~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Andrew Bennetts
  • Date: 2007-10-29 08:34:38 UTC
  • mto: (2535.4.22 streaming-smart-fetch)
  • mto: This revision was merged to the branch mainline in revision 2981.
  • Revision ID: andrew.bennetts@canonical.com-20071029083438-ke1vsv97dvgrvup5
Improve some docstrings.

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
116
151
        self._dirstate = None
117
152
        self._inventory = None
118
153
        #-------------
119
 
        self._setup_directory_is_tree_reference()
120
 
        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
154
 
126
155
    @needs_tree_write_lock
127
156
    def _add(self, files, ids, kinds):
142
171
            state.add(f, file_id, kind, None, '')
143
172
        self._make_dirty(reset_inventory=True)
144
173
 
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
174
    def _make_dirty(self, reset_inventory):
150
175
        """Make the tree state dirty.
151
176
 
159
184
    @needs_tree_write_lock
160
185
    def add_reference(self, sub_tree):
161
186
        # use standard implementation, which calls back to self._add
162
 
        #
 
187
        # 
163
188
        # So we don't store the reference_revision in the working dirstate,
164
 
        # it's just recorded at the moment of commit.
 
189
        # it's just recorded at the moment of commit. 
165
190
        self._add_reference(sub_tree)
166
191
 
167
192
    def break_lock(self):
203
228
 
204
229
    def _comparison_data(self, entry, path):
205
230
        kind, executable, stat_value = \
206
 
            WorkingTree._comparison_data(self, entry, path)
 
231
            WorkingTree3._comparison_data(self, entry, path)
207
232
        # it looks like a plain directory, but it's really a reference -- see
208
233
        # also kind()
209
 
        if (self._repo_supports_tree_reference and kind == 'directory'
210
 
            and entry is not None and entry.kind == 'tree-reference'):
 
234
        if (self._repo_supports_tree_reference and
 
235
            kind == 'directory' and
 
236
            self._directory_is_tree_reference(path)):
211
237
            kind = 'tree-reference'
212
238
        return kind, executable, stat_value
213
239
 
215
241
    def commit(self, message=None, revprops=None, *args, **kwargs):
216
242
        # mark the tree as dirty post commit - commit
217
243
        # can change the current versioned list by doing deletes.
218
 
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
 
244
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
219
245
        self._make_dirty(reset_inventory=True)
220
246
        return result
221
247
 
239
265
            return self._dirstate
240
266
        local_path = self.bzrdir.get_workingtree_transport(None
241
267
            ).local_abspath('dirstate')
242
 
        self._dirstate = dirstate.DirState.on_file(local_path,
243
 
            self._sha1_provider(), self._worth_saving_limit())
 
268
        self._dirstate = dirstate.DirState.on_file(local_path)
244
269
        return self._dirstate
245
270
 
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
271
    def filter_unversioned_files(self, paths):
267
272
        """Filter out paths that are versioned.
268
273
 
300
305
 
301
306
    def _generate_inventory(self):
302
307
        """Create and set self.inventory from the dirstate object.
303
 
 
 
308
        
304
309
        This is relatively expensive: we have to walk the entire dirstate.
305
310
        Ideally we would not, and can deprecate this function.
306
311
        """
310
315
        state._read_dirblocks_if_needed()
311
316
        root_key, current_entry = self._get_entry(path='')
312
317
        current_id = root_key[2]
313
 
        if not (current_entry[0][0] == 'd'): # directory
314
 
            raise AssertionError(current_entry)
 
318
        assert current_entry[0][0] == 'd' # directory
315
319
        inv = Inventory(root_id=current_id)
316
320
        # Turn some things into local variables
317
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
350
354
                    # add this entry to the parent map.
351
355
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
352
356
                elif kind == 'tree-reference':
353
 
                    if not self._repo_supports_tree_reference:
354
 
                        raise errors.UnsupportedOperation(
355
 
                            self._generate_inventory,
356
 
                            self.branch.repository)
 
357
                    assert self._repo_supports_tree_reference, \
 
358
                        "repository of %r " \
 
359
                        "doesn't support tree references " \
 
360
                        "required by entry %r" \
 
361
                        % (self, name)
357
362
                    inv_entry.reference_revision = link_or_sha1 or None
358
363
                elif kind != 'symlink':
359
364
                    raise AssertionError("unknown kind %r" % kind)
360
365
                # 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,))
 
366
                assert file_id not in inv_byid, ('file_id %s already in'
 
367
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                assert name_unicode not in parent_ie.children
367
369
                inv_byid[file_id] = inv_entry
368
370
                parent_ie.children[name_unicode] = inv_entry
369
371
        self._inventory = inv
374
376
        If either file_id or path is supplied, it is used as the key to lookup.
375
377
        If both are supplied, the fastest lookup is used, and an error is
376
378
        raised if they do not both point at the same row.
377
 
 
 
379
        
378
380
        :param file_id: An optional unicode file_id to be looked up.
379
381
        :param path: An optional unicode path to be looked up.
380
382
        :return: The dirstate row tuple for path/file_id, or (None, None)
398
400
        state = self.current_dirstate()
399
401
        if stat_value is None:
400
402
            try:
401
 
                stat_value = osutils.lstat(file_abspath)
 
403
                stat_value = os.lstat(file_abspath)
402
404
            except OSError, e:
403
405
                if e.errno == errno.ENOENT:
404
406
                    return None
405
407
                else:
406
408
                    raise
407
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
408
 
            stat_value=stat_value)
 
409
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
410
                                          stat_value=stat_value)
409
411
        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
 
412
            return link_or_sha1
420
413
        return None
421
414
 
422
 
    def _get_root_inventory(self):
 
415
    def _get_inventory(self):
423
416
        """Get the inventory for the tree. This is only valid within a lock."""
424
417
        if 'evil' in debug.debug_flags:
425
418
            trace.mutter_callsite(2,
430
423
        self._generate_inventory()
431
424
        return self._inventory
432
425
 
433
 
    @deprecated_method(deprecated_in((2, 5, 0)))
434
 
    def _get_inventory(self):
435
 
        return self.root_inventory
436
 
 
437
426
    inventory = property(_get_inventory,
438
427
                         doc="Inventory of this Tree")
439
428
 
440
 
    root_inventory = property(_get_root_inventory,
441
 
        "Root inventory of this tree")
442
 
 
443
429
    @needs_read_lock
444
430
    def get_parent_ids(self):
445
431
        """See Tree.get_parent_ids.
446
 
 
 
432
        
447
433
        This implementation requests the ids list from the dirstate file.
448
434
        """
449
435
        return self.current_dirstate().get_parent_ids()
471
457
        return osutils.lexists(pathjoin(
472
458
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
473
459
 
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
460
    @needs_read_lock
480
461
    def id2path(self, file_id):
481
462
        "Convert a file-id to a path."
486
467
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
487
468
        return path_utf8.decode('utf8')
488
469
 
489
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
490
 
        entry = self._get_entry(path=path)
491
 
        if entry == (None, None):
492
 
            return False # Missing entries are not executable
493
 
        return entry[1][0][3] # Executable?
494
 
 
495
 
    def is_executable(self, file_id, path=None):
496
 
        """Test if a file is executable or not.
497
 
 
498
 
        Note: The caller is expected to take a read-lock before calling this.
499
 
        """
500
 
        if not self._supports_executable():
 
470
    if not osutils.supports_executable():
 
471
        def is_executable(self, file_id, path=None):
 
472
            """Test if a file is executable or not.
 
473
 
 
474
            Note: The caller is expected to take a read-lock before calling this.
 
475
            """
501
476
            entry = self._get_entry(file_id=file_id, path=path)
502
477
            if entry == (None, None):
503
478
                return False
504
479
            return entry[1][0][3]
505
 
        else:
506
 
            self._must_be_locked()
 
480
    else:
 
481
        def is_executable(self, file_id, path=None):
 
482
            """Test if a file is executable or not.
 
483
 
 
484
            Note: The caller is expected to take a read-lock before calling this.
 
485
            """
507
486
            if not path:
508
487
                path = self.id2path(file_id)
509
 
            mode = osutils.lstat(self.abspath(path)).st_mode
 
488
            mode = os.lstat(self.abspath(path)).st_mode
510
489
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
511
490
 
512
 
    def all_file_ids(self):
513
 
        """See Tree.iter_all_file_ids"""
514
 
        self._must_be_locked()
515
 
        result = set()
516
 
        for key, tree_details in self.current_dirstate()._iter_entries():
517
 
            if tree_details[0][0] in ('a', 'r'): # relocated
518
 
                continue
519
 
            result.add(key[2])
520
 
        return result
521
 
 
522
491
    @needs_read_lock
523
492
    def __iter__(self):
524
493
        """Iterate through file_ids for this tree.
537
506
        return iter(result)
538
507
 
539
508
    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
509
        for key, tree_details in self.current_dirstate()._iter_entries():
545
510
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
546
511
                # not relevant to the working tree
548
513
            if not key[1]:
549
514
                # the root is not a reference.
550
515
                continue
551
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
516
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
552
517
            try:
553
 
                if self._kind(relpath) == 'tree-reference':
554
 
                    yield relpath, key[2]
 
518
                if self._kind(path) == 'tree-reference':
 
519
                    yield path, key[2]
555
520
            except errors.NoSuchFile:
556
521
                # path is missing on disk.
557
522
                continue
558
523
 
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
524
    def kind(self, file_id):
566
525
        """Return the kind of a file.
567
526
 
571
530
        Note: The caller is expected to take a read-lock before calling this.
572
531
        """
573
532
        relpath = self.id2path(file_id)
574
 
        if relpath is None:
575
 
            raise AssertionError(
576
 
                "path for id {%s} is None!" % file_id)
 
533
        assert relpath != None, \
 
534
            "path for id {%s} is None!" % file_id
577
535
        return self._kind(relpath)
578
536
 
579
537
    def _kind(self, relpath):
580
538
        abspath = self.abspath(relpath)
581
539
        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'
 
540
        if (self._repo_supports_tree_reference and
 
541
            kind == 'directory' and
 
542
            self._directory_is_tree_reference(relpath)):
 
543
            kind = 'tree-reference'
587
544
        return kind
588
545
 
589
546
    @needs_read_lock
596
553
            return _mod_revision.NULL_REVISION
597
554
 
598
555
    def lock_read(self):
599
 
        """See Branch.lock_read, and WorkingTree.unlock.
600
 
 
601
 
        :return: A bzrlib.lock.LogicalLockResult.
602
 
        """
 
556
        """See Branch.lock_read, and WorkingTree.unlock."""
603
557
        self.branch.lock_read()
604
558
        try:
605
559
            self._control_files.lock_read()
618
572
        except:
619
573
            self.branch.unlock()
620
574
            raise
621
 
        return LogicalLockResult(self.unlock)
622
575
 
623
576
    def _lock_self_write(self):
624
577
        """This should be called after the branch is locked."""
639
592
        except:
640
593
            self.branch.unlock()
641
594
            raise
642
 
        return LogicalLockResult(self.unlock)
643
595
 
644
596
    def lock_tree_write(self):
645
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
646
 
 
647
 
        :return: A bzrlib.lock.LogicalLockResult.
648
 
        """
 
597
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
649
598
        self.branch.lock_read()
650
 
        return self._lock_self_write()
 
599
        self._lock_self_write()
651
600
 
652
601
    def lock_write(self):
653
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
654
 
 
655
 
        :return: A bzrlib.lock.LogicalLockResult.
656
 
        """
 
602
        """See MutableTree.lock_write, and WorkingTree.unlock."""
657
603
        self.branch.lock_write()
658
 
        return self._lock_self_write()
 
604
        self._lock_self_write()
659
605
 
660
606
    @needs_tree_write_lock
661
607
    def move(self, from_paths, to_dir, after=False):
663
609
        result = []
664
610
        if not from_paths:
665
611
            return result
 
612
 
666
613
        state = self.current_dirstate()
667
 
        if isinstance(from_paths, basestring):
668
 
            raise ValueError()
 
614
 
 
615
        assert not isinstance(from_paths, basestring)
669
616
        to_dir_utf8 = to_dir.encode('utf8')
670
617
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
671
618
        id_index = state._get_id_index()
692
639
 
693
640
        if self._inventory is not None:
694
641
            update_inventory = True
695
 
            inv = self.root_inventory
 
642
            inv = self.inventory
 
643
            to_dir_ie = inv[to_dir_id]
696
644
            to_dir_id = to_entry[0][2]
697
 
            to_dir_ie = inv[to_dir_id]
698
645
        else:
699
646
            update_inventory = False
700
647
 
731
678
            from_entry = self._get_entry(path=from_rel)
732
679
            if from_entry == (None, None):
733
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
734
 
                    errors.NotVersionedError(path=from_rel))
 
681
                    errors.NotVersionedError(path=str(from_rel)))
735
682
 
736
683
            from_id = from_entry[0][2]
737
684
            to_rel = pathjoin(to_dir, from_tail)
764
711
                if from_missing: # implicitly just update our path mapping
765
712
                    move_file = False
766
713
                elif not after:
767
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
768
716
 
769
717
            rollbacks = []
770
718
            def rollback_rename():
826
774
                if minikind == 'd':
827
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
828
776
                        """Recursively update all entries in this dirblock."""
829
 
                        if from_dir == '':
830
 
                            raise AssertionError("renaming root not supported")
 
777
                        assert from_dir != '', "renaming root not supported"
831
778
                        from_key = (from_dir, '')
832
779
                        from_block_idx, present = \
833
780
                            state._find_block_index_from_key(from_key)
846
793
 
847
794
                        # Grab a copy since move_one may update the list.
848
795
                        for entry in from_block[1][:]:
849
 
                            if not (entry[0][0] == from_dir):
850
 
                                raise AssertionError()
 
796
                            assert entry[0][0] == from_dir
851
797
                            cur_details = entry[1][0]
852
798
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
853
799
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
878
824
                rollback_rename()
879
825
                raise
880
826
            result.append((from_rel, to_rel))
881
 
            state._mark_modified()
 
827
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
882
828
            self._make_dirty(reset_inventory=False)
883
829
 
884
830
        return result
894
840
    @needs_read_lock
895
841
    def path2id(self, path):
896
842
        """Return the id for path in this tree."""
897
 
        if isinstance(path, list):
898
 
            if path == []:
899
 
                path = [""]
900
 
            path = osutils.pathjoin(*path)
901
843
        path = path.strip('/')
902
844
        entry = self._get_entry(path=path)
903
845
        if entry == (None, None):
916
858
        for tree in trees:
917
859
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
918
860
                parents):
919
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
920
 
                    trees, require_versioned)
 
861
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
921
862
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
922
863
        # -- make all paths utf8 --
923
864
        paths_utf8 = set()
981
922
                    all_versioned = False
982
923
                    break
983
924
            if not all_versioned:
984
 
                raise errors.PathsNotVersionedError(
985
 
                    [p.decode('utf-8') for p in paths])
 
925
                raise errors.PathsNotVersionedError(paths)
986
926
        # -- remove redundancy in supplied paths to prevent over-scanning --
987
927
        search_paths = osutils.minimum_path_selection(paths)
988
 
        # sketch:
 
928
        # sketch: 
989
929
        # for all search_indexs in each path at or under each element of
990
930
        # search_paths, if the detail is relocated: add the id, and add the
991
931
        # relocated path as one to search if its not searched already. If the
1037
977
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
1038
978
            for dir_name in split_paths:
1039
979
                if dir_name not in found_dir_names:
1040
 
                    raise errors.PathsNotVersionedError(
1041
 
                        [p.decode('utf-8') for p in paths])
 
980
                    raise errors.PathsNotVersionedError(paths)
1042
981
 
1043
982
        for dir_name_id, trees_info in found.iteritems():
1044
983
            for index in search_indexes:
1048
987
 
1049
988
    def read_working_inventory(self):
1050
989
        """Read the working inventory.
1051
 
 
 
990
        
1052
991
        This is a meaningless operation for dirstate, but we obey it anyhow.
1053
992
        """
1054
 
        return self.root_inventory
 
993
        return self.inventory
1055
994
 
1056
995
    @needs_read_lock
1057
996
    def revision_tree(self, revision_id):
1072
1011
    def set_last_revision(self, new_revision):
1073
1012
        """Change the last revision in the working tree."""
1074
1013
        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.")
 
1014
        if new_revision in (NULL_REVISION, None):
 
1015
            assert len(parents) < 2, (
 
1016
                "setting the last parent to none with a pending merge is "
 
1017
                "unsupported.")
1080
1018
            self.set_parent_ids([])
1081
1019
        else:
1082
1020
            self.set_parent_ids([new_revision] + parents[1:],
1085
1023
    @needs_tree_write_lock
1086
1024
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1087
1025
        """Set the parent ids to revision_ids.
1088
 
 
 
1026
        
1089
1027
        See also set_parent_trees. This api will try to retrieve the tree data
1090
1028
        for each element of revision_ids from the trees repository. If you have
1091
1029
        tree data already available, it is more efficient to use
1123
1061
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1124
1062
        real_trees = []
1125
1063
        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
1064
        # convert absent trees to the null tree, which we convert back to
1133
1065
        # missing on access.
1134
1066
        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
1067
            _mod_revision.check_not_reserved_id(rev_id)
1142
1068
            if tree is not None:
1143
1069
                real_trees.append((rev_id, tree))
1144
1070
            else:
1145
1071
                real_trees.append((rev_id,
1146
 
                    self.branch.repository.revision_tree(
1147
 
                        _mod_revision.NULL_REVISION)))
 
1072
                    self.branch.repository.revision_tree(None)))
1148
1073
                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)
 
1074
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
1174
1075
        self._make_dirty(reset_inventory=False)
1175
1076
 
1176
1077
    def _set_root_id(self, file_id):
1188
1089
        :param path: The path.
1189
1090
        :param stat_result: The stat result being looked up.
1190
1091
        """
1191
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1092
        state = self.current_dirstate()
 
1093
        # XXX: should we make the path be passed in as utf8 ?
 
1094
        entry = state._get_entry(0, path_utf8=cache_utf8.encode(path))
 
1095
        tree_details = entry[1][0]
 
1096
        packed_stat = dirstate.pack_stat(stat_result)
 
1097
        if tree_details[4] == packed_stat:
 
1098
            return tree_details[1]
 
1099
        else:
 
1100
            return None
1192
1101
 
1193
1102
    @needs_read_lock
1194
1103
    def supports_tree_reference(self):
1196
1105
 
1197
1106
    def unlock(self):
1198
1107
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1108
        # do non-implementation specific cleanup
 
1109
        self._cleanup()
 
1110
 
1199
1111
        if self._control_files._lock_count == 1:
1200
 
            # do non-implementation specific cleanup
1201
 
            self._cleanup()
1202
 
 
1203
1112
            # eventually we should do signature checking during read locks for
1204
1113
            # dirstate updates.
1205
1114
            if self._control_files._lock_mode == 'w':
1268
1177
                # just forget the whole block.
1269
1178
                entry_index = 0
1270
1179
                while entry_index < len(block[1]):
 
1180
                    # Mark this file id as having been removed
1271
1181
                    entry = block[1][entry_index]
1272
 
                    if entry[1][0][0] in 'ar':
1273
 
                        # don't remove absent or renamed entries
 
1182
                    ids_to_unversion.discard(entry[0][2])
 
1183
                    if (entry[1][0][0] == 'a'
 
1184
                        or not state._make_absent(entry)):
1274
1185
                        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
1186
                # go to the next block. (At the moment we dont delete empty
1282
1187
                # dirblocks)
1283
1188
                block_index += 1
1304
1209
        # have to change the legacy inventory too.
1305
1210
        if self._inventory is not None:
1306
1211
            for file_id in file_ids:
1307
 
                if self._inventory.has_id(file_id):
1308
 
                    self._inventory.remove_recursive_id(file_id)
1309
 
 
1310
 
    @needs_tree_write_lock
1311
 
    def rename_one(self, from_rel, to_rel, after=False):
1312
 
        """See WorkingTree.rename_one"""
1313
 
        self.flush()
1314
 
        super(DirStateWorkingTree, self).rename_one(from_rel, to_rel, after)
1315
 
 
1316
 
    @needs_tree_write_lock
1317
 
    def apply_inventory_delta(self, changes):
1318
 
        """See MutableTree.apply_inventory_delta"""
1319
 
        state = self.current_dirstate()
1320
 
        state.update_by_delta(changes)
1321
 
        self._make_dirty(reset_inventory=True)
1322
 
 
1323
 
    def update_basis_by_delta(self, new_revid, delta):
1324
 
        """See MutableTree.update_basis_by_delta."""
1325
 
        if self.last_revision() == new_revid:
1326
 
            raise AssertionError()
1327
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1212
                self._inventory.remove_recursive_id(file_id)
1328
1213
 
1329
1214
    @needs_read_lock
1330
1215
    def _validate(self):
1333
1218
    @needs_tree_write_lock
1334
1219
    def _write_inventory(self, inv):
1335
1220
        """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:
 
1221
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1222
        self.current_dirstate().set_state_from_inventory(inv)
 
1223
        self._make_dirty(reset_inventory=False)
 
1224
        if self._inventory is not None:
1350
1225
            self._inventory = inv
1351
1226
        self.flush()
1352
1227
 
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)
1477
 
 
1478
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1479
 
                   accelerator_tree=None, hardlink=False):
 
1228
 
 
1229
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1230
    """The first consolidated dirstate working tree format.
 
1231
 
 
1232
    This format:
 
1233
        - exists within a metadir controlling .bzr
 
1234
        - includes an explicit version marker for the workingtree control
 
1235
          files, separate from the BzrDir format
 
1236
        - modifies the hash cache format
 
1237
        - is new in bzr 0.15
 
1238
        - uses a LockDir to guard access to it.
 
1239
    """
 
1240
 
 
1241
    upgrade_recommended = False
 
1242
 
 
1243
    def get_format_string(self):
 
1244
        """See WorkingTreeFormat.get_format_string()."""
 
1245
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1246
 
 
1247
    def get_format_description(self):
 
1248
        """See WorkingTreeFormat.get_format_description()."""
 
1249
        return "Working tree format 4"
 
1250
 
 
1251
    def initialize(self, a_bzrdir, revision_id=None):
1480
1252
        """See WorkingTreeFormat.initialize().
1481
1253
 
1482
1254
        :param revision_id: allows creating a working tree at a different
1483
 
            revision than the branch is at.
1484
 
        :param accelerator_tree: A tree which can be used for retrieving file
1485
 
            contents more quickly than the revision tree, i.e. a workingtree.
1486
 
            The revision tree will be used for cases where accelerator_tree's
1487
 
            content is different.
1488
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1489
 
            where possible.
 
1255
        revision than the branch is at.
1490
1256
 
1491
1257
        These trees get an initial random root id, if their repository supports
1492
1258
        rich root data, TREE_ROOT otherwise.
1497
1263
        control_files = self._open_control_files(a_bzrdir)
1498
1264
        control_files.create_lock()
1499
1265
        control_files.lock_write()
1500
 
        transport.put_bytes('format', self.as_string(),
1501
 
            mode=a_bzrdir._get_file_mode())
1502
 
        if from_branch is not None:
1503
 
            branch = from_branch
1504
 
        else:
1505
 
            branch = a_bzrdir.open_branch()
 
1266
        control_files.put_utf8('format', self.get_format_string())
 
1267
        branch = a_bzrdir.open_branch()
1506
1268
        if revision_id is None:
1507
1269
            revision_id = branch.last_revision()
1508
1270
        local_path = transport.local_abspath('dirstate')
1510
1272
        state = dirstate.DirState.initialize(local_path)
1511
1273
        state.unlock()
1512
1274
        del state
1513
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1275
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1514
1276
                         branch,
1515
1277
                         _format=self,
1516
1278
                         _bzrdir=a_bzrdir,
1518
1280
        wt._new_tree()
1519
1281
        wt.lock_tree_write()
1520
1282
        try:
1521
 
            self._init_custom_control_files(wt)
1522
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1283
            if revision_id in (None, NULL_REVISION):
1523
1284
                if branch.repository.supports_rich_root():
1524
1285
                    wt._set_root_id(generate_ids.gen_root_id())
1525
1286
                else:
1526
1287
                    wt._set_root_id(ROOT_ID)
1527
1288
                wt.flush()
1528
 
            basis = None
1529
 
            # frequently, we will get here due to branching.  The accelerator
1530
 
            # tree will be the tree from the branch, so the desired basis
1531
 
            # tree will often be a parent of the accelerator tree.
1532
 
            if accelerator_tree is not None:
1533
 
                try:
1534
 
                    basis = accelerator_tree.revision_tree(revision_id)
1535
 
                except errors.NoSuchRevision:
1536
 
                    pass
1537
 
            if basis is None:
1538
 
                basis = branch.repository.revision_tree(revision_id)
1539
 
            if revision_id == _mod_revision.NULL_REVISION:
1540
 
                parents_list = []
1541
 
            else:
1542
 
                parents_list = [(revision_id, basis)]
 
1289
            wt.set_last_revision(revision_id)
 
1290
            wt.flush()
 
1291
            basis = wt.basis_tree()
1543
1292
            basis.lock_read()
1544
 
            try:
1545
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1293
            # if the basis has a root id we have to use that; otherwise we use
 
1294
            # a new random one
 
1295
            basis_root_id = basis.get_root_id()
 
1296
            if basis_root_id is not None:
 
1297
                wt._set_root_id(basis_root_id)
1546
1298
                wt.flush()
1547
 
                # if the basis has a root id we have to use that; otherwise we
1548
 
                # use a new random one
1549
 
                basis_root_id = basis.get_root_id()
1550
 
                if basis_root_id is not None:
1551
 
                    wt._set_root_id(basis_root_id)
1552
 
                    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
 
                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)
1568
 
            finally:
1569
 
                basis.unlock()
 
1299
            transform.build_tree(basis, wt)
 
1300
            basis.unlock()
1570
1301
        finally:
1571
1302
            control_files.unlock()
1572
1303
            wt.unlock()
1573
1304
        return wt
1574
1305
 
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
1306
    def _open(self, a_bzrdir, control_files):
1599
1307
        """Open the tree itself.
1600
1308
 
1601
1309
        :param a_bzrdir: the dir for the tree.
1602
1310
        :param control_files: the control files for the tree.
1603
1311
        """
1604
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1312
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1605
1313
                           branch=a_bzrdir.open_branch(),
1606
1314
                           _format=self,
1607
1315
                           _bzrdir=a_bzrdir,
1608
1316
                           _control_files=control_files)
1609
1317
 
1610
1318
    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
1319
        # please test against something that will let us do tree references
1616
 
        return controldir.format_registry.make_bzrdir(
1617
 
            'development-subtree')
 
1320
        return bzrdir.format_registry.make_bzrdir(
 
1321
            'dirstate-with-subtree')
1618
1322
 
1619
1323
    _matchingbzrdir = property(__get_matchingbzrdir)
1620
1324
 
1621
1325
 
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
 
    """
 
1326
class DirStateRevisionTree(Tree):
 
1327
    """A revision tree pulling the inventory from a dirstate."""
1709
1328
 
1710
1329
    def __init__(self, dirstate, revision_id, repository):
1711
1330
        self._dirstate = dirstate
1714
1333
        self._inventory = None
1715
1334
        self._locked = 0
1716
1335
        self._dirstate_locked = False
1717
 
        self._repo_supports_tree_reference = getattr(
1718
 
            repository._format, "supports_tree_reference",
1719
 
            False)
1720
1336
 
1721
1337
    def __repr__(self):
1722
1338
        return "<%s of %s in %s>" % \
1725
1341
    def annotate_iter(self, file_id,
1726
1342
                      default_revision=_mod_revision.CURRENT_REVISION):
1727
1343
        """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]
 
1344
        w = self._get_weave(file_id)
 
1345
        return w.annotate_iter(self.inventory[file_id].revision)
1731
1346
 
 
1347
    def _get_ancestors(self, default_revision):
 
1348
        return set(self._repository.get_ancestry(self._revision_id,
 
1349
                                                 topo_sorted=False))
1732
1350
    def _comparison_data(self, entry, path):
1733
1351
        """See Tree._comparison_data."""
1734
1352
        if entry is None:
1751
1369
    def get_root_id(self):
1752
1370
        return self.path2id('')
1753
1371
 
1754
 
    def id2path(self, file_id):
1755
 
        "Convert a file-id to a path."
1756
 
        entry = self._get_entry(file_id=file_id)
1757
 
        if entry == (None, None):
1758
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1759
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1760
 
        return path_utf8.decode('utf8')
1761
 
 
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
1372
    def _get_parent_index(self):
1771
1373
        """Return the index in the dirstate referenced by this tree."""
1772
1374
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1777
1379
        If either file_id or path is supplied, it is used as the key to lookup.
1778
1380
        If both are supplied, the fastest lookup is used, and an error is
1779
1381
        raised if they do not both point at the same row.
1780
 
 
 
1382
        
1781
1383
        :param file_id: An optional unicode file_id to be looked up.
1782
1384
        :param path: An optional unicode path to be looked up.
1783
1385
        :return: The dirstate row tuple for path/file_id, or (None, None)
1787
1389
        if path is not None:
1788
1390
            path = path.encode('utf8')
1789
1391
        parent_index = self._get_parent_index()
1790
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
1791
 
            path_utf8=path)
 
1392
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
1792
1393
 
1793
1394
    def _generate_inventory(self):
1794
1395
        """Create and set self.inventory from the dirstate object.
1798
1399
 
1799
1400
        This is relatively expensive: we have to walk the entire dirstate.
1800
1401
        """
1801
 
        if not self._locked:
1802
 
            raise AssertionError(
1803
 
                'cannot generate inventory of an unlocked '
1804
 
                'dirstate revision tree')
 
1402
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1403
            'dirstate revision tree'
1805
1404
        # separate call for profiling - makes it clear where the costs are.
1806
1405
        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()))
 
1406
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1407
            'parent %s has disappeared from %s' % (
 
1408
            self._revision_id, self._dirstate.get_parent_ids())
1811
1409
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1812
1410
        # This is identical now to the WorkingTree _generate_inventory except
1813
1411
        # for the tree index use.
1814
1412
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1815
1413
        current_id = root_key[2]
1816
 
        if current_entry[parent_index][0] != 'd':
1817
 
            raise AssertionError()
 
1414
        assert current_entry[parent_index][0] == 'd'
1818
1415
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1819
1416
        inv.root.revision = current_entry[parent_index][4]
1820
1417
        # Turn some things into local variables
1851
1448
                elif kind == 'directory':
1852
1449
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1853
1450
                elif kind == 'symlink':
 
1451
                    inv_entry.executable = False
 
1452
                    inv_entry.text_size = None
1854
1453
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1855
1454
                elif kind == 'tree-reference':
1856
1455
                    inv_entry.reference_revision = fingerprint or None
1858
1457
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1859
1458
                            % entry)
1860
1459
                # 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,))
 
1460
                assert file_id not in inv_byid
 
1461
                assert name_unicode not in parent_ie.children
1867
1462
                inv_byid[file_id] = inv_entry
1868
1463
                parent_ie.children[name_unicode] = inv_entry
1869
1464
        self._inventory = inv
1876
1471
        # Make sure the file exists
1877
1472
        entry = self._get_entry(file_id, path=path)
1878
1473
        if entry == (None, None): # do we raise?
1879
 
            raise errors.NoSuchId(self, file_id)
 
1474
            return None
1880
1475
        parent_index = self._get_parent_index()
1881
1476
        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
 
1477
        return self._repository.get_revision(last_changed_revision).timestamp
1887
1478
 
1888
1479
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1889
1480
        entry = self._get_entry(file_id=file_id, path=path)
1893
1484
            return parent_details[1]
1894
1485
        return None
1895
1486
 
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
 
1487
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
 
1488
    def get_weave(self, file_id):
 
1489
        return self._get_weave(file_id)
 
1490
 
 
1491
    def _get_weave(self, file_id):
 
1492
        return self._repository.weave_store.get_weave(file_id,
 
1493
                self._repository.get_transaction())
1900
1494
 
1901
1495
    def get_file(self, file_id, path=None):
1902
1496
        return StringIO(self.get_file_text(file_id))
1903
1497
 
 
1498
    def get_file_lines(self, file_id):
 
1499
        ie = self.inventory[file_id]
 
1500
        return self._get_weave(file_id).get_lines(ie.revision)
 
1501
 
1904
1502
    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
 
1503
        return self.inventory[file_id].text_size
1908
1504
 
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
 
1505
    def get_file_text(self, file_id):
 
1506
        return ''.join(self.get_file_lines(file_id))
1922
1507
 
1923
1508
    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
 
1509
        return self.inventory[file_id].reference_revision
1926
1510
 
1927
1511
    def iter_files_bytes(self, desired_files):
1928
1512
        """See Tree.iter_files_bytes.
1938
1522
                                       identifier))
1939
1523
        return self._repository.iter_files_bytes(repo_desired_files)
1940
1524
 
1941
 
    def get_symlink_target(self, file_id, path=None):
 
1525
    def get_symlink_target(self, file_id):
1942
1526
        entry = self._get_entry(file_id=file_id)
1943
1527
        parent_index = self._get_parent_index()
1944
1528
        if entry[1][parent_index][0] != 'l':
1945
1529
            return None
1946
1530
        else:
1947
 
            target = entry[1][parent_index][1]
1948
 
            target = target.decode('utf8')
1949
 
            return target
 
1531
            # At present, none of the tree implementations supports non-ascii
 
1532
            # symlink targets. So we will just assume that the dirstate path is
 
1533
            # correct.
 
1534
            return entry[1][parent_index][1]
1950
1535
 
1951
1536
    def get_revision_id(self):
1952
1537
        """Return the revision id for this tree."""
1953
1538
        return self._revision_id
1954
1539
 
1955
 
    def _get_root_inventory(self):
 
1540
    def _get_inventory(self):
1956
1541
        if self._inventory is not None:
1957
1542
            return self._inventory
1958
1543
        self._must_be_locked()
1959
1544
        self._generate_inventory()
1960
1545
        return self._inventory
1961
1546
 
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
1547
    inventory = property(_get_inventory,
1970
1548
                         doc="Inventory of this Tree")
1971
1549
 
1977
1555
        return bool(self.path2id(filename))
1978
1556
 
1979
1557
    def kind(self, file_id):
1980
 
        entry = self._get_entry(file_id=file_id)[1]
1981
 
        if entry is None:
1982
 
            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]]
1985
 
 
1986
 
    def stored_kind(self, file_id):
1987
 
        """See Tree.stored_kind"""
1988
 
        return self.kind(file_id)
 
1558
        return self.inventory[file_id].kind
1989
1559
 
1990
1560
    def path_content_summary(self, path):
1991
1561
        """See Tree.path_content_summary."""
1992
 
        inv, inv_file_id = self._path2inv_file_id(path)
1993
 
        if inv_file_id is None:
 
1562
        id = self.inventory.path2id(path)
 
1563
        if id is None:
1994
1564
            return ('missing', None, None, None)
1995
 
        entry = inv[inv_file_id]
 
1565
        entry = self._inventory[id]
1996
1566
        kind = entry.kind
1997
1567
        if kind == 'file':
1998
1568
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
2002
1572
            return (kind, None, None, None)
2003
1573
 
2004
1574
    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]
 
1575
        ie = self.inventory[file_id]
2007
1576
        if ie.kind != "file":
2008
 
            return False
 
1577
            return None
2009
1578
        return ie.executable
2010
1579
 
2011
 
    def is_locked(self):
2012
 
        return self._locked
2013
 
 
2014
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1580
    def list_files(self, include_root=False):
2015
1581
        # We use a standard implementation, because DirStateRevisionTree is
2016
1582
        # 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:
 
1583
        inv = self._get_inventory()
 
1584
        entries = inv.iter_entries()
 
1585
        if self.inventory.root is not None and not include_root:
2028
1586
            entries.next()
2029
1587
        for path, entry in entries:
2030
1588
            yield path, 'V', entry.kind, entry.file_id, entry
2031
1589
 
2032
1590
    def lock_read(self):
2033
 
        """Lock the tree for a set of operations.
2034
 
 
2035
 
        :return: A bzrlib.lock.LogicalLockResult.
2036
 
        """
 
1591
        """Lock the tree for a set of operations."""
2037
1592
        if not self._locked:
2038
1593
            self._repository.lock_read()
2039
1594
            if self._dirstate._lock_token is None:
2040
1595
                self._dirstate.lock_read()
2041
1596
                self._dirstate_locked = True
2042
1597
        self._locked += 1
2043
 
        return LogicalLockResult(self.unlock)
2044
1598
 
2045
1599
    def _must_be_locked(self):
2046
1600
        if not self._locked:
2050
1604
    def path2id(self, path):
2051
1605
        """Return the id for path in this tree."""
2052
1606
        # 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
1607
        entry = self._get_entry(path=path)
2058
1608
        if entry == (None, None):
2059
1609
            return None
2071
1621
                self._dirstate_locked = False
2072
1622
            self._repository.unlock()
2073
1623
 
2074
 
    @needs_read_lock
2075
 
    def supports_tree_reference(self):
2076
 
        return self._repo_supports_tree_reference
2077
 
 
2078
1624
    def walkdirs(self, prefix=""):
2079
1625
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
2080
 
        # implementation based on an inventory.
 
1626
        # implementation based on an inventory.  
2081
1627
        # This should be cleaned up to use the much faster Dirstate code
2082
1628
        # So for now, we just build up the parent inventory, and extract
2083
1629
        # it the same way RevisionTree does.
2084
1630
        _directory = 'directory'
2085
 
        inv = self._get_root_inventory()
 
1631
        inv = self._get_inventory()
2086
1632
        top_id = inv.path2id(prefix)
2087
1633
        if top_id is None:
2088
1634
            pending = []
2112
1658
 
2113
1659
class InterDirStateTree(InterTree):
2114
1660
    """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
 
1661
    
 
1662
    This is used only when both trees are in the dirstate working file, and 
 
1663
    the source is any parent within the dirstate, and the destination is 
2118
1664
    the current working tree of the same dirstate.
2119
1665
    """
2120
1666
    # this could be generalized to allow comparisons between any trees in the
2129
1675
    def make_source_parent_tree(source, target):
2130
1676
        """Change the source tree into a parent of the target."""
2131
1677
        revid = source.commit('record tree')
2132
 
        target.branch.fetch(source.branch, revid)
 
1678
        target.branch.repository.fetch(source.branch.repository, revid)
2133
1679
        target.set_parent_ids([revid])
2134
1680
        return target.basis_tree(), target
2135
1681
 
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
1682
    _matching_from_tree_format = WorkingTreeFormat4()
2154
1683
    _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
2161
 
 
2162
 
    def iter_changes(self, include_unchanged=False,
 
1684
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1685
 
 
1686
    def _iter_changes(self, include_unchanged=False,
2163
1687
                      specific_files=None, pb=None, extra_trees=[],
2164
1688
                      require_versioned=True, want_unversioned=False):
2165
1689
        """Return the changes from source to target.
2166
1690
 
2167
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1691
        :return: An iterator that yields tuples. See InterTree._iter_changes
2168
1692
            for details.
2169
1693
        :param specific_files: An optional list of file paths to restrict the
2170
1694
            comparison to. When mapping filenames to ids, all matches in all
2181
1705
            output. An unversioned file is defined as one with (False, False)
2182
1706
            for the versioned pair.
2183
1707
        """
 
1708
        utf8_decode = cache_utf8._utf8_decode
 
1709
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1710
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1711
        # NB: show_status depends on being able to pass in non-versioned files
 
1712
        # and report them as unknown
2184
1713
        # TODO: handle extra trees in the dirstate.
2185
 
        if (extra_trees or specific_files == []):
 
1714
        # TODO: handle comparisons as an empty tree as a different special
 
1715
        # case? mbp 20070226
 
1716
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
 
1717
            or specific_files == []):
2186
1718
            # we can't fast-path these cases (yet)
2187
 
            return super(InterDirStateTree, self).iter_changes(
 
1719
            for f in super(InterDirStateTree, self)._iter_changes(
2188
1720
                include_unchanged, specific_files, pb, extra_trees,
2189
 
                require_versioned, want_unversioned=want_unversioned)
 
1721
                require_versioned, want_unversioned=want_unversioned):
 
1722
                yield f
 
1723
            return
2190
1724
        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))
 
1725
        assert (self.source._revision_id in parent_ids), \
 
1726
                "revision {%s} is not stored in {%s}, but %s " \
 
1727
                "can only be used for trees stored in the dirstate" \
 
1728
                % (self.source._revision_id, self.target, self._iter_changes)
2197
1729
        target_index = 0
2198
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1730
        if self.source._revision_id == NULL_REVISION:
2199
1731
            source_index = None
2200
1732
            indices = (target_index,)
2201
1733
        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))
 
1734
            assert (self.source._revision_id in parent_ids), \
 
1735
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1736
                self.source._revision_id, parent_ids)
2206
1737
            source_index = 1 + parent_ids.index(self.source._revision_id)
2207
 
            indices = (source_index, target_index)
 
1738
            indices = (source_index,target_index)
2208
1739
        # -- make all specific_files utf8 --
2209
1740
        if specific_files:
2210
1741
            specific_files_utf8 = set()
2211
1742
            for path in specific_files:
2212
 
                # Note, if there are many specific files, using cache_utf8
2213
 
                # would be good here.
2214
1743
                specific_files_utf8.add(path.encode('utf8'))
2215
1744
            specific_files = specific_files_utf8
2216
1745
        else:
2217
1746
            specific_files = set([''])
2218
1747
        # -- specific_files is now a utf8 path set --
2219
 
 
2220
1748
        # -- get the state object and prepare it.
2221
1749
        state = self.target.current_dirstate()
2222
1750
        state._read_dirblocks_if_needed()
 
1751
        def _entries_for_path(path):
 
1752
            """Return a list with all the entries that match path for all ids.
 
1753
            """
 
1754
            dirname, basename = os.path.split(path)
 
1755
            key = (dirname, basename, '')
 
1756
            block_index, present = state._find_block_index_from_key(key)
 
1757
            if not present:
 
1758
                # the block which should contain path is absent.
 
1759
                return []
 
1760
            result = []
 
1761
            block = state._dirblocks[block_index][1]
 
1762
            entry_index, _ = state._find_entry_index(key, block)
 
1763
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1764
            while (entry_index < len(block) and
 
1765
                block[entry_index][0][0:2] == key[0:2]):
 
1766
                result.append(block[entry_index])
 
1767
                entry_index += 1
 
1768
            return result
2223
1769
        if require_versioned:
2224
1770
            # -- check all supplied paths are versioned in a search tree. --
2225
 
            not_versioned = []
 
1771
            all_versioned = True
2226
1772
            for path in specific_files:
2227
 
                path_entries = state._entries_for_path(path)
 
1773
                path_entries = _entries_for_path(path)
2228
1774
                if not path_entries:
2229
1775
                    # this specified path is not present at all: error
2230
 
                    not_versioned.append(path.decode('utf-8'))
2231
 
                    continue
 
1776
                    all_versioned = False
 
1777
                    break
2232
1778
                found_versioned = False
2233
1779
                # for each id at this path
2234
1780
                for entry in path_entries:
2241
1787
                if not found_versioned:
2242
1788
                    # none of the indexes was not 'absent' at all ids for this
2243
1789
                    # path.
2244
 
                    not_versioned.append(path.decode('utf-8'))
2245
 
            if len(not_versioned) > 0:
2246
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1790
                    all_versioned = False
 
1791
                    break
 
1792
            if not all_versioned:
 
1793
                raise errors.PathsNotVersionedError(specific_files)
2247
1794
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2248
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1795
        search_specific_files = set()
 
1796
        for path in specific_files:
 
1797
            other_specific_files = specific_files.difference(set([path]))
 
1798
            if not osutils.is_inside_any(other_specific_files, path):
 
1799
                # this is a top level path, we must check it.
 
1800
                search_specific_files.add(path)
 
1801
        # sketch: 
 
1802
        # compare source_index and target_index at or under each element of search_specific_files.
 
1803
        # follow the following comparison table. Note that we only want to do diff operations when
 
1804
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1805
        # for the target.
 
1806
        # cases:
 
1807
        # 
 
1808
        # Source | Target | disk | action
 
1809
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1810
        #        |        |      | diff check on source-target
 
1811
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1812
        #        |        |      | ???
 
1813
        #   r    |  a     |      | add source to search
 
1814
        #   r    |  a     |  a   | 
 
1815
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1816
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1817
        #   a    | fdlt   |      | add new id
 
1818
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1819
        #   a    |  a     |      | not present in either tree, skip
 
1820
        #   a    |  a     |  a   | not present in any tree, skip
 
1821
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1822
        #        |        |      | may not be selected by the users list of paths.
 
1823
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1824
        #        |        |      | may not be selected by the users list of paths.
 
1825
        #  fdlt  | fdlt   |      | content in both: diff them
 
1826
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1827
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1828
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1829
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1830
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1831
        #        |        |      | this id at the other path.
 
1832
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1833
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1834
        #        |        |      | this id at the other path.
 
1835
 
 
1836
        # for all search_indexs in each path at or under each element of
 
1837
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1838
        # relocated path as one to search if its not searched already. If the
 
1839
        # detail is not relocated, add the id.
 
1840
        searched_specific_files = set()
 
1841
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1842
        # Using a list so that we can access the values and change them in
 
1843
        # nested scope. Each one is [path, file_id, entry]
 
1844
        last_source_parent = [None, None]
 
1845
        last_target_parent = [None, None]
2249
1846
 
2250
1847
        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()
 
1848
 
 
1849
        # Just a sentry, so that _process_entry can say that this
 
1850
        # record is handled, but isn't interesting to process (unchanged)
 
1851
        uninteresting = object()
 
1852
 
 
1853
 
 
1854
        old_dirname_to_file_id = {}
 
1855
        new_dirname_to_file_id = {}
 
1856
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1857
        #       keeping a cache of directories that we have seen.
 
1858
 
 
1859
        def _process_entry(entry, path_info):
 
1860
            """Compare an entry and real disk to generate delta information.
 
1861
 
 
1862
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1863
                the path of entry. If None, then the path is considered absent.
 
1864
                (Perhaps we should pass in a concrete entry for this ?)
 
1865
                Basename is returned as a utf8 string because we expect this
 
1866
                tuple will be ignored, and don't want to take the time to
 
1867
                decode.
 
1868
            :return: None if these don't match
 
1869
                     A tuple of information about the change, or
 
1870
                     the object 'uninteresting' if these match, but are
 
1871
                     basically identical.
 
1872
            """
 
1873
            if source_index is None:
 
1874
                source_details = NULL_PARENT_DETAILS
 
1875
            else:
 
1876
                source_details = entry[1][source_index]
 
1877
            target_details = entry[1][target_index]
 
1878
            target_minikind = target_details[0]
 
1879
            if path_info is not None and target_minikind in 'fdlt':
 
1880
                assert target_index == 0
 
1881
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1882
                                                  stat_value=path_info[3])
 
1883
                # The entry may have been modified by update_entry
 
1884
                target_details = entry[1][target_index]
 
1885
                target_minikind = target_details[0]
 
1886
            else:
 
1887
                link_or_sha1 = None
 
1888
            file_id = entry[0][2]
 
1889
            source_minikind = source_details[0]
 
1890
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1891
                # claimed content in both: diff
 
1892
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1893
                #        |        |      | diff check on source-target
 
1894
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1895
                #        |        |      | ???
 
1896
                if source_minikind in 'r':
 
1897
                    # add the source to the search path to find any children it
 
1898
                    # has.  TODO ? : only add if it is a container ?
 
1899
                    if not osutils.is_inside_any(searched_specific_files,
 
1900
                                                 source_details[1]):
 
1901
                        search_specific_files.add(source_details[1])
 
1902
                    # generate the old path; this is needed for stating later
 
1903
                    # as well.
 
1904
                    old_path = source_details[1]
 
1905
                    old_dirname, old_basename = os.path.split(old_path)
 
1906
                    path = pathjoin(entry[0][0], entry[0][1])
 
1907
                    old_entry = state._get_entry(source_index,
 
1908
                                                 path_utf8=old_path)
 
1909
                    # update the source details variable to be the real
 
1910
                    # location.
 
1911
                    source_details = old_entry[1][source_index]
 
1912
                    source_minikind = source_details[0]
 
1913
                else:
 
1914
                    old_dirname = entry[0][0]
 
1915
                    old_basename = entry[0][1]
 
1916
                    old_path = path = None
 
1917
                if path_info is None:
 
1918
                    # the file is missing on disk, show as removed.
 
1919
                    content_change = True
 
1920
                    target_kind = None
 
1921
                    target_exec = False
 
1922
                else:
 
1923
                    # source and target are both versioned and disk file is present.
 
1924
                    target_kind = path_info[2]
 
1925
                    if target_kind == 'directory':
 
1926
                        if path is None:
 
1927
                            old_path = path = pathjoin(old_dirname, old_basename)
 
1928
                        new_dirname_to_file_id[path] = file_id
 
1929
                        if source_minikind != 'd':
 
1930
                            content_change = True
 
1931
                        else:
 
1932
                            # directories have no fingerprint
 
1933
                            content_change = False
 
1934
                        target_exec = False
 
1935
                    elif target_kind == 'file':
 
1936
                        if source_minikind != 'f':
 
1937
                            content_change = True
 
1938
                        else:
 
1939
                            # We could check the size, but we already have the
 
1940
                            # sha1 hash.
 
1941
                            content_change = (link_or_sha1 != source_details[1])
 
1942
                        # Target details is updated at update_entry time
 
1943
                        if use_filesystem_for_exec:
 
1944
                            # We don't need S_ISREG here, because we are sure
 
1945
                            # we are dealing with a file.
 
1946
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1947
                        else:
 
1948
                            target_exec = target_details[3]
 
1949
                    elif target_kind == 'symlink':
 
1950
                        if source_minikind != 'l':
 
1951
                            content_change = True
 
1952
                        else:
 
1953
                            content_change = (link_or_sha1 != source_details[1])
 
1954
                        target_exec = False
 
1955
                    elif target_kind == 'tree-reference':
 
1956
                        if source_minikind != 't':
 
1957
                            content_change = True
 
1958
                        else:
 
1959
                            content_change = False
 
1960
                        target_exec = False
 
1961
                    else:
 
1962
                        raise Exception, "unknown kind %s" % path_info[2]
 
1963
                if source_minikind == 'd':
 
1964
                    if path is None:
 
1965
                        old_path = path = pathjoin(old_dirname, old_basename)
 
1966
                    old_dirname_to_file_id[old_path] = file_id
 
1967
                # parent id is the entry for the path in the target tree
 
1968
                if old_dirname == last_source_parent[0]:
 
1969
                    source_parent_id = last_source_parent[1]
 
1970
                else:
 
1971
                    try:
 
1972
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
1973
                    except KeyError:
 
1974
                        source_parent_entry = state._get_entry(source_index,
 
1975
                                                               path_utf8=old_dirname)
 
1976
                        source_parent_id = source_parent_entry[0][2]
 
1977
                    if source_parent_id == entry[0][2]:
 
1978
                        # This is the root, so the parent is None
 
1979
                        source_parent_id = None
 
1980
                    else:
 
1981
                        last_source_parent[0] = old_dirname
 
1982
                        last_source_parent[1] = source_parent_id
 
1983
                new_dirname = entry[0][0]
 
1984
                if new_dirname == last_target_parent[0]:
 
1985
                    target_parent_id = last_target_parent[1]
 
1986
                else:
 
1987
                    try:
 
1988
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
1989
                    except KeyError:
 
1990
                        # TODO: We don't always need to do the lookup, because the
 
1991
                        #       parent entry will be the same as the source entry.
 
1992
                        target_parent_entry = state._get_entry(target_index,
 
1993
                                                               path_utf8=new_dirname)
 
1994
                        assert target_parent_entry != (None, None), (
 
1995
                            "Could not find target parent in wt: %s\nparent of: %s"
 
1996
                            % (new_dirname, entry))
 
1997
                        target_parent_id = target_parent_entry[0][2]
 
1998
                    if target_parent_id == entry[0][2]:
 
1999
                        # This is the root, so the parent is None
 
2000
                        target_parent_id = None
 
2001
                    else:
 
2002
                        last_target_parent[0] = new_dirname
 
2003
                        last_target_parent[1] = target_parent_id
 
2004
 
 
2005
                source_exec = source_details[3]
 
2006
                if (include_unchanged
 
2007
                    or content_change
 
2008
                    or source_parent_id != target_parent_id
 
2009
                    or old_basename != entry[0][1]
 
2010
                    or source_exec != target_exec
 
2011
                    ):
 
2012
                    if old_path is None:
 
2013
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2014
                        old_path_u = utf8_decode(old_path)[0]
 
2015
                        path_u = old_path_u
 
2016
                    else:
 
2017
                        old_path_u = utf8_decode(old_path)[0]
 
2018
                        if old_path == path:
 
2019
                            path_u = old_path_u
 
2020
                        else:
 
2021
                            path_u = utf8_decode(path)[0]
 
2022
                    source_kind = _minikind_to_kind[source_minikind]
 
2023
                    return (entry[0][2],
 
2024
                           (old_path_u, path_u),
 
2025
                           content_change,
 
2026
                           (True, True),
 
2027
                           (source_parent_id, target_parent_id),
 
2028
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2029
                           (source_kind, target_kind),
 
2030
                           (source_exec, target_exec))
 
2031
                else:
 
2032
                    return uninteresting
 
2033
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2034
                # looks like a new file
 
2035
                if path_info is not None:
 
2036
                    path = pathjoin(entry[0][0], entry[0][1])
 
2037
                    # parent id is the entry for the path in the target tree
 
2038
                    # TODO: these are the same for an entire directory: cache em.
 
2039
                    parent_id = state._get_entry(target_index,
 
2040
                                                 path_utf8=entry[0][0])[0][2]
 
2041
                    if parent_id == entry[0][2]:
 
2042
                        parent_id = None
 
2043
                    if use_filesystem_for_exec:
 
2044
                        # We need S_ISREG here, because we aren't sure if this
 
2045
                        # is a file or not.
 
2046
                        target_exec = bool(
 
2047
                            stat.S_ISREG(path_info[3].st_mode)
 
2048
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2049
                    else:
 
2050
                        target_exec = target_details[3]
 
2051
                    return (entry[0][2],
 
2052
                           (None, utf8_decode(path)[0]),
 
2053
                           True,
 
2054
                           (False, True),
 
2055
                           (None, parent_id),
 
2056
                           (None, utf8_decode(entry[0][1])[0]),
 
2057
                           (None, path_info[2]),
 
2058
                           (None, target_exec))
 
2059
                else:
 
2060
                    # but its not on disk: we deliberately treat this as just
 
2061
                    # never-present. (Why ?! - RBC 20070224)
 
2062
                    pass
 
2063
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2064
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2065
                # if its still on disk, *and* theres no other entry at this
 
2066
                # path [we dont know this in this routine at the moment -
 
2067
                # perhaps we should change this - then it would be an unknown.
 
2068
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2069
                # parent id is the entry for the path in the target tree
 
2070
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2071
                if parent_id == entry[0][2]:
 
2072
                    parent_id = None
 
2073
                return (entry[0][2],
 
2074
                       (utf8_decode(old_path)[0], None),
 
2075
                       True,
 
2076
                       (True, False),
 
2077
                       (parent_id, None),
 
2078
                       (utf8_decode(entry[0][1])[0], None),
 
2079
                       (_minikind_to_kind[source_minikind], None),
 
2080
                       (source_details[3], None))
 
2081
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2082
                # a rename; could be a true rename, or a rename inherited from
 
2083
                # a renamed parent. TODO: handle this efficiently. Its not
 
2084
                # common case to rename dirs though, so a correct but slow
 
2085
                # implementation will do.
 
2086
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2087
                    search_specific_files.add(target_details[1])
 
2088
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2089
                # neither of the selected trees contain this file,
 
2090
                # so skip over it. This is not currently directly tested, but
 
2091
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2092
                pass
 
2093
            else:
 
2094
                raise AssertionError("don't know how to compare "
 
2095
                    "source_minikind=%r, target_minikind=%r"
 
2096
                    % (source_minikind, target_minikind))
 
2097
                ## import pdb;pdb.set_trace()
 
2098
            return None
 
2099
 
 
2100
        while search_specific_files:
 
2101
            # TODO: the pending list should be lexically sorted?  the
 
2102
            # interface doesn't require it.
 
2103
            current_root = search_specific_files.pop()
 
2104
            current_root_unicode = current_root.decode('utf8')
 
2105
            searched_specific_files.add(current_root)
 
2106
            # process the entries for this containing directory: the rest will be
 
2107
            # found by their parents recursively.
 
2108
            root_entries = _entries_for_path(current_root)
 
2109
            root_abspath = self.target.abspath(current_root_unicode)
 
2110
            try:
 
2111
                root_stat = os.lstat(root_abspath)
 
2112
            except OSError, e:
 
2113
                if e.errno == errno.ENOENT:
 
2114
                    # the path does not exist: let _process_entry know that.
 
2115
                    root_dir_info = None
 
2116
                else:
 
2117
                    # some other random error: hand it up.
 
2118
                    raise
 
2119
            else:
 
2120
                root_dir_info = ('', current_root,
 
2121
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2122
                    root_abspath)
 
2123
                if root_dir_info[2] == 'directory':
 
2124
                    if self.target._directory_is_tree_reference(
 
2125
                        current_root.decode('utf8')):
 
2126
                        root_dir_info = root_dir_info[:2] + \
 
2127
                            ('tree-reference',) + root_dir_info[3:]
 
2128
 
 
2129
            if not root_entries and not root_dir_info:
 
2130
                # this specified path is not present at all, skip it.
 
2131
                continue
 
2132
            path_handled = False
 
2133
            for entry in root_entries:
 
2134
                result = _process_entry(entry, root_dir_info)
 
2135
                if result is not None:
 
2136
                    path_handled = True
 
2137
                    if result is not uninteresting:
 
2138
                        yield result
 
2139
            if want_unversioned and not path_handled and root_dir_info:
 
2140
                new_executable = bool(
 
2141
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2142
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2143
                yield (None,
 
2144
                       (None, current_root_unicode),
 
2145
                       True,
 
2146
                       (False, False),
 
2147
                       (None, None),
 
2148
                       (None, splitpath(current_root_unicode)[-1]),
 
2149
                       (None, root_dir_info[2]),
 
2150
                       (None, new_executable)
 
2151
                      )
 
2152
            initial_key = (current_root, '', '')
 
2153
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2154
            if block_index == 0:
 
2155
                # we have processed the total root already, but because the
 
2156
                # initial key matched it we should skip it here.
 
2157
                block_index +=1
 
2158
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2159
                current_dir_info = None
 
2160
            else:
 
2161
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2162
                try:
 
2163
                    current_dir_info = dir_iterator.next()
 
2164
                except OSError, e:
 
2165
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2166
                    # python 2.5 has e.errno == EINVAL,
 
2167
                    #            and e.winerror == ERROR_DIRECTORY
 
2168
                    e_winerror = getattr(e, 'winerror', None)
 
2169
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2170
                    # there may be directories in the inventory even though
 
2171
                    # this path is not a file on disk: so mark it as end of
 
2172
                    # iterator
 
2173
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2174
                        current_dir_info = None
 
2175
                    elif (sys.platform == 'win32'
 
2176
                          and (e.errno in win_errors
 
2177
                               or e_winerror in win_errors)):
 
2178
                        current_dir_info = None
 
2179
                    else:
 
2180
                        raise
 
2181
                else:
 
2182
                    if current_dir_info[0][0] == '':
 
2183
                        # remove .bzr from iteration
 
2184
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2185
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2186
                        del current_dir_info[1][bzr_index]
 
2187
            # walk until both the directory listing and the versioned metadata
 
2188
            # are exhausted. 
 
2189
            if (block_index < len(state._dirblocks) and
 
2190
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2191
                current_block = state._dirblocks[block_index]
 
2192
            else:
 
2193
                current_block = None
 
2194
            while (current_dir_info is not None or
 
2195
                   current_block is not None):
 
2196
                if (current_dir_info and current_block
 
2197
                    and current_dir_info[0][0] != current_block[0]):
 
2198
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2199
                        # filesystem data refers to paths not covered by the dirblock.
 
2200
                        # this has two possibilities:
 
2201
                        # A) it is versioned but empty, so there is no block for it
 
2202
                        # B) it is not versioned.
 
2203
 
 
2204
                        # if (A) then we need to recurse into it to check for
 
2205
                        # new unknown files or directories.
 
2206
                        # if (B) then we should ignore it, because we don't
 
2207
                        # recurse into unknown directories.
 
2208
                        path_index = 0
 
2209
                        while path_index < len(current_dir_info[1]):
 
2210
                                current_path_info = current_dir_info[1][path_index]
 
2211
                                if want_unversioned:
 
2212
                                    if current_path_info[2] == 'directory':
 
2213
                                        if self.target._directory_is_tree_reference(
 
2214
                                            current_path_info[0].decode('utf8')):
 
2215
                                            current_path_info = current_path_info[:2] + \
 
2216
                                                ('tree-reference',) + current_path_info[3:]
 
2217
                                    new_executable = bool(
 
2218
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2219
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2220
                                    yield (None,
 
2221
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2222
                                        True,
 
2223
                                        (False, False),
 
2224
                                        (None, None),
 
2225
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2226
                                        (None, current_path_info[2]),
 
2227
                                        (None, new_executable))
 
2228
                                # dont descend into this unversioned path if it is
 
2229
                                # a dir
 
2230
                                if current_path_info[2] in ('directory',
 
2231
                                                            'tree-reference'):
 
2232
                                    del current_dir_info[1][path_index]
 
2233
                                    path_index -= 1
 
2234
                                path_index += 1
 
2235
 
 
2236
                        # This dir info has been handled, go to the next
 
2237
                        try:
 
2238
                            current_dir_info = dir_iterator.next()
 
2239
                        except StopIteration:
 
2240
                            current_dir_info = None
 
2241
                    else:
 
2242
                        # We have a dirblock entry for this location, but there
 
2243
                        # is no filesystem path for this. This is most likely
 
2244
                        # because a directory was removed from the disk.
 
2245
                        # We don't have to report the missing directory,
 
2246
                        # because that should have already been handled, but we
 
2247
                        # need to handle all of the files that are contained
 
2248
                        # within.
 
2249
                        for current_entry in current_block[1]:
 
2250
                            # entry referring to file not present on disk.
 
2251
                            # advance the entry only, after processing.
 
2252
                            result = _process_entry(current_entry, None)
 
2253
                            if result is not None:
 
2254
                                if result is not uninteresting:
 
2255
                                    yield result
 
2256
                        block_index +=1
 
2257
                        if (block_index < len(state._dirblocks) and
 
2258
                            osutils.is_inside(current_root,
 
2259
                                              state._dirblocks[block_index][0])):
 
2260
                            current_block = state._dirblocks[block_index]
 
2261
                        else:
 
2262
                            current_block = None
 
2263
                    continue
 
2264
                entry_index = 0
 
2265
                if current_block and entry_index < len(current_block[1]):
 
2266
                    current_entry = current_block[1][entry_index]
 
2267
                else:
 
2268
                    current_entry = None
 
2269
                advance_entry = True
 
2270
                path_index = 0
 
2271
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2272
                    current_path_info = current_dir_info[1][path_index]
 
2273
                    if current_path_info[2] == 'directory':
 
2274
                        if self.target._directory_is_tree_reference(
 
2275
                            current_path_info[0].decode('utf8')):
 
2276
                            current_path_info = current_path_info[:2] + \
 
2277
                                ('tree-reference',) + current_path_info[3:]
 
2278
                else:
 
2279
                    current_path_info = None
 
2280
                advance_path = True
 
2281
                path_handled = False
 
2282
                while (current_entry is not None or
 
2283
                    current_path_info is not None):
 
2284
                    if current_entry is None:
 
2285
                        # the check for path_handled when the path is adnvaced
 
2286
                        # will yield this path if needed.
 
2287
                        pass
 
2288
                    elif current_path_info is None:
 
2289
                        # no path is fine: the per entry code will handle it.
 
2290
                        result = _process_entry(current_entry, current_path_info)
 
2291
                        if result is not None:
 
2292
                            if result is not uninteresting:
 
2293
                                yield result
 
2294
                    elif (current_entry[0][1] != current_path_info[1]
 
2295
                          or current_entry[1][target_index][0] in 'ar'):
 
2296
                        # The current path on disk doesn't match the dirblock
 
2297
                        # record. Either the dirblock is marked as absent, or
 
2298
                        # the file on disk is not present at all in the
 
2299
                        # dirblock. Either way, report about the dirblock
 
2300
                        # entry, and let other code handle the filesystem one.
 
2301
 
 
2302
                        # Compare the basename for these files to determine
 
2303
                        # which comes first
 
2304
                        if current_path_info[1] < current_entry[0][1]:
 
2305
                            # extra file on disk: pass for now, but only
 
2306
                            # increment the path, not the entry
 
2307
                            advance_entry = False
 
2308
                        else:
 
2309
                            # entry referring to file not present on disk.
 
2310
                            # advance the entry only, after processing.
 
2311
                            result = _process_entry(current_entry, None)
 
2312
                            if result is not None:
 
2313
                                if result is not uninteresting:
 
2314
                                    yield result
 
2315
                            advance_path = False
 
2316
                    else:
 
2317
                        result = _process_entry(current_entry, current_path_info)
 
2318
                        if result is not None:
 
2319
                            path_handled = True
 
2320
                            if result is not uninteresting:
 
2321
                                yield result
 
2322
                    if advance_entry and current_entry is not None:
 
2323
                        entry_index += 1
 
2324
                        if entry_index < len(current_block[1]):
 
2325
                            current_entry = current_block[1][entry_index]
 
2326
                        else:
 
2327
                            current_entry = None
 
2328
                    else:
 
2329
                        advance_entry = True # reset the advance flaga
 
2330
                    if advance_path and current_path_info is not None:
 
2331
                        if not path_handled:
 
2332
                            # unversioned in all regards
 
2333
                            if want_unversioned:
 
2334
                                new_executable = bool(
 
2335
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2336
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2337
                                yield (None,
 
2338
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2339
                                    True,
 
2340
                                    (False, False),
 
2341
                                    (None, None),
 
2342
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2343
                                    (None, current_path_info[2]),
 
2344
                                    (None, new_executable))
 
2345
                            # dont descend into this unversioned path if it is
 
2346
                            # a dir
 
2347
                            if current_path_info[2] in ('directory'):
 
2348
                                del current_dir_info[1][path_index]
 
2349
                                path_index -= 1
 
2350
                        # dont descend the disk iterator into any tree 
 
2351
                        # paths.
 
2352
                        if current_path_info[2] == 'tree-reference':
 
2353
                            del current_dir_info[1][path_index]
 
2354
                            path_index -= 1
 
2355
                        path_index += 1
 
2356
                        if path_index < len(current_dir_info[1]):
 
2357
                            current_path_info = current_dir_info[1][path_index]
 
2358
                            if current_path_info[2] == 'directory':
 
2359
                                if self.target._directory_is_tree_reference(
 
2360
                                    current_path_info[0].decode('utf8')):
 
2361
                                    current_path_info = current_path_info[:2] + \
 
2362
                                        ('tree-reference',) + current_path_info[3:]
 
2363
                        else:
 
2364
                            current_path_info = None
 
2365
                        path_handled = False
 
2366
                    else:
 
2367
                        advance_path = True # reset the advance flagg.
 
2368
                if current_block is not None:
 
2369
                    block_index += 1
 
2370
                    if (block_index < len(state._dirblocks) and
 
2371
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2372
                        current_block = state._dirblocks[block_index]
 
2373
                    else:
 
2374
                        current_block = None
 
2375
                if current_dir_info is not None:
 
2376
                    try:
 
2377
                        current_dir_info = dir_iterator.next()
 
2378
                    except StopIteration:
 
2379
                        current_dir_info = None
 
2380
 
2255
2381
 
2256
2382
    @staticmethod
2257
2383
    def is_compatible(source, target):
2258
2384
        # the target must be a dirstate working tree
2259
 
        if not isinstance(target, DirStateWorkingTree):
 
2385
        if not isinstance(target, WorkingTree4):
2260
2386
            return False
2261
 
        # the source must be a revtree or dirstate rev tree.
 
2387
        # the source must be a revtreee or dirstate rev tree.
2262
2388
        if not isinstance(source,
2263
2389
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2264
2390
            return False
2265
2391
        # the source revid must be in the target dirstate
2266
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2392
        if not (source._revision_id == NULL_REVISION or
2267
2393
            source._revision_id in target.get_parent_ids()):
2268
 
            # TODO: what about ghosts? it may well need to
 
2394
            # TODO: what about ghosts? it may well need to 
2269
2395
            # check for them explicitly.
2270
2396
            return False
2271
2397
        return True
2281
2407
 
2282
2408
    def convert(self, tree):
2283
2409
        # 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
 
2410
        # on-unlock behaviours, and so that noone else diddles with the 
2285
2411
        # tree during upgrade.
2286
2412
        tree._control_files.lock_write()
2287
2413
        try:
2313
2439
 
2314
2440
    def update_format(self, tree):
2315
2441
        """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())
 
2442
        tree._control_files.put_utf8('format',
 
2443
            self.target_format.get_format_string())