~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: John Arbash Meinel
  • Date: 2008-03-20 15:10:05 UTC
  • mto: This revision was merged to the branch mainline in revision 3299.
  • Revision ID: john@arbash-meinel.com-20080320151005-z9lajjy69m20of17
uncommit --local in an unbound branch raises the same exception as commit --local

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2012 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
22
22
WorkingTree.open(dir).
23
23
"""
24
24
 
25
 
from __future__ import absolute_import
26
 
 
27
25
from cStringIO import StringIO
28
26
import os
29
27
import sys
30
28
 
31
29
from bzrlib.lazy_import import lazy_import
32
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
33
34
import errno
 
35
import itertools
 
36
import operator
34
37
import stat
 
38
from time import time
 
39
import warnings
35
40
 
 
41
import bzrlib
36
42
from bzrlib import (
37
43
    bzrdir,
38
44
    cache_utf8,
39
 
    config,
40
45
    conflicts as _mod_conflicts,
41
 
    controldir,
42
46
    debug,
 
47
    delta,
43
48
    dirstate,
44
49
    errors,
45
 
    filters as _mod_filters,
46
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
47
54
    osutils,
48
55
    revision as _mod_revision,
49
56
    revisiontree,
 
57
    textui,
50
58
    trace,
51
59
    transform,
52
 
    views,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
53
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
54
67
""")
55
68
 
 
69
from bzrlib import symbol_versioning
56
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
57
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
58
 
from bzrlib.lock import LogicalLockResult
59
 
from bzrlib.lockable_files import LockableFiles
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
60
73
from bzrlib.lockdir import LockDir
61
 
from bzrlib.mutabletree import (
62
 
    MutableTree,
63
 
    needs_tree_write_lock,
64
 
    )
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
65
76
from bzrlib.osutils import (
66
77
    file_kind,
67
78
    isdir,
 
79
    normpath,
68
80
    pathjoin,
 
81
    rand_chars,
69
82
    realpath,
70
83
    safe_unicode,
71
 
    )
72
 
from bzrlib.symbol_versioning import (
73
 
    deprecated_in,
74
 
    deprecated_method,
75
 
    )
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
76
87
from bzrlib.transport.local import LocalTransport
77
 
from bzrlib.tree import (
78
 
    InterTree,
79
 
    InventoryTree,
80
 
    )
81
 
from bzrlib.workingtree import (
82
 
    InventoryWorkingTree,
83
 
    WorkingTree,
84
 
    WorkingTreeFormatMetaDir,
85
 
    )
86
 
 
87
 
 
88
 
class DirStateWorkingTree(InventoryWorkingTree):
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
89
119
 
90
120
    def __init__(self, basedir,
91
121
                 branch,
101
131
        """
102
132
        self._format = _format
103
133
        self.bzrdir = _bzrdir
 
134
        assert isinstance(basedir, basestring), \
 
135
            "base directory %r is not a string" % basedir
104
136
        basedir = safe_unicode(basedir)
105
 
        trace.mutter("opening working tree %r", basedir)
 
137
        mutter("opening working tree %r", basedir)
106
138
        self._branch = branch
 
139
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
140
            "branch %r is not a Branch" % self.branch
107
141
        self.basedir = realpath(basedir)
108
142
        # if branch is at our basedir and is a format 6 or less
109
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
110
146
        self._control_files = _control_files
111
 
        self._transport = self._control_files._transport
112
147
        self._dirty = None
113
148
        #-------------
114
149
        # during a read or write lock these objects are set, and are
118
153
        #-------------
119
154
        self._setup_directory_is_tree_reference()
120
155
        self._detect_case_handling()
121
 
        self._rules_searcher = None
122
 
        self.views = self._make_views()
123
 
        #--- allow tests to select the dirstate iter_changes implementation
124
 
        self._iter_changes = dirstate._process_entry
125
156
 
126
157
    @needs_tree_write_lock
127
158
    def _add(self, files, ids, kinds):
142
173
            state.add(f, file_id, kind, None, '')
143
174
        self._make_dirty(reset_inventory=True)
144
175
 
145
 
    def _get_check_refs(self):
146
 
        """Return the references needed to perform a check of this tree."""
147
 
        return [('trees', self.last_revision())]
148
 
 
149
176
    def _make_dirty(self, reset_inventory):
150
177
        """Make the tree state dirty.
151
178
 
159
186
    @needs_tree_write_lock
160
187
    def add_reference(self, sub_tree):
161
188
        # use standard implementation, which calls back to self._add
162
 
        #
 
189
        # 
163
190
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
191
        # it's just recorded at the moment of commit. 
165
192
        self._add_reference(sub_tree)
166
193
 
167
194
    def break_lock(self):
203
230
 
204
231
    def _comparison_data(self, entry, path):
205
232
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
233
            WorkingTree3._comparison_data(self, entry, path)
207
234
        # it looks like a plain directory, but it's really a reference -- see
208
235
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            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)):
211
239
            kind = 'tree-reference'
212
240
        return kind, executable, stat_value
213
241
 
215
243
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
244
        # mark the tree as dirty post commit - commit
217
245
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
246
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
247
        self._make_dirty(reset_inventory=True)
220
248
        return result
221
249
 
239
267
            return self._dirstate
240
268
        local_path = self.bzrdir.get_workingtree_transport(None
241
269
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
270
        self._dirstate = dirstate.DirState.on_file(local_path)
244
271
        return self._dirstate
245
272
 
246
 
    def _sha1_provider(self):
247
 
        """A function that returns a SHA1Provider suitable for this tree.
248
 
 
249
 
        :return: None if content filtering is not supported by this tree.
250
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
251
 
          form of files, i.e. after read filters are applied.
252
 
        """
253
 
        if self.supports_content_filtering():
254
 
            return ContentFilterAwareSHA1Provider(self)
255
 
        else:
256
 
            return None
257
 
 
258
 
    def _worth_saving_limit(self):
259
 
        """How many hash changes are ok before we must save the dirstate.
260
 
 
261
 
        :return: an integer. -1 means never save.
262
 
        """
263
 
        conf = self.get_config_stack()
264
 
        return conf.get('bzr.workingtree.worth_saving_limit')
265
 
 
266
273
    def filter_unversioned_files(self, paths):
267
274
        """Filter out paths that are versioned.
268
275
 
300
307
 
301
308
    def _generate_inventory(self):
302
309
        """Create and set self.inventory from the dirstate object.
303
 
 
 
310
        
304
311
        This is relatively expensive: we have to walk the entire dirstate.
305
312
        Ideally we would not, and can deprecate this function.
306
313
        """
310
317
        state._read_dirblocks_if_needed()
311
318
        root_key, current_entry = self._get_entry(path='')
312
319
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
320
        assert current_entry[0][0] == 'd' # directory
315
321
        inv = Inventory(root_id=current_id)
316
322
        # Turn some things into local variables
317
323
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
356
                    # add this entry to the parent map.
351
357
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
358
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            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)
357
364
                    inv_entry.reference_revision = link_or_sha1 or None
358
365
                elif kind != 'symlink':
359
366
                    raise AssertionError("unknown kind %r" % kind)
360
367
                # These checks cost us around 40ms on a 55k entry tree
361
 
                if file_id in inv_byid:
362
 
                    raise AssertionError('file_id %s already in'
363
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
364
 
                if name_unicode in parent_ie.children:
365
 
                    raise AssertionError('name %r already in parent'
366
 
                        % (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
367
371
                inv_byid[file_id] = inv_entry
368
372
                parent_ie.children[name_unicode] = inv_entry
369
373
        self._inventory = inv
374
378
        If either file_id or path is supplied, it is used as the key to lookup.
375
379
        If both are supplied, the fastest lookup is used, and an error is
376
380
        raised if they do not both point at the same row.
377
 
 
 
381
        
378
382
        :param file_id: An optional unicode file_id to be looked up.
379
383
        :param path: An optional unicode path to be looked up.
380
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
398
402
        state = self.current_dirstate()
399
403
        if stat_value is None:
400
404
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
405
                stat_value = os.lstat(file_abspath)
402
406
            except OSError, e:
403
407
                if e.errno == errno.ENOENT:
404
408
                    return None
405
409
                else:
406
410
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
409
413
        if entry[1][0][0] == 'f':
410
 
            if link_or_sha1 is None:
411
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
412
 
                try:
413
 
                    sha1 = osutils.sha_file(file_obj)
414
 
                finally:
415
 
                    file_obj.close()
416
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
417
 
                return sha1
418
 
            else:
419
 
                return link_or_sha1
 
414
            return link_or_sha1
420
415
        return None
421
416
 
422
 
    def _get_root_inventory(self):
 
417
    def _get_inventory(self):
423
418
        """Get the inventory for the tree. This is only valid within a lock."""
424
419
        if 'evil' in debug.debug_flags:
425
420
            trace.mutter_callsite(2,
430
425
        self._generate_inventory()
431
426
        return self._inventory
432
427
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
428
    inventory = property(_get_inventory,
438
429
                         doc="Inventory of this Tree")
439
430
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
431
    @needs_read_lock
444
432
    def get_parent_ids(self):
445
433
        """See Tree.get_parent_ids.
446
 
 
 
434
        
447
435
        This implementation requests the ids list from the dirstate file.
448
436
        """
449
437
        return self.current_dirstate().get_parent_ids()
471
459
        return osutils.lexists(pathjoin(
472
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
461
 
474
 
    def has_or_had_id(self, file_id):
475
 
        state = self.current_dirstate()
476
 
        row, parents = self._get_entry(file_id=file_id)
477
 
        return row is not None
478
 
 
479
462
    @needs_read_lock
480
463
    def id2path(self, file_id):
481
464
        "Convert a file-id to a path."
492
475
            return False # Missing entries are not executable
493
476
        return entry[1][0][3] # Executable?
494
477
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
497
481
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
501
484
            entry = self._get_entry(file_id=file_id, path=path)
502
485
            if entry == (None, None):
503
486
                return False
504
487
            return entry[1][0][3]
505
 
        else:
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
506
497
            self._must_be_locked()
507
498
            if not path:
508
499
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
500
            mode = os.lstat(self.abspath(path)).st_mode
510
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
502
 
512
503
    def all_file_ids(self):
537
528
        return iter(result)
538
529
 
539
530
    def iter_references(self):
540
 
        if not self._repo_supports_tree_reference:
541
 
            # When the repo doesn't support references, we will have nothing to
542
 
            # return
543
 
            return
544
531
        for key, tree_details in self.current_dirstate()._iter_entries():
545
532
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
533
                # not relevant to the working tree
548
535
            if not key[1]:
549
536
                # the root is not a reference.
550
537
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
538
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
539
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
540
                if self._kind(path) == 'tree-reference':
 
541
                    yield path, key[2]
555
542
            except errors.NoSuchFile:
556
543
                # path is missing on disk.
557
544
                continue
558
545
 
559
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
560
 
        """See MutableTree._observed_sha1."""
561
 
        state = self.current_dirstate()
562
 
        entry = self._get_entry(file_id=file_id, path=path)
563
 
        state._observed_sha1(entry, sha1, statvalue)
564
 
 
565
546
    def kind(self, file_id):
566
547
        """Return the kind of a file.
567
548
 
571
552
        Note: The caller is expected to take a read-lock before calling this.
572
553
        """
573
554
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
555
        assert relpath != None, \
 
556
            "path for id {%s} is None!" % file_id
577
557
        return self._kind(relpath)
578
558
 
579
559
    def _kind(self, relpath):
580
560
        abspath = self.abspath(relpath)
581
561
        kind = file_kind(abspath)
582
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
583
 
            entry = self._get_entry(path=relpath)
584
 
            if entry[1] is not None:
585
 
                if entry[1][0][0] == 't':
586
 
                    kind = 'tree-reference'
 
562
        if (self._repo_supports_tree_reference and
 
563
            kind == 'directory' and
 
564
            self._directory_is_tree_reference(relpath)):
 
565
            kind = 'tree-reference'
587
566
        return kind
588
567
 
589
568
    @needs_read_lock
596
575
            return _mod_revision.NULL_REVISION
597
576
 
598
577
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
578
        """See Branch.lock_read, and WorkingTree.unlock."""
603
579
        self.branch.lock_read()
604
580
        try:
605
581
            self._control_files.lock_read()
618
594
        except:
619
595
            self.branch.unlock()
620
596
            raise
621
 
        return LogicalLockResult(self.unlock)
622
597
 
623
598
    def _lock_self_write(self):
624
599
        """This should be called after the branch is locked."""
639
614
        except:
640
615
            self.branch.unlock()
641
616
            raise
642
 
        return LogicalLockResult(self.unlock)
643
617
 
644
618
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
619
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
620
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
621
        self._lock_self_write()
651
622
 
652
623
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
624
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
625
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
626
        self._lock_self_write()
659
627
 
660
628
    @needs_tree_write_lock
661
629
    def move(self, from_paths, to_dir, after=False):
663
631
        result = []
664
632
        if not from_paths:
665
633
            return result
 
634
 
666
635
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
636
 
 
637
        assert not isinstance(from_paths, basestring)
669
638
        to_dir_utf8 = to_dir.encode('utf8')
670
639
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
640
        id_index = state._get_id_index()
692
661
 
693
662
        if self._inventory is not None:
694
663
            update_inventory = True
695
 
            inv = self.root_inventory
 
664
            inv = self.inventory
 
665
            to_dir_ie = inv[to_dir_id]
696
666
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
667
        else:
699
668
            update_inventory = False
700
669
 
731
700
            from_entry = self._get_entry(path=from_rel)
732
701
            if from_entry == (None, None):
733
702
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
703
                    errors.NotVersionedError(path=str(from_rel)))
735
704
 
736
705
            from_id = from_entry[0][2]
737
706
            to_rel = pathjoin(to_dir, from_tail)
826
795
                if minikind == 'd':
827
796
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
797
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
798
                        assert from_dir != '', "renaming root not supported"
831
799
                        from_key = (from_dir, '')
832
800
                        from_block_idx, present = \
833
801
                            state._find_block_index_from_key(from_key)
846
814
 
847
815
                        # Grab a copy since move_one may update the list.
848
816
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
817
                            assert entry[0][0] == from_dir
851
818
                            cur_details = entry[1][0]
852
819
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
820
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
878
845
                rollback_rename()
879
846
                raise
880
847
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
848
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
849
            self._make_dirty(reset_inventory=False)
883
850
 
884
851
        return result
894
861
    @needs_read_lock
895
862
    def path2id(self, path):
896
863
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
864
        path = path.strip('/')
902
865
        entry = self._get_entry(path=path)
903
866
        if entry == (None, None):
916
879
        for tree in trees:
917
880
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
881
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
882
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
883
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
884
        # -- make all paths utf8 --
923
885
        paths_utf8 = set()
981
943
                    all_versioned = False
982
944
                    break
983
945
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
946
                raise errors.PathsNotVersionedError(paths)
986
947
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
948
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
949
        # sketch: 
989
950
        # for all search_indexs in each path at or under each element of
990
951
        # search_paths, if the detail is relocated: add the id, and add the
991
952
        # relocated path as one to search if its not searched already. If the
1037
998
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
999
            for dir_name in split_paths:
1039
1000
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
1001
                    raise errors.PathsNotVersionedError(paths)
1042
1002
 
1043
1003
        for dir_name_id, trees_info in found.iteritems():
1044
1004
            for index in search_indexes:
1048
1008
 
1049
1009
    def read_working_inventory(self):
1050
1010
        """Read the working inventory.
1051
 
 
 
1011
        
1052
1012
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
1013
        """
1054
 
        return self.root_inventory
 
1014
        return self.inventory
1055
1015
 
1056
1016
    @needs_read_lock
1057
1017
    def revision_tree(self, revision_id):
1072
1032
    def set_last_revision(self, new_revision):
1073
1033
        """Change the last revision in the working tree."""
1074
1034
        parents = self.get_parent_ids()
1075
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1076
 
            if len(parents) >= 2:
1077
 
                raise AssertionError(
1078
 
                    "setting the last parent to none with a pending merge is "
1079
 
                    "unsupported.")
 
1035
        if new_revision in (NULL_REVISION, None):
 
1036
            assert len(parents) < 2, (
 
1037
                "setting the last parent to none with a pending merge is "
 
1038
                "unsupported.")
1080
1039
            self.set_parent_ids([])
1081
1040
        else:
1082
1041
            self.set_parent_ids([new_revision] + parents[1:],
1085
1044
    @needs_tree_write_lock
1086
1045
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1046
        """Set the parent ids to revision_ids.
1088
 
 
 
1047
        
1089
1048
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1049
        for each element of revision_ids from the trees repository. If you have
1091
1050
        tree data already available, it is more efficient to use
1123
1082
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1083
        real_trees = []
1125
1084
        ghosts = []
1126
 
 
1127
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1128
 
        graph = self.branch.repository.get_graph()
1129
 
        heads = graph.heads(parent_ids)
1130
 
        accepted_revisions = set()
1131
 
 
1132
1085
        # convert absent trees to the null tree, which we convert back to
1133
1086
        # missing on access.
1134
1087
        for rev_id, tree in parents_list:
1135
 
            if len(accepted_revisions) > 0:
1136
 
                # we always accept the first tree
1137
 
                if rev_id in accepted_revisions or rev_id not in heads:
1138
 
                    # We have already included either this tree, or its
1139
 
                    # descendent, so we skip it.
1140
 
                    continue
1141
1088
            _mod_revision.check_not_reserved_id(rev_id)
1142
1089
            if tree is not None:
1143
1090
                real_trees.append((rev_id, tree))
1144
1091
            else:
1145
1092
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1093
                    self.branch.repository.revision_tree(None)))
1148
1094
                ghosts.append(rev_id)
1149
 
            accepted_revisions.add(rev_id)
1150
 
        updated = False
1151
 
        if (len(real_trees) == 1
1152
 
            and not ghosts
1153
 
            and self.branch.repository._format.fast_deltas
1154
 
            and isinstance(real_trees[0][1],
1155
 
                revisiontree.InventoryRevisionTree)
1156
 
            and self.get_parent_ids()):
1157
 
            rev_id, rev_tree = real_trees[0]
1158
 
            basis_id = self.get_parent_ids()[0]
1159
 
            # There are times when basis_tree won't be in
1160
 
            # self.branch.repository, (switch, for example)
1161
 
            try:
1162
 
                basis_tree = self.branch.repository.revision_tree(basis_id)
1163
 
            except errors.NoSuchRevision:
1164
 
                # Fall back to the set_parent_trees(), since we can't use
1165
 
                # _make_delta if we can't get the RevisionTree
1166
 
                pass
1167
 
            else:
1168
 
                delta = rev_tree.root_inventory._make_delta(
1169
 
                    basis_tree.root_inventory)
1170
 
                dirstate.update_basis_by_delta(delta, rev_id)
1171
 
                updated = True
1172
 
        if not updated:
1173
 
            dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1095
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1096
        self._make_dirty(reset_inventory=False)
1175
1097
 
1176
1098
    def _set_root_id(self, file_id):
1196
1118
 
1197
1119
    def unlock(self):
1198
1120
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1121
        # do non-implementation specific cleanup
 
1122
        self._cleanup()
 
1123
 
1199
1124
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1125
            # eventually we should do signature checking during read locks for
1204
1126
            # dirstate updates.
1205
1127
            if self._control_files._lock_mode == 'w':
1268
1190
                # just forget the whole block.
1269
1191
                entry_index = 0
1270
1192
                while entry_index < len(block[1]):
 
1193
                    # Mark this file id as having been removed
1271
1194
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1195
                    ids_to_unversion.discard(entry[0][2])
 
1196
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1197
                                               # entries
 
1198
                        or not state._make_absent(entry)):
1274
1199
                        entry_index += 1
1275
 
                    else:
1276
 
                        # Mark this file id as having been removed
1277
 
                        ids_to_unversion.discard(entry[0][2])
1278
 
                        if not state._make_absent(entry):
1279
 
                            # The block has not shrunk.
1280
 
                            entry_index += 1
1281
1200
                # go to the next block. (At the moment we dont delete empty
1282
1201
                # dirblocks)
1283
1202
                block_index += 1
1304
1223
        # have to change the legacy inventory too.
1305
1224
        if self._inventory is not None:
1306
1225
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
 
1226
                self._inventory.remove_recursive_id(file_id)
1309
1227
 
1310
1228
    @needs_tree_write_lock
1311
1229
    def rename_one(self, from_rel, to_rel, after=False):
1312
1230
        """See WorkingTree.rename_one"""
1313
1231
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
 
1232
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1315
1233
 
1316
1234
    @needs_tree_write_lock
1317
1235
    def apply_inventory_delta(self, changes):
1322
1240
 
1323
1241
    def update_basis_by_delta(self, new_revid, delta):
1324
1242
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
 
1243
        assert self.last_revision() != new_revid
1327
1244
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1328
1245
 
1329
1246
    @needs_read_lock
1333
1250
    @needs_tree_write_lock
1334
1251
    def _write_inventory(self, inv):
1335
1252
        """Write inventory as the current inventory."""
1336
 
        if self._dirty:
1337
 
            raise AssertionError("attempting to write an inventory when the "
1338
 
                "dirstate is dirty will lose pending changes")
1339
 
        had_inventory = self._inventory is not None
1340
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1341
 
        # working inventory. We do this because self.inventory may be inv, or
1342
 
        # may have been modified, and either case would prevent a clean delta
1343
 
        # being created.
1344
 
        self._inventory = None
1345
 
        # generate a delta,
1346
 
        delta = inv._make_delta(self.root_inventory)
1347
 
        # and apply it.
1348
 
        self.apply_inventory_delta(delta)
1349
 
        if had_inventory:
 
1253
        assert not self._dirty, ("attempting to write an inventory when the "
 
1254
            "dirstate is dirty will cause data loss")
 
1255
        self.current_dirstate().set_state_from_inventory(inv)
 
1256
        self._make_dirty(reset_inventory=False)
 
1257
        if self._inventory is not None:
1350
1258
            self._inventory = inv
1351
1259
        self.flush()
1352
1260
 
1353
 
    @needs_tree_write_lock
1354
 
    def reset_state(self, revision_ids=None):
1355
 
        """Reset the state of the working tree.
1356
 
 
1357
 
        This does a hard-reset to a last-known-good state. This is a way to
1358
 
        fix if something got corrupted (like the .bzr/checkout/dirstate file)
1359
 
        """
1360
 
        if revision_ids is None:
1361
 
            revision_ids = self.get_parent_ids()
1362
 
        if not revision_ids:
1363
 
            base_tree = self.branch.repository.revision_tree(
1364
 
                _mod_revision.NULL_REVISION)
1365
 
            trees = []
1366
 
        else:
1367
 
            trees = zip(revision_ids,
1368
 
                        self.branch.repository.revision_trees(revision_ids))
1369
 
            base_tree = trees[0][1]
1370
 
        state = self.current_dirstate()
1371
 
        # We don't support ghosts yet
1372
 
        state.set_state_from_scratch(base_tree.root_inventory, trees, [])
1373
 
 
1374
 
 
1375
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1376
 
 
1377
 
    def __init__(self, tree):
1378
 
        self.tree = tree
1379
 
 
1380
 
    def sha1(self, abspath):
1381
 
        """See dirstate.SHA1Provider.sha1()."""
1382
 
        filters = self.tree._content_filter_stack(
1383
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1384
 
        return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
1385
 
 
1386
 
    def stat_and_sha1(self, abspath):
1387
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1388
 
        filters = self.tree._content_filter_stack(
1389
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1390
 
        file_obj = file(abspath, 'rb', 65000)
1391
 
        try:
1392
 
            statvalue = os.fstat(file_obj.fileno())
1393
 
            if filters:
1394
 
                file_obj = _mod_filters.filtered_input_file(file_obj, filters)
1395
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1396
 
        finally:
1397
 
            file_obj.close()
1398
 
        return statvalue, sha1
1399
 
 
1400
 
 
1401
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1402
 
    """Dirstate working tree that supports content filtering.
1403
 
 
1404
 
    The dirstate holds the hash and size of the canonical form of the file, 
1405
 
    and most methods must return that.
1406
 
    """
1407
 
 
1408
 
    def _file_content_summary(self, path, stat_result):
1409
 
        # This is to support the somewhat obsolete path_content_summary method
1410
 
        # with content filtering: see
1411
 
        # <https://bugs.launchpad.net/bzr/+bug/415508>.
1412
 
        #
1413
 
        # If the dirstate cache is up to date and knows the hash and size,
1414
 
        # return that.
1415
 
        # Otherwise if there are no content filters, return the on-disk size
1416
 
        # and leave the hash blank.
1417
 
        # Otherwise, read and filter the on-disk file and use its size and
1418
 
        # hash.
1419
 
        #
1420
 
        # The dirstate doesn't store the size of the canonical form so we
1421
 
        # can't trust it for content-filtered trees.  We just return None.
1422
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1423
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1424
 
        return ('file', None, executable, dirstate_sha1)
1425
 
 
1426
 
 
1427
 
class WorkingTree4(DirStateWorkingTree):
1428
 
    """This is the Format 4 working tree.
1429
 
 
1430
 
    This differs from WorkingTree by:
1431
 
     - Having a consolidated internal dirstate, stored in a
1432
 
       randomly-accessible sorted file on disk.
1433
 
     - Not having a regular inventory attribute.  One can be synthesized
1434
 
       on demand but this is expensive and should be avoided.
1435
 
 
1436
 
    This is new in bzr 0.15.
1437
 
    """
1438
 
 
1439
 
 
1440
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1441
 
    """This is the Format 5 working tree.
1442
 
 
1443
 
    This differs from WorkingTree4 by:
1444
 
     - Supporting content filtering.
1445
 
 
1446
 
    This is new in bzr 1.11.
1447
 
    """
1448
 
 
1449
 
 
1450
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1451
 
    """This is the Format 6 working tree.
1452
 
 
1453
 
    This differs from WorkingTree5 by:
1454
 
     - Supporting a current view that may mask the set of files in a tree
1455
 
       impacted by most user operations.
1456
 
 
1457
 
    This is new in bzr 1.14.
1458
 
    """
1459
 
 
1460
 
    def _make_views(self):
1461
 
        return views.PathBasedViews(self)
1462
 
 
1463
 
 
1464
 
class DirStateWorkingTreeFormat(WorkingTreeFormatMetaDir):
1465
 
 
1466
 
    missing_parent_conflicts = True
1467
 
 
1468
 
    supports_versioned_directories = True
1469
 
 
1470
 
    _lock_class = LockDir
1471
 
    _lock_file_name = 'lock'
1472
 
 
1473
 
    def _open_control_files(self, a_bzrdir):
1474
 
        transport = a_bzrdir.get_workingtree_transport(None)
1475
 
        return LockableFiles(transport, self._lock_file_name,
1476
 
                             self._lock_class)
 
1261
 
 
1262
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1263
    """The first consolidated dirstate working tree format.
 
1264
 
 
1265
    This format:
 
1266
        - exists within a metadir controlling .bzr
 
1267
        - includes an explicit version marker for the workingtree control
 
1268
          files, separate from the BzrDir format
 
1269
        - modifies the hash cache format
 
1270
        - is new in bzr 0.15
 
1271
        - uses a LockDir to guard access to it.
 
1272
    """
 
1273
 
 
1274
    upgrade_recommended = False
 
1275
 
 
1276
    def get_format_string(self):
 
1277
        """See WorkingTreeFormat.get_format_string()."""
 
1278
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1279
 
 
1280
    def get_format_description(self):
 
1281
        """See WorkingTreeFormat.get_format_description()."""
 
1282
        return "Working tree format 4"
1477
1283
 
1478
1284
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
1285
                   accelerator_tree=None, hardlink=False):
1480
1286
        """See WorkingTreeFormat.initialize().
1481
1287
 
1482
1288
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
 
1289
        revision than the branch is at.
1484
1290
        :param accelerator_tree: A tree which can be used for retrieving file
1485
1291
            contents more quickly than the revision tree, i.e. a workingtree.
1486
1292
            The revision tree will be used for cases where accelerator_tree's
1497
1303
        control_files = self._open_control_files(a_bzrdir)
1498
1304
        control_files.create_lock()
1499
1305
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
 
1306
        control_files.put_utf8('format', self.get_format_string())
1502
1307
        if from_branch is not None:
1503
1308
            branch = from_branch
1504
1309
        else:
1510
1315
        state = dirstate.DirState.initialize(local_path)
1511
1316
        state.unlock()
1512
1317
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1318
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1319
                         branch,
1515
1320
                         _format=self,
1516
1321
                         _bzrdir=a_bzrdir,
1518
1323
        wt._new_tree()
1519
1324
        wt.lock_tree_write()
1520
1325
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1326
            if revision_id in (None, NULL_REVISION):
1523
1327
                if branch.repository.supports_rich_root():
1524
1328
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1329
                else:
1536
1340
                    pass
1537
1341
            if basis is None:
1538
1342
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
 
1343
            if revision_id == NULL_REVISION:
1540
1344
                parents_list = []
1541
1345
            else:
1542
1346
                parents_list = [(revision_id, basis)]
1550
1354
                if basis_root_id is not None:
1551
1355
                    wt._set_root_id(basis_root_id)
1552
1356
                    wt.flush()
1553
 
                if wt.supports_content_filtering():
1554
 
                    # The original tree may not have the same content filters
1555
 
                    # applied so we can't safely build the inventory delta from
1556
 
                    # the source tree.
1557
 
                    delta_from_tree = False
1558
 
                else:
1559
 
                    delta_from_tree = True
1560
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1561
 
                # because wt4.apply_inventory_delta does not mutate the input
1562
 
                # inventory entries.
1563
1357
                transform.build_tree(basis, wt, accelerator_tree,
1564
 
                                     hardlink=hardlink,
1565
 
                                     delta_from_tree=delta_from_tree)
1566
 
                for hook in MutableTree.hooks['post_build_tree']:
1567
 
                    hook(wt)
 
1358
                                     hardlink=hardlink)
1568
1359
            finally:
1569
1360
                basis.unlock()
1570
1361
        finally:
1572
1363
            wt.unlock()
1573
1364
        return wt
1574
1365
 
1575
 
    def _init_custom_control_files(self, wt):
1576
 
        """Subclasses with custom control files should override this method.
1577
 
 
1578
 
        The working tree and control files are locked for writing when this
1579
 
        method is called.
1580
 
 
1581
 
        :param wt: the WorkingTree object
1582
 
        """
1583
 
 
1584
 
    def open(self, a_bzrdir, _found=False):
1585
 
        """Return the WorkingTree object for a_bzrdir
1586
 
 
1587
 
        _found is a private parameter, do not use it. It is used to indicate
1588
 
               if format probing has already been done.
1589
 
        """
1590
 
        if not _found:
1591
 
            # we are being called directly and must probe.
1592
 
            raise NotImplementedError
1593
 
        if not isinstance(a_bzrdir.transport, LocalTransport):
1594
 
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1595
 
        wt = self._open(a_bzrdir, self._open_control_files(a_bzrdir))
1596
 
        return wt
1597
 
 
1598
1366
    def _open(self, a_bzrdir, control_files):
1599
1367
        """Open the tree itself.
1600
1368
 
1601
1369
        :param a_bzrdir: the dir for the tree.
1602
1370
        :param control_files: the control files for the tree.
1603
1371
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1372
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1373
                           branch=a_bzrdir.open_branch(),
1606
1374
                           _format=self,
1607
1375
                           _bzrdir=a_bzrdir,
1608
1376
                           _control_files=control_files)
1609
1377
 
1610
1378
    def __get_matchingbzrdir(self):
1611
 
        return self._get_matchingbzrdir()
1612
 
 
1613
 
    def _get_matchingbzrdir(self):
1614
 
        """Overrideable method to get a bzrdir for testing."""
1615
1379
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1380
        return bzrdir.format_registry.make_bzrdir(
 
1381
            'dirstate-with-subtree')
1618
1382
 
1619
1383
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1384
 
1621
1385
 
1622
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1623
 
    """The first consolidated dirstate working tree format.
1624
 
 
1625
 
    This format:
1626
 
        - exists within a metadir controlling .bzr
1627
 
        - includes an explicit version marker for the workingtree control
1628
 
          files, separate from the ControlDir format
1629
 
        - modifies the hash cache format
1630
 
        - is new in bzr 0.15
1631
 
        - uses a LockDir to guard access to it.
1632
 
    """
1633
 
 
1634
 
    upgrade_recommended = False
1635
 
 
1636
 
    _tree_class = WorkingTree4
1637
 
 
1638
 
    @classmethod
1639
 
    def get_format_string(cls):
1640
 
        """See WorkingTreeFormat.get_format_string()."""
1641
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1642
 
 
1643
 
    def get_format_description(self):
1644
 
        """See WorkingTreeFormat.get_format_description()."""
1645
 
        return "Working tree format 4"
1646
 
 
1647
 
 
1648
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1649
 
    """WorkingTree format supporting content filtering.
1650
 
    """
1651
 
 
1652
 
    upgrade_recommended = False
1653
 
 
1654
 
    _tree_class = WorkingTree5
1655
 
 
1656
 
    @classmethod
1657
 
    def get_format_string(cls):
1658
 
        """See WorkingTreeFormat.get_format_string()."""
1659
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1660
 
 
1661
 
    def get_format_description(self):
1662
 
        """See WorkingTreeFormat.get_format_description()."""
1663
 
        return "Working tree format 5"
1664
 
 
1665
 
    def supports_content_filtering(self):
1666
 
        return True
1667
 
 
1668
 
 
1669
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1670
 
    """WorkingTree format supporting views.
1671
 
    """
1672
 
 
1673
 
    upgrade_recommended = False
1674
 
 
1675
 
    _tree_class = WorkingTree6
1676
 
 
1677
 
    @classmethod
1678
 
    def get_format_string(cls):
1679
 
        """See WorkingTreeFormat.get_format_string()."""
1680
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1681
 
 
1682
 
    def get_format_description(self):
1683
 
        """See WorkingTreeFormat.get_format_description()."""
1684
 
        return "Working tree format 6"
1685
 
 
1686
 
    def _init_custom_control_files(self, wt):
1687
 
        """Subclasses with custom control files should override this method."""
1688
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1689
 
 
1690
 
    def supports_content_filtering(self):
1691
 
        return True
1692
 
 
1693
 
    def supports_views(self):
1694
 
        return True
1695
 
 
1696
 
    def _get_matchingbzrdir(self):
1697
 
        """Overrideable method to get a bzrdir for testing."""
1698
 
        # We use 'development-subtree' instead of '2a', because we have a
1699
 
        # few tests that want to test tree references
1700
 
        return bzrdir.format_registry.make_bzrdir('development-subtree')
1701
 
 
1702
 
 
1703
 
class DirStateRevisionTree(InventoryTree):
1704
 
    """A revision tree pulling the inventory from a dirstate.
1705
 
    
1706
 
    Note that this is one of the historical (ie revision) trees cached in the
1707
 
    dirstate for easy access, not the workingtree.
1708
 
    """
 
1386
class DirStateRevisionTree(Tree):
 
1387
    """A revision tree pulling the inventory from a dirstate."""
1709
1388
 
1710
1389
    def __init__(self, dirstate, revision_id, repository):
1711
1390
        self._dirstate = dirstate
1714
1393
        self._inventory = None
1715
1394
        self._locked = 0
1716
1395
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1396
 
1721
1397
    def __repr__(self):
1722
1398
        return "<%s of %s in %s>" % \
1725
1401
    def annotate_iter(self, file_id,
1726
1402
                      default_revision=_mod_revision.CURRENT_REVISION):
1727
1403
        """See Tree.annotate_iter"""
1728
 
        text_key = (file_id, self.get_file_revision(file_id))
1729
 
        annotations = self._repository.texts.annotate(text_key)
1730
 
        return [(key[-1], line) for (key, line) in annotations]
 
1404
        w = self._get_weave(file_id)
 
1405
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1406
 
 
1407
    def _get_ancestors(self, default_revision):
 
1408
        return set(self._repository.get_ancestry(self._revision_id,
 
1409
                                                 topo_sorted=False))
1732
1410
    def _comparison_data(self, entry, path):
1733
1411
        """See Tree._comparison_data."""
1734
1412
        if entry is None:
1759
1437
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
1438
        return path_utf8.decode('utf8')
1761
1439
 
1762
 
    def iter_references(self):
1763
 
        if not self._repo_supports_tree_reference:
1764
 
            # When the repo doesn't support references, we will have nothing to
1765
 
            # return
1766
 
            return iter([])
1767
 
        # Otherwise, fall back to the default implementation
1768
 
        return super(DirStateRevisionTree, self).iter_references()
1769
 
 
1770
1440
    def _get_parent_index(self):
1771
1441
        """Return the index in the dirstate referenced by this tree."""
1772
1442
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1447
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1448
        If both are supplied, the fastest lookup is used, and an error is
1779
1449
        raised if they do not both point at the same row.
1780
 
 
 
1450
        
1781
1451
        :param file_id: An optional unicode file_id to be looked up.
1782
1452
        :param path: An optional unicode path to be looked up.
1783
1453
        :return: The dirstate row tuple for path/file_id, or (None, None)
1787
1457
        if path is not None:
1788
1458
            path = path.encode('utf8')
1789
1459
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1460
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1461
 
1793
1462
    def _generate_inventory(self):
1794
1463
        """Create and set self.inventory from the dirstate object.
1798
1467
 
1799
1468
        This is relatively expensive: we have to walk the entire dirstate.
1800
1469
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1470
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1471
            'dirstate revision tree'
1805
1472
        # separate call for profiling - makes it clear where the costs are.
1806
1473
        self._dirstate._read_dirblocks_if_needed()
1807
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1808
 
            raise AssertionError(
1809
 
                'parent %s has disappeared from %s' % (
1810
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1474
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1475
            'parent %s has disappeared from %s' % (
 
1476
            self._revision_id, self._dirstate.get_parent_ids())
1811
1477
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1478
        # This is identical now to the WorkingTree _generate_inventory except
1813
1479
        # for the tree index use.
1814
1480
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1481
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1482
        assert current_entry[parent_index][0] == 'd'
1818
1483
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1484
        inv.root.revision = current_entry[parent_index][4]
1820
1485
        # Turn some things into local variables
1851
1516
                elif kind == 'directory':
1852
1517
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1518
                elif kind == 'symlink':
 
1519
                    inv_entry.executable = False
 
1520
                    inv_entry.text_size = None
1854
1521
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1522
                elif kind == 'tree-reference':
1856
1523
                    inv_entry.reference_revision = fingerprint or None
1858
1525
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1526
                            % entry)
1860
1527
                # These checks cost us around 40ms on a 55k entry tree
1861
 
                if file_id in inv_byid:
1862
 
                    raise AssertionError('file_id %s already in'
1863
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1864
 
                if name_unicode in parent_ie.children:
1865
 
                    raise AssertionError('name %r already in parent'
1866
 
                        % (name_unicode,))
 
1528
                assert file_id not in inv_byid
 
1529
                assert name_unicode not in parent_ie.children
1867
1530
                inv_byid[file_id] = inv_entry
1868
1531
                parent_ie.children[name_unicode] = inv_entry
1869
1532
        self._inventory = inv
1876
1539
        # Make sure the file exists
1877
1540
        entry = self._get_entry(file_id, path=path)
1878
1541
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1542
            return None
1880
1543
        parent_index = self._get_parent_index()
1881
1544
        last_changed_revision = entry[1][parent_index][4]
1882
 
        try:
1883
 
            rev = self._repository.get_revision(last_changed_revision)
1884
 
        except errors.NoSuchRevision:
1885
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1886
 
        return rev.timestamp
 
1545
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1546
 
1888
1547
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1548
        entry = self._get_entry(file_id=file_id, path=path)
1893
1552
            return parent_details[1]
1894
1553
        return None
1895
1554
 
1896
 
    @needs_read_lock
1897
 
    def get_file_revision(self, file_id):
1898
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1899
 
        return inv[inv_file_id].revision
 
1555
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1556
    def get_weave(self, file_id):
 
1557
        return self._get_weave(file_id)
 
1558
 
 
1559
    def _get_weave(self, file_id):
 
1560
        return self._repository.weave_store.get_weave(file_id,
 
1561
                self._repository.get_transaction())
1900
1562
 
1901
1563
    def get_file(self, file_id, path=None):
1902
1564
        return StringIO(self.get_file_text(file_id))
1903
1565
 
 
1566
    def get_file_lines(self, file_id):
 
1567
        entry = self._get_entry(file_id=file_id)[1]
 
1568
        if entry == None:
 
1569
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1570
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1571
 
1904
1572
    def get_file_size(self, file_id):
1905
 
        """See Tree.get_file_size"""
1906
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1907
 
        return inv[inv_file_id].text_size
 
1573
        return self.inventory[file_id].text_size
1908
1574
 
1909
 
    def get_file_text(self, file_id, path=None):
1910
 
        content = None
1911
 
        for _, content_iter in self.iter_files_bytes([(file_id, None)]):
1912
 
            if content is not None:
1913
 
                raise AssertionError('iter_files_bytes returned'
1914
 
                    ' too many entries')
1915
 
            # For each entry returned by iter_files_bytes, we must consume the
1916
 
            # content_iter before we step the files iterator.
1917
 
            content = ''.join(content_iter)
1918
 
        if content is None:
1919
 
            raise AssertionError('iter_files_bytes did not return'
1920
 
                ' the requested data')
1921
 
        return content
 
1575
    def get_file_text(self, file_id):
 
1576
        return ''.join(self.get_file_lines(file_id))
1922
1577
 
1923
1578
    def get_reference_revision(self, file_id, path=None):
1924
 
        inv, inv_file_id = self._unpack_file_id(file_id)
1925
 
        return inv[inv_file_id].reference_revision
 
1579
        return self.inventory[file_id].reference_revision
1926
1580
 
1927
1581
    def iter_files_bytes(self, desired_files):
1928
1582
        """See Tree.iter_files_bytes.
1938
1592
                                       identifier))
1939
1593
        return self._repository.iter_files_bytes(repo_desired_files)
1940
1594
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1595
    def get_symlink_target(self, file_id):
1942
1596
        entry = self._get_entry(file_id=file_id)
1943
1597
        parent_index = self._get_parent_index()
1944
1598
        if entry[1][parent_index][0] != 'l':
1945
1599
            return None
1946
1600
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1601
            # At present, none of the tree implementations supports non-ascii
 
1602
            # symlink targets. So we will just assume that the dirstate path is
 
1603
            # correct.
 
1604
            return entry[1][parent_index][1]
1950
1605
 
1951
1606
    def get_revision_id(self):
1952
1607
        """Return the revision id for this tree."""
1953
1608
        return self._revision_id
1954
1609
 
1955
 
    def _get_root_inventory(self):
 
1610
    def _get_inventory(self):
1956
1611
        if self._inventory is not None:
1957
1612
            return self._inventory
1958
1613
        self._must_be_locked()
1959
1614
        self._generate_inventory()
1960
1615
        return self._inventory
1961
1616
 
1962
 
    root_inventory = property(_get_root_inventory,
1963
 
                         doc="Inventory of this Tree")
1964
 
 
1965
 
    @deprecated_method(deprecated_in((2, 5, 0)))
1966
 
    def _get_inventory(self):
1967
 
        return self.root_inventory
1968
 
 
1969
1617
    inventory = property(_get_inventory,
1970
1618
                         doc="Inventory of this Tree")
1971
1619
 
1978
1626
 
1979
1627
    def kind(self, file_id):
1980
1628
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
 
1629
        if entry == None:
1982
1630
            raise errors.NoSuchId(tree=self, file_id=file_id)
1983
 
        parent_index = self._get_parent_index()
1984
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
 
1631
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
1985
1632
 
1986
1633
    def stored_kind(self, file_id):
1987
1634
        """See Tree.stored_kind"""
1989
1636
 
1990
1637
    def path_content_summary(self, path):
1991
1638
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
 
1639
        id = self.inventory.path2id(path)
 
1640
        if id is None:
1994
1641
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
 
1642
        entry = self._inventory[id]
1996
1643
        kind = entry.kind
1997
1644
        if kind == 'file':
1998
1645
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
2002
1649
            return (kind, None, None, None)
2003
1650
 
2004
1651
    def is_executable(self, file_id, path=None):
2005
 
        inv, inv_file_id = self._unpack_file_id(file_id)
2006
 
        ie = inv[inv_file_id]
 
1652
        ie = self.inventory[file_id]
2007
1653
        if ie.kind != "file":
2008
 
            return False
 
1654
            return None
2009
1655
        return ie.executable
2010
1656
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1657
    def list_files(self, include_root=False):
2015
1658
        # We use a standard implementation, because DirStateRevisionTree is
2016
1659
        # dealing with one of the parents of the current state
2017
 
        if from_dir is None:
2018
 
            inv = self.root_inventory
2019
 
            from_dir_id = None
2020
 
        else:
2021
 
            inv, from_dir_id = self._path2inv_file_id(from_dir)
2022
 
            if from_dir_id is None:
2023
 
                # Directory not versioned
2024
 
                return
2025
 
        # FIXME: Support nested trees
2026
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
2027
 
        if inv.root is not None and not include_root and from_dir is None:
 
1660
        inv = self._get_inventory()
 
1661
        entries = inv.iter_entries()
 
1662
        if self.inventory.root is not None and not include_root:
2028
1663
            entries.next()
2029
1664
        for path, entry in entries:
2030
1665
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1666
 
2032
1667
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1668
        """Lock the tree for a set of operations."""
2037
1669
        if not self._locked:
2038
1670
            self._repository.lock_read()
2039
1671
            if self._dirstate._lock_token is None:
2040
1672
                self._dirstate.lock_read()
2041
1673
                self._dirstate_locked = True
2042
1674
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1675
 
2045
1676
    def _must_be_locked(self):
2046
1677
        if not self._locked:
2050
1681
    def path2id(self, path):
2051
1682
        """Return the id for path in this tree."""
2052
1683
        # lookup by path: faster than splitting and walking the ivnentory.
2053
 
        if isinstance(path, list):
2054
 
            if path == []:
2055
 
                path = [""]
2056
 
            path = osutils.pathjoin(*path)
2057
1684
        entry = self._get_entry(path=path)
2058
1685
        if entry == (None, None):
2059
1686
            return None
2071
1698
                self._dirstate_locked = False
2072
1699
            self._repository.unlock()
2073
1700
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1701
    def walkdirs(self, prefix=""):
2079
1702
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1703
        # implementation based on an inventory.  
2081
1704
        # This should be cleaned up to use the much faster Dirstate code
2082
1705
        # So for now, we just build up the parent inventory, and extract
2083
1706
        # it the same way RevisionTree does.
2084
1707
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1708
        inv = self._get_inventory()
2086
1709
        top_id = inv.path2id(prefix)
2087
1710
        if top_id is None:
2088
1711
            pending = []
2112
1735
 
2113
1736
class InterDirStateTree(InterTree):
2114
1737
    """Fast path optimiser for changes_from with dirstate trees.
2115
 
 
2116
 
    This is used only when both trees are in the dirstate working file, and
2117
 
    the source is any parent within the dirstate, and the destination is
 
1738
    
 
1739
    This is used only when both trees are in the dirstate working file, and 
 
1740
    the source is any parent within the dirstate, and the destination is 
2118
1741
    the current working tree of the same dirstate.
2119
1742
    """
2120
1743
    # this could be generalized to allow comparisons between any trees in the
2129
1752
    def make_source_parent_tree(source, target):
2130
1753
        """Change the source tree into a parent of the target."""
2131
1754
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1755
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1756
        target.set_parent_ids([revid])
2134
1757
        return target.basis_tree(), target
2135
1758
 
2136
 
    @classmethod
2137
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
2138
 
        result = klass.make_source_parent_tree(source, target)
2139
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2140
 
        return result
2141
 
 
2142
 
    @classmethod
2143
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2144
 
                                                  target):
2145
 
        from bzrlib.tests.test__dirstate_helpers import \
2146
 
            compiled_dirstate_helpers_feature
2147
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2148
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2149
 
        result = klass.make_source_parent_tree(source, target)
2150
 
        result[1]._iter_changes = ProcessEntryC
2151
 
        return result
2152
 
 
2153
1759
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1760
    _matching_to_tree_format = WorkingTreeFormat4()
2155
 
 
2156
 
    @classmethod
2157
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2158
 
        # This method shouldn't be called, because we have python and C
2159
 
        # specific flavours.
2160
 
        raise NotImplementedError
 
1761
    _test_mutable_trees_to_test_trees = make_source_parent_tree
2161
1762
 
2162
1763
    def iter_changes(self, include_unchanged=False,
2163
1764
                      specific_files=None, pb=None, extra_trees=[],
2181
1782
            output. An unversioned file is defined as one with (False, False)
2182
1783
            for the versioned pair.
2183
1784
        """
 
1785
        utf8_decode = cache_utf8._utf8_decode
 
1786
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1787
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1788
        # NB: show_status depends on being able to pass in non-versioned files
 
1789
        # and report them as unknown
2184
1790
        # TODO: handle extra trees in the dirstate.
2185
1791
        if (extra_trees or specific_files == []):
2186
1792
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1793
            for f in super(InterDirStateTree, self).iter_changes(
2188
1794
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1795
                require_versioned, want_unversioned=want_unversioned):
 
1796
                yield f
 
1797
            return
2190
1798
        parent_ids = self.target.get_parent_ids()
2191
 
        if not (self.source._revision_id in parent_ids
2192
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2193
 
            raise AssertionError(
2194
 
                "revision {%s} is not stored in {%s}, but %s "
2195
 
                "can only be used for trees stored in the dirstate"
2196
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1799
        assert (self.source._revision_id in parent_ids
 
1800
                or self.source._revision_id == NULL_REVISION), \
 
1801
                "revision {%s} is not stored in {%s}, but %s " \
 
1802
                "can only be used for trees stored in the dirstate" \
 
1803
                % (self.source._revision_id, self.target, self.iter_changes)
2197
1804
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1805
        if self.source._revision_id == NULL_REVISION:
2199
1806
            source_index = None
2200
1807
            indices = (target_index,)
2201
1808
        else:
2202
 
            if not (self.source._revision_id in parent_ids):
2203
 
                raise AssertionError(
2204
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2205
 
                    self.source._revision_id, parent_ids))
 
1809
            assert (self.source._revision_id in parent_ids), \
 
1810
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1811
                self.source._revision_id, parent_ids)
2206
1812
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
1813
            indices = (source_index, target_index)
2208
1814
        # -- make all specific_files utf8 --
2209
1815
        if specific_files:
2210
1816
            specific_files_utf8 = set()
2211
1817
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1818
                specific_files_utf8.add(path.encode('utf8'))
2215
1819
            specific_files = specific_files_utf8
2216
1820
        else:
2217
1821
            specific_files = set([''])
2218
1822
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1823
        # -- get the state object and prepare it.
2221
1824
        state = self.target.current_dirstate()
2222
1825
        state._read_dirblocks_if_needed()
 
1826
        def _entries_for_path(path):
 
1827
            """Return a list with all the entries that match path for all ids.
 
1828
            """
 
1829
            dirname, basename = os.path.split(path)
 
1830
            key = (dirname, basename, '')
 
1831
            block_index, present = state._find_block_index_from_key(key)
 
1832
            if not present:
 
1833
                # the block which should contain path is absent.
 
1834
                return []
 
1835
            result = []
 
1836
            block = state._dirblocks[block_index][1]
 
1837
            entry_index, _ = state._find_entry_index(key, block)
 
1838
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1839
            while (entry_index < len(block) and
 
1840
                block[entry_index][0][0:2] == key[0:2]):
 
1841
                result.append(block[entry_index])
 
1842
                entry_index += 1
 
1843
            return result
2223
1844
        if require_versioned:
2224
1845
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1846
            all_versioned = True
2226
1847
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1848
                path_entries = _entries_for_path(path)
2228
1849
                if not path_entries:
2229
1850
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1851
                    all_versioned = False
 
1852
                    break
2232
1853
                found_versioned = False
2233
1854
                # for each id at this path
2234
1855
                for entry in path_entries:
2241
1862
                if not found_versioned:
2242
1863
                    # none of the indexes was not 'absent' at all ids for this
2243
1864
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1865
                    all_versioned = False
 
1866
                    break
 
1867
            if not all_versioned:
 
1868
                raise errors.PathsNotVersionedError(specific_files)
2247
1869
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1870
        search_specific_files = set()
 
1871
        for path in specific_files:
 
1872
            other_specific_files = specific_files.difference(set([path]))
 
1873
            if not osutils.is_inside_any(other_specific_files, path):
 
1874
                # this is a top level path, we must check it.
 
1875
                search_specific_files.add(path)
 
1876
        # sketch: 
 
1877
        # compare source_index and target_index at or under each element of search_specific_files.
 
1878
        # follow the following comparison table. Note that we only want to do diff operations when
 
1879
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1880
        # for the target.
 
1881
        # cases:
 
1882
        # 
 
1883
        # Source | Target | disk | action
 
1884
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1885
        #        |        |      | diff check on source-target
 
1886
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1887
        #        |        |      | ???
 
1888
        #   r    |  a     |      | add source to search
 
1889
        #   r    |  a     |  a   | 
 
1890
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1891
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1892
        #   a    | fdlt   |      | add new id
 
1893
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1894
        #   a    |  a     |      | not present in either tree, skip
 
1895
        #   a    |  a     |  a   | not present in any tree, skip
 
1896
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1897
        #        |        |      | may not be selected by the users list of paths.
 
1898
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1899
        #        |        |      | may not be selected by the users list of paths.
 
1900
        #  fdlt  | fdlt   |      | content in both: diff them
 
1901
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1902
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1903
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1904
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1905
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1906
        #        |        |      | this id at the other path.
 
1907
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1908
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1909
        #        |        |      | this id at the other path.
 
1910
 
 
1911
        # for all search_indexs in each path at or under each element of
 
1912
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1913
        # relocated path as one to search if its not searched already. If the
 
1914
        # detail is not relocated, add the id.
 
1915
        searched_specific_files = set()
 
1916
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1917
        # Using a list so that we can access the values and change them in
 
1918
        # nested scope. Each one is [path, file_id, entry]
 
1919
        last_source_parent = [None, None]
 
1920
        last_target_parent = [None, None]
2249
1921
 
2250
1922
        use_filesystem_for_exec = (sys.platform != 'win32')
2251
 
        iter_changes = self.target._iter_changes(include_unchanged,
2252
 
            use_filesystem_for_exec, search_specific_files, state,
2253
 
            source_index, target_index, want_unversioned, self.target)
2254
 
        return iter_changes.iter_changes()
 
1923
 
 
1924
        # Just a sentry, so that _process_entry can say that this
 
1925
        # record is handled, but isn't interesting to process (unchanged)
 
1926
        uninteresting = object()
 
1927
 
 
1928
 
 
1929
        old_dirname_to_file_id = {}
 
1930
        new_dirname_to_file_id = {}
 
1931
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1932
        #       keeping a cache of directories that we have seen.
 
1933
 
 
1934
        def _process_entry(entry, path_info):
 
1935
            """Compare an entry and real disk to generate delta information.
 
1936
 
 
1937
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1938
                the path of entry. If None, then the path is considered absent.
 
1939
                (Perhaps we should pass in a concrete entry for this ?)
 
1940
                Basename is returned as a utf8 string because we expect this
 
1941
                tuple will be ignored, and don't want to take the time to
 
1942
                decode.
 
1943
            :return: None if these don't match
 
1944
                     A tuple of information about the change, or
 
1945
                     the object 'uninteresting' if these match, but are
 
1946
                     basically identical.
 
1947
            """
 
1948
            if source_index is None:
 
1949
                source_details = NULL_PARENT_DETAILS
 
1950
            else:
 
1951
                source_details = entry[1][source_index]
 
1952
            target_details = entry[1][target_index]
 
1953
            target_minikind = target_details[0]
 
1954
            if path_info is not None and target_minikind in 'fdlt':
 
1955
                assert target_index == 0
 
1956
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1957
                                                  stat_value=path_info[3])
 
1958
                # The entry may have been modified by update_entry
 
1959
                target_details = entry[1][target_index]
 
1960
                target_minikind = target_details[0]
 
1961
            else:
 
1962
                link_or_sha1 = None
 
1963
            file_id = entry[0][2]
 
1964
            source_minikind = source_details[0]
 
1965
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1966
                # claimed content in both: diff
 
1967
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1968
                #        |        |      | diff check on source-target
 
1969
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1970
                #        |        |      | ???
 
1971
                if source_minikind in 'r':
 
1972
                    # add the source to the search path to find any children it
 
1973
                    # has.  TODO ? : only add if it is a container ?
 
1974
                    if not osutils.is_inside_any(searched_specific_files,
 
1975
                                                 source_details[1]):
 
1976
                        search_specific_files.add(source_details[1])
 
1977
                    # generate the old path; this is needed for stating later
 
1978
                    # as well.
 
1979
                    old_path = source_details[1]
 
1980
                    old_dirname, old_basename = os.path.split(old_path)
 
1981
                    path = pathjoin(entry[0][0], entry[0][1])
 
1982
                    old_entry = state._get_entry(source_index,
 
1983
                                                 path_utf8=old_path)
 
1984
                    # update the source details variable to be the real
 
1985
                    # location.
 
1986
                    if old_entry == (None, None):
 
1987
                        raise errors.CorruptDirstate(state._filename,
 
1988
                            "entry '%s/%s' is considered renamed from %r"
 
1989
                            " but source does not exist\n"
 
1990
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
1991
                    source_details = old_entry[1][source_index]
 
1992
                    source_minikind = source_details[0]
 
1993
                else:
 
1994
                    old_dirname = entry[0][0]
 
1995
                    old_basename = entry[0][1]
 
1996
                    old_path = path = None
 
1997
                if path_info is None:
 
1998
                    # the file is missing on disk, show as removed.
 
1999
                    content_change = True
 
2000
                    target_kind = None
 
2001
                    target_exec = False
 
2002
                else:
 
2003
                    # source and target are both versioned and disk file is present.
 
2004
                    target_kind = path_info[2]
 
2005
                    if target_kind == 'directory':
 
2006
                        if path is None:
 
2007
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2008
                        new_dirname_to_file_id[path] = file_id
 
2009
                        if source_minikind != 'd':
 
2010
                            content_change = True
 
2011
                        else:
 
2012
                            # directories have no fingerprint
 
2013
                            content_change = False
 
2014
                        target_exec = False
 
2015
                    elif target_kind == 'file':
 
2016
                        if source_minikind != 'f':
 
2017
                            content_change = True
 
2018
                        else:
 
2019
                            # We could check the size, but we already have the
 
2020
                            # sha1 hash.
 
2021
                            content_change = (link_or_sha1 != source_details[1])
 
2022
                        # Target details is updated at update_entry time
 
2023
                        if use_filesystem_for_exec:
 
2024
                            # We don't need S_ISREG here, because we are sure
 
2025
                            # we are dealing with a file.
 
2026
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2027
                        else:
 
2028
                            target_exec = target_details[3]
 
2029
                    elif target_kind == 'symlink':
 
2030
                        if source_minikind != 'l':
 
2031
                            content_change = True
 
2032
                        else:
 
2033
                            content_change = (link_or_sha1 != source_details[1])
 
2034
                        target_exec = False
 
2035
                    elif target_kind == 'tree-reference':
 
2036
                        if source_minikind != 't':
 
2037
                            content_change = True
 
2038
                        else:
 
2039
                            content_change = False
 
2040
                        target_exec = False
 
2041
                    else:
 
2042
                        raise Exception, "unknown kind %s" % path_info[2]
 
2043
                if source_minikind == 'd':
 
2044
                    if path is None:
 
2045
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2046
                    old_dirname_to_file_id[old_path] = file_id
 
2047
                # parent id is the entry for the path in the target tree
 
2048
                if old_dirname == last_source_parent[0]:
 
2049
                    source_parent_id = last_source_parent[1]
 
2050
                else:
 
2051
                    try:
 
2052
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2053
                    except KeyError:
 
2054
                        source_parent_entry = state._get_entry(source_index,
 
2055
                                                               path_utf8=old_dirname)
 
2056
                        source_parent_id = source_parent_entry[0][2]
 
2057
                    if source_parent_id == entry[0][2]:
 
2058
                        # This is the root, so the parent is None
 
2059
                        source_parent_id = None
 
2060
                    else:
 
2061
                        last_source_parent[0] = old_dirname
 
2062
                        last_source_parent[1] = source_parent_id
 
2063
                new_dirname = entry[0][0]
 
2064
                if new_dirname == last_target_parent[0]:
 
2065
                    target_parent_id = last_target_parent[1]
 
2066
                else:
 
2067
                    try:
 
2068
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2069
                    except KeyError:
 
2070
                        # TODO: We don't always need to do the lookup, because the
 
2071
                        #       parent entry will be the same as the source entry.
 
2072
                        target_parent_entry = state._get_entry(target_index,
 
2073
                                                               path_utf8=new_dirname)
 
2074
                        assert target_parent_entry != (None, None), (
 
2075
                            "Could not find target parent in wt: %s\nparent of: %s"
 
2076
                            % (new_dirname, entry))
 
2077
                        target_parent_id = target_parent_entry[0][2]
 
2078
                    if target_parent_id == entry[0][2]:
 
2079
                        # This is the root, so the parent is None
 
2080
                        target_parent_id = None
 
2081
                    else:
 
2082
                        last_target_parent[0] = new_dirname
 
2083
                        last_target_parent[1] = target_parent_id
 
2084
 
 
2085
                source_exec = source_details[3]
 
2086
                if (include_unchanged
 
2087
                    or content_change
 
2088
                    or source_parent_id != target_parent_id
 
2089
                    or old_basename != entry[0][1]
 
2090
                    or source_exec != target_exec
 
2091
                    ):
 
2092
                    if old_path is None:
 
2093
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2094
                        old_path_u = utf8_decode(old_path)[0]
 
2095
                        path_u = old_path_u
 
2096
                    else:
 
2097
                        old_path_u = utf8_decode(old_path)[0]
 
2098
                        if old_path == path:
 
2099
                            path_u = old_path_u
 
2100
                        else:
 
2101
                            path_u = utf8_decode(path)[0]
 
2102
                    source_kind = _minikind_to_kind[source_minikind]
 
2103
                    return (entry[0][2],
 
2104
                           (old_path_u, path_u),
 
2105
                           content_change,
 
2106
                           (True, True),
 
2107
                           (source_parent_id, target_parent_id),
 
2108
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2109
                           (source_kind, target_kind),
 
2110
                           (source_exec, target_exec))
 
2111
                else:
 
2112
                    return uninteresting
 
2113
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2114
                # looks like a new file
 
2115
                if path_info is not None:
 
2116
                    path = pathjoin(entry[0][0], entry[0][1])
 
2117
                    # parent id is the entry for the path in the target tree
 
2118
                    # TODO: these are the same for an entire directory: cache em.
 
2119
                    parent_id = state._get_entry(target_index,
 
2120
                                                 path_utf8=entry[0][0])[0][2]
 
2121
                    if parent_id == entry[0][2]:
 
2122
                        parent_id = None
 
2123
                    if use_filesystem_for_exec:
 
2124
                        # We need S_ISREG here, because we aren't sure if this
 
2125
                        # is a file or not.
 
2126
                        target_exec = bool(
 
2127
                            stat.S_ISREG(path_info[3].st_mode)
 
2128
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2129
                    else:
 
2130
                        target_exec = target_details[3]
 
2131
                    return (entry[0][2],
 
2132
                           (None, utf8_decode(path)[0]),
 
2133
                           True,
 
2134
                           (False, True),
 
2135
                           (None, parent_id),
 
2136
                           (None, utf8_decode(entry[0][1])[0]),
 
2137
                           (None, path_info[2]),
 
2138
                           (None, target_exec))
 
2139
                else:
 
2140
                    # but its not on disk: we deliberately treat this as just
 
2141
                    # never-present. (Why ?! - RBC 20070224)
 
2142
                    pass
 
2143
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2144
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2145
                # if its still on disk, *and* theres no other entry at this
 
2146
                # path [we dont know this in this routine at the moment -
 
2147
                # perhaps we should change this - then it would be an unknown.
 
2148
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2149
                # parent id is the entry for the path in the target tree
 
2150
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2151
                if parent_id == entry[0][2]:
 
2152
                    parent_id = None
 
2153
                return (entry[0][2],
 
2154
                       (utf8_decode(old_path)[0], None),
 
2155
                       True,
 
2156
                       (True, False),
 
2157
                       (parent_id, None),
 
2158
                       (utf8_decode(entry[0][1])[0], None),
 
2159
                       (_minikind_to_kind[source_minikind], None),
 
2160
                       (source_details[3], None))
 
2161
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2162
                # a rename; could be a true rename, or a rename inherited from
 
2163
                # a renamed parent. TODO: handle this efficiently. Its not
 
2164
                # common case to rename dirs though, so a correct but slow
 
2165
                # implementation will do.
 
2166
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2167
                    search_specific_files.add(target_details[1])
 
2168
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2169
                # neither of the selected trees contain this file,
 
2170
                # so skip over it. This is not currently directly tested, but
 
2171
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2172
                pass
 
2173
            else:
 
2174
                raise AssertionError("don't know how to compare "
 
2175
                    "source_minikind=%r, target_minikind=%r"
 
2176
                    % (source_minikind, target_minikind))
 
2177
                ## import pdb;pdb.set_trace()
 
2178
            return None
 
2179
 
 
2180
        while search_specific_files:
 
2181
            # TODO: the pending list should be lexically sorted?  the
 
2182
            # interface doesn't require it.
 
2183
            current_root = search_specific_files.pop()
 
2184
            current_root_unicode = current_root.decode('utf8')
 
2185
            searched_specific_files.add(current_root)
 
2186
            # process the entries for this containing directory: the rest will be
 
2187
            # found by their parents recursively.
 
2188
            root_entries = _entries_for_path(current_root)
 
2189
            root_abspath = self.target.abspath(current_root_unicode)
 
2190
            try:
 
2191
                root_stat = os.lstat(root_abspath)
 
2192
            except OSError, e:
 
2193
                if e.errno == errno.ENOENT:
 
2194
                    # the path does not exist: let _process_entry know that.
 
2195
                    root_dir_info = None
 
2196
                else:
 
2197
                    # some other random error: hand it up.
 
2198
                    raise
 
2199
            else:
 
2200
                root_dir_info = ('', current_root,
 
2201
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2202
                    root_abspath)
 
2203
                if root_dir_info[2] == 'directory':
 
2204
                    if self.target._directory_is_tree_reference(
 
2205
                        current_root.decode('utf8')):
 
2206
                        root_dir_info = root_dir_info[:2] + \
 
2207
                            ('tree-reference',) + root_dir_info[3:]
 
2208
 
 
2209
            if not root_entries and not root_dir_info:
 
2210
                # this specified path is not present at all, skip it.
 
2211
                continue
 
2212
            path_handled = False
 
2213
            for entry in root_entries:
 
2214
                result = _process_entry(entry, root_dir_info)
 
2215
                if result is not None:
 
2216
                    path_handled = True
 
2217
                    if result is not uninteresting:
 
2218
                        yield result
 
2219
            if want_unversioned and not path_handled and root_dir_info:
 
2220
                new_executable = bool(
 
2221
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2222
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2223
                yield (None,
 
2224
                       (None, current_root_unicode),
 
2225
                       True,
 
2226
                       (False, False),
 
2227
                       (None, None),
 
2228
                       (None, splitpath(current_root_unicode)[-1]),
 
2229
                       (None, root_dir_info[2]),
 
2230
                       (None, new_executable)
 
2231
                      )
 
2232
            initial_key = (current_root, '', '')
 
2233
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2234
            if block_index == 0:
 
2235
                # we have processed the total root already, but because the
 
2236
                # initial key matched it we should skip it here.
 
2237
                block_index +=1
 
2238
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2239
                current_dir_info = None
 
2240
            else:
 
2241
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2242
                try:
 
2243
                    current_dir_info = dir_iterator.next()
 
2244
                except OSError, e:
 
2245
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2246
                    # python 2.5 has e.errno == EINVAL,
 
2247
                    #            and e.winerror == ERROR_DIRECTORY
 
2248
                    e_winerror = getattr(e, 'winerror', None)
 
2249
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2250
                    # there may be directories in the inventory even though
 
2251
                    # this path is not a file on disk: so mark it as end of
 
2252
                    # iterator
 
2253
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2254
                        current_dir_info = None
 
2255
                    elif (sys.platform == 'win32'
 
2256
                          and (e.errno in win_errors
 
2257
                               or e_winerror in win_errors)):
 
2258
                        current_dir_info = None
 
2259
                    else:
 
2260
                        raise
 
2261
                else:
 
2262
                    if current_dir_info[0][0] == '':
 
2263
                        # remove .bzr from iteration
 
2264
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2265
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2266
                        del current_dir_info[1][bzr_index]
 
2267
            # walk until both the directory listing and the versioned metadata
 
2268
            # are exhausted. 
 
2269
            if (block_index < len(state._dirblocks) and
 
2270
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2271
                current_block = state._dirblocks[block_index]
 
2272
            else:
 
2273
                current_block = None
 
2274
            while (current_dir_info is not None or
 
2275
                   current_block is not None):
 
2276
                if (current_dir_info and current_block
 
2277
                    and current_dir_info[0][0] != current_block[0]):
 
2278
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2279
                        # filesystem data refers to paths not covered by the dirblock.
 
2280
                        # this has two possibilities:
 
2281
                        # A) it is versioned but empty, so there is no block for it
 
2282
                        # B) it is not versioned.
 
2283
 
 
2284
                        # if (A) then we need to recurse into it to check for
 
2285
                        # new unknown files or directories.
 
2286
                        # if (B) then we should ignore it, because we don't
 
2287
                        # recurse into unknown directories.
 
2288
                        path_index = 0
 
2289
                        while path_index < len(current_dir_info[1]):
 
2290
                                current_path_info = current_dir_info[1][path_index]
 
2291
                                if want_unversioned:
 
2292
                                    if current_path_info[2] == 'directory':
 
2293
                                        if self.target._directory_is_tree_reference(
 
2294
                                            current_path_info[0].decode('utf8')):
 
2295
                                            current_path_info = current_path_info[:2] + \
 
2296
                                                ('tree-reference',) + current_path_info[3:]
 
2297
                                    new_executable = bool(
 
2298
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2299
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2300
                                    yield (None,
 
2301
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2302
                                        True,
 
2303
                                        (False, False),
 
2304
                                        (None, None),
 
2305
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2306
                                        (None, current_path_info[2]),
 
2307
                                        (None, new_executable))
 
2308
                                # dont descend into this unversioned path if it is
 
2309
                                # a dir
 
2310
                                if current_path_info[2] in ('directory',
 
2311
                                                            'tree-reference'):
 
2312
                                    del current_dir_info[1][path_index]
 
2313
                                    path_index -= 1
 
2314
                                path_index += 1
 
2315
 
 
2316
                        # This dir info has been handled, go to the next
 
2317
                        try:
 
2318
                            current_dir_info = dir_iterator.next()
 
2319
                        except StopIteration:
 
2320
                            current_dir_info = None
 
2321
                    else:
 
2322
                        # We have a dirblock entry for this location, but there
 
2323
                        # is no filesystem path for this. This is most likely
 
2324
                        # because a directory was removed from the disk.
 
2325
                        # We don't have to report the missing directory,
 
2326
                        # because that should have already been handled, but we
 
2327
                        # need to handle all of the files that are contained
 
2328
                        # within.
 
2329
                        for current_entry in current_block[1]:
 
2330
                            # entry referring to file not present on disk.
 
2331
                            # advance the entry only, after processing.
 
2332
                            result = _process_entry(current_entry, None)
 
2333
                            if result is not None:
 
2334
                                if result is not uninteresting:
 
2335
                                    yield result
 
2336
                        block_index +=1
 
2337
                        if (block_index < len(state._dirblocks) and
 
2338
                            osutils.is_inside(current_root,
 
2339
                                              state._dirblocks[block_index][0])):
 
2340
                            current_block = state._dirblocks[block_index]
 
2341
                        else:
 
2342
                            current_block = None
 
2343
                    continue
 
2344
                entry_index = 0
 
2345
                if current_block and entry_index < len(current_block[1]):
 
2346
                    current_entry = current_block[1][entry_index]
 
2347
                else:
 
2348
                    current_entry = None
 
2349
                advance_entry = True
 
2350
                path_index = 0
 
2351
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2352
                    current_path_info = current_dir_info[1][path_index]
 
2353
                    if current_path_info[2] == 'directory':
 
2354
                        if self.target._directory_is_tree_reference(
 
2355
                            current_path_info[0].decode('utf8')):
 
2356
                            current_path_info = current_path_info[:2] + \
 
2357
                                ('tree-reference',) + current_path_info[3:]
 
2358
                else:
 
2359
                    current_path_info = None
 
2360
                advance_path = True
 
2361
                path_handled = False
 
2362
                while (current_entry is not None or
 
2363
                    current_path_info is not None):
 
2364
                    if current_entry is None:
 
2365
                        # the check for path_handled when the path is adnvaced
 
2366
                        # will yield this path if needed.
 
2367
                        pass
 
2368
                    elif current_path_info is None:
 
2369
                        # no path is fine: the per entry code will handle it.
 
2370
                        result = _process_entry(current_entry, current_path_info)
 
2371
                        if result is not None:
 
2372
                            if result is not uninteresting:
 
2373
                                yield result
 
2374
                    elif (current_entry[0][1] != current_path_info[1]
 
2375
                          or current_entry[1][target_index][0] in 'ar'):
 
2376
                        # The current path on disk doesn't match the dirblock
 
2377
                        # record. Either the dirblock is marked as absent, or
 
2378
                        # the file on disk is not present at all in the
 
2379
                        # dirblock. Either way, report about the dirblock
 
2380
                        # entry, and let other code handle the filesystem one.
 
2381
 
 
2382
                        # Compare the basename for these files to determine
 
2383
                        # which comes first
 
2384
                        if current_path_info[1] < current_entry[0][1]:
 
2385
                            # extra file on disk: pass for now, but only
 
2386
                            # increment the path, not the entry
 
2387
                            advance_entry = False
 
2388
                        else:
 
2389
                            # entry referring to file not present on disk.
 
2390
                            # advance the entry only, after processing.
 
2391
                            result = _process_entry(current_entry, None)
 
2392
                            if result is not None:
 
2393
                                if result is not uninteresting:
 
2394
                                    yield result
 
2395
                            advance_path = False
 
2396
                    else:
 
2397
                        result = _process_entry(current_entry, current_path_info)
 
2398
                        if result is not None:
 
2399
                            path_handled = True
 
2400
                            if result is not uninteresting:
 
2401
                                yield result
 
2402
                    if advance_entry and current_entry is not None:
 
2403
                        entry_index += 1
 
2404
                        if entry_index < len(current_block[1]):
 
2405
                            current_entry = current_block[1][entry_index]
 
2406
                        else:
 
2407
                            current_entry = None
 
2408
                    else:
 
2409
                        advance_entry = True # reset the advance flaga
 
2410
                    if advance_path and current_path_info is not None:
 
2411
                        if not path_handled:
 
2412
                            # unversioned in all regards
 
2413
                            if want_unversioned:
 
2414
                                new_executable = bool(
 
2415
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2416
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2417
                                yield (None,
 
2418
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2419
                                    True,
 
2420
                                    (False, False),
 
2421
                                    (None, None),
 
2422
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2423
                                    (None, current_path_info[2]),
 
2424
                                    (None, new_executable))
 
2425
                            # dont descend into this unversioned path if it is
 
2426
                            # a dir
 
2427
                            if current_path_info[2] in ('directory'):
 
2428
                                del current_dir_info[1][path_index]
 
2429
                                path_index -= 1
 
2430
                        # dont descend the disk iterator into any tree 
 
2431
                        # paths.
 
2432
                        if current_path_info[2] == 'tree-reference':
 
2433
                            del current_dir_info[1][path_index]
 
2434
                            path_index -= 1
 
2435
                        path_index += 1
 
2436
                        if path_index < len(current_dir_info[1]):
 
2437
                            current_path_info = current_dir_info[1][path_index]
 
2438
                            if current_path_info[2] == 'directory':
 
2439
                                if self.target._directory_is_tree_reference(
 
2440
                                    current_path_info[0].decode('utf8')):
 
2441
                                    current_path_info = current_path_info[:2] + \
 
2442
                                        ('tree-reference',) + current_path_info[3:]
 
2443
                        else:
 
2444
                            current_path_info = None
 
2445
                        path_handled = False
 
2446
                    else:
 
2447
                        advance_path = True # reset the advance flagg.
 
2448
                if current_block is not None:
 
2449
                    block_index += 1
 
2450
                    if (block_index < len(state._dirblocks) and
 
2451
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2452
                        current_block = state._dirblocks[block_index]
 
2453
                    else:
 
2454
                        current_block = None
 
2455
                if current_dir_info is not None:
 
2456
                    try:
 
2457
                        current_dir_info = dir_iterator.next()
 
2458
                    except StopIteration:
 
2459
                        current_dir_info = None
 
2460
 
2255
2461
 
2256
2462
    @staticmethod
2257
2463
    def is_compatible(source, target):
2258
2464
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2465
        if not isinstance(target, WorkingTree4):
2260
2466
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2467
        # the source must be a revtreee or dirstate rev tree.
2262
2468
        if not isinstance(source,
2263
2469
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2470
            return False
2265
2471
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2472
        if not (source._revision_id == NULL_REVISION or
2267
2473
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2474
            # TODO: what about ghosts? it may well need to 
2269
2475
            # check for them explicitly.
2270
2476
            return False
2271
2477
        return True
2281
2487
 
2282
2488
    def convert(self, tree):
2283
2489
        # lock the control files not the tree, so that we dont get tree
2284
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2490
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2491
        # tree during upgrade.
2286
2492
        tree._control_files.lock_write()
2287
2493
        try:
2313
2519
 
2314
2520
    def update_format(self, tree):
2315
2521
        """Change the format marker."""
2316
 
        tree._transport.put_bytes('format',
2317
 
            self.target_format.as_string(),
2318
 
            mode=tree.bzrdir._get_file_mode())
2319
 
 
2320
 
 
2321
 
class Converter4to5(object):
2322
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2323
 
 
2324
 
    def __init__(self):
2325
 
        self.target_format = WorkingTreeFormat5()
2326
 
 
2327
 
    def convert(self, tree):
2328
 
        # lock the control files not the tree, so that we don't get tree
2329
 
        # on-unlock behaviours, and so that no-one else diddles with the
2330
 
        # tree during upgrade.
2331
 
        tree._control_files.lock_write()
2332
 
        try:
2333
 
            self.update_format(tree)
2334
 
        finally:
2335
 
            tree._control_files.unlock()
2336
 
 
2337
 
    def update_format(self, tree):
2338
 
        """Change the format marker."""
2339
 
        tree._transport.put_bytes('format',
2340
 
            self.target_format.as_string(),
2341
 
            mode=tree.bzrdir._get_file_mode())
2342
 
 
2343
 
 
2344
 
class Converter4or5to6(object):
2345
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2346
 
 
2347
 
    def __init__(self):
2348
 
        self.target_format = WorkingTreeFormat6()
2349
 
 
2350
 
    def convert(self, tree):
2351
 
        # lock the control files not the tree, so that we don't get tree
2352
 
        # on-unlock behaviours, and so that no-one else diddles with the
2353
 
        # tree during upgrade.
2354
 
        tree._control_files.lock_write()
2355
 
        try:
2356
 
            self.init_custom_control_files(tree)
2357
 
            self.update_format(tree)
2358
 
        finally:
2359
 
            tree._control_files.unlock()
2360
 
 
2361
 
    def init_custom_control_files(self, tree):
2362
 
        """Initialize custom control files."""
2363
 
        tree._transport.put_bytes('views', '',
2364
 
            mode=tree.bzrdir._get_file_mode())
2365
 
 
2366
 
    def update_format(self, tree):
2367
 
        """Change the format marker."""
2368
 
        tree._transport.put_bytes('format',
2369
 
            self.target_format.as_string(),
2370
 
            mode=tree.bzrdir._get_file_mode())
 
2522
        tree._control_files.put_utf8('format',
 
2523
            self.target_format.get_format_string())