index, _ = state._find_entry_index(key, block)
 
1777
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1778
            while (entry_index < len(block) and
 
1779
                block[entry_index][0][0:2] == key[0:2]):
 
1780
                result.append(block[entry_index])
 
1781
                entry_index += 1
 
1782
            return result
2062
1783
        if require_versioned:
2063
1784
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1785
            all_versioned = True
2065
1786
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1787
                path_entries = _entries_for_path(path)
2067
1788
                if not path_entries:
2068
1789
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1790
                    all_versioned = False
 
1791
                    break
2071
1792
                found_versioned = False
2072
1793
                # for each id at this path
2073
1794
                for entry in path_entries:
2080
1801
                if not found_versioned:
2081
1802
                    # none of the indexes was not 'absent' at all ids for this
2082
1803
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1804
                    all_versioned = False
 
1805
                    break
 
1806
            if not all_versioned:
 
1807
                raise errors.PathsNotVersionedError(specific_files)
2086
1808
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1809
        search_specific_files = set()
 
1810
        for path in specific_files:
 
1811
            other_specific_files = specific_files.difference(set([path]))
 
1812
            if not osutils.is_inside_any(other_specific_files, path):
 
1813
                # this is a top level path, we must check it.
 
1814
                search_specific_files.add(path)
 
1815
        # sketch: 
 
1816
        # compare source_index and target_index at or under each element of search_specific_files.
 
1817
        # follow the following comparison table. Note that we only want to do diff operations when
 
1818
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1819
        # for the target.
 
1820
        # cases:
 
1821
        # 
 
1822
        # Source | Target | disk | action
 
1823
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1824
        #        |        |      | diff check on source-target
 
1825
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1826
        #        |        |      | ???
 
1827
        #   r    |  a     |      | add source to search
 
1828
        #   r    |  a     |  a   | 
 
1829
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1830
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1831
        #   a    | fdlt   |      | add new id
 
1832
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1833
        #   a    |  a     |      | not present in either tree, skip
 
1834
        #   a    |  a     |  a   | not present in any tree, skip
 
1835
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1836
        #        |        |      | may not be selected by the users list of paths.
 
1837
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1838
        #        |        |      | may not be selected by the users list of paths.
 
1839
        #  fdlt  | fdlt   |      | content in both: diff them
 
1840
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1841
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1842
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1843
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1844
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1845
        #        |        |      | this id at the other path.
 
1846
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1847
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1848
        #        |        |      | this id at the other path.
 
1849
 
 
1850
        # for all search_indexs in each path at or under each element of
 
1851
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1852
        # relocated path as one to search if its not searched already. If the
 
1853
        # detail is not relocated, add the id.
 
1854
        searched_specific_files = set()
 
1855
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1856
        # Using a list so that we can access the values and change them in
 
1857
        # nested scope. Each one is [path, file_id, entry]
 
1858
        last_source_parent = [None, None]
 
1859
        last_target_parent = [None, None]
2088
1860
 
2089
1861
        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()
 
1862
 
 
1863
        # Just a sentry, so that _process_entry can say that this
 
1864
        # record is handled, but isn't interesting to process (unchanged)
 
1865
        uninteresting = object()
 
1866
 
 
1867
 
 
1868
        old_dirname_to_file_id = {}
 
1869
        new_dirname_to_file_id = {}
 
1870
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1871
        #       keeping a cache of directories that we have seen.
 
1872
 
 
1873
        def _process_entry(entry, path_info):
 
1874
            """Compare an entry and real disk to generate delta information.
 
1875
 
 
1876
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1877
                the path of entry. If None, then the path is considered absent.
 
1878
                (Perhaps we should pass in a concrete entry for this ?)
 
1879
                Basename is returned as a utf8 string because we expect this
 
1880
                tuple will be ignored, and don't want to take the time to
 
1881
                decode.
 
1882
            :return: None if these don't match
 
1883
                     A tuple of information about the change, or
 
1884
                     the object 'uninteresting' if these match, but are
 
1885
                     basically identical.
 
1886
            """
 
1887
            if source_index is None:
 
1888
                source_details = NULL_PARENT_DETAILS
 
1889
            else:
 
1890
                source_details = entry[1][source_index]
 
1891
            target_details = entry[1][target_index]
 
1892
            target_minikind = target_details[0]
 
1893
            if path_info is not None and target_minikind in 'fdlt':
 
1894
                assert target_index == 0
 
1895
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1896
                                                  stat_value=path_info[3])
 
1897
                # The entry may have been modified by update_entry
 
1898
                target_details = entry[1][target_index]
 
1899
                target_minikind = target_details[0]
 
1900
            else:
 
1901
                link_or_sha1 = None
 
1902
            file_id = entry[0][2]
 
1903
            source_minikind = source_details[0]
 
1904
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1905
                # claimed content in both: diff
 
1906
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1907
                #        |        |      | diff check on source-target
 
1908
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1909
                #        |        |      | ???
 
1910
                if source_minikind in 'r':
 
1911
                    # add the source to the search path to find any children it
 
1912
                    # has.  TODO ? : only add if it is a container ?
 
1913
                    if not osutils.is_inside_any(searched_specific_files,
 
1914
                                                 source_details[1]):
 
1915
                        search_specific_files.add(source_details[1])
 
1916
                    # generate the old path; this is needed for stating later
 
1917
                    # as well.
 
1918
                    old_path = source_details[1]
 
1919
                    old_dirname, old_basename = os.path.split(old_path)
 
1920
                    path = pathjoin(entry[0][0], entry[0][1])
 
1921
                    old_entry = state._get_entry(source_index,
 
1922
                                                 path_utf8=old_path)
 
1923
                    # update the source details variable to be the real
 
1924
                    # location.
 
1925
                    source_details = old_entry[1][source_index]
 
1926
                    source_minikind = source_details[0]
 
1927
                else:
 
1928
                    old_dirname = entry[0][0]
 
1929
                    old_basename = entry[0][1]
 
1930
                    old_path = path = None
 
1931
                if path_info is None:
 
1932
                    # the file is missing on disk, show as removed.
 
1933
                    content_change = True
 
1934
                    target_kind = None
 
1935
                    target_exec = False
 
1936
                else:
 
1937
                    # source and target are both versioned and disk file is present.
 
1938
                    target_kind = path_info[2]
 
1939
                    if target_kind == 'directory':
 
1940
                        if path is None:
 
1941
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1942
                        new_dirname_to_file_id[path] = file_id
 
1943
                        if source_minikind != 'd':
 
1944
                            content_change = True
 
1945
                        else:
 
1946
                            # directories have no fingerprint
 
1947
                            content_change = False
 
1948
                        target_exec = False
 
1949
                    elif target_kind == 'file':
 
1950
                        if source_minikind != 'f':
 
1951
                            content_change = True
 
1952
                        else:
 
1953
                            # We could check the size, but we already have the
 
1954
                            # sha1 hash.
 
1955
                            content_change = (link_or_sha1 != source_details[1])
 
1956
                        # Target details is updated at update_entry time
 
1957
                        if use_filesystem_for_exec:
 
1958
                            # We don't need S_ISREG here, because we are sure
 
1959
                            # we are dealing with a file.
 
1960
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1961
                        else:
 
1962
                            target_exec = target_details[3]
 
1963
                    elif target_kind == 'symlink':
 
1964
                        if source_minikind != 'l':
 
1965
                            content_change = True
 
1966
                        else:
 
1967
                            content_change = (link_or_sha1 != source_details[1])
 
1968
                        target_exec = False
 
1969
                    elif target_kind == 'tree-reference':
 
1970
                        if source_minikind != 't':
 
1971
                            content_change = True
 
1972
                        else:
 
1973
                            content_change = False
 
1974
                        target_exec = False
 
1975
                    else:
 
1976
                        raise Exception, "unknown kind %s" % path_info[2]
 
1977
                if source_minikind == 'd':
 
1978
                    if path is None:
 
1979
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1980
                    old_dirname_to_file_id[old_path] = file_id
 
1981
                # parent id is the entry for the path in the target tree
 
1982
                if old_dirname == last_source_parent[0]:
 
1983
                    source_parent_id = last_source_parent[1]
 
1984
                else:
 
1985
                    try:
 
1986
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1987
                    except KeyError:
 
1988
                        source_parent_entry = state._get_entry(source_index,
 
1989
                                                               path_utf8=old_dirname)
 
1990
                        source_parent_id = source_parent_entry[0][2]
 
1991
                    if source_parent_id == entry[0][2]:
 
1992
                        # This is the root, so the parent is None
 
1993
                        source_parent_id = None
 
1994
                    else:
 
1995
                        last_source_parent[0] = old_dirname
 
1996
                        last_source_parent[1] = source_parent_id
 
1997
                new_dirname = entry[0][0]
 
1998
                if new_dirname == last_target_parent[0]:
 
1999
                    target_parent_id = last_target_parent[1]
 
2000
                else:
 
2001
                    try:
 
2002
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2003
                    except KeyError:
 
2004
                        # TODO: We don't always need to do the lookup, because the
 
2005
                        #       parent entry will be the same as the source entry.
 
2006
                        target_parent_entry = state._get_entry(target_index,
 
2007
                                                               path_utf8=new_dirname)
 
2008
                        assert target_parent_entry != (None, None), (
 
2009
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2010
                            % (new_dirname, entry))
 
2011
                        target_parent_id = target_parent_entry[0][2]
 
2012
                    if target_parent_id == entry[0][2]:
 
2013
                        # This is the root, so the parent is None
 
2014
                        target_parent_id = None
 
2015
                    else:
 
2016
                        last_target_parent[0] = new_dirname
 
2017
                        last_target_parent[1] = target_parent_id
 
2018
 
 
2019
                source_exec = source_details[3]
 
2020
                if (include_unchanged
 
2021
                    or content_change
 
2022
                    or source_parent_id != target_parent_id
 
2023
                    or old_basename != entry[0][1]
 
2024
                    or source_exec != target_exec
 
2025
                    ):
 
2026
                    if old_path is None:
 
2027
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2028
                        old_path_u = utf8_decode(old_path)[0]
 
2029
                        path_u = old_path_u
 
2030
                    else:
 
2031
                        old_path_u = utf8_decode(old_path)[0]
 
2032
                        if old_path == path:
 
2033
                            path_u = old_path_u
 
2034
                        else:
 
2035
                            path_u = utf8_decode(path)[0]
 
2036
                    source_kind = _minikind_to_kind[source_minikind]
 
2037
                    return (entry[0][2],
 
2038
                           (old_path_u, path_u),
 
2039
                           content_change,
 
2040
                           (True, True),
 
2041
                           (source_parent_id, target_parent_id),
 
2042
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2043
                           (source_kind, target_kind),
 
2044
                           (source_exec, target_exec))
 
2045
                else:
 
2046
                    return uninteresting
 
2047
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2048
                # looks like a new file
 
2049
                if path_info is not None:
 
2050
                    path = pathjoin(entry[0][0], entry[0][1])
 
2051
                    # parent id is the entry for the path in the target tree
 
2052
                    # TODO: these are the same for an entire directory: cache em.
 
2053
                    parent_id = state._get_entry(target_index,
 
2054
                                                 path_utf8=entry[0][0])[0][2]
 
2055
                    if parent_id == entry[0][2]:
 
2056
                        parent_id = None
 
2057
                    if use_filesystem_for_exec:
 
2058
                        # We need S_ISREG here, because we aren't sure if this
 
2059
                        # is a file or not.
 
2060
                        target_exec = bool(
 
2061
                            stat.S_ISREG(path_info[3].st_mode)
 
2062
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2063
                    else:
 
2064
                        target_exec = target_details[3]
 
2065
                    return (entry[0][2],
 
2066
                           (None, utf8_decode(path)[0]),
 
2067
                           True,
 
2068
                           (False, True),
 
2069
                           (None, parent_id),
 
2070
                           (None, utf8_decode(entry[0][1])[0]),
 
2071
                           (None, path_info[2]),
 
2072
                           (None, target_exec))
 
2073
                else:
 
2074
                    # but its not on disk: we deliberately treat this as just
 
2075
                    # never-present. (Why ?! - RBC 20070224)
 
2076
                    pass
 
2077
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2078
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2079
                # if its still on disk, *and* theres no other entry at this
 
2080
                # path [we dont know this in this routine at the moment -
 
2081
                # perhaps we should change this - then it would be an unknown.
 
2082
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2083
                # parent id is the entry for the path in the target tree
 
2084
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2085
                if parent_id == entry[0][2]:
 
2086
                    parent_id = None
 
2087
                return (entry[0][2],
 
2088
                       (utf8_decode(old_path)[0], None),
 
2089
                       True,
 
2090
                       (True, False),
 
2091
                       (parent_id, None),
 
2092
                       (utf8_decode(entry[0][1])[0], None),
 
2093
                       (_minikind_to_kind[source_minikind], None),
 
2094
                       (source_details[3], None))
 
2095
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2096
                # a rename; could be a true rename, or a rename inherited from
 
2097
                # a renamed parent. TODO: handle this efficiently. Its not
 
2098
                # common case to rename dirs though, so a correct but slow
 
2099
                # implementation will do.
 
2100
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2101
                    search_specific_files.add(target_details[1])
 
2102
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2103
                # neither of the selected trees contain this file,
 
2104
                # so skip over it. This is not currently directly tested, but
 
2105
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2106
                pass
 
2107
            else:
 
2108
                raise AssertionError("don't know how to compare "
 
2109
                    "source_minikind=%r, target_minikind=%r"
 
2110
                    % (source_minikind, target_minikind))
 
2111
                ## import pdb;pdb.set_trace()
 
2112
            return None
 
2113
 
 
2114
        while search_specific_files:
 
2115
            # TODO: the pending list should be lexically sorted?  the
 
2116
            # interface doesn't require it.
 
2117
            current_root = search_specific_files.pop()
 
2118
            current_root_unicode = current_root.decode('utf8')
 
2119
            searched_specific_files.add(current_root)
 
2120
            # process the entries for this containing directory: the rest will be
 
2121
            # found by their parents recursively.
 
2122
            root_entries = _entries_for_path(current_root)
 
2123
            root_abspath = self.target.abspath(current_root_unicode)
 
2124
            try:
 
2125
                root_stat = os.lstat(root_abspath)
 
2126
            except OSError, e:
 
2127
                if e.errno == errno.ENOENT:
 
2128
                    # the path does not exist: let _process_entry know that.
 
2129
                    root_dir_info = None
 
2130
                else:
 
2131
                    # some other random error: hand it up.
 
2132
                    raise
 
2133
            else:
 
2134
                root_dir_info = ('', current_root,
 
2135
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2136
                    root_abspath)
 
2137
                if root_dir_info[2] == 'directory':
 
2138
                    if self.target._directory_is_tree_reference(
 
2139
                        current_root.decode('utf8')):
 
2140
                        root_dir_info = root_dir_info[:2] + \
 
2141
                            ('tree-reference',) + root_dir_info[3:]
 
2142
 
 
2143
            if not root_entries and not root_dir_info:
 
2144
                # this specified path is not present at all, skip it.
 
2145
                continue
 
2146
            path_handled = False
 
2147
            for entry in root_entries:
 
2148
                result = _process_entry(entry, root_dir_info)
 
2149
                if result is not None:
 
2150
                    path_handled = True
 
2151
                    if result is not uninteresting:
 
2152
                        yield result
 
2153
            if want_unversioned and not path_handled and root_dir_info:
 
2154
                new_executable = bool(
 
2155
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2156
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2157
                yield (None,
 
2158
                       (None, current_root_unicode),
 
2159
                       True,
 
2160
                       (False, False),
 
2161
                       (None, None),
 
2162
                       (None, splitpath(current_root_unicode)[-1]),
 
2163
                       (None, root_dir_info[2]),
 
2164
                       (None, new_executable)
 
2165
                      )
 
2166
            initial_key = (current_root, '', '')
 
2167
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2168
            if block_index == 0:
 
2169
                # we have processed the total root already, but because the
 
2170
                # initial key matched it we should skip it here.
 
2171
                block_index +=1
 
2172
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2173
                current_dir_info = None
 
2174
            else:
 
2175
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2176
                try:
 
2177
                    current_dir_info = dir_iterator.next()
 
2178
                except OSError, e:
 
2179
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2180
                    # python 2.5 has e.errno == EINVAL,
 
2181
                    #            and e.winerror == ERROR_DIRECTORY
 
2182
                    e_winerror = getattr(e, 'winerror', None)
 
2183
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2184
                    # there may be directories in the inventory even though
 
2185
                    # this path is not a file on disk: so mark it as end of
 
2186
                    # iterator
 
2187
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2188
                        current_dir_info = None
 
2189
                    elif (sys.platform == 'win32'
 
2190
                          and (e.errno in win_errors
 
2191
                               or e_winerror in win_errors)):
 
2192
                        current_dir_info = None
 
2193
                    else:
 
2194
                        raise
 
2195
                else:
 
2196
                    if current_dir_info[0][0] == '':
 
2197
                        # remove .bzr from iteration
 
2198
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2199
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2200
                        del current_dir_info[1][bzr_index]
 
2201
            # walk until both the directory listing and the versioned metadata
 
2202
            # are exhausted. 
 
2203
            if (block_index < len(state._dirblocks) and
 
2204
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2205
                current_block = state._dirblocks[block_index]
 
2206
            else:
 
2207
                current_block = None
 
2208
            while (current_dir_info is not None or
 
2209
                   current_block is not None):
 
2210
                if (current_dir_info and current_block
 
2211
                    and current_dir_info[0][0] != current_block[0]):
 
2212
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2213
                        # filesystem data refers to paths not covered by the dirblock.
 
2214
                        # this has two possibilities:
 
2215
                        # A) it is versioned but empty, so there is no block for it
 
2216
                        # B) it is not versioned.
 
2217
 
 
2218
                        # if (A) then we need to recurse into it to check for
 
2219
                        # new unknown files or directories.
 
2220
                        # if (B) then we should ignore it, because we don't
 
2221
                        # recurse into unknown directories.
 
2222
                        path_index = 0
 
2223
                        while path_index < len(current_dir_info[1]):
 
2224
                                current_path_info = current_dir_info[1][path_index]
 
2225
                                if want_unversioned:
 
2226
                                    if current_path_info[2] == 'directory':
 
2227
                                        if self.target._directory_is_tree_reference(
 
2228
                                            current_path_info[0].decode('utf8')):
 
2229
                                            current_path_info = current_path_info[:2] + \
 
2230
                                                ('tree-reference',) + current_path_info[3:]
 
2231
                                    new_executable = bool(
 
2232
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2233
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2234
                                    yield (None,
 
2235
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2236
                                        True,
 
2237
                                        (False, False),
 
2238
                                        (None, None),
 
2239
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2240
                                        (None, current_path_info[2]),
 
2241
                                        (None, new_executable))
 
2242
                                # dont descend into this unversioned path if it is
 
2243
                                # a dir
 
2244
                                if current_path_info[2] in ('directory',
 
2245
                                                            'tree-reference'):
 
2246
                                    del current_dir_info[1][path_index]
 
2247
                                    path_index -= 1
 
2248
                                path_index += 1
 
2249
 
 
2250
                        # This dir info has been handled, go to the next
 
2251
                        try:
 
2252
                            current_dir_info = dir_iterator.next()
 
2253
                        except StopIteration:
 
2254
                            current_dir_info = None
 
2255
                    else:
 
2256
                        # We have a dirblock entry for this location, but there
 
2257
                        # is no filesystem path for this. This is most likely
 
2258
                        # because a directory was removed from the disk.
 
2259
                        # We don't have to report the missing directory,
 
2260
                        # because that should have already been handled, but we
 
2261
                        # need to handle all of the files that are contained
 
2262
                        # within.
 
2263
                        for current_entry in current_block[1]:
 
2264
                            # entry referring to file not present on disk.
 
2265
                            # advance the entry only, after processing.
 
2266
                            result = _process_entry(current_entry, None)
 
2267
                            if result is not None:
 
2268
                                if result is not uninteresting:
 
2269
                                    yield result
 
2270
                        block_index +=1
 
2271
                        if (block_index < len(state._dirblocks) and
 
2272
                            osutils.is_inside(current_root,
 
2273
                                              state._dirblocks[block_index][0])):
 
2274
                            current_block = state._dirblocks[block_index]
 
2275
                        else:
 
2276
                            current_block = None
 
2277
                    continue
 
2278
                entry_index = 0
 
2279
                if current_block and entry_index < len(current_block[1]):
 
2280
                    current_entry = current_block[1][entry_index]
 
2281
                else:
 
2282
                    current_entry = None
 
2283
                advance_entry = True
 
2284
                path_index = 0
 
2285
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2286
                    current_path_info = current_dir_info[1][path_index]
 
2287
                    if current_path_info[2] == 'directory':
 
2288
                        if self.target._directory_is_tree_reference(
 
2289
                            current_path_info[0].decode('utf8')):
 
2290
                            current_path_info = current_path_info[:2] + \
 
2291
                                ('tree-reference',) + current_path_info[3:]
 
2292
                else:
 
2293
                    current_path_info = None
 
2294
                advance_path = True
 
2295
                path_handled = False
 
2296
                while (current_entry is not None or
 
2297
                    current_path_info is not None):
 
2298
                    if current_entry is None:
 
2299
                        # the check for path_handled when the path is adnvaced
 
2300
                        # will yield this path if needed.
 
2301
                        pass
 
2302
                    elif current_path_info is None:
 
2303
                        # no path is fine: the per entry code will handle it.
 
2304
                        result = _process_entry(current_entry, current_path_info)
 
2305
                        if result is not None:
 
2306
                            if result is not uninteresting:
 
2307
                                yield result
 
2308
                    elif (current_entry[0][1] != current_path_info[1]
 
2309
                          or current_entry[1][target_index][0] in 'ar'):
 
2310
                        # The current path on disk doesn't match the dirblock
 
2311
                        # record. Either the dirblock is marked as absent, or
 
2312
                        # the file on disk is not present at all in the
 
2313
                        # dirblock. Either way, report about the dirblock
 
2314
                        # entry, and let other code handle the filesystem one.
 
2315
 
 
2316
                        # Compare the basename for these files to determine
 
2317
                        # which comes first
 
2318
                        if current_path_info[1] < current_entry[0][1]:
 
2319
                            # extra file on disk: pass for now, but only
 
2320
                            # increment the path, not the entry
 
2321
                            advance_entry = False
 
2322
                        else:
 
2323
                            # entry referring to file not present on disk.
 
2324
                            # advance the entry only, after processing.
 
2325
                            result = _process_entry(current_entry, None)
 
2326
                            if result is not None:
 
2327
                                if result is not uninteresting:
 
2328
                                    yield result
 
2329
                            advance_path = False
 
2330
                    else:
 
2331
                        result = _process_entry(current_entry, current_path_info)
 
2332
                        if result is not None:
 
2333
                            path_handled = True
 
2334
                            if result is not uninteresting:
 
2335
                                yield result
 
2336
                    if advance_entry and current_entry is not None:
 
2337
                        entry_index += 1
 
2338
                        if entry_index < len(current_block[1]):
 
2339
                            current_entry = current_block[1][entry_index]
 
2340
                        else:
 
2341
                            current_entry = None
 
2342
                    else:
 
2343
                        advance_entry = True # reset the advance flaga
 
2344
                    if advance_path and current_path_info is not None:
 
2345
                        if not path_handled:
 
2346
                            # unversioned in all regards
 
2347
                            if want_unversioned:
 
2348
                                new_executable = bool(
 
2349
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2350
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2351
                                yield (None,
 
2352
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2353
                                    True,
 
2354
                                    (False, False),
 
2355
                                    (None, None),
 
2356
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2357
                                    (None, current_path_info[2]),
 
2358
                                    (None, new_executable))
 
2359
                            # dont descend into this unversioned path if it is
 
2360
                            # a dir
 
2361
                            if current_path_info[2] in ('directory'):
 
2362
                                del current_dir_info[1][path_index]
 
2363
                                path_index -= 1
 
2364
                        # dont descend the disk iterator into any tree 
 
2365
                        # paths.
 
2366
                        if current_path_info[2] == 'tree-reference':
 
2367
                            del current_dir_info[1][path_index]
 
2368
                            path_index -= 1
 
2369
                        path_index += 1
 
2370
                        if path_index < len(current_dir_info[1]):
 
2371
                            current_path_info = current_dir_info[1][path_index]
 
2372
                            if current_path_info[2] == 'directory':
 
2373
                                if self.target._directory_is_tree_reference(
 
2374
                                    current_path_info[0].decode('utf8')):
 
2375
                                    current_path_info = current_path_info[:2] + \
 
2376
                                        ('tree-reference',) + current_path_info[3:]
 
2377
                        else:
 
2378
                            current_path_info = None
 
2379
                        path_handled = False
 
2380
                    else:
 
2381
                        advance_path = True # reset the advance flagg.
 
2382
                if current_block is not None:
 
2383
                    block_index += 1
 
2384
                    if (block_index < len(state._dirblocks) and
 
2385
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2386
                        current_block = state._dirblocks[block_index]
 
2387
                    else:
 
2388
                        current_block = None
 
2389
                if current_dir_info is not None:
 
2390
                    try:
 
2391
                        current_dir_info = dir_iterator.next()
 
2392
                    except StopIteration:
 
2393
                        current_dir_info = None
 
2394
 
2094
2395
 
2095
2396
    @staticmethod
2096
2397
    def is_compatible(source, target):
2097
2398
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2399
        if not isinstance(target, WorkingTree4):
2099
2400
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2401
        # the source must be a revtreee or dirstate rev tree.
2101
2402
        if not isinstance(source,
2102
2403
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2404
            return False
2104
2405
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2406
        if not (source._revision_id == NULL_REVISION or
2106
2407
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2408
            # TODO: what about ghosts? it may well need to 
2108
2409
            # check for them explicitly.
2109
2410
            return False
2110
2411
        return True
2120
2421
 
2121
2422
    def convert(self, tree):
2122
2423
        # 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
 
2424
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2425
        # tree during upgrade.
2125
2426
        tree._control_files.lock_write()
2126
2427
        try:
2152
2453
 
2153
2454
    def update_format(self, tree):
2154
2455
        """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())
 
2456
        tree._control_files.put_utf8('format',
 
2457
            self.target_format.get_format_string())