~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Patch Queue Manager
  • Date: 2015-09-30 16:43:21 UTC
  • mfrom: (6603.2.2 fix-keep-dirty)
  • Revision ID: pqm@pqm.ubuntu.com-20150930164321-ct2v2qnmvimqt8qf
(vila) Avoid associating dirty patch headers with the previous file in the
 patch. (Colin Watson)

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
1
# Copyright (C) 2007-2012 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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
 
25
27
from cStringIO import StringIO
26
28
import os
27
29
import sys
28
30
 
29
31
from bzrlib.lazy_import import lazy_import
30
32
lazy_import(globals(), """
31
 
from bisect import bisect_left
32
 
import collections
33
 
from copy import deepcopy
34
33
import errno
35
 
import itertools
36
 
import operator
37
34
import stat
38
 
from time import time
39
 
import warnings
40
35
 
41
 
import bzrlib
42
36
from bzrlib import (
43
37
    bzrdir,
44
38
    cache_utf8,
 
39
    config,
45
40
    conflicts as _mod_conflicts,
 
41
    controldir,
46
42
    debug,
47
 
    delta,
48
43
    dirstate,
49
44
    errors,
 
45
    filters as _mod_filters,
50
46
    generate_ids,
51
 
    globbing,
52
 
    ignores,
53
 
    merge,
54
47
    osutils,
55
48
    revision as _mod_revision,
56
49
    revisiontree,
57
 
    textui,
58
50
    trace,
59
51
    transform,
60
 
    urlutils,
61
 
    xml5,
62
 
    xml6,
 
52
    views,
63
53
    )
64
 
import bzrlib.branch
65
 
from bzrlib.transport import get_transport
66
 
import bzrlib.ui
67
54
""")
68
55
 
69
 
from bzrlib import symbol_versioning
70
56
from bzrlib.decorators import needs_read_lock, needs_write_lock
71
 
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
72
 
from bzrlib.lockable_files import LockableFiles, TransportLock
 
57
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
58
from bzrlib.lock import LogicalLockResult
 
59
from bzrlib.lockable_files import LockableFiles
73
60
from bzrlib.lockdir import LockDir
74
 
import bzrlib.mutabletree
75
 
from bzrlib.mutabletree import needs_tree_write_lock
 
61
from bzrlib.mutabletree import (
 
62
    MutableTree,
 
63
    needs_tree_write_lock,
 
64
    )
76
65
from bzrlib.osutils import (
77
66
    file_kind,
78
67
    isdir,
79
 
    normpath,
80
68
    pathjoin,
81
 
    rand_chars,
82
69
    realpath,
83
70
    safe_unicode,
84
 
    splitpath,
85
 
    )
86
 
from bzrlib.trace import mutter, note
 
71
    )
 
72
from bzrlib.symbol_versioning import (
 
73
    deprecated_in,
 
74
    deprecated_method,
 
75
    )
87
76
from bzrlib.transport.local import LocalTransport
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
 
    """
 
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):
119
89
 
120
90
    def __init__(self, basedir,
121
91
                 branch,
131
101
        """
132
102
        self._format = _format
133
103
        self.bzrdir = _bzrdir
134
 
        assert isinstance(basedir, basestring), \
135
 
            "base directory %r is not a string" % basedir
136
104
        basedir = safe_unicode(basedir)
137
 
        mutter("opening working tree %r", basedir)
 
105
        trace.mutter("opening working tree %r", basedir)
138
106
        self._branch = branch
139
 
        assert isinstance(self.branch, bzrlib.branch.Branch), \
140
 
            "branch %r is not a Branch" % self.branch
141
107
        self.basedir = realpath(basedir)
142
108
        # if branch is at our basedir and is a format 6 or less
143
109
        # 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
146
110
        self._control_files = _control_files
 
111
        self._transport = self._control_files._transport
147
112
        self._dirty = None
148
113
        #-------------
149
114
        # during a read or write lock these objects are set, and are
151
116
        self._dirstate = None
152
117
        self._inventory = None
153
118
        #-------------
 
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
154
125
 
155
126
    @needs_tree_write_lock
156
127
    def _add(self, files, ids, kinds):
171
142
            state.add(f, file_id, kind, None, '')
172
143
        self._make_dirty(reset_inventory=True)
173
144
 
 
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
 
174
149
    def _make_dirty(self, reset_inventory):
175
150
        """Make the tree state dirty.
176
151
 
184
159
    @needs_tree_write_lock
185
160
    def add_reference(self, sub_tree):
186
161
        # use standard implementation, which calls back to self._add
187
 
        # 
 
162
        #
188
163
        # So we don't store the reference_revision in the working dirstate,
189
 
        # it's just recorded at the moment of commit. 
 
164
        # it's just recorded at the moment of commit.
190
165
        self._add_reference(sub_tree)
191
166
 
192
167
    def break_lock(self):
228
203
 
229
204
    def _comparison_data(self, entry, path):
230
205
        kind, executable, stat_value = \
231
 
            WorkingTree3._comparison_data(self, entry, path)
 
206
            WorkingTree._comparison_data(self, entry, path)
232
207
        # it looks like a plain directory, but it's really a reference -- see
233
208
        # also kind()
234
 
        if (self._repo_supports_tree_reference and
235
 
            kind == 'directory' and
236
 
            self._directory_is_tree_reference(path)):
 
209
        if (self._repo_supports_tree_reference and kind == 'directory'
 
210
            and entry is not None and entry.kind == 'tree-reference'):
237
211
            kind = 'tree-reference'
238
212
        return kind, executable, stat_value
239
213
 
241
215
    def commit(self, message=None, revprops=None, *args, **kwargs):
242
216
        # mark the tree as dirty post commit - commit
243
217
        # can change the current versioned list by doing deletes.
244
 
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
218
        result = WorkingTree.commit(self, message, revprops, *args, **kwargs)
245
219
        self._make_dirty(reset_inventory=True)
246
220
        return result
247
221
 
265
239
            return self._dirstate
266
240
        local_path = self.bzrdir.get_workingtree_transport(None
267
241
            ).local_abspath('dirstate')
268
 
        self._dirstate = dirstate.DirState.on_file(local_path)
 
242
        self._dirstate = dirstate.DirState.on_file(local_path,
 
243
            self._sha1_provider(), self._worth_saving_limit())
269
244
        return self._dirstate
270
245
 
 
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
 
271
266
    def filter_unversioned_files(self, paths):
272
267
        """Filter out paths that are versioned.
273
268
 
305
300
 
306
301
    def _generate_inventory(self):
307
302
        """Create and set self.inventory from the dirstate object.
308
 
        
 
303
 
309
304
        This is relatively expensive: we have to walk the entire dirstate.
310
305
        Ideally we would not, and can deprecate this function.
311
306
        """
315
310
        state._read_dirblocks_if_needed()
316
311
        root_key, current_entry = self._get_entry(path='')
317
312
        current_id = root_key[2]
318
 
        assert current_entry[0][0] == 'd' # directory
 
313
        if not (current_entry[0][0] == 'd'): # directory
 
314
            raise AssertionError(current_entry)
319
315
        inv = Inventory(root_id=current_id)
320
316
        # Turn some things into local variables
321
317
        minikind_to_kind = dirstate.DirState._minikind_to_kind
354
350
                    # add this entry to the parent map.
355
351
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
356
352
                elif kind == 'tree-reference':
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)
 
353
                    if not self._repo_supports_tree_reference:
 
354
                        raise errors.UnsupportedOperation(
 
355
                            self._generate_inventory,
 
356
                            self.branch.repository)
362
357
                    inv_entry.reference_revision = link_or_sha1 or None
363
358
                elif kind != 'symlink':
364
359
                    raise AssertionError("unknown kind %r" % kind)
365
360
                # These checks cost us around 40ms on a 55k entry tree
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
 
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,))
369
367
                inv_byid[file_id] = inv_entry
370
368
                parent_ie.children[name_unicode] = inv_entry
371
369
        self._inventory = inv
376
374
        If either file_id or path is supplied, it is used as the key to lookup.
377
375
        If both are supplied, the fastest lookup is used, and an error is
378
376
        raised if they do not both point at the same row.
379
 
        
 
377
 
380
378
        :param file_id: An optional unicode file_id to be looked up.
381
379
        :param path: An optional unicode path to be looked up.
382
380
        :return: The dirstate row tuple for path/file_id, or (None, None)
400
398
        state = self.current_dirstate()
401
399
        if stat_value is None:
402
400
            try:
403
 
                stat_value = os.lstat(file_abspath)
 
401
                stat_value = osutils.lstat(file_abspath)
404
402
            except OSError, e:
405
403
                if e.errno == errno.ENOENT:
406
404
                    return None
407
405
                else:
408
406
                    raise
409
 
        link_or_sha1 = state.update_entry(entry, file_abspath,
410
 
                                          stat_value=stat_value)
 
407
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
 
408
            stat_value=stat_value)
411
409
        if entry[1][0][0] == 'f':
412
 
            return link_or_sha1
 
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
413
420
        return None
414
421
 
415
 
    def _get_inventory(self):
 
422
    def _get_root_inventory(self):
416
423
        """Get the inventory for the tree. This is only valid within a lock."""
417
424
        if 'evil' in debug.debug_flags:
418
425
            trace.mutter_callsite(2,
423
430
        self._generate_inventory()
424
431
        return self._inventory
425
432
 
 
433
    @deprecated_method(deprecated_in((2, 5, 0)))
 
434
    def _get_inventory(self):
 
435
        return self.root_inventory
 
436
 
426
437
    inventory = property(_get_inventory,
427
438
                         doc="Inventory of this Tree")
428
439
 
 
440
    root_inventory = property(_get_root_inventory,
 
441
        "Root inventory of this tree")
 
442
 
429
443
    @needs_read_lock
430
444
    def get_parent_ids(self):
431
445
        """See Tree.get_parent_ids.
432
 
        
 
446
 
433
447
        This implementation requests the ids list from the dirstate file.
434
448
        """
435
449
        return self.current_dirstate().get_parent_ids()
457
471
        return osutils.lexists(pathjoin(
458
472
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
459
473
 
 
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
 
460
479
    @needs_read_lock
461
480
    def id2path(self, file_id):
462
481
        "Convert a file-id to a path."
473
492
            return False # Missing entries are not executable
474
493
        return entry[1][0][3] # Executable?
475
494
 
476
 
    if not osutils.supports_executable():
477
 
        def is_executable(self, file_id, path=None):
478
 
            """Test if a file is executable or not.
 
495
    def is_executable(self, file_id, path=None):
 
496
        """Test if a file is executable or not.
479
497
 
480
 
            Note: The caller is expected to take a read-lock before calling this.
481
 
            """
 
498
        Note: The caller is expected to take a read-lock before calling this.
 
499
        """
 
500
        if not self._supports_executable():
482
501
            entry = self._get_entry(file_id=file_id, path=path)
483
502
            if entry == (None, None):
484
503
                return False
485
504
            return entry[1][0][3]
486
 
 
487
 
        _is_executable_from_path_and_stat = \
488
 
            _is_executable_from_path_and_stat_from_basis
489
 
    else:
490
 
        def is_executable(self, file_id, path=None):
491
 
            """Test if a file is executable or not.
492
 
 
493
 
            Note: The caller is expected to take a read-lock before calling this.
494
 
            """
 
505
        else:
 
506
            self._must_be_locked()
495
507
            if not path:
496
508
                path = self.id2path(file_id)
497
 
            mode = os.lstat(self.abspath(path)).st_mode
 
509
            mode = osutils.lstat(self.abspath(path)).st_mode
498
510
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
499
511
 
 
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
 
500
522
    @needs_read_lock
501
523
    def __iter__(self):
502
524
        """Iterate through file_ids for this tree.
515
537
        return iter(result)
516
538
 
517
539
    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
518
544
        for key, tree_details in self.current_dirstate()._iter_entries():
519
545
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
520
546
                # not relevant to the working tree
522
548
            if not key[1]:
523
549
                # the root is not a reference.
524
550
                continue
525
 
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
551
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
526
552
            try:
527
 
                if self._kind(path) == 'tree-reference':
528
 
                    yield path, key[2]
 
553
                if self._kind(relpath) == 'tree-reference':
 
554
                    yield relpath, key[2]
529
555
            except errors.NoSuchFile:
530
556
                # path is missing on disk.
531
557
                continue
532
558
 
 
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
 
533
565
    def kind(self, file_id):
534
566
        """Return the kind of a file.
535
567
 
539
571
        Note: The caller is expected to take a read-lock before calling this.
540
572
        """
541
573
        relpath = self.id2path(file_id)
542
 
        assert relpath != None, \
543
 
            "path for id {%s} is None!" % file_id
 
574
        if relpath is None:
 
575
            raise AssertionError(
 
576
                "path for id {%s} is None!" % file_id)
544
577
        return self._kind(relpath)
545
578
 
546
579
    def _kind(self, relpath):
547
580
        abspath = self.abspath(relpath)
548
581
        kind = file_kind(abspath)
549
 
        if (self._repo_supports_tree_reference and
550
 
            kind == 'directory' and
551
 
            self._directory_is_tree_reference(relpath)):
552
 
            kind = 'tree-reference'
 
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'
553
587
        return kind
554
588
 
555
589
    @needs_read_lock
562
596
            return _mod_revision.NULL_REVISION
563
597
 
564
598
    def lock_read(self):
565
 
        """See Branch.lock_read, and WorkingTree.unlock."""
 
599
        """See Branch.lock_read, and WorkingTree.unlock.
 
600
 
 
601
        :return: A bzrlib.lock.LogicalLockResult.
 
602
        """
566
603
        self.branch.lock_read()
567
604
        try:
568
605
            self._control_files.lock_read()
581
618
        except:
582
619
            self.branch.unlock()
583
620
            raise
 
621
        return LogicalLockResult(self.unlock)
584
622
 
585
623
    def _lock_self_write(self):
586
624
        """This should be called after the branch is locked."""
601
639
        except:
602
640
            self.branch.unlock()
603
641
            raise
 
642
        return LogicalLockResult(self.unlock)
604
643
 
605
644
    def lock_tree_write(self):
606
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
645
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
 
646
 
 
647
        :return: A bzrlib.lock.LogicalLockResult.
 
648
        """
607
649
        self.branch.lock_read()
608
 
        self._lock_self_write()
 
650
        return self._lock_self_write()
609
651
 
610
652
    def lock_write(self):
611
 
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
653
        """See MutableTree.lock_write, and WorkingTree.unlock.
 
654
 
 
655
        :return: A bzrlib.lock.LogicalLockResult.
 
656
        """
612
657
        self.branch.lock_write()
613
 
        self._lock_self_write()
 
658
        return self._lock_self_write()
614
659
 
615
660
    @needs_tree_write_lock
616
661
    def move(self, from_paths, to_dir, after=False):
618
663
        result = []
619
664
        if not from_paths:
620
665
            return result
621
 
 
622
666
        state = self.current_dirstate()
623
 
 
624
 
        assert not isinstance(from_paths, basestring)
 
667
        if isinstance(from_paths, basestring):
 
668
            raise ValueError()
625
669
        to_dir_utf8 = to_dir.encode('utf8')
626
670
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
627
671
        id_index = state._get_id_index()
648
692
 
649
693
        if self._inventory is not None:
650
694
            update_inventory = True
651
 
            inv = self.inventory
 
695
            inv = self.root_inventory
 
696
            to_dir_id = to_entry[0][2]
652
697
            to_dir_ie = inv[to_dir_id]
653
 
            to_dir_id = to_entry[0][2]
654
698
        else:
655
699
            update_inventory = False
656
700
 
687
731
            from_entry = self._get_entry(path=from_rel)
688
732
            if from_entry == (None, None):
689
733
                raise errors.BzrMoveFailedError(from_rel,to_dir,
690
 
                    errors.NotVersionedError(path=str(from_rel)))
 
734
                    errors.NotVersionedError(path=from_rel))
691
735
 
692
736
            from_id = from_entry[0][2]
693
737
            to_rel = pathjoin(to_dir, from_tail)
720
764
                if from_missing: # implicitly just update our path mapping
721
765
                    move_file = False
722
766
                elif not after:
723
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
724
 
                        extra="(Use --after to update the Bazaar id)")
 
767
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
725
768
 
726
769
            rollbacks = []
727
770
            def rollback_rename():
783
826
                if minikind == 'd':
784
827
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
785
828
                        """Recursively update all entries in this dirblock."""
786
 
                        assert from_dir != '', "renaming root not supported"
 
829
                        if from_dir == '':
 
830
                            raise AssertionError("renaming root not supported")
787
831
                        from_key = (from_dir, '')
788
832
                        from_block_idx, present = \
789
833
                            state._find_block_index_from_key(from_key)
802
846
 
803
847
                        # Grab a copy since move_one may update the list.
804
848
                        for entry in from_block[1][:]:
805
 
                            assert entry[0][0] == from_dir
 
849
                            if not (entry[0][0] == from_dir):
 
850
                                raise AssertionError()
806
851
                            cur_details = entry[1][0]
807
852
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
808
853
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
833
878
                rollback_rename()
834
879
                raise
835
880
            result.append((from_rel, to_rel))
836
 
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
881
            state._mark_modified()
837
882
            self._make_dirty(reset_inventory=False)
838
883
 
839
884
        return result
849
894
    @needs_read_lock
850
895
    def path2id(self, path):
851
896
        """Return the id for path in this tree."""
 
897
        if isinstance(path, list):
 
898
            if path == []:
 
899
                path = [""]
 
900
            path = osutils.pathjoin(*path)
852
901
        path = path.strip('/')
853
902
        entry = self._get_entry(path=path)
854
903
        if entry == (None, None):
867
916
        for tree in trees:
868
917
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
869
918
                parents):
870
 
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
919
                return super(DirStateWorkingTree, self).paths2ids(paths,
 
920
                    trees, require_versioned)
871
921
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
872
922
        # -- make all paths utf8 --
873
923
        paths_utf8 = set()
931
981
                    all_versioned = False
932
982
                    break
933
983
            if not all_versioned:
934
 
                raise errors.PathsNotVersionedError(paths)
 
984
                raise errors.PathsNotVersionedError(
 
985
                    [p.decode('utf-8') for p in paths])
935
986
        # -- remove redundancy in supplied paths to prevent over-scanning --
936
987
        search_paths = osutils.minimum_path_selection(paths)
937
 
        # sketch: 
 
988
        # sketch:
938
989
        # for all search_indexs in each path at or under each element of
939
990
        # search_paths, if the detail is relocated: add the id, and add the
940
991
        # relocated path as one to search if its not searched already. If the
986
1037
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
987
1038
            for dir_name in split_paths:
988
1039
                if dir_name not in found_dir_names:
989
 
                    raise errors.PathsNotVersionedError(paths)
 
1040
                    raise errors.PathsNotVersionedError(
 
1041
                        [p.decode('utf-8') for p in paths])
990
1042
 
991
1043
        for dir_name_id, trees_info in found.iteritems():
992
1044
            for index in search_indexes:
996
1048
 
997
1049
    def read_working_inventory(self):
998
1050
        """Read the working inventory.
999
 
        
 
1051
 
1000
1052
        This is a meaningless operation for dirstate, but we obey it anyhow.
1001
1053
        """
1002
 
        return self.inventory
 
1054
        return self.root_inventory
1003
1055
 
1004
1056
    @needs_read_lock
1005
1057
    def revision_tree(self, revision_id):
1020
1072
    def set_last_revision(self, new_revision):
1021
1073
        """Change the last revision in the working tree."""
1022
1074
        parents = self.get_parent_ids()
1023
 
        if new_revision in (NULL_REVISION, None):
1024
 
            assert len(parents) < 2, (
1025
 
                "setting the last parent to none with a pending merge is "
1026
 
                "unsupported.")
 
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.")
1027
1080
            self.set_parent_ids([])
1028
1081
        else:
1029
1082
            self.set_parent_ids([new_revision] + parents[1:],
1032
1085
    @needs_tree_write_lock
1033
1086
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1034
1087
        """Set the parent ids to revision_ids.
1035
 
        
 
1088
 
1036
1089
        See also set_parent_trees. This api will try to retrieve the tree data
1037
1090
        for each element of revision_ids from the trees repository. If you have
1038
1091
        tree data already available, it is more efficient to use
1070
1123
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1071
1124
        real_trees = []
1072
1125
        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
 
1073
1132
        # convert absent trees to the null tree, which we convert back to
1074
1133
        # missing on access.
1075
1134
        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
1076
1141
            _mod_revision.check_not_reserved_id(rev_id)
1077
1142
            if tree is not None:
1078
1143
                real_trees.append((rev_id, tree))
1079
1144
            else:
1080
1145
                real_trees.append((rev_id,
1081
 
                    self.branch.repository.revision_tree(None)))
 
1146
                    self.branch.repository.revision_tree(
 
1147
                        _mod_revision.NULL_REVISION)))
1082
1148
                ghosts.append(rev_id)
1083
 
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
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)
1084
1174
        self._make_dirty(reset_inventory=False)
1085
1175
 
1086
1176
    def _set_root_id(self, file_id):
1106
1196
 
1107
1197
    def unlock(self):
1108
1198
        """Unlock in format 4 trees needs to write the entire dirstate."""
1109
 
        # do non-implementation specific cleanup
1110
 
        self._cleanup()
1111
 
 
1112
1199
        if self._control_files._lock_count == 1:
 
1200
            # do non-implementation specific cleanup
 
1201
            self._cleanup()
 
1202
 
1113
1203
            # eventually we should do signature checking during read locks for
1114
1204
            # dirstate updates.
1115
1205
            if self._control_files._lock_mode == 'w':
1178
1268
                # just forget the whole block.
1179
1269
                entry_index = 0
1180
1270
                while entry_index < len(block[1]):
1181
 
                    # Mark this file id as having been removed
1182
1271
                    entry = block[1][entry_index]
1183
 
                    ids_to_unversion.discard(entry[0][2])
1184
 
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
1185
 
                                               # entries
1186
 
                        or not state._make_absent(entry)):
 
1272
                    if entry[1][0][0] in 'ar':
 
1273
                        # don't remove absent or renamed entries
1187
1274
                        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
1188
1281
                # go to the next block. (At the moment we dont delete empty
1189
1282
                # dirblocks)
1190
1283
                block_index += 1
1211
1304
        # have to change the legacy inventory too.
1212
1305
        if self._inventory is not None:
1213
1306
            for file_id in file_ids:
1214
 
                self._inventory.remove_recursive_id(file_id)
 
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)
1215
1328
 
1216
1329
    @needs_read_lock
1217
1330
    def _validate(self):
1220
1333
    @needs_tree_write_lock
1221
1334
    def _write_inventory(self, inv):
1222
1335
        """Write inventory as the current inventory."""
1223
 
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
1224
 
        self.current_dirstate().set_state_from_inventory(inv)
1225
 
        self._make_dirty(reset_inventory=False)
1226
 
        if self._inventory is not None:
 
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:
1227
1350
            self._inventory = inv
1228
1351
        self.flush()
1229
1352
 
1230
 
 
1231
 
class WorkingTreeFormat4(WorkingTreeFormat3):
1232
 
    """The first consolidated dirstate working tree format.
1233
 
 
1234
 
    This format:
1235
 
        - exists within a metadir controlling .bzr
1236
 
        - includes an explicit version marker for the workingtree control
1237
 
          files, separate from the BzrDir format
1238
 
        - modifies the hash cache format
1239
 
        - is new in bzr 0.15
1240
 
        - uses a LockDir to guard access to it.
1241
 
    """
1242
 
 
1243
 
    upgrade_recommended = False
1244
 
 
1245
 
    def get_format_string(self):
1246
 
        """See WorkingTreeFormat.get_format_string()."""
1247
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1248
 
 
1249
 
    def get_format_description(self):
1250
 
        """See WorkingTreeFormat.get_format_description()."""
1251
 
        return "Working tree format 4"
1252
 
 
1253
 
    def initialize(self, a_bzrdir, revision_id=None):
 
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):
1254
1480
        """See WorkingTreeFormat.initialize().
1255
1481
 
1256
1482
        :param revision_id: allows creating a working tree at a different
1257
 
        revision than the branch is at.
 
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.
1258
1490
 
1259
1491
        These trees get an initial random root id, if their repository supports
1260
1492
        rich root data, TREE_ROOT otherwise.
1265
1497
        control_files = self._open_control_files(a_bzrdir)
1266
1498
        control_files.create_lock()
1267
1499
        control_files.lock_write()
1268
 
        control_files.put_utf8('format', self.get_format_string())
1269
 
        branch = a_bzrdir.open_branch()
 
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()
1270
1506
        if revision_id is None:
1271
1507
            revision_id = branch.last_revision()
1272
1508
        local_path = transport.local_abspath('dirstate')
1274
1510
        state = dirstate.DirState.initialize(local_path)
1275
1511
        state.unlock()
1276
1512
        del state
1277
 
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1513
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1278
1514
                         branch,
1279
1515
                         _format=self,
1280
1516
                         _bzrdir=a_bzrdir,
1282
1518
        wt._new_tree()
1283
1519
        wt.lock_tree_write()
1284
1520
        try:
1285
 
            if revision_id in (None, NULL_REVISION):
 
1521
            self._init_custom_control_files(wt)
 
1522
            if revision_id in (None, _mod_revision.NULL_REVISION):
1286
1523
                if branch.repository.supports_rich_root():
1287
1524
                    wt._set_root_id(generate_ids.gen_root_id())
1288
1525
                else:
1289
1526
                    wt._set_root_id(ROOT_ID)
1290
1527
                wt.flush()
1291
 
            wt.set_last_revision(revision_id)
1292
 
            wt.flush()
1293
 
            basis = wt.basis_tree()
 
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)]
1294
1543
            basis.lock_read()
1295
 
            # if the basis has a root id we have to use that; otherwise we use
1296
 
            # a new random one
1297
 
            basis_root_id = basis.get_root_id()
1298
 
            if basis_root_id is not None:
1299
 
                wt._set_root_id(basis_root_id)
 
1544
            try:
 
1545
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
1300
1546
                wt.flush()
1301
 
            transform.build_tree(basis, wt)
1302
 
            basis.unlock()
 
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()
1303
1570
        finally:
1304
1571
            control_files.unlock()
1305
1572
            wt.unlock()
1306
1573
        return wt
1307
1574
 
 
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
 
1308
1598
    def _open(self, a_bzrdir, control_files):
1309
1599
        """Open the tree itself.
1310
1600
 
1311
1601
        :param a_bzrdir: the dir for the tree.
1312
1602
        :param control_files: the control files for the tree.
1313
1603
        """
1314
 
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1604
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
1315
1605
                           branch=a_bzrdir.open_branch(),
1316
1606
                           _format=self,
1317
1607
                           _bzrdir=a_bzrdir,
1318
1608
                           _control_files=control_files)
1319
1609
 
1320
1610
    def __get_matchingbzrdir(self):
 
1611
        return self._get_matchingbzrdir()
 
1612
 
 
1613
    def _get_matchingbzrdir(self):
 
1614
        """Overrideable method to get a bzrdir for testing."""
1321
1615
        # please test against something that will let us do tree references
1322
 
        return bzrdir.format_registry.make_bzrdir(
1323
 
            'dirstate-with-subtree')
 
1616
        return controldir.format_registry.make_bzrdir(
 
1617
            'development-subtree')
1324
1618
 
1325
1619
    _matchingbzrdir = property(__get_matchingbzrdir)
1326
1620
 
1327
1621
 
1328
 
class DirStateRevisionTree(Tree):
1329
 
    """A revision tree pulling the inventory from a dirstate."""
 
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
    """
1330
1709
 
1331
1710
    def __init__(self, dirstate, revision_id, repository):
1332
1711
        self._dirstate = dirstate
1335
1714
        self._inventory = None
1336
1715
        self._locked = 0
1337
1716
        self._dirstate_locked = False
 
1717
        self._repo_supports_tree_reference = getattr(
 
1718
            repository._format, "supports_tree_reference",
 
1719
            False)
1338
1720
 
1339
1721
    def __repr__(self):
1340
1722
        return "<%s of %s in %s>" % \
1343
1725
    def annotate_iter(self, file_id,
1344
1726
                      default_revision=_mod_revision.CURRENT_REVISION):
1345
1727
        """See Tree.annotate_iter"""
1346
 
        w = self._get_weave(file_id)
1347
 
        return w.annotate_iter(self.inventory[file_id].revision)
 
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]
1348
1731
 
1349
 
    def _get_ancestors(self, default_revision):
1350
 
        return set(self._repository.get_ancestry(self._revision_id,
1351
 
                                                 topo_sorted=False))
1352
1732
    def _comparison_data(self, entry, path):
1353
1733
        """See Tree._comparison_data."""
1354
1734
        if entry is None:
1371
1751
    def get_root_id(self):
1372
1752
        return self.path2id('')
1373
1753
 
 
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
 
1374
1770
    def _get_parent_index(self):
1375
1771
        """Return the index in the dirstate referenced by this tree."""
1376
1772
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1381
1777
        If either file_id or path is supplied, it is used as the key to lookup.
1382
1778
        If both are supplied, the fastest lookup is used, and an error is
1383
1779
        raised if they do not both point at the same row.
1384
 
        
 
1780
 
1385
1781
        :param file_id: An optional unicode file_id to be looked up.
1386
1782
        :param path: An optional unicode path to be looked up.
1387
1783
        :return: The dirstate row tuple for path/file_id, or (None, None)
1391
1787
        if path is not None:
1392
1788
            path = path.encode('utf8')
1393
1789
        parent_index = self._get_parent_index()
1394
 
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1790
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id,
 
1791
            path_utf8=path)
1395
1792
 
1396
1793
    def _generate_inventory(self):
1397
1794
        """Create and set self.inventory from the dirstate object.
1401
1798
 
1402
1799
        This is relatively expensive: we have to walk the entire dirstate.
1403
1800
        """
1404
 
        assert self._locked, 'cannot generate inventory of an unlocked '\
1405
 
            'dirstate revision tree'
 
1801
        if not self._locked:
 
1802
            raise AssertionError(
 
1803
                'cannot generate inventory of an unlocked '
 
1804
                'dirstate revision tree')
1406
1805
        # separate call for profiling - makes it clear where the costs are.
1407
1806
        self._dirstate._read_dirblocks_if_needed()
1408
 
        assert self._revision_id in self._dirstate.get_parent_ids(), \
1409
 
            'parent %s has disappeared from %s' % (
1410
 
            self._revision_id, self._dirstate.get_parent_ids())
 
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()))
1411
1811
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1412
1812
        # This is identical now to the WorkingTree _generate_inventory except
1413
1813
        # for the tree index use.
1414
1814
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1415
1815
        current_id = root_key[2]
1416
 
        assert current_entry[parent_index][0] == 'd'
 
1816
        if current_entry[parent_index][0] != 'd':
 
1817
            raise AssertionError()
1417
1818
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1418
1819
        inv.root.revision = current_entry[parent_index][4]
1419
1820
        # Turn some things into local variables
1450
1851
                elif kind == 'directory':
1451
1852
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1452
1853
                elif kind == 'symlink':
1453
 
                    inv_entry.executable = False
1454
 
                    inv_entry.text_size = None
1455
1854
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1456
1855
                elif kind == 'tree-reference':
1457
1856
                    inv_entry.reference_revision = fingerprint or None
1459
1858
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1460
1859
                            % entry)
1461
1860
                # These checks cost us around 40ms on a 55k entry tree
1462
 
                assert file_id not in inv_byid
1463
 
                assert name_unicode not in parent_ie.children
 
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,))
1464
1867
                inv_byid[file_id] = inv_entry
1465
1868
                parent_ie.children[name_unicode] = inv_entry
1466
1869
        self._inventory = inv
1473
1876
        # Make sure the file exists
1474
1877
        entry = self._get_entry(file_id, path=path)
1475
1878
        if entry == (None, None): # do we raise?
1476
 
            return None
 
1879
            raise errors.NoSuchId(self, file_id)
1477
1880
        parent_index = self._get_parent_index()
1478
1881
        last_changed_revision = entry[1][parent_index][4]
1479
 
        return self._repository.get_revision(last_changed_revision).timestamp
 
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
1480
1887
 
1481
1888
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1482
1889
        entry = self._get_entry(file_id=file_id, path=path)
1486
1893
            return parent_details[1]
1487
1894
        return None
1488
1895
 
1489
 
    @symbol_versioning.deprecated_method(symbol_versioning.zero_ninety)
1490
 
    def get_weave(self, file_id):
1491
 
        return self._get_weave(file_id)
1492
 
 
1493
 
    def _get_weave(self, file_id):
1494
 
        return self._repository.weave_store.get_weave(file_id,
1495
 
                self._repository.get_transaction())
 
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
1496
1900
 
1497
1901
    def get_file(self, file_id, path=None):
1498
1902
        return StringIO(self.get_file_text(file_id))
1499
1903
 
1500
 
    def get_file_lines(self, file_id):
1501
 
        ie = self.inventory[file_id]
1502
 
        return self._get_weave(file_id).get_lines(ie.revision)
1503
 
 
1504
1904
    def get_file_size(self, file_id):
1505
 
        return self.inventory[file_id].text_size
 
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
1506
1908
 
1507
 
    def get_file_text(self, file_id):
1508
 
        return ''.join(self.get_file_lines(file_id))
 
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
1509
1922
 
1510
1923
    def get_reference_revision(self, file_id, path=None):
1511
 
        return self.inventory[file_id].reference_revision
 
1924
        inv, inv_file_id = self._unpack_file_id(file_id)
 
1925
        return inv[inv_file_id].reference_revision
1512
1926
 
1513
1927
    def iter_files_bytes(self, desired_files):
1514
1928
        """See Tree.iter_files_bytes.
1524
1938
                                       identifier))
1525
1939
        return self._repository.iter_files_bytes(repo_desired_files)
1526
1940
 
1527
 
    def get_symlink_target(self, file_id):
 
1941
    def get_symlink_target(self, file_id, path=None):
1528
1942
        entry = self._get_entry(file_id=file_id)
1529
1943
        parent_index = self._get_parent_index()
1530
1944
        if entry[1][parent_index][0] != 'l':
1531
1945
            return None
1532
1946
        else:
1533
 
            # At present, none of the tree implementations supports non-ascii
1534
 
            # symlink targets. So we will just assume that the dirstate path is
1535
 
            # correct.
1536
 
            return entry[1][parent_index][1]
 
1947
            target = entry[1][parent_index][1]
 
1948
            target = target.decode('utf8')
 
1949
            return target
1537
1950
 
1538
1951
    def get_revision_id(self):
1539
1952
        """Return the revision id for this tree."""
1540
1953
        return self._revision_id
1541
1954
 
1542
 
    def _get_inventory(self):
 
1955
    def _get_root_inventory(self):
1543
1956
        if self._inventory is not None:
1544
1957
            return self._inventory
1545
1958
        self._must_be_locked()
1546
1959
        self._generate_inventory()
1547
1960
        return self._inventory
1548
1961
 
 
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
 
1549
1969
    inventory = property(_get_inventory,
1550
1970
                         doc="Inventory of this Tree")
1551
1971
 
1557
1977
        return bool(self.path2id(filename))
1558
1978
 
1559
1979
    def kind(self, file_id):
1560
 
        return self.inventory[file_id].kind
 
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)
1561
1989
 
1562
1990
    def path_content_summary(self, path):
1563
1991
        """See Tree.path_content_summary."""
1564
 
        id = self.inventory.path2id(path)
1565
 
        if id is None:
 
1992
        inv, inv_file_id = self._path2inv_file_id(path)
 
1993
        if inv_file_id is None:
1566
1994
            return ('missing', None, None, None)
1567
 
        entry = self._inventory[id]
 
1995
        entry = inv[inv_file_id]
1568
1996
        kind = entry.kind
1569
1997
        if kind == 'file':
1570
1998
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1574
2002
            return (kind, None, None, None)
1575
2003
 
1576
2004
    def is_executable(self, file_id, path=None):
1577
 
        ie = self.inventory[file_id]
 
2005
        inv, inv_file_id = self._unpack_file_id(file_id)
 
2006
        ie = inv[inv_file_id]
1578
2007
        if ie.kind != "file":
1579
 
            return None
 
2008
            return False
1580
2009
        return ie.executable
1581
2010
 
1582
 
    def list_files(self, include_root=False):
 
2011
    def is_locked(self):
 
2012
        return self._locked
 
2013
 
 
2014
    def list_files(self, include_root=False, from_dir=None, recursive=True):
1583
2015
        # We use a standard implementation, because DirStateRevisionTree is
1584
2016
        # dealing with one of the parents of the current state
1585
 
        inv = self._get_inventory()
1586
 
        entries = inv.iter_entries()
1587
 
        if self.inventory.root is not None and not include_root:
 
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:
1588
2028
            entries.next()
1589
2029
        for path, entry in entries:
1590
2030
            yield path, 'V', entry.kind, entry.file_id, entry
1591
2031
 
1592
2032
    def lock_read(self):
1593
 
        """Lock the tree for a set of operations."""
 
2033
        """Lock the tree for a set of operations.
 
2034
 
 
2035
        :return: A bzrlib.lock.LogicalLockResult.
 
2036
        """
1594
2037
        if not self._locked:
1595
2038
            self._repository.lock_read()
1596
2039
            if self._dirstate._lock_token is None:
1597
2040
                self._dirstate.lock_read()
1598
2041
                self._dirstate_locked = True
1599
2042
        self._locked += 1
 
2043
        return LogicalLockResult(self.unlock)
1600
2044
 
1601
2045
    def _must_be_locked(self):
1602
2046
        if not self._locked:
1606
2050
    def path2id(self, path):
1607
2051
        """Return the id for path in this tree."""
1608
2052
        # 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)
1609
2057
        entry = self._get_entry(path=path)
1610
2058
        if entry == (None, None):
1611
2059
            return None
1623
2071
                self._dirstate_locked = False
1624
2072
            self._repository.unlock()
1625
2073
 
 
2074
    @needs_read_lock
 
2075
    def supports_tree_reference(self):
 
2076
        return self._repo_supports_tree_reference
 
2077
 
1626
2078
    def walkdirs(self, prefix=""):
1627
2079
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1628
 
        # implementation based on an inventory.  
 
2080
        # implementation based on an inventory.
1629
2081
        # This should be cleaned up to use the much faster Dirstate code
1630
2082
        # So for now, we just build up the parent inventory, and extract
1631
2083
        # it the same way RevisionTree does.
1632
2084
        _directory = 'directory'
1633
 
        inv = self._get_inventory()
 
2085
        inv = self._get_root_inventory()
1634
2086
        top_id = inv.path2id(prefix)
1635
2087
        if top_id is None:
1636
2088
            pending = []
1660
2112
 
1661
2113
class InterDirStateTree(InterTree):
1662
2114
    """Fast path optimiser for changes_from with dirstate trees.
1663
 
    
1664
 
    This is used only when both trees are in the dirstate working file, and 
1665
 
    the source is any parent within the dirstate, and the destination is 
 
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
1666
2118
    the current working tree of the same dirstate.
1667
2119
    """
1668
2120
    # this could be generalized to allow comparisons between any trees in the
1677
2129
    def make_source_parent_tree(source, target):
1678
2130
        """Change the source tree into a parent of the target."""
1679
2131
        revid = source.commit('record tree')
1680
 
        target.branch.repository.fetch(source.branch.repository, revid)
 
2132
        target.branch.fetch(source.branch, revid)
1681
2133
        target.set_parent_ids([revid])
1682
2134
        return target.basis_tree(), target
1683
2135
 
 
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
 
1684
2153
    _matching_from_tree_format = WorkingTreeFormat4()
1685
2154
    _matching_to_tree_format = WorkingTreeFormat4()
1686
 
    _test_mutable_trees_to_test_trees = make_source_parent_tree
1687
 
 
1688
 
    def _iter_changes(self, include_unchanged=False,
 
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,
1689
2163
                      specific_files=None, pb=None, extra_trees=[],
1690
2164
                      require_versioned=True, want_unversioned=False):
1691
2165
        """Return the changes from source to target.
1692
2166
 
1693
 
        :return: An iterator that yields tuples. See InterTree._iter_changes
 
2167
        :return: An iterator that yields tuples. See InterTree.iter_changes
1694
2168
            for details.
1695
2169
        :param specific_files: An optional list of file paths to restrict the
1696
2170
            comparison to. When mapping filenames to ids, all matches in all
1707
2181
            output. An unversioned file is defined as one with (False, False)
1708
2182
            for the versioned pair.
1709
2183
        """
1710
 
        utf8_decode = cache_utf8._utf8_decode
1711
 
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
1712
 
        cmp_by_dirs = dirstate.cmp_by_dirs
1713
 
        # NB: show_status depends on being able to pass in non-versioned files
1714
 
        # and report them as unknown
1715
2184
        # TODO: handle extra trees in the dirstate.
1716
 
        # TODO: handle comparisons as an empty tree as a different special
1717
 
        # case? mbp 20070226
1718
 
        if (extra_trees or (self.source._revision_id == NULL_REVISION)
1719
 
            or specific_files == []):
 
2185
        if (extra_trees or specific_files == []):
1720
2186
            # we can't fast-path these cases (yet)
1721
 
            for f in super(InterDirStateTree, self)._iter_changes(
 
2187
            return super(InterDirStateTree, self).iter_changes(
1722
2188
                include_unchanged, specific_files, pb, extra_trees,
1723
 
                require_versioned, want_unversioned=want_unversioned):
1724
 
                yield f
1725
 
            return
 
2189
                require_versioned, want_unversioned=want_unversioned)
1726
2190
        parent_ids = self.target.get_parent_ids()
1727
 
        assert (self.source._revision_id in parent_ids), \
1728
 
                "revision {%s} is not stored in {%s}, but %s " \
1729
 
                "can only be used for trees stored in the dirstate" \
1730
 
                % (self.source._revision_id, self.target, self._iter_changes)
 
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))
1731
2197
        target_index = 0
1732
 
        if self.source._revision_id == NULL_REVISION:
 
2198
        if self.source._revision_id == _mod_revision.NULL_REVISION:
1733
2199
            source_index = None
1734
2200
            indices = (target_index,)
1735
2201
        else:
1736
 
            assert (self.source._revision_id in parent_ids), \
1737
 
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
1738
 
                self.source._revision_id, parent_ids)
 
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))
1739
2206
            source_index = 1 + parent_ids.index(self.source._revision_id)
1740
 
            indices = (source_index,target_index)
 
2207
            indices = (source_index, target_index)
1741
2208
        # -- make all specific_files utf8 --
1742
2209
        if specific_files:
1743
2210
            specific_files_utf8 = set()
1744
2211
            for path in specific_files:
 
2212
                # Note, if there are many specific files, using cache_utf8
 
2213
                # would be good here.
1745
2214
                specific_files_utf8.add(path.encode('utf8'))
1746
2215
            specific_files = specific_files_utf8
1747
2216
        else:
1748
2217
            specific_files = set([''])
1749
2218
        # -- specific_files is now a utf8 path set --
 
2219
 
1750
2220
        # -- get the state object and prepare it.
1751
2221
        state = self.target.current_dirstate()
1752
2222
        state._read_dirblocks_if_needed()
1753
 
        def _entries_for_path(path):
1754
 
            """Return a list with all the entries that match path for all ids.
1755
 
            """
1756
 
            dirname, basename = os.path.split(path)
1757
 
            key = (dirname, basename, '')
1758
 
            block_index, present = state._find_block_index_from_key(key)
1759
 
            if not present:
1760
 
                # the block which should contain path is absent.
1761
 
                return []
1762
 
            result = []
1763
 
            block = state._dirblocks[block_index][1]
1764
 
            entry_index, _ = state._find_entry_index(key, block)
1765
 
            # we may need to look at multiple entries at this path: walk while the specific_files match.
1766
 
            while (entry_index < len(block) and
1767
 
                block[entry_index][0][0:2] == key[0:2]):
1768
 
                result.append(block[entry_index])
1769
 
                entry_index += 1
1770
 
            return result
1771
2223
        if require_versioned:
1772
2224
            # -- check all supplied paths are versioned in a search tree. --
1773
 
            all_versioned = True
 
2225
            not_versioned = []
1774
2226
            for path in specific_files:
1775
 
                path_entries = _entries_for_path(path)
 
2227
                path_entries = state._entries_for_path(path)
1776
2228
                if not path_entries:
1777
2229
                    # this specified path is not present at all: error
1778
 
                    all_versioned = False
1779
 
                    break
 
2230
                    not_versioned.append(path.decode('utf-8'))
 
2231
                    continue
1780
2232
                found_versioned = False
1781
2233
                # for each id at this path
1782
2234
                for entry in path_entries:
1789
2241
                if not found_versioned:
1790
2242
                    # none of the indexes was not 'absent' at all ids for this
1791
2243
                    # path.
1792
 
                    all_versioned = False
1793
 
                    break
1794
 
            if not all_versioned:
1795
 
                raise errors.PathsNotVersionedError(specific_files)
 
2244
                    not_versioned.append(path.decode('utf-8'))
 
2245
            if len(not_versioned) > 0:
 
2246
                raise errors.PathsNotVersionedError(not_versioned)
1796
2247
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
1797
 
        search_specific_files = set()
1798
 
        for path in specific_files:
1799
 
            other_specific_files = specific_files.difference(set([path]))
1800
 
            if not osutils.is_inside_any(other_specific_files, path):
1801
 
                # this is a top level path, we must check it.
1802
 
                search_specific_files.add(path)
1803
 
        # sketch: 
1804
 
        # compare source_index and target_index at or under each element of search_specific_files.
1805
 
        # follow the following comparison table. Note that we only want to do diff operations when
1806
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1807
 
        # for the target.
1808
 
        # cases:
1809
 
        # 
1810
 
        # Source | Target | disk | action
1811
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1812
 
        #        |        |      | diff check on source-target
1813
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1814
 
        #        |        |      | ???
1815
 
        #   r    |  a     |      | add source to search
1816
 
        #   r    |  a     |  a   | 
1817
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1818
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1819
 
        #   a    | fdlt   |      | add new id
1820
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1821
 
        #   a    |  a     |      | not present in either tree, skip
1822
 
        #   a    |  a     |  a   | not present in any tree, skip
1823
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1824
 
        #        |        |      | may not be selected by the users list of paths.
1825
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1826
 
        #        |        |      | may not be selected by the users list of paths.
1827
 
        #  fdlt  | fdlt   |      | content in both: diff them
1828
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1829
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1830
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1831
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1832
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1833
 
        #        |        |      | this id at the other path.
1834
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1835
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1836
 
        #        |        |      | this id at the other path.
1837
 
 
1838
 
        # for all search_indexs in each path at or under each element of
1839
 
        # search_specific_files, if the detail is relocated: add the id, and add the
1840
 
        # relocated path as one to search if its not searched already. If the
1841
 
        # detail is not relocated, add the id.
1842
 
        searched_specific_files = set()
1843
 
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
1844
 
        # Using a list so that we can access the values and change them in
1845
 
        # nested scope. Each one is [path, file_id, entry]
1846
 
        last_source_parent = [None, None]
1847
 
        last_target_parent = [None, None]
 
2248
        search_specific_files = osutils.minimum_path_selection(specific_files)
1848
2249
 
1849
2250
        use_filesystem_for_exec = (sys.platform != 'win32')
1850
 
 
1851
 
        # Just a sentry, so that _process_entry can say that this
1852
 
        # record is handled, but isn't interesting to process (unchanged)
1853
 
        uninteresting = object()
1854
 
 
1855
 
 
1856
 
        old_dirname_to_file_id = {}
1857
 
        new_dirname_to_file_id = {}
1858
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1859
 
        #       keeping a cache of directories that we have seen.
1860
 
 
1861
 
        def _process_entry(entry, path_info):
1862
 
            """Compare an entry and real disk to generate delta information.
1863
 
 
1864
 
            :param path_info: top_relpath, basename, kind, lstat, abspath for
1865
 
                the path of entry. If None, then the path is considered absent.
1866
 
                (Perhaps we should pass in a concrete entry for this ?)
1867
 
                Basename is returned as a utf8 string because we expect this
1868
 
                tuple will be ignored, and don't want to take the time to
1869
 
                decode.
1870
 
            :return: None if these don't match
1871
 
                     A tuple of information about the change, or
1872
 
                     the object 'uninteresting' if these match, but are
1873
 
                     basically identical.
1874
 
            """
1875
 
            if source_index is None:
1876
 
                source_details = NULL_PARENT_DETAILS
1877
 
            else:
1878
 
                source_details = entry[1][source_index]
1879
 
            target_details = entry[1][target_index]
1880
 
            target_minikind = target_details[0]
1881
 
            if path_info is not None and target_minikind in 'fdlt':
1882
 
                assert target_index == 0
1883
 
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
1884
 
                                                  stat_value=path_info[3])
1885
 
                # The entry may have been modified by update_entry
1886
 
                target_details = entry[1][target_index]
1887
 
                target_minikind = target_details[0]
1888
 
            else:
1889
 
                link_or_sha1 = None
1890
 
            file_id = entry[0][2]
1891
 
            source_minikind = source_details[0]
1892
 
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
1893
 
                # claimed content in both: diff
1894
 
                #   r    | fdlt   |      | add source to search, add id path move and perform
1895
 
                #        |        |      | diff check on source-target
1896
 
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
1897
 
                #        |        |      | ???
1898
 
                if source_minikind in 'r':
1899
 
                    # add the source to the search path to find any children it
1900
 
                    # has.  TODO ? : only add if it is a container ?
1901
 
                    if not osutils.is_inside_any(searched_specific_files,
1902
 
                                                 source_details[1]):
1903
 
                        search_specific_files.add(source_details[1])
1904
 
                    # generate the old path; this is needed for stating later
1905
 
                    # as well.
1906
 
                    old_path = source_details[1]
1907
 
                    old_dirname, old_basename = os.path.split(old_path)
1908
 
                    path = pathjoin(entry[0][0], entry[0][1])
1909
 
                    old_entry = state._get_entry(source_index,
1910
 
                                                 path_utf8=old_path)
1911
 
                    # update the source details variable to be the real
1912
 
                    # location.
1913
 
                    source_details = old_entry[1][source_index]
1914
 
                    source_minikind = source_details[0]
1915
 
                else:
1916
 
                    old_dirname = entry[0][0]
1917
 
                    old_basename = entry[0][1]
1918
 
                    old_path = path = None
1919
 
                if path_info is None:
1920
 
                    # the file is missing on disk, show as removed.
1921
 
                    content_change = True
1922
 
                    target_kind = None
1923
 
                    target_exec = False
1924
 
                else:
1925
 
                    # source and target are both versioned and disk file is present.
1926
 
                    target_kind = path_info[2]
1927
 
                    if target_kind == 'directory':
1928
 
                        if path is None:
1929
 
                            old_path = path = pathjoin(old_dirname, old_basename)
1930
 
                        new_dirname_to_file_id[path] = file_id
1931
 
                        if source_minikind != 'd':
1932
 
                            content_change = True
1933
 
                        else:
1934
 
                            # directories have no fingerprint
1935
 
                            content_change = False
1936
 
                        target_exec = False
1937
 
                    elif target_kind == 'file':
1938
 
                        if source_minikind != 'f':
1939
 
                            content_change = True
1940
 
                        else:
1941
 
                            # We could check the size, but we already have the
1942
 
                            # sha1 hash.
1943
 
                            content_change = (link_or_sha1 != source_details[1])
1944
 
                        # Target details is updated at update_entry time
1945
 
                        if use_filesystem_for_exec:
1946
 
                            # We don't need S_ISREG here, because we are sure
1947
 
                            # we are dealing with a file.
1948
 
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
1949
 
                        else:
1950
 
                            target_exec = target_details[3]
1951
 
                    elif target_kind == 'symlink':
1952
 
                        if source_minikind != 'l':
1953
 
                            content_change = True
1954
 
                        else:
1955
 
                            content_change = (link_or_sha1 != source_details[1])
1956
 
                        target_exec = False
1957
 
                    elif target_kind == 'tree-reference':
1958
 
                        if source_minikind != 't':
1959
 
                            content_change = True
1960
 
                        else:
1961
 
                            content_change = False
1962
 
                        target_exec = False
1963
 
                    else:
1964
 
                        raise Exception, "unknown kind %s" % path_info[2]
1965
 
                if source_minikind == 'd':
1966
 
                    if path is None:
1967
 
                        old_path = path = pathjoin(old_dirname, old_basename)
1968
 
                    old_dirname_to_file_id[old_path] = file_id
1969
 
                # parent id is the entry for the path in the target tree
1970
 
                if old_dirname == last_source_parent[0]:
1971
 
                    source_parent_id = last_source_parent[1]
1972
 
                else:
1973
 
                    try:
1974
 
                        source_parent_id = old_dirname_to_file_id[old_dirname]
1975
 
                    except KeyError:
1976
 
                        source_parent_entry = state._get_entry(source_index,
1977
 
                                                               path_utf8=old_dirname)
1978
 
                        source_parent_id = source_parent_entry[0][2]
1979
 
                    if source_parent_id == entry[0][2]:
1980
 
                        # This is the root, so the parent is None
1981
 
                        source_parent_id = None
1982
 
                    else:
1983
 
                        last_source_parent[0] = old_dirname
1984
 
                        last_source_parent[1] = source_parent_id
1985
 
                new_dirname = entry[0][0]
1986
 
                if new_dirname == last_target_parent[0]:
1987
 
                    target_parent_id = last_target_parent[1]
1988
 
                else:
1989
 
                    try:
1990
 
                        target_parent_id = new_dirname_to_file_id[new_dirname]
1991
 
                    except KeyError:
1992
 
                        # TODO: We don't always need to do the lookup, because the
1993
 
                        #       parent entry will be the same as the source entry.
1994
 
                        target_parent_entry = state._get_entry(target_index,
1995
 
                                                               path_utf8=new_dirname)
1996
 
                        assert target_parent_entry != (None, None), (
1997
 
                            "Could not find target parent in wt: %s\nparent of: %s"
1998
 
                            % (new_dirname, entry))
1999
 
                        target_parent_id = target_parent_entry[0][2]
2000
 
                    if target_parent_id == entry[0][2]:
2001
 
                        # This is the root, so the parent is None
2002
 
                        target_parent_id = None
2003
 
                    else:
2004
 
                        last_target_parent[0] = new_dirname
2005
 
                        last_target_parent[1] = target_parent_id
2006
 
 
2007
 
                source_exec = source_details[3]
2008
 
                if (include_unchanged
2009
 
                    or content_change
2010
 
                    or source_parent_id != target_parent_id
2011
 
                    or old_basename != entry[0][1]
2012
 
                    or source_exec != target_exec
2013
 
                    ):
2014
 
                    if old_path is None:
2015
 
                        old_path = path = pathjoin(old_dirname, old_basename)
2016
 
                        old_path_u = utf8_decode(old_path)[0]
2017
 
                        path_u = old_path_u
2018
 
                    else:
2019
 
                        old_path_u = utf8_decode(old_path)[0]
2020
 
                        if old_path == path:
2021
 
                            path_u = old_path_u
2022
 
                        else:
2023
 
                            path_u = utf8_decode(path)[0]
2024
 
                    source_kind = _minikind_to_kind[source_minikind]
2025
 
                    return (entry[0][2],
2026
 
                           (old_path_u, path_u),
2027
 
                           content_change,
2028
 
                           (True, True),
2029
 
                           (source_parent_id, target_parent_id),
2030
 
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
2031
 
                           (source_kind, target_kind),
2032
 
                           (source_exec, target_exec))
2033
 
                else:
2034
 
                    return uninteresting
2035
 
            elif source_minikind in 'a' and target_minikind in 'fdlt':
2036
 
                # looks like a new file
2037
 
                if path_info is not None:
2038
 
                    path = pathjoin(entry[0][0], entry[0][1])
2039
 
                    # parent id is the entry for the path in the target tree
2040
 
                    # TODO: these are the same for an entire directory: cache em.
2041
 
                    parent_id = state._get_entry(target_index,
2042
 
                                                 path_utf8=entry[0][0])[0][2]
2043
 
                    if parent_id == entry[0][2]:
2044
 
                        parent_id = None
2045
 
                    if use_filesystem_for_exec:
2046
 
                        # We need S_ISREG here, because we aren't sure if this
2047
 
                        # is a file or not.
2048
 
                        target_exec = bool(
2049
 
                            stat.S_ISREG(path_info[3].st_mode)
2050
 
                            and stat.S_IEXEC & path_info[3].st_mode)
2051
 
                    else:
2052
 
                        target_exec = target_details[3]
2053
 
                    return (entry[0][2],
2054
 
                           (None, utf8_decode(path)[0]),
2055
 
                           True,
2056
 
                           (False, True),
2057
 
                           (None, parent_id),
2058
 
                           (None, utf8_decode(entry[0][1])[0]),
2059
 
                           (None, path_info[2]),
2060
 
                           (None, target_exec))
2061
 
                else:
2062
 
                    # but its not on disk: we deliberately treat this as just
2063
 
                    # never-present. (Why ?! - RBC 20070224)
2064
 
                    pass
2065
 
            elif source_minikind in 'fdlt' and target_minikind in 'a':
2066
 
                # unversioned, possibly, or possibly not deleted: we dont care.
2067
 
                # if its still on disk, *and* theres no other entry at this
2068
 
                # path [we dont know this in this routine at the moment -
2069
 
                # perhaps we should change this - then it would be an unknown.
2070
 
                old_path = pathjoin(entry[0][0], entry[0][1])
2071
 
                # parent id is the entry for the path in the target tree
2072
 
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
2073
 
                if parent_id == entry[0][2]:
2074
 
                    parent_id = None
2075
 
                return (entry[0][2],
2076
 
                       (utf8_decode(old_path)[0], None),
2077
 
                       True,
2078
 
                       (True, False),
2079
 
                       (parent_id, None),
2080
 
                       (utf8_decode(entry[0][1])[0], None),
2081
 
                       (_minikind_to_kind[source_minikind], None),
2082
 
                       (source_details[3], None))
2083
 
            elif source_minikind in 'fdlt' and target_minikind in 'r':
2084
 
                # a rename; could be a true rename, or a rename inherited from
2085
 
                # a renamed parent. TODO: handle this efficiently. Its not
2086
 
                # common case to rename dirs though, so a correct but slow
2087
 
                # implementation will do.
2088
 
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
2089
 
                    search_specific_files.add(target_details[1])
2090
 
            elif source_minikind in 'ra' and target_minikind in 'ra':
2091
 
                # neither of the selected trees contain this file,
2092
 
                # so skip over it. This is not currently directly tested, but
2093
 
                # is indirectly via test_too_much.TestCommands.test_conflicts.
2094
 
                pass
2095
 
            else:
2096
 
                raise AssertionError("don't know how to compare "
2097
 
                    "source_minikind=%r, target_minikind=%r"
2098
 
                    % (source_minikind, target_minikind))
2099
 
                ## import pdb;pdb.set_trace()
2100
 
            return None
2101
 
 
2102
 
        while search_specific_files:
2103
 
            # TODO: the pending list should be lexically sorted?  the
2104
 
            # interface doesn't require it.
2105
 
            current_root = search_specific_files.pop()
2106
 
            current_root_unicode = current_root.decode('utf8')
2107
 
            searched_specific_files.add(current_root)
2108
 
            # process the entries for this containing directory: the rest will be
2109
 
            # found by their parents recursively.
2110
 
            root_entries = _entries_for_path(current_root)
2111
 
            root_abspath = self.target.abspath(current_root_unicode)
2112
 
            try:
2113
 
                root_stat = os.lstat(root_abspath)
2114
 
            except OSError, e:
2115
 
                if e.errno == errno.ENOENT:
2116
 
                    # the path does not exist: let _process_entry know that.
2117
 
                    root_dir_info = None
2118
 
                else:
2119
 
                    # some other random error: hand it up.
2120
 
                    raise
2121
 
            else:
2122
 
                root_dir_info = ('', current_root,
2123
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
2124
 
                    root_abspath)
2125
 
                if root_dir_info[2] == 'directory':
2126
 
                    if self.target._directory_is_tree_reference(
2127
 
                        current_root.decode('utf8')):
2128
 
                        root_dir_info = root_dir_info[:2] + \
2129
 
                            ('tree-reference',) + root_dir_info[3:]
2130
 
 
2131
 
            if not root_entries and not root_dir_info:
2132
 
                # this specified path is not present at all, skip it.
2133
 
                continue
2134
 
            path_handled = False
2135
 
            for entry in root_entries:
2136
 
                result = _process_entry(entry, root_dir_info)
2137
 
                if result is not None:
2138
 
                    path_handled = True
2139
 
                    if result is not uninteresting:
2140
 
                        yield result
2141
 
            if want_unversioned and not path_handled and root_dir_info:
2142
 
                new_executable = bool(
2143
 
                    stat.S_ISREG(root_dir_info[3].st_mode)
2144
 
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
2145
 
                yield (None,
2146
 
                       (None, current_root_unicode),
2147
 
                       True,
2148
 
                       (False, False),
2149
 
                       (None, None),
2150
 
                       (None, splitpath(current_root_unicode)[-1]),
2151
 
                       (None, root_dir_info[2]),
2152
 
                       (None, new_executable)
2153
 
                      )
2154
 
            initial_key = (current_root, '', '')
2155
 
            block_index, _ = state._find_block_index_from_key(initial_key)
2156
 
            if block_index == 0:
2157
 
                # we have processed the total root already, but because the
2158
 
                # initial key matched it we should skip it here.
2159
 
                block_index +=1
2160
 
            if root_dir_info and root_dir_info[2] == 'tree-reference':
2161
 
                current_dir_info = None
2162
 
            else:
2163
 
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
2164
 
                try:
2165
 
                    current_dir_info = dir_iterator.next()
2166
 
                except OSError, e:
2167
 
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
2168
 
                    # python 2.5 has e.errno == EINVAL,
2169
 
                    #            and e.winerror == ERROR_DIRECTORY
2170
 
                    e_winerror = getattr(e, 'winerror', None)
2171
 
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
2172
 
                    # there may be directories in the inventory even though
2173
 
                    # this path is not a file on disk: so mark it as end of
2174
 
                    # iterator
2175
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
2176
 
                        current_dir_info = None
2177
 
                    elif (sys.platform == 'win32'
2178
 
                          and (e.errno in win_errors
2179
 
                               or e_winerror in win_errors)):
2180
 
                        current_dir_info = None
2181
 
                    else:
2182
 
                        raise
2183
 
                else:
2184
 
                    if current_dir_info[0][0] == '':
2185
 
                        # remove .bzr from iteration
2186
 
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
2187
 
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
2188
 
                        del current_dir_info[1][bzr_index]
2189
 
            # walk until both the directory listing and the versioned metadata
2190
 
            # are exhausted. 
2191
 
            if (block_index < len(state._dirblocks) and
2192
 
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2193
 
                current_block = state._dirblocks[block_index]
2194
 
            else:
2195
 
                current_block = None
2196
 
            while (current_dir_info is not None or
2197
 
                   current_block is not None):
2198
 
                if (current_dir_info and current_block
2199
 
                    and current_dir_info[0][0] != current_block[0]):
2200
 
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
2201
 
                        # filesystem data refers to paths not covered by the dirblock.
2202
 
                        # this has two possibilities:
2203
 
                        # A) it is versioned but empty, so there is no block for it
2204
 
                        # B) it is not versioned.
2205
 
 
2206
 
                        # if (A) then we need to recurse into it to check for
2207
 
                        # new unknown files or directories.
2208
 
                        # if (B) then we should ignore it, because we don't
2209
 
                        # recurse into unknown directories.
2210
 
                        path_index = 0
2211
 
                        while path_index < len(current_dir_info[1]):
2212
 
                                current_path_info = current_dir_info[1][path_index]
2213
 
                                if want_unversioned:
2214
 
                                    if current_path_info[2] == 'directory':
2215
 
                                        if self.target._directory_is_tree_reference(
2216
 
                                            current_path_info[0].decode('utf8')):
2217
 
                                            current_path_info = current_path_info[:2] + \
2218
 
                                                ('tree-reference',) + current_path_info[3:]
2219
 
                                    new_executable = bool(
2220
 
                                        stat.S_ISREG(current_path_info[3].st_mode)
2221
 
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
2222
 
                                    yield (None,
2223
 
                                        (None, utf8_decode(current_path_info[0])[0]),
2224
 
                                        True,
2225
 
                                        (False, False),
2226
 
                                        (None, None),
2227
 
                                        (None, utf8_decode(current_path_info[1])[0]),
2228
 
                                        (None, current_path_info[2]),
2229
 
                                        (None, new_executable))
2230
 
                                # dont descend into this unversioned path if it is
2231
 
                                # a dir
2232
 
                                if current_path_info[2] in ('directory',
2233
 
                                                            'tree-reference'):
2234
 
                                    del current_dir_info[1][path_index]
2235
 
                                    path_index -= 1
2236
 
                                path_index += 1
2237
 
 
2238
 
                        # This dir info has been handled, go to the next
2239
 
                        try:
2240
 
                            current_dir_info = dir_iterator.next()
2241
 
                        except StopIteration:
2242
 
                            current_dir_info = None
2243
 
                    else:
2244
 
                        # We have a dirblock entry for this location, but there
2245
 
                        # is no filesystem path for this. This is most likely
2246
 
                        # because a directory was removed from the disk.
2247
 
                        # We don't have to report the missing directory,
2248
 
                        # because that should have already been handled, but we
2249
 
                        # need to handle all of the files that are contained
2250
 
                        # within.
2251
 
                        for current_entry in current_block[1]:
2252
 
                            # entry referring to file not present on disk.
2253
 
                            # advance the entry only, after processing.
2254
 
                            result = _process_entry(current_entry, None)
2255
 
                            if result is not None:
2256
 
                                if result is not uninteresting:
2257
 
                                    yield result
2258
 
                        block_index +=1
2259
 
                        if (block_index < len(state._dirblocks) and
2260
 
                            osutils.is_inside(current_root,
2261
 
                                              state._dirblocks[block_index][0])):
2262
 
                            current_block = state._dirblocks[block_index]
2263
 
                        else:
2264
 
                            current_block = None
2265
 
                    continue
2266
 
                entry_index = 0
2267
 
                if current_block and entry_index < len(current_block[1]):
2268
 
                    current_entry = current_block[1][entry_index]
2269
 
                else:
2270
 
                    current_entry = None
2271
 
                advance_entry = True
2272
 
                path_index = 0
2273
 
                if current_dir_info and path_index < len(current_dir_info[1]):
2274
 
                    current_path_info = current_dir_info[1][path_index]
2275
 
                    if current_path_info[2] == 'directory':
2276
 
                        if self.target._directory_is_tree_reference(
2277
 
                            current_path_info[0].decode('utf8')):
2278
 
                            current_path_info = current_path_info[:2] + \
2279
 
                                ('tree-reference',) + current_path_info[3:]
2280
 
                else:
2281
 
                    current_path_info = None
2282
 
                advance_path = True
2283
 
                path_handled = False
2284
 
                while (current_entry is not None or
2285
 
                    current_path_info is not None):
2286
 
                    if current_entry is None:
2287
 
                        # the check for path_handled when the path is adnvaced
2288
 
                        # will yield this path if needed.
2289
 
                        pass
2290
 
                    elif current_path_info is None:
2291
 
                        # no path is fine: the per entry code will handle it.
2292
 
                        result = _process_entry(current_entry, current_path_info)
2293
 
                        if result is not None:
2294
 
                            if result is not uninteresting:
2295
 
                                yield result
2296
 
                    elif (current_entry[0][1] != current_path_info[1]
2297
 
                          or current_entry[1][target_index][0] in 'ar'):
2298
 
                        # The current path on disk doesn't match the dirblock
2299
 
                        # record. Either the dirblock is marked as absent, or
2300
 
                        # the file on disk is not present at all in the
2301
 
                        # dirblock. Either way, report about the dirblock
2302
 
                        # entry, and let other code handle the filesystem one.
2303
 
 
2304
 
                        # Compare the basename for these files to determine
2305
 
                        # which comes first
2306
 
                        if current_path_info[1] < current_entry[0][1]:
2307
 
                            # extra file on disk: pass for now, but only
2308
 
                            # increment the path, not the entry
2309
 
                            advance_entry = False
2310
 
                        else:
2311
 
                            # entry referring to file not present on disk.
2312
 
                            # advance the entry only, after processing.
2313
 
                            result = _process_entry(current_entry, None)
2314
 
                            if result is not None:
2315
 
                                if result is not uninteresting:
2316
 
                                    yield result
2317
 
                            advance_path = False
2318
 
                    else:
2319
 
                        result = _process_entry(current_entry, current_path_info)
2320
 
                        if result is not None:
2321
 
                            path_handled = True
2322
 
                            if result is not uninteresting:
2323
 
                                yield result
2324
 
                    if advance_entry and current_entry is not None:
2325
 
                        entry_index += 1
2326
 
                        if entry_index < len(current_block[1]):
2327
 
                            current_entry = current_block[1][entry_index]
2328
 
                        else:
2329
 
                            current_entry = None
2330
 
                    else:
2331
 
                        advance_entry = True # reset the advance flaga
2332
 
                    if advance_path and current_path_info is not None:
2333
 
                        if not path_handled:
2334
 
                            # unversioned in all regards
2335
 
                            if want_unversioned:
2336
 
                                new_executable = bool(
2337
 
                                    stat.S_ISREG(current_path_info[3].st_mode)
2338
 
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
2339
 
                                yield (None,
2340
 
                                    (None, utf8_decode(current_path_info[0])[0]),
2341
 
                                    True,
2342
 
                                    (False, False),
2343
 
                                    (None, None),
2344
 
                                    (None, utf8_decode(current_path_info[1])[0]),
2345
 
                                    (None, current_path_info[2]),
2346
 
                                    (None, new_executable))
2347
 
                            # dont descend into this unversioned path if it is
2348
 
                            # a dir
2349
 
                            if current_path_info[2] in ('directory'):
2350
 
                                del current_dir_info[1][path_index]
2351
 
                                path_index -= 1
2352
 
                        # dont descend the disk iterator into any tree 
2353
 
                        # paths.
2354
 
                        if current_path_info[2] == 'tree-reference':
2355
 
                            del current_dir_info[1][path_index]
2356
 
                            path_index -= 1
2357
 
                        path_index += 1
2358
 
                        if path_index < len(current_dir_info[1]):
2359
 
                            current_path_info = current_dir_info[1][path_index]
2360
 
                            if current_path_info[2] == 'directory':
2361
 
                                if self.target._directory_is_tree_reference(
2362
 
                                    current_path_info[0].decode('utf8')):
2363
 
                                    current_path_info = current_path_info[:2] + \
2364
 
                                        ('tree-reference',) + current_path_info[3:]
2365
 
                        else:
2366
 
                            current_path_info = None
2367
 
                        path_handled = False
2368
 
                    else:
2369
 
                        advance_path = True # reset the advance flagg.
2370
 
                if current_block is not None:
2371
 
                    block_index += 1
2372
 
                    if (block_index < len(state._dirblocks) and
2373
 
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
2374
 
                        current_block = state._dirblocks[block_index]
2375
 
                    else:
2376
 
                        current_block = None
2377
 
                if current_dir_info is not None:
2378
 
                    try:
2379
 
                        current_dir_info = dir_iterator.next()
2380
 
                    except StopIteration:
2381
 
                        current_dir_info = None
2382
 
 
 
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()
2383
2255
 
2384
2256
    @staticmethod
2385
2257
    def is_compatible(source, target):
2386
2258
        # the target must be a dirstate working tree
2387
 
        if not isinstance(target, WorkingTree4):
 
2259
        if not isinstance(target, DirStateWorkingTree):
2388
2260
            return False
2389
 
        # the source must be a revtreee or dirstate rev tree.
 
2261
        # the source must be a revtree or dirstate rev tree.
2390
2262
        if not isinstance(source,
2391
2263
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2392
2264
            return False
2393
2265
        # the source revid must be in the target dirstate
2394
 
        if not (source._revision_id == NULL_REVISION or
 
2266
        if not (source._revision_id == _mod_revision.NULL_REVISION or
2395
2267
            source._revision_id in target.get_parent_ids()):
2396
 
            # TODO: what about ghosts? it may well need to 
 
2268
            # TODO: what about ghosts? it may well need to
2397
2269
            # check for them explicitly.
2398
2270
            return False
2399
2271
        return True
2409
2281
 
2410
2282
    def convert(self, tree):
2411
2283
        # lock the control files not the tree, so that we dont get tree
2412
 
        # on-unlock behaviours, and so that noone else diddles with the 
 
2284
        # on-unlock behaviours, and so that noone else diddles with the
2413
2285
        # tree during upgrade.
2414
2286
        tree._control_files.lock_write()
2415
2287
        try:
2441
2313
 
2442
2314
    def update_format(self, tree):
2443
2315
        """Change the format marker."""
2444
 
        tree._control_files.put_utf8('format',
2445
 
            self.target_format.get_format_string())
 
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())