/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-05-07 16:48:14 UTC
  • mto: This revision was merged to the branch mainline in revision 2485.
  • Revision ID: robertc@robertcollins.net-20070507164814-wpagonutf4b5cf8s
Move HACKING to docs/developers/HACKING and adjust Makefile to accomodate this.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
38
 
    debug,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
39
47
    dirstate,
40
48
    errors,
41
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
 
    revision as _mod_revision,
44
55
    revisiontree,
45
 
    trace,
 
56
    textui,
46
57
    transform,
47
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
48
61
    )
49
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
50
64
import bzrlib.ui
51
65
""")
52
66
 
 
67
from bzrlib import symbol_versioning
53
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
56
 
from bzrlib.lock import LogicalLockResult
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
57
73
from bzrlib.mutabletree import needs_tree_write_lock
58
74
from bzrlib.osutils import (
59
75
    file_kind,
60
76
    isdir,
 
77
    normpath,
61
78
    pathjoin,
 
79
    rand_chars,
62
80
    realpath,
63
81
    safe_unicode,
 
82
    splitpath,
64
83
    )
65
 
from bzrlib.trace import mutter
 
84
from bzrlib.trace import mutter, note
66
85
from bzrlib.transport.local import LocalTransport
67
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
68
95
from bzrlib.tree import Tree
69
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
70
97
 
71
98
 
72
 
class DirStateWorkingTree(WorkingTree3):
 
99
# This is the Windows equivalent of ENOTDIR
 
100
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
101
# just an error code.
 
102
ERROR_PATH_NOT_FOUND = 3
 
103
ERROR_DIRECTORY = 267
 
104
 
 
105
 
 
106
class WorkingTree4(WorkingTree3):
 
107
    """This is the Format 4 working tree.
 
108
 
 
109
    This differs from WorkingTree3 by:
 
110
     - Having a consolidated internal dirstate, stored in a
 
111
       randomly-accessible sorted file on disk.
 
112
     - Not having a regular inventory attribute.  One can be synthesized 
 
113
       on demand but this is expensive and should be avoided.
 
114
 
 
115
    This is new in bzr 0.15.
 
116
    """
 
117
 
73
118
    def __init__(self, basedir,
74
119
                 branch,
75
120
                 _control_files=None,
84
129
        """
85
130
        self._format = _format
86
131
        self.bzrdir = _bzrdir
 
132
        from bzrlib.trace import note, mutter
 
133
        assert isinstance(basedir, basestring), \
 
134
            "base directory %r is not a string" % basedir
87
135
        basedir = safe_unicode(basedir)
88
136
        mutter("opening working tree %r", basedir)
89
137
        self._branch = branch
 
138
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
139
            "branch %r is not a Branch" % self.branch
90
140
        self.basedir = realpath(basedir)
91
141
        # if branch is at our basedir and is a format 6 or less
92
142
        # assume all other formats have their own control files.
 
143
        assert isinstance(_control_files, LockableFiles), \
 
144
            "_control_files must be a LockableFiles, not %r" % _control_files
93
145
        self._control_files = _control_files
94
 
        self._transport = self._control_files._transport
95
146
        self._dirty = None
96
147
        #-------------
97
148
        # during a read or write lock these objects are set, and are
99
150
        self._dirstate = None
100
151
        self._inventory = None
101
152
        #-------------
102
 
        self._setup_directory_is_tree_reference()
103
 
        self._detect_case_handling()
104
 
        self._rules_searcher = None
105
 
        self.views = self._make_views()
106
 
        #--- allow tests to select the dirstate iter_changes implementation
107
 
        self._iter_changes = dirstate._process_entry
108
153
 
109
154
    @needs_tree_write_lock
110
155
    def _add(self, files, ids, kinds):
112
157
        state = self.current_dirstate()
113
158
        for f, file_id, kind in zip(files, ids, kinds):
114
159
            f = f.strip('/')
 
160
            assert '//' not in f
 
161
            assert '..' not in f
115
162
            if self.path2id(f):
116
163
                # special case tree root handling.
117
164
                if f == '' and self.path2id(f) == ROOT_ID:
138
185
    @needs_tree_write_lock
139
186
    def add_reference(self, sub_tree):
140
187
        # use standard implementation, which calls back to self._add
141
 
        #
 
188
        # 
142
189
        # So we don't store the reference_revision in the working dirstate,
143
 
        # it's just recorded at the moment of commit.
 
190
        # it's just recorded at the moment of commit. 
144
191
        self._add_reference(sub_tree)
145
192
 
146
193
    def break_lock(self):
185
232
            WorkingTree3._comparison_data(self, entry, path)
186
233
        # it looks like a plain directory, but it's really a reference -- see
187
234
        # also kind()
188
 
        if (self._repo_supports_tree_reference and kind == 'directory'
189
 
            and entry is not None and entry.kind == 'tree-reference'):
 
235
        if (self._repo_supports_tree_reference and
 
236
            kind == 'directory' and
 
237
            self._directory_is_tree_reference(path)):
190
238
            kind = 'tree-reference'
191
239
        return kind, executable, stat_value
192
240
 
218
266
            return self._dirstate
219
267
        local_path = self.bzrdir.get_workingtree_transport(None
220
268
            ).local_abspath('dirstate')
221
 
        self._dirstate = dirstate.DirState.on_file(local_path,
222
 
            self._sha1_provider())
 
269
        self._dirstate = dirstate.DirState.on_file(local_path)
223
270
        return self._dirstate
224
271
 
225
 
    def _sha1_provider(self):
226
 
        """A function that returns a SHA1Provider suitable for this tree.
227
 
 
228
 
        :return: None if content filtering is not supported by this tree.
229
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
230
 
          form of files, i.e. after read filters are applied.
231
 
        """
232
 
        if self.supports_content_filtering():
233
 
            return ContentFilterAwareSHA1Provider(self)
234
 
        else:
235
 
            return None
 
272
    def _directory_is_tree_reference(self, relpath):
 
273
        # as a special case, if a directory contains control files then 
 
274
        # it's a tree reference, except that the root of the tree is not
 
275
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
276
        # TODO: We could ask all the control formats whether they
 
277
        # recognize this directory, but at the moment there's no cheap api
 
278
        # to do that.  Since we probably can only nest bzr checkouts and
 
279
        # they always use this name it's ok for now.  -- mbp 20060306
 
280
        #
 
281
        # FIXME: There is an unhandled case here of a subdirectory
 
282
        # containing .bzr but not a branch; that will probably blow up
 
283
        # when you try to commit it.  It might happen if there is a
 
284
        # checkout in a subdirectory.  This can be avoided by not adding
 
285
        # it.  mbp 20070306
236
286
 
237
287
    def filter_unversioned_files(self, paths):
238
288
        """Filter out paths that are versioned.
271
321
 
272
322
    def _generate_inventory(self):
273
323
        """Create and set self.inventory from the dirstate object.
274
 
 
 
324
        
275
325
        This is relatively expensive: we have to walk the entire dirstate.
276
326
        Ideally we would not, and can deprecate this function.
277
327
        """
281
331
        state._read_dirblocks_if_needed()
282
332
        root_key, current_entry = self._get_entry(path='')
283
333
        current_id = root_key[2]
284
 
        if not (current_entry[0][0] == 'd'): # directory
285
 
            raise AssertionError(current_entry)
 
334
        assert current_entry[0][0] == 'd' # directory
286
335
        inv = Inventory(root_id=current_id)
287
336
        # Turn some things into local variables
288
337
        minikind_to_kind = dirstate.DirState._minikind_to_kind
321
370
                    # add this entry to the parent map.
322
371
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
323
372
                elif kind == 'tree-reference':
324
 
                    if not self._repo_supports_tree_reference:
325
 
                        raise errors.UnsupportedOperation(
326
 
                            self._generate_inventory,
327
 
                            self.branch.repository)
 
373
                    assert self._repo_supports_tree_reference, \
 
374
                        "repository of %r " \
 
375
                        "doesn't support tree references " \
 
376
                        "required by entry %r" \
 
377
                        % (self, name)
328
378
                    inv_entry.reference_revision = link_or_sha1 or None
329
379
                elif kind != 'symlink':
330
380
                    raise AssertionError("unknown kind %r" % kind)
331
381
                # These checks cost us around 40ms on a 55k entry tree
332
 
                if file_id in inv_byid:
333
 
                    raise AssertionError('file_id %s already in'
334
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
335
 
                if name_unicode in parent_ie.children:
336
 
                    raise AssertionError('name %r already in parent'
337
 
                        % (name_unicode,))
 
382
                assert file_id not in inv_byid, ('file_id %s already in'
 
383
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
384
                assert name_unicode not in parent_ie.children
338
385
                inv_byid[file_id] = inv_entry
339
386
                parent_ie.children[name_unicode] = inv_entry
340
387
        self._inventory = inv
345
392
        If either file_id or path is supplied, it is used as the key to lookup.
346
393
        If both are supplied, the fastest lookup is used, and an error is
347
394
        raised if they do not both point at the same row.
348
 
 
 
395
        
349
396
        :param file_id: An optional unicode file_id to be looked up.
350
397
        :param path: An optional unicode path to be looked up.
351
398
        :return: The dirstate row tuple for path/file_id, or (None, None)
360
407
    def get_file_sha1(self, file_id, path=None, stat_value=None):
361
408
        # check file id is valid unconditionally.
362
409
        entry = self._get_entry(file_id=file_id, path=path)
363
 
        if entry[0] is None:
364
 
            raise errors.NoSuchId(self, file_id)
 
410
        assert entry[0] is not None, 'what error should this raise'
365
411
        if path is None:
366
412
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
367
413
 
368
414
        file_abspath = self.abspath(path)
369
415
        state = self.current_dirstate()
370
 
        if stat_value is None:
371
 
            try:
372
 
                stat_value = os.lstat(file_abspath)
373
 
            except OSError, e:
374
 
                if e.errno == errno.ENOENT:
375
 
                    return None
376
 
                else:
377
 
                    raise
378
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
379
 
            stat_value=stat_value)
 
416
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
417
                                          stat_value=stat_value)
380
418
        if entry[1][0][0] == 'f':
381
 
            if link_or_sha1 is None:
382
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
383
 
                try:
384
 
                    sha1 = osutils.sha_file(file_obj)
385
 
                finally:
386
 
                    file_obj.close()
387
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
388
 
                return sha1
389
 
            else:
390
 
                return link_or_sha1
 
419
            return link_or_sha1
391
420
        return None
392
421
 
393
422
    def _get_inventory(self):
394
423
        """Get the inventory for the tree. This is only valid within a lock."""
395
 
        if 'evil' in debug.debug_flags:
396
 
            trace.mutter_callsite(2,
397
 
                "accessing .inventory forces a size of tree translation.")
398
424
        if self._inventory is not None:
399
425
            return self._inventory
400
426
        self._must_be_locked()
407
433
    @needs_read_lock
408
434
    def get_parent_ids(self):
409
435
        """See Tree.get_parent_ids.
410
 
 
 
436
        
411
437
        This implementation requests the ids list from the dirstate file.
412
438
        """
413
439
        return self.current_dirstate().get_parent_ids()
429
455
 
430
456
    def has_id(self, file_id):
431
457
        state = self.current_dirstate()
 
458
        file_id = osutils.safe_file_id(file_id)
432
459
        row, parents = self._get_entry(file_id=file_id)
433
460
        if row is None:
434
461
            return False
435
462
        return osutils.lexists(pathjoin(
436
463
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
437
464
 
438
 
    def has_or_had_id(self, file_id):
439
 
        state = self.current_dirstate()
440
 
        row, parents = self._get_entry(file_id=file_id)
441
 
        return row is not None
442
 
 
443
465
    @needs_read_lock
444
466
    def id2path(self, file_id):
445
 
        "Convert a file-id to a path."
 
467
        file_id = osutils.safe_file_id(file_id)
446
468
        state = self.current_dirstate()
447
469
        entry = self._get_entry(file_id=file_id)
448
470
        if entry == (None, None):
450
472
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
451
473
        return path_utf8.decode('utf8')
452
474
 
453
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
454
 
        entry = self._get_entry(path=path)
455
 
        if entry == (None, None):
456
 
            return False # Missing entries are not executable
457
 
        return entry[1][0][3] # Executable?
458
 
 
459
475
    if not osutils.supports_executable():
 
476
        @needs_read_lock
460
477
        def is_executable(self, file_id, path=None):
461
 
            """Test if a file is executable or not.
462
 
 
463
 
            Note: The caller is expected to take a read-lock before calling this.
464
 
            """
 
478
            file_id = osutils.safe_file_id(file_id)
465
479
            entry = self._get_entry(file_id=file_id, path=path)
466
480
            if entry == (None, None):
467
481
                return False
468
482
            return entry[1][0][3]
469
 
 
470
 
        _is_executable_from_path_and_stat = \
471
 
            _is_executable_from_path_and_stat_from_basis
472
483
    else:
 
484
        @needs_read_lock
473
485
        def is_executable(self, file_id, path=None):
474
 
            """Test if a file is executable or not.
475
 
 
476
 
            Note: The caller is expected to take a read-lock before calling this.
477
 
            """
478
 
            self._must_be_locked()
479
486
            if not path:
 
487
                file_id = osutils.safe_file_id(file_id)
480
488
                path = self.id2path(file_id)
481
489
            mode = os.lstat(self.abspath(path)).st_mode
482
490
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
483
491
 
484
 
    def all_file_ids(self):
485
 
        """See Tree.iter_all_file_ids"""
486
 
        self._must_be_locked()
487
 
        result = set()
488
 
        for key, tree_details in self.current_dirstate()._iter_entries():
489
 
            if tree_details[0][0] in ('a', 'r'): # relocated
490
 
                continue
491
 
            result.add(key[2])
492
 
        return result
493
 
 
494
492
    @needs_read_lock
495
493
    def __iter__(self):
496
494
        """Iterate through file_ids for this tree.
509
507
        return iter(result)
510
508
 
511
509
    def iter_references(self):
512
 
        if not self._repo_supports_tree_reference:
513
 
            # When the repo doesn't support references, we will have nothing to
514
 
            # return
515
 
            return
516
510
        for key, tree_details in self.current_dirstate()._iter_entries():
517
511
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
518
512
                # not relevant to the working tree
520
514
            if not key[1]:
521
515
                # the root is not a reference.
522
516
                continue
523
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
517
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
524
518
            try:
525
 
                if self._kind(relpath) == 'tree-reference':
526
 
                    yield relpath, key[2]
 
519
                if self._kind(path) == 'tree-reference':
 
520
                    yield path, key[2]
527
521
            except errors.NoSuchFile:
528
522
                # path is missing on disk.
529
523
                continue
530
524
 
531
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
532
 
        """See MutableTree._observed_sha1."""
533
 
        state = self.current_dirstate()
534
 
        entry = self._get_entry(file_id=file_id, path=path)
535
 
        state._observed_sha1(entry, sha1, statvalue)
536
 
 
 
525
    @needs_read_lock
537
526
    def kind(self, file_id):
538
527
        """Return the kind of a file.
539
528
 
540
529
        This is always the actual kind that's on disk, regardless of what it
541
530
        was added as.
542
 
 
543
 
        Note: The caller is expected to take a read-lock before calling this.
544
531
        """
545
532
        relpath = self.id2path(file_id)
546
 
        if relpath is None:
547
 
            raise AssertionError(
548
 
                "path for id {%s} is None!" % file_id)
 
533
        assert relpath != None, \
 
534
            "path for id {%s} is None!" % file_id
549
535
        return self._kind(relpath)
550
536
 
551
537
    def _kind(self, relpath):
552
538
        abspath = self.abspath(relpath)
553
539
        kind = file_kind(abspath)
554
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
555
 
            entry = self._get_entry(path=relpath)
556
 
            if entry[1] is not None:
557
 
                if entry[1][0][0] == 't':
558
 
                    kind = 'tree-reference'
 
540
        if (self._repo_supports_tree_reference and
 
541
            kind == 'directory' and
 
542
            self._directory_is_tree_reference(relpath)):
 
543
            kind = 'tree-reference'
559
544
        return kind
560
545
 
561
546
    @needs_read_lock
565
550
        if parent_ids:
566
551
            return parent_ids[0]
567
552
        else:
568
 
            return _mod_revision.NULL_REVISION
 
553
            return None
569
554
 
570
555
    def lock_read(self):
571
 
        """See Branch.lock_read, and WorkingTree.unlock.
572
 
 
573
 
        :return: A bzrlib.lock.LogicalLockResult.
574
 
        """
 
556
        """See Branch.lock_read, and WorkingTree.unlock."""
575
557
        self.branch.lock_read()
576
558
        try:
577
559
            self._control_files.lock_read()
590
572
        except:
591
573
            self.branch.unlock()
592
574
            raise
593
 
        return LogicalLockResult(self.unlock)
594
575
 
595
576
    def _lock_self_write(self):
596
577
        """This should be called after the branch is locked."""
611
592
        except:
612
593
            self.branch.unlock()
613
594
            raise
614
 
        return LogicalLockResult(self.unlock)
615
595
 
616
596
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: A bzrlib.lock.LogicalLockResult.
620
 
        """
 
597
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
621
598
        self.branch.lock_read()
622
 
        return self._lock_self_write()
 
599
        self._lock_self_write()
623
600
 
624
601
    def lock_write(self):
625
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
626
 
 
627
 
        :return: A bzrlib.lock.LogicalLockResult.
628
 
        """
 
602
        """See MutableTree.lock_write, and WorkingTree.unlock."""
629
603
        self.branch.lock_write()
630
 
        return self._lock_self_write()
 
604
        self._lock_self_write()
631
605
 
632
606
    @needs_tree_write_lock
633
607
    def move(self, from_paths, to_dir, after=False):
635
609
        result = []
636
610
        if not from_paths:
637
611
            return result
 
612
 
638
613
        state = self.current_dirstate()
639
 
        if isinstance(from_paths, basestring):
640
 
            raise ValueError()
 
614
 
 
615
        assert not isinstance(from_paths, basestring)
641
616
        to_dir_utf8 = to_dir.encode('utf8')
642
617
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
643
618
        id_index = state._get_id_index()
665
640
        if self._inventory is not None:
666
641
            update_inventory = True
667
642
            inv = self.inventory
 
643
            to_dir_ie = inv[to_dir_id]
668
644
            to_dir_id = to_entry[0][2]
669
 
            to_dir_ie = inv[to_dir_id]
670
645
        else:
671
646
            update_inventory = False
672
647
 
703
678
            from_entry = self._get_entry(path=from_rel)
704
679
            if from_entry == (None, None):
705
680
                raise errors.BzrMoveFailedError(from_rel,to_dir,
706
 
                    errors.NotVersionedError(path=from_rel))
 
681
                    errors.NotVersionedError(path=str(from_rel)))
707
682
 
708
683
            from_id = from_entry[0][2]
709
684
            to_rel = pathjoin(to_dir, from_tail)
736
711
                if from_missing: # implicitly just update our path mapping
737
712
                    move_file = False
738
713
                elif not after:
739
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
714
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
715
                        extra="(Use --after to update the Bazaar id)")
740
716
 
741
717
            rollbacks = []
742
718
            def rollback_rename():
798
774
                if minikind == 'd':
799
775
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
800
776
                        """Recursively update all entries in this dirblock."""
801
 
                        if from_dir == '':
802
 
                            raise AssertionError("renaming root not supported")
 
777
                        assert from_dir != '', "renaming root not supported"
803
778
                        from_key = (from_dir, '')
804
779
                        from_block_idx, present = \
805
780
                            state._find_block_index_from_key(from_key)
818
793
 
819
794
                        # Grab a copy since move_one may update the list.
820
795
                        for entry in from_block[1][:]:
821
 
                            if not (entry[0][0] == from_dir):
822
 
                                raise AssertionError()
 
796
                            assert entry[0][0] == from_dir
823
797
                            cur_details = entry[1][0]
824
798
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
825
799
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
884
858
        for tree in trees:
885
859
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
886
860
                parents):
887
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
888
 
                    trees, require_versioned)
 
861
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
889
862
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
890
863
        # -- make all paths utf8 --
891
864
        paths_utf8 = set()
951
924
            if not all_versioned:
952
925
                raise errors.PathsNotVersionedError(paths)
953
926
        # -- remove redundancy in supplied paths to prevent over-scanning --
954
 
        search_paths = osutils.minimum_path_selection(paths)
955
 
        # sketch:
 
927
        search_paths = set()
 
928
        for path in paths:
 
929
            other_paths = paths.difference(set([path]))
 
930
            if not osutils.is_inside_any(other_paths, path):
 
931
                # this is a top level path, we must check it.
 
932
                search_paths.add(path)
 
933
        # sketch: 
956
934
        # for all search_indexs in each path at or under each element of
957
935
        # search_paths, if the detail is relocated: add the id, and add the
958
936
        # relocated path as one to search if its not searched already. If the
1014
992
 
1015
993
    def read_working_inventory(self):
1016
994
        """Read the working inventory.
1017
 
 
 
995
        
1018
996
        This is a meaningless operation for dirstate, but we obey it anyhow.
1019
997
        """
1020
998
        return self.inventory
1025
1003
 
1026
1004
        WorkingTree4 supplies revision_trees for any basis tree.
1027
1005
        """
 
1006
        revision_id = osutils.safe_revision_id(revision_id)
1028
1007
        dirstate = self.current_dirstate()
1029
1008
        parent_ids = dirstate.get_parent_ids()
1030
1009
        if revision_id not in parent_ids:
1037
1016
    @needs_tree_write_lock
1038
1017
    def set_last_revision(self, new_revision):
1039
1018
        """Change the last revision in the working tree."""
 
1019
        new_revision = osutils.safe_revision_id(new_revision)
1040
1020
        parents = self.get_parent_ids()
1041
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1042
 
            if len(parents) >= 2:
1043
 
                raise AssertionError(
1044
 
                    "setting the last parent to none with a pending merge is "
1045
 
                    "unsupported.")
 
1021
        if new_revision in (NULL_REVISION, None):
 
1022
            assert len(parents) < 2, (
 
1023
                "setting the last parent to none with a pending merge is "
 
1024
                "unsupported.")
1046
1025
            self.set_parent_ids([])
1047
1026
        else:
1048
1027
            self.set_parent_ids([new_revision] + parents[1:],
1051
1030
    @needs_tree_write_lock
1052
1031
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1053
1032
        """Set the parent ids to revision_ids.
1054
 
 
 
1033
        
1055
1034
        See also set_parent_trees. This api will try to retrieve the tree data
1056
1035
        for each element of revision_ids from the trees repository. If you have
1057
1036
        tree data already available, it is more efficient to use
1061
1040
        :param revision_ids: The revision_ids to set as the parent ids of this
1062
1041
            working tree. Any of these may be ghosts.
1063
1042
        """
 
1043
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1064
1044
        trees = []
1065
1045
        for revision_id in revision_ids:
1066
1046
            try:
1072
1052
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1073
1053
                revtree = None
1074
1054
            trees.append((revision_id, revtree))
 
1055
        self.current_dirstate()._validate()
1075
1056
        self.set_parent_trees(trees,
1076
1057
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1058
        self.current_dirstate()._validate()
1077
1059
 
1078
1060
    @needs_tree_write_lock
1079
1061
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1084
1066
            parent tree - i.e. a ghost.
1085
1067
        """
1086
1068
        dirstate = self.current_dirstate()
 
1069
        dirstate._validate()
1087
1070
        if len(parents_list) > 0:
1088
1071
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1089
1072
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1090
1073
        real_trees = []
1091
1074
        ghosts = []
1092
 
 
1093
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1094
 
        graph = self.branch.repository.get_graph()
1095
 
        heads = graph.heads(parent_ids)
1096
 
        accepted_revisions = set()
1097
 
 
1098
1075
        # convert absent trees to the null tree, which we convert back to
1099
1076
        # missing on access.
1100
1077
        for rev_id, tree in parents_list:
1101
 
            if len(accepted_revisions) > 0:
1102
 
                # we always accept the first tree
1103
 
                if rev_id in accepted_revisions or rev_id not in heads:
1104
 
                    # We have already included either this tree, or its
1105
 
                    # descendent, so we skip it.
1106
 
                    continue
1107
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1078
            rev_id = osutils.safe_revision_id(rev_id)
1108
1079
            if tree is not None:
1109
1080
                real_trees.append((rev_id, tree))
1110
1081
            else:
1111
1082
                real_trees.append((rev_id,
1112
 
                    self.branch.repository.revision_tree(
1113
 
                        _mod_revision.NULL_REVISION)))
 
1083
                    self.branch.repository.revision_tree(None)))
1114
1084
                ghosts.append(rev_id)
1115
 
            accepted_revisions.add(rev_id)
 
1085
        dirstate._validate()
1116
1086
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1087
        dirstate._validate()
1117
1088
        self._make_dirty(reset_inventory=False)
 
1089
        dirstate._validate()
1118
1090
 
1119
1091
    def _set_root_id(self, file_id):
1120
1092
        """See WorkingTree.set_root_id."""
1123
1095
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1124
1096
            self._make_dirty(reset_inventory=True)
1125
1097
 
1126
 
    def _sha_from_stat(self, path, stat_result):
1127
 
        """Get a sha digest from the tree's stat cache.
1128
 
 
1129
 
        The default implementation assumes no stat cache is present.
1130
 
 
1131
 
        :param path: The path.
1132
 
        :param stat_result: The stat result being looked up.
1133
 
        """
1134
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1135
 
 
1136
1098
    @needs_read_lock
1137
1099
    def supports_tree_reference(self):
1138
1100
        return self._repo_supports_tree_reference
1139
1101
 
1140
1102
    def unlock(self):
1141
1103
        """Unlock in format 4 trees needs to write the entire dirstate."""
1142
 
        # do non-implementation specific cleanup
1143
 
        self._cleanup()
1144
 
 
1145
1104
        if self._control_files._lock_count == 1:
1146
1105
            # eventually we should do signature checking during read locks for
1147
1106
            # dirstate updates.
1178
1137
            return
1179
1138
        state = self.current_dirstate()
1180
1139
        state._read_dirblocks_if_needed()
1181
 
        ids_to_unversion = set(file_ids)
 
1140
        ids_to_unversion = set()
 
1141
        for file_id in file_ids:
 
1142
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1182
1143
        paths_to_unversion = set()
1183
1144
        # sketch:
1184
1145
        # check if the root is to be unversioned, if so, assert for now.
1211
1172
                # just forget the whole block.
1212
1173
                entry_index = 0
1213
1174
                while entry_index < len(block[1]):
 
1175
                    # Mark this file id as having been removed
1214
1176
                    entry = block[1][entry_index]
1215
 
                    if entry[1][0][0] in 'ar':
1216
 
                        # don't remove absent or renamed entries
 
1177
                    ids_to_unversion.discard(entry[0][2])
 
1178
                    if (entry[1][0][0] == 'a'
 
1179
                        or not state._make_absent(entry)):
1217
1180
                        entry_index += 1
1218
 
                    else:
1219
 
                        # Mark this file id as having been removed
1220
 
                        ids_to_unversion.discard(entry[0][2])
1221
 
                        if not state._make_absent(entry):
1222
 
                            # The block has not shrunk.
1223
 
                            entry_index += 1
1224
1181
                # go to the next block. (At the moment we dont delete empty
1225
1182
                # dirblocks)
1226
1183
                block_index += 1
1249
1206
            for file_id in file_ids:
1250
1207
                self._inventory.remove_recursive_id(file_id)
1251
1208
 
1252
 
    @needs_tree_write_lock
1253
 
    def rename_one(self, from_rel, to_rel, after=False):
1254
 
        """See WorkingTree.rename_one"""
1255
 
        self.flush()
1256
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1257
 
 
1258
 
    @needs_tree_write_lock
1259
 
    def apply_inventory_delta(self, changes):
1260
 
        """See MutableTree.apply_inventory_delta"""
1261
 
        state = self.current_dirstate()
1262
 
        state.update_by_delta(changes)
1263
 
        self._make_dirty(reset_inventory=True)
1264
 
 
1265
 
    def update_basis_by_delta(self, new_revid, delta):
1266
 
        """See MutableTree.update_basis_by_delta."""
1267
 
        if self.last_revision() == new_revid:
1268
 
            raise AssertionError()
1269
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1270
 
 
1271
1209
    @needs_read_lock
1272
1210
    def _validate(self):
1273
1211
        self._dirstate._validate()
1275
1213
    @needs_tree_write_lock
1276
1214
    def _write_inventory(self, inv):
1277
1215
        """Write inventory as the current inventory."""
1278
 
        if self._dirty:
1279
 
            raise AssertionError("attempting to write an inventory when the "
1280
 
                "dirstate is dirty will lose pending changes")
1281
 
        had_inventory = self._inventory is not None
1282
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1283
 
        # working inventory. We do this because self.inventory may be inv, or
1284
 
        # may have been modified, and either case would prevent a clean delta
1285
 
        # being created.
1286
 
        self._inventory = None
1287
 
        # generate a delta,
1288
 
        delta = inv._make_delta(self.inventory)
1289
 
        # and apply it.
1290
 
        self.apply_inventory_delta(delta)
1291
 
        if had_inventory:
 
1216
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1217
        self.current_dirstate().set_state_from_inventory(inv)
 
1218
        self._make_dirty(reset_inventory=False)
 
1219
        if self._inventory is not None:
1292
1220
            self._inventory = inv
1293
1221
        self.flush()
1294
1222
 
1295
1223
 
1296
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1297
 
 
1298
 
    def __init__(self, tree):
1299
 
        self.tree = tree
1300
 
 
1301
 
    def sha1(self, abspath):
1302
 
        """See dirstate.SHA1Provider.sha1()."""
1303
 
        filters = self.tree._content_filter_stack(
1304
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1305
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1306
 
 
1307
 
    def stat_and_sha1(self, abspath):
1308
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1309
 
        filters = self.tree._content_filter_stack(
1310
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1311
 
        file_obj = file(abspath, 'rb', 65000)
1312
 
        try:
1313
 
            statvalue = os.fstat(file_obj.fileno())
1314
 
            if filters:
1315
 
                file_obj = filtered_input_file(file_obj, filters)
1316
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1317
 
        finally:
1318
 
            file_obj.close()
1319
 
        return statvalue, sha1
1320
 
 
1321
 
 
1322
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1323
 
    """Dirstate working tree that supports content filtering.
1324
 
 
1325
 
    The dirstate holds the hash and size of the canonical form of the file, 
1326
 
    and most methods must return that.
1327
 
    """
1328
 
 
1329
 
    def _file_content_summary(self, path, stat_result):
1330
 
        # This is to support the somewhat obsolete path_content_summary method
1331
 
        # with content filtering: see
1332
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1333
 
        #
1334
 
        # If the dirstate cache is up to date and knows the hash and size,
1335
 
        # return that.
1336
 
        # Otherwise if there are no content filters, return the on-disk size
1337
 
        # and leave the hash blank.
1338
 
        # Otherwise, read and filter the on-disk file and use its size and
1339
 
        # hash.
1340
 
        #
1341
 
        # The dirstate doesn't store the size of the canonical form so we
1342
 
        # can't trust it for content-filtered trees.  We just return None.
1343
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1344
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1345
 
        return ('file', None, executable, dirstate_sha1)
1346
 
 
1347
 
 
1348
 
class WorkingTree4(DirStateWorkingTree):
1349
 
    """This is the Format 4 working tree.
1350
 
 
1351
 
    This differs from WorkingTree3 by:
1352
 
     - Having a consolidated internal dirstate, stored in a
1353
 
       randomly-accessible sorted file on disk.
1354
 
     - Not having a regular inventory attribute.  One can be synthesized
1355
 
       on demand but this is expensive and should be avoided.
1356
 
 
1357
 
    This is new in bzr 0.15.
1358
 
    """
1359
 
 
1360
 
 
1361
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1362
 
    """This is the Format 5 working tree.
1363
 
 
1364
 
    This differs from WorkingTree4 by:
1365
 
     - Supporting content filtering.
1366
 
 
1367
 
    This is new in bzr 1.11.
1368
 
    """
1369
 
 
1370
 
 
1371
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1372
 
    """This is the Format 6 working tree.
1373
 
 
1374
 
    This differs from WorkingTree5 by:
1375
 
     - Supporting a current view that may mask the set of files in a tree
1376
 
       impacted by most user operations.
1377
 
 
1378
 
    This is new in bzr 1.14.
1379
 
    """
1380
 
 
1381
 
    def _make_views(self):
1382
 
        return views.PathBasedViews(self)
1383
 
 
1384
 
 
1385
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1386
 
 
1387
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1388
 
                   accelerator_tree=None, hardlink=False):
 
1224
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1225
    """The first consolidated dirstate working tree format.
 
1226
 
 
1227
    This format:
 
1228
        - exists within a metadir controlling .bzr
 
1229
        - includes an explicit version marker for the workingtree control
 
1230
          files, separate from the BzrDir format
 
1231
        - modifies the hash cache format
 
1232
        - is new in bzr 0.15
 
1233
        - uses a LockDir to guard access to it.
 
1234
    """
 
1235
 
 
1236
    upgrade_recommended = False
 
1237
 
 
1238
    def get_format_string(self):
 
1239
        """See WorkingTreeFormat.get_format_string()."""
 
1240
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1241
 
 
1242
    def get_format_description(self):
 
1243
        """See WorkingTreeFormat.get_format_description()."""
 
1244
        return "Working tree format 4"
 
1245
 
 
1246
    def initialize(self, a_bzrdir, revision_id=None):
1389
1247
        """See WorkingTreeFormat.initialize().
1390
1248
 
1391
1249
        :param revision_id: allows creating a working tree at a different
1392
1250
        revision than the branch is at.
1393
 
        :param accelerator_tree: A tree which can be used for retrieving file
1394
 
            contents more quickly than the revision tree, i.e. a workingtree.
1395
 
            The revision tree will be used for cases where accelerator_tree's
1396
 
            content is different.
1397
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1398
 
            where possible.
1399
1251
 
1400
1252
        These trees get an initial random root id, if their repository supports
1401
1253
        rich root data, TREE_ROOT otherwise.
1402
1254
        """
 
1255
        revision_id = osutils.safe_revision_id(revision_id)
1403
1256
        if not isinstance(a_bzrdir.transport, LocalTransport):
1404
1257
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1405
1258
        transport = a_bzrdir.get_workingtree_transport(self)
1406
1259
        control_files = self._open_control_files(a_bzrdir)
1407
1260
        control_files.create_lock()
1408
1261
        control_files.lock_write()
1409
 
        transport.put_bytes('format', self.get_format_string(),
1410
 
            mode=a_bzrdir._get_file_mode())
1411
 
        if from_branch is not None:
1412
 
            branch = from_branch
1413
 
        else:
1414
 
            branch = a_bzrdir.open_branch()
 
1262
        control_files.put_utf8('format', self.get_format_string())
 
1263
        branch = a_bzrdir.open_branch()
1415
1264
        if revision_id is None:
1416
1265
            revision_id = branch.last_revision()
1417
1266
        local_path = transport.local_abspath('dirstate')
1419
1268
        state = dirstate.DirState.initialize(local_path)
1420
1269
        state.unlock()
1421
1270
        del state
1422
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1271
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1423
1272
                         branch,
1424
1273
                         _format=self,
1425
1274
                         _bzrdir=a_bzrdir,
1426
1275
                         _control_files=control_files)
1427
1276
        wt._new_tree()
1428
1277
        wt.lock_tree_write()
 
1278
        wt.current_dirstate()._validate()
1429
1279
        try:
1430
 
            self._init_custom_control_files(wt)
1431
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
 
1280
            if revision_id in (None, NULL_REVISION):
1432
1281
                if branch.repository.supports_rich_root():
1433
1282
                    wt._set_root_id(generate_ids.gen_root_id())
1434
1283
                else:
1435
1284
                    wt._set_root_id(ROOT_ID)
1436
1285
                wt.flush()
1437
 
            basis = None
1438
 
            # frequently, we will get here due to branching.  The accelerator
1439
 
            # tree will be the tree from the branch, so the desired basis
1440
 
            # tree will often be a parent of the accelerator tree.
1441
 
            if accelerator_tree is not None:
1442
 
                try:
1443
 
                    basis = accelerator_tree.revision_tree(revision_id)
1444
 
                except errors.NoSuchRevision:
1445
 
                    pass
1446
 
            if basis is None:
1447
 
                basis = branch.repository.revision_tree(revision_id)
1448
 
            if revision_id == _mod_revision.NULL_REVISION:
1449
 
                parents_list = []
1450
 
            else:
1451
 
                parents_list = [(revision_id, basis)]
 
1286
                wt.current_dirstate()._validate()
 
1287
            wt.set_last_revision(revision_id)
 
1288
            wt.flush()
 
1289
            basis = wt.basis_tree()
1452
1290
            basis.lock_read()
1453
 
            try:
1454
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1291
            # if the basis has a root id we have to use that; otherwise we use
 
1292
            # a new random one
 
1293
            basis_root_id = basis.get_root_id()
 
1294
            if basis_root_id is not None:
 
1295
                wt._set_root_id(basis_root_id)
1455
1296
                wt.flush()
1456
 
                # if the basis has a root id we have to use that; otherwise we
1457
 
                # use a new random one
1458
 
                basis_root_id = basis.get_root_id()
1459
 
                if basis_root_id is not None:
1460
 
                    wt._set_root_id(basis_root_id)
1461
 
                    wt.flush()
1462
 
                if wt.supports_content_filtering():
1463
 
                    # The original tree may not have the same content filters
1464
 
                    # applied so we can't safely build the inventory delta from
1465
 
                    # the source tree.
1466
 
                    delta_from_tree = False
1467
 
                else:
1468
 
                    delta_from_tree = True
1469
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1470
 
                # because wt4.apply_inventory_delta does not mutate the input
1471
 
                # inventory entries.
1472
 
                transform.build_tree(basis, wt, accelerator_tree,
1473
 
                                     hardlink=hardlink,
1474
 
                                     delta_from_tree=delta_from_tree)
1475
 
            finally:
1476
 
                basis.unlock()
 
1297
            transform.build_tree(basis, wt)
 
1298
            basis.unlock()
1477
1299
        finally:
1478
1300
            control_files.unlock()
1479
1301
            wt.unlock()
1480
1302
        return wt
1481
1303
 
1482
 
    def _init_custom_control_files(self, wt):
1483
 
        """Subclasses with custom control files should override this method.
1484
 
 
1485
 
        The working tree and control files are locked for writing when this
1486
 
        method is called.
1487
 
 
1488
 
        :param wt: the WorkingTree object
1489
 
        """
1490
 
 
1491
1304
    def _open(self, a_bzrdir, control_files):
1492
1305
        """Open the tree itself.
1493
1306
 
1494
1307
        :param a_bzrdir: the dir for the tree.
1495
1308
        :param control_files: the control files for the tree.
1496
1309
        """
1497
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1310
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1498
1311
                           branch=a_bzrdir.open_branch(),
1499
1312
                           _format=self,
1500
1313
                           _bzrdir=a_bzrdir,
1501
1314
                           _control_files=control_files)
1502
1315
 
1503
1316
    def __get_matchingbzrdir(self):
1504
 
        return self._get_matchingbzrdir()
1505
 
 
1506
 
    def _get_matchingbzrdir(self):
1507
 
        """Overrideable method to get a bzrdir for testing."""
1508
1317
        # please test against something that will let us do tree references
1509
1318
        return bzrdir.format_registry.make_bzrdir(
1510
1319
            'dirstate-with-subtree')
1512
1321
    _matchingbzrdir = property(__get_matchingbzrdir)
1513
1322
 
1514
1323
 
1515
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1516
 
    """The first consolidated dirstate working tree format.
1517
 
 
1518
 
    This format:
1519
 
        - exists within a metadir controlling .bzr
1520
 
        - includes an explicit version marker for the workingtree control
1521
 
          files, separate from the BzrDir format
1522
 
        - modifies the hash cache format
1523
 
        - is new in bzr 0.15
1524
 
        - uses a LockDir to guard access to it.
1525
 
    """
1526
 
 
1527
 
    upgrade_recommended = False
1528
 
 
1529
 
    _tree_class = WorkingTree4
1530
 
 
1531
 
    def get_format_string(self):
1532
 
        """See WorkingTreeFormat.get_format_string()."""
1533
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1534
 
 
1535
 
    def get_format_description(self):
1536
 
        """See WorkingTreeFormat.get_format_description()."""
1537
 
        return "Working tree format 4"
1538
 
 
1539
 
 
1540
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1541
 
    """WorkingTree format supporting content filtering.
1542
 
    """
1543
 
 
1544
 
    upgrade_recommended = False
1545
 
 
1546
 
    _tree_class = WorkingTree5
1547
 
 
1548
 
    def get_format_string(self):
1549
 
        """See WorkingTreeFormat.get_format_string()."""
1550
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1551
 
 
1552
 
    def get_format_description(self):
1553
 
        """See WorkingTreeFormat.get_format_description()."""
1554
 
        return "Working tree format 5"
1555
 
 
1556
 
    def supports_content_filtering(self):
1557
 
        return True
1558
 
 
1559
 
 
1560
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1561
 
    """WorkingTree format supporting views.
1562
 
    """
1563
 
 
1564
 
    upgrade_recommended = False
1565
 
 
1566
 
    _tree_class = WorkingTree6
1567
 
 
1568
 
    def get_format_string(self):
1569
 
        """See WorkingTreeFormat.get_format_string()."""
1570
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1571
 
 
1572
 
    def get_format_description(self):
1573
 
        """See WorkingTreeFormat.get_format_description()."""
1574
 
        return "Working tree format 6"
1575
 
 
1576
 
    def _init_custom_control_files(self, wt):
1577
 
        """Subclasses with custom control files should override this method."""
1578
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1579
 
 
1580
 
    def supports_content_filtering(self):
1581
 
        return True
1582
 
 
1583
 
    def supports_views(self):
1584
 
        return True
1585
 
 
1586
 
 
1587
1324
class DirStateRevisionTree(Tree):
1588
 
    """A revision tree pulling the inventory from a dirstate.
1589
 
    
1590
 
    Note that this is one of the historical (ie revision) trees cached in the
1591
 
    dirstate for easy access, not the workingtree.
1592
 
    """
 
1325
    """A revision tree pulling the inventory from a dirstate."""
1593
1326
 
1594
1327
    def __init__(self, dirstate, revision_id, repository):
1595
1328
        self._dirstate = dirstate
1596
 
        self._revision_id = revision_id
 
1329
        self._revision_id = osutils.safe_revision_id(revision_id)
1597
1330
        self._repository = repository
1598
1331
        self._inventory = None
1599
1332
        self._locked = 0
1600
1333
        self._dirstate_locked = False
1601
 
        self._repo_supports_tree_reference = getattr(
1602
 
            repository._format, "supports_tree_reference",
1603
 
            False)
1604
1334
 
1605
1335
    def __repr__(self):
1606
1336
        return "<%s of %s in %s>" % \
1607
1337
            (self.__class__.__name__, self._revision_id, self._dirstate)
1608
1338
 
1609
 
    def annotate_iter(self, file_id,
1610
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1339
    def annotate_iter(self, file_id):
1611
1340
        """See Tree.annotate_iter"""
1612
 
        text_key = (file_id, self.inventory[file_id].revision)
1613
 
        annotations = self._repository.texts.annotate(text_key)
1614
 
        return [(key[-1], line) for (key, line) in annotations]
 
1341
        w = self._repository.weave_store.get_weave(file_id,
 
1342
                           self._repository.get_transaction())
 
1343
        return w.annotate_iter(self.inventory[file_id].revision)
1615
1344
 
1616
 
    def _get_ancestors(self, default_revision):
1617
 
        return set(self._repository.get_ancestry(self._revision_id,
1618
 
                                                 topo_sorted=False))
1619
1345
    def _comparison_data(self, entry, path):
1620
1346
        """See Tree._comparison_data."""
1621
1347
        if entry is None:
1638
1364
    def get_root_id(self):
1639
1365
        return self.path2id('')
1640
1366
 
1641
 
    def id2path(self, file_id):
1642
 
        "Convert a file-id to a path."
1643
 
        entry = self._get_entry(file_id=file_id)
1644
 
        if entry == (None, None):
1645
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1646
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1647
 
        return path_utf8.decode('utf8')
1648
 
 
1649
 
    def iter_references(self):
1650
 
        if not self._repo_supports_tree_reference:
1651
 
            # When the repo doesn't support references, we will have nothing to
1652
 
            # return
1653
 
            return iter([])
1654
 
        # Otherwise, fall back to the default implementation
1655
 
        return super(DirStateRevisionTree, self).iter_references()
1656
 
 
1657
1367
    def _get_parent_index(self):
1658
1368
        """Return the index in the dirstate referenced by this tree."""
1659
1369
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1664
1374
        If either file_id or path is supplied, it is used as the key to lookup.
1665
1375
        If both are supplied, the fastest lookup is used, and an error is
1666
1376
        raised if they do not both point at the same row.
1667
 
 
 
1377
        
1668
1378
        :param file_id: An optional unicode file_id to be looked up.
1669
1379
        :param path: An optional unicode path to be looked up.
1670
1380
        :return: The dirstate row tuple for path/file_id, or (None, None)
1671
1381
        """
1672
1382
        if file_id is None and path is None:
1673
1383
            raise errors.BzrError('must supply file_id or path')
 
1384
        file_id = osutils.safe_file_id(file_id)
1674
1385
        if path is not None:
1675
1386
            path = path.encode('utf8')
1676
1387
        parent_index = self._get_parent_index()
1684
1395
 
1685
1396
        This is relatively expensive: we have to walk the entire dirstate.
1686
1397
        """
1687
 
        if not self._locked:
1688
 
            raise AssertionError(
1689
 
                'cannot generate inventory of an unlocked '
1690
 
                'dirstate revision tree')
 
1398
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1399
            'dirstate revision tree'
1691
1400
        # separate call for profiling - makes it clear where the costs are.
1692
1401
        self._dirstate._read_dirblocks_if_needed()
1693
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1694
 
            raise AssertionError(
1695
 
                'parent %s has disappeared from %s' % (
1696
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1402
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1403
            'parent %s has disappeared from %s' % (
 
1404
            self._revision_id, self._dirstate.get_parent_ids())
1697
1405
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1698
1406
        # This is identical now to the WorkingTree _generate_inventory except
1699
1407
        # for the tree index use.
1700
1408
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1701
1409
        current_id = root_key[2]
1702
 
        if current_entry[parent_index][0] != 'd':
1703
 
            raise AssertionError()
 
1410
        assert current_entry[parent_index][0] == 'd'
1704
1411
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1705
1412
        inv.root.revision = current_entry[parent_index][4]
1706
1413
        # Turn some things into local variables
1746
1453
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1747
1454
                            % entry)
1748
1455
                # These checks cost us around 40ms on a 55k entry tree
1749
 
                if file_id in inv_byid:
1750
 
                    raise AssertionError('file_id %s already in'
1751
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1752
 
                if name_unicode in parent_ie.children:
1753
 
                    raise AssertionError('name %r already in parent'
1754
 
                        % (name_unicode,))
 
1456
                assert file_id not in inv_byid
 
1457
                assert name_unicode not in parent_ie.children
1755
1458
                inv_byid[file_id] = inv_entry
1756
1459
                parent_ie.children[name_unicode] = inv_entry
1757
1460
        self._inventory = inv
1767
1470
            return None
1768
1471
        parent_index = self._get_parent_index()
1769
1472
        last_changed_revision = entry[1][parent_index][4]
1770
 
        try:
1771
 
            rev = self._repository.get_revision(last_changed_revision)
1772
 
        except errors.NoSuchRevision:
1773
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1774
 
        return rev.timestamp
 
1473
        return self._repository.get_revision(last_changed_revision).timestamp
1775
1474
 
1776
1475
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1777
1476
        entry = self._get_entry(file_id=file_id, path=path)
1781
1480
            return parent_details[1]
1782
1481
        return None
1783
1482
 
1784
 
    def get_file(self, file_id, path=None):
 
1483
    def get_file(self, file_id):
1785
1484
        return StringIO(self.get_file_text(file_id))
1786
1485
 
 
1486
    def get_file_lines(self, file_id):
 
1487
        ie = self.inventory[file_id]
 
1488
        return self._repository.weave_store.get_weave(file_id,
 
1489
                self._repository.get_transaction()).get_lines(ie.revision)
 
1490
 
1787
1491
    def get_file_size(self, file_id):
1788
 
        """See Tree.get_file_size"""
1789
1492
        return self.inventory[file_id].text_size
1790
1493
 
1791
 
    def get_file_text(self, file_id, path=None):
1792
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1793
 
        return ''.join(content)
 
1494
    def get_file_text(self, file_id):
 
1495
        return ''.join(self.get_file_lines(file_id))
1794
1496
 
1795
1497
    def get_reference_revision(self, file_id, path=None):
1796
1498
        return self.inventory[file_id].reference_revision
1797
1499
 
1798
 
    def iter_files_bytes(self, desired_files):
1799
 
        """See Tree.iter_files_bytes.
1800
 
 
1801
 
        This version is implemented on top of Repository.iter_files_bytes"""
1802
 
        parent_index = self._get_parent_index()
1803
 
        repo_desired_files = []
1804
 
        for file_id, identifier in desired_files:
1805
 
            entry = self._get_entry(file_id)
1806
 
            if entry == (None, None):
1807
 
                raise errors.NoSuchId(self, file_id)
1808
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1809
 
                                       identifier))
1810
 
        return self._repository.iter_files_bytes(repo_desired_files)
1811
 
 
1812
1500
    def get_symlink_target(self, file_id):
1813
1501
        entry = self._get_entry(file_id=file_id)
1814
1502
        parent_index = self._get_parent_index()
1815
1503
        if entry[1][parent_index][0] != 'l':
1816
1504
            return None
1817
1505
        else:
1818
 
            target = entry[1][parent_index][1]
1819
 
            target = target.decode('utf8')
1820
 
            return target
 
1506
            # At present, none of the tree implementations supports non-ascii
 
1507
            # symlink targets. So we will just assume that the dirstate path is
 
1508
            # correct.
 
1509
            return entry[1][parent_index][1]
1821
1510
 
1822
1511
    def get_revision_id(self):
1823
1512
        """Return the revision id for this tree."""
1841
1530
        return bool(self.path2id(filename))
1842
1531
 
1843
1532
    def kind(self, file_id):
1844
 
        entry = self._get_entry(file_id=file_id)[1]
1845
 
        if entry is None:
1846
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1847
 
        parent_index = self._get_parent_index()
1848
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1849
 
 
1850
 
    def stored_kind(self, file_id):
1851
 
        """See Tree.stored_kind"""
1852
 
        return self.kind(file_id)
1853
 
 
1854
 
    def path_content_summary(self, path):
1855
 
        """See Tree.path_content_summary."""
1856
 
        id = self.inventory.path2id(path)
1857
 
        if id is None:
1858
 
            return ('missing', None, None, None)
1859
 
        entry = self._inventory[id]
1860
 
        kind = entry.kind
1861
 
        if kind == 'file':
1862
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1863
 
        elif kind == 'symlink':
1864
 
            return (kind, None, None, entry.symlink_target)
1865
 
        else:
1866
 
            return (kind, None, None, None)
 
1533
        return self.inventory[file_id].kind
1867
1534
 
1868
1535
    def is_executable(self, file_id, path=None):
1869
1536
        ie = self.inventory[file_id]
1871
1538
            return None
1872
1539
        return ie.executable
1873
1540
 
1874
 
    def is_locked(self):
1875
 
        return self._locked
1876
 
 
1877
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1541
    def list_files(self, include_root=False):
1878
1542
        # We use a standard implementation, because DirStateRevisionTree is
1879
1543
        # dealing with one of the parents of the current state
1880
1544
        inv = self._get_inventory()
1881
 
        if from_dir is None:
1882
 
            from_dir_id = None
1883
 
        else:
1884
 
            from_dir_id = inv.path2id(from_dir)
1885
 
            if from_dir_id is None:
1886
 
                # Directory not versioned
1887
 
                return
1888
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1889
 
        if inv.root is not None and not include_root and from_dir is None:
 
1545
        entries = inv.iter_entries()
 
1546
        if self.inventory.root is not None and not include_root:
1890
1547
            entries.next()
1891
1548
        for path, entry in entries:
1892
1549
            yield path, 'V', entry.kind, entry.file_id, entry
1893
1550
 
1894
1551
    def lock_read(self):
1895
 
        """Lock the tree for a set of operations.
1896
 
 
1897
 
        :return: A bzrlib.lock.LogicalLockResult.
1898
 
        """
 
1552
        """Lock the tree for a set of operations."""
1899
1553
        if not self._locked:
1900
1554
            self._repository.lock_read()
1901
1555
            if self._dirstate._lock_token is None:
1902
1556
                self._dirstate.lock_read()
1903
1557
                self._dirstate_locked = True
1904
1558
        self._locked += 1
1905
 
        return LogicalLockResult(self.unlock)
1906
1559
 
1907
1560
    def _must_be_locked(self):
1908
1561
        if not self._locked:
1929
1582
                self._dirstate_locked = False
1930
1583
            self._repository.unlock()
1931
1584
 
1932
 
    @needs_read_lock
1933
 
    def supports_tree_reference(self):
1934
 
        return self._repo_supports_tree_reference
1935
 
 
1936
1585
    def walkdirs(self, prefix=""):
1937
1586
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1938
 
        # implementation based on an inventory.
 
1587
        # implementation based on an inventory.  
1939
1588
        # This should be cleaned up to use the much faster Dirstate code
1940
1589
        # So for now, we just build up the parent inventory, and extract
1941
1590
        # it the same way RevisionTree does.
1970
1619
 
1971
1620
class InterDirStateTree(InterTree):
1972
1621
    """Fast path optimiser for changes_from with dirstate trees.
1973
 
 
1974
 
    This is used only when both trees are in the dirstate working file, and
1975
 
    the source is any parent within the dirstate, and the destination is
 
1622
    
 
1623
    This is used only when both trees are in the dirstate working file, and 
 
1624
    the source is any parent within the dirstate, and the destination is 
1976
1625
    the current working tree of the same dirstate.
1977
1626
    """
1978
1627
    # this could be generalized to allow comparisons between any trees in the
1991
1640
        target.set_parent_ids([revid])
1992
1641
        return target.basis_tree(), target
1993
1642
 
1994
 
    @classmethod
1995
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1996
 
        result = klass.make_source_parent_tree(source, target)
1997
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1998
 
        return result
1999
 
 
2000
 
    @classmethod
2001
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2002
 
                                                  target):
2003
 
        from bzrlib.tests.test__dirstate_helpers import \
2004
 
            compiled_dirstate_helpers_feature
2005
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2006
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2007
 
        result = klass.make_source_parent_tree(source, target)
2008
 
        result[1]._iter_changes = ProcessEntryC
2009
 
        return result
2010
 
 
2011
1643
    _matching_from_tree_format = WorkingTreeFormat4()
2012
1644
    _matching_to_tree_format = WorkingTreeFormat4()
2013
 
 
2014
 
    @classmethod
2015
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2016
 
        # This method shouldn't be called, because we have python and C
2017
 
        # specific flavours.
2018
 
        raise NotImplementedError
2019
 
 
2020
 
    def iter_changes(self, include_unchanged=False,
 
1645
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1646
 
 
1647
    def _iter_changes(self, include_unchanged=False,
2021
1648
                      specific_files=None, pb=None, extra_trees=[],
2022
1649
                      require_versioned=True, want_unversioned=False):
2023
1650
        """Return the changes from source to target.
2024
1651
 
2025
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1652
        :return: An iterator that yields tuples. See InterTree._iter_changes
2026
1653
            for details.
2027
1654
        :param specific_files: An optional list of file paths to restrict the
2028
1655
            comparison to. When mapping filenames to ids, all matches in all
2039
1666
            output. An unversioned file is defined as one with (False, False)
2040
1667
            for the versioned pair.
2041
1668
        """
 
1669
        utf8_decode_or_none = cache_utf8._utf8_decode_with_None
 
1670
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1671
        # NB: show_status depends on being able to pass in non-versioned files
 
1672
        # and report them as unknown
2042
1673
        # TODO: handle extra trees in the dirstate.
2043
 
        if (extra_trees or specific_files == []):
 
1674
        # TODO: handle comparisons as an empty tree as a different special
 
1675
        # case? mbp 20070226
 
1676
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2044
1677
            # we can't fast-path these cases (yet)
2045
 
            return super(InterDirStateTree, self).iter_changes(
 
1678
            for f in super(InterDirStateTree, self)._iter_changes(
2046
1679
                include_unchanged, specific_files, pb, extra_trees,
2047
 
                require_versioned, want_unversioned=want_unversioned)
 
1680
                require_versioned, want_unversioned=want_unversioned):
 
1681
                yield f
 
1682
            return
2048
1683
        parent_ids = self.target.get_parent_ids()
2049
 
        if not (self.source._revision_id in parent_ids
2050
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2051
 
            raise AssertionError(
2052
 
                "revision {%s} is not stored in {%s}, but %s "
2053
 
                "can only be used for trees stored in the dirstate"
2054
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1684
        assert (self.source._revision_id in parent_ids), \
 
1685
                "revision {%s} is not stored in {%s}, but %s " \
 
1686
                "can only be used for trees stored in the dirstate" \
 
1687
                % (self.source._revision_id, self.target, self._iter_changes)
2055
1688
        target_index = 0
2056
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1689
        if self.source._revision_id == NULL_REVISION:
2057
1690
            source_index = None
2058
1691
            indices = (target_index,)
2059
1692
        else:
2060
 
            if not (self.source._revision_id in parent_ids):
2061
 
                raise AssertionError(
2062
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2063
 
                    self.source._revision_id, parent_ids))
 
1693
            assert (self.source._revision_id in parent_ids), \
 
1694
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1695
                self.source._revision_id, parent_ids)
2064
1696
            source_index = 1 + parent_ids.index(self.source._revision_id)
2065
 
            indices = (source_index, target_index)
 
1697
            indices = (source_index,target_index)
2066
1698
        # -- make all specific_files utf8 --
2067
1699
        if specific_files:
2068
1700
            specific_files_utf8 = set()
2069
1701
            for path in specific_files:
2070
 
                # Note, if there are many specific files, using cache_utf8
2071
 
                # would be good here.
2072
1702
                specific_files_utf8.add(path.encode('utf8'))
2073
1703
            specific_files = specific_files_utf8
2074
1704
        else:
2075
1705
            specific_files = set([''])
2076
1706
        # -- specific_files is now a utf8 path set --
2077
 
 
2078
1707
        # -- get the state object and prepare it.
2079
1708
        state = self.target.current_dirstate()
2080
1709
        state._read_dirblocks_if_needed()
 
1710
        def _entries_for_path(path):
 
1711
            """Return a list with all the entries that match path for all ids.
 
1712
            """
 
1713
            dirname, basename = os.path.split(path)
 
1714
            key = (dirname, basename, '')
 
1715
            block_index, present = state._find_block_index_from_key(key)
 
1716
            if not present:
 
1717
                # the block which should contain path is absent.
 
1718
                return []
 
1719
            result = []
 
1720
            block = state._dirblocks[block_index][1]
 
1721
            entry_index, _ = state._find_entry_index(key, block)
 
1722
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1723
            while (entry_index < len(block) and
 
1724
                block[entry_index][0][0:2] == key[0:2]):
 
1725
                result.append(block[entry_index])
 
1726
                entry_index += 1
 
1727
            return result
2081
1728
        if require_versioned:
2082
1729
            # -- check all supplied paths are versioned in a search tree. --
2083
 
            not_versioned = []
 
1730
            all_versioned = True
2084
1731
            for path in specific_files:
2085
 
                path_entries = state._entries_for_path(path)
 
1732
                path_entries = _entries_for_path(path)
2086
1733
                if not path_entries:
2087
1734
                    # this specified path is not present at all: error
2088
 
                    not_versioned.append(path)
2089
 
                    continue
 
1735
                    all_versioned = False
 
1736
                    break
2090
1737
                found_versioned = False
2091
1738
                # for each id at this path
2092
1739
                for entry in path_entries:
2099
1746
                if not found_versioned:
2100
1747
                    # none of the indexes was not 'absent' at all ids for this
2101
1748
                    # path.
2102
 
                    not_versioned.append(path)
2103
 
            if len(not_versioned) > 0:
2104
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1749
                    all_versioned = False
 
1750
                    break
 
1751
            if not all_versioned:
 
1752
                raise errors.PathsNotVersionedError(specific_files)
2105
1753
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2106
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1754
        search_specific_files = set()
 
1755
        for path in specific_files:
 
1756
            other_specific_files = specific_files.difference(set([path]))
 
1757
            if not osutils.is_inside_any(other_specific_files, path):
 
1758
                # this is a top level path, we must check it.
 
1759
                search_specific_files.add(path)
 
1760
        # sketch: 
 
1761
        # compare source_index and target_index at or under each element of search_specific_files.
 
1762
        # follow the following comparison table. Note that we only want to do diff operations when
 
1763
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1764
        # for the target.
 
1765
        # cases:
 
1766
        # 
 
1767
        # Source | Target | disk | action
 
1768
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1769
        #        |        |      | diff check on source-target
 
1770
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1771
        #        |        |      | ???
 
1772
        #   r    |  a     |      | add source to search
 
1773
        #   r    |  a     |  a   | 
 
1774
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1775
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1776
        #   a    | fdlt   |      | add new id
 
1777
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1778
        #   a    |  a     |      | not present in either tree, skip
 
1779
        #   a    |  a     |  a   | not present in any tree, skip
 
1780
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1781
        #        |        |      | may not be selected by the users list of paths.
 
1782
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1783
        #        |        |      | may not be selected by the users list of paths.
 
1784
        #  fdlt  | fdlt   |      | content in both: diff them
 
1785
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1786
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1787
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1788
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1789
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1790
        #        |        |      | this id at the other path.
 
1791
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1792
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1793
        #        |        |      | this id at the other path.
 
1794
 
 
1795
        # for all search_indexs in each path at or under each element of
 
1796
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1797
        # relocated path as one to search if its not searched already. If the
 
1798
        # detail is not relocated, add the id.
 
1799
        searched_specific_files = set()
 
1800
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1801
        # Using a list so that we can access the values and change them in
 
1802
        # nested scope. Each one is [path, file_id, entry]
 
1803
        last_source_parent = [None, None, None]
 
1804
        last_target_parent = [None, None, None]
2107
1805
 
2108
1806
        use_filesystem_for_exec = (sys.platform != 'win32')
2109
 
        iter_changes = self.target._iter_changes(include_unchanged,
2110
 
            use_filesystem_for_exec, search_specific_files, state,
2111
 
            source_index, target_index, want_unversioned, self.target)
2112
 
        return iter_changes.iter_changes()
 
1807
 
 
1808
        def _process_entry(entry, path_info):
 
1809
            """Compare an entry and real disk to generate delta information.
 
1810
 
 
1811
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1812
                the path of entry. If None, then the path is considered absent.
 
1813
                (Perhaps we should pass in a concrete entry for this ?)
 
1814
                Basename is returned as a utf8 string because we expect this
 
1815
                tuple will be ignored, and don't want to take the time to
 
1816
                decode.
 
1817
            """
 
1818
            if source_index is None:
 
1819
                source_details = NULL_PARENT_DETAILS
 
1820
            else:
 
1821
                source_details = entry[1][source_index]
 
1822
            target_details = entry[1][target_index]
 
1823
            target_minikind = target_details[0]
 
1824
            if path_info is not None and target_minikind in 'fdlt':
 
1825
                assert target_index == 0
 
1826
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1827
                                                  stat_value=path_info[3])
 
1828
                # The entry may have been modified by update_entry
 
1829
                target_details = entry[1][target_index]
 
1830
                target_minikind = target_details[0]
 
1831
            else:
 
1832
                link_or_sha1 = None
 
1833
            source_minikind = source_details[0]
 
1834
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1835
                # claimed content in both: diff
 
1836
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1837
                #        |        |      | diff check on source-target
 
1838
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1839
                #        |        |      | ???
 
1840
                if source_minikind in 'r':
 
1841
                    # add the source to the search path to find any children it
 
1842
                    # has.  TODO ? : only add if it is a container ?
 
1843
                    if not osutils.is_inside_any(searched_specific_files,
 
1844
                                                 source_details[1]):
 
1845
                        search_specific_files.add(source_details[1])
 
1846
                    # generate the old path; this is needed for stating later
 
1847
                    # as well.
 
1848
                    old_path = source_details[1]
 
1849
                    old_dirname, old_basename = os.path.split(old_path)
 
1850
                    path = pathjoin(entry[0][0], entry[0][1])
 
1851
                    old_entry = state._get_entry(source_index,
 
1852
                                                 path_utf8=old_path)
 
1853
                    # update the source details variable to be the real
 
1854
                    # location.
 
1855
                    source_details = old_entry[1][source_index]
 
1856
                    source_minikind = source_details[0]
 
1857
                else:
 
1858
                    old_dirname = entry[0][0]
 
1859
                    old_basename = entry[0][1]
 
1860
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1861
                if path_info is None:
 
1862
                    # the file is missing on disk, show as removed.
 
1863
                    content_change = True
 
1864
                    target_kind = None
 
1865
                    target_exec = False
 
1866
                else:
 
1867
                    # source and target are both versioned and disk file is present.
 
1868
                    target_kind = path_info[2]
 
1869
                    if target_kind == 'directory':
 
1870
                        if source_minikind != 'd':
 
1871
                            content_change = True
 
1872
                        else:
 
1873
                            # directories have no fingerprint
 
1874
                            content_change = False
 
1875
                        target_exec = False
 
1876
                    elif target_kind == 'file':
 
1877
                        if source_minikind != 'f':
 
1878
                            content_change = True
 
1879
                        else:
 
1880
                            # We could check the size, but we already have the
 
1881
                            # sha1 hash.
 
1882
                            content_change = (link_or_sha1 != source_details[1])
 
1883
                        # Target details is updated at update_entry time
 
1884
                        if use_filesystem_for_exec:
 
1885
                            # We don't need S_ISREG here, because we are sure
 
1886
                            # we are dealing with a file.
 
1887
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1888
                        else:
 
1889
                            target_exec = target_details[3]
 
1890
                    elif target_kind == 'symlink':
 
1891
                        if source_minikind != 'l':
 
1892
                            content_change = True
 
1893
                        else:
 
1894
                            content_change = (link_or_sha1 != source_details[1])
 
1895
                        target_exec = False
 
1896
                    elif target_kind == 'tree-reference':
 
1897
                        if source_minikind != 't':
 
1898
                            content_change = True
 
1899
                        else:
 
1900
                            content_change = False
 
1901
                        target_exec = False
 
1902
                    else:
 
1903
                        raise Exception, "unknown kind %s" % path_info[2]
 
1904
                # parent id is the entry for the path in the target tree
 
1905
                if old_dirname == last_source_parent[0]:
 
1906
                    source_parent_id = last_source_parent[1]
 
1907
                else:
 
1908
                    source_parent_entry = state._get_entry(source_index,
 
1909
                                                           path_utf8=old_dirname)
 
1910
                    source_parent_id = source_parent_entry[0][2]
 
1911
                    if source_parent_id == entry[0][2]:
 
1912
                        # This is the root, so the parent is None
 
1913
                        source_parent_id = None
 
1914
                    else:
 
1915
                        last_source_parent[0] = old_dirname
 
1916
                        last_source_parent[1] = source_parent_id
 
1917
                        last_source_parent[2] = source_parent_entry
 
1918
                new_dirname = entry[0][0]
 
1919
                if new_dirname == last_target_parent[0]:
 
1920
                    target_parent_id = last_target_parent[1]
 
1921
                else:
 
1922
                    # TODO: We don't always need to do the lookup, because the
 
1923
                    #       parent entry will be the same as the source entry.
 
1924
                    target_parent_entry = state._get_entry(target_index,
 
1925
                                                           path_utf8=new_dirname)
 
1926
                    assert target_parent_entry != (None, None), (
 
1927
                        "Could not find target parent in wt: %s\nparent of: %s"
 
1928
                        % (new_dirname, entry))
 
1929
                    target_parent_id = target_parent_entry[0][2]
 
1930
                    if target_parent_id == entry[0][2]:
 
1931
                        # This is the root, so the parent is None
 
1932
                        target_parent_id = None
 
1933
                    else:
 
1934
                        last_target_parent[0] = new_dirname
 
1935
                        last_target_parent[1] = target_parent_id
 
1936
                        last_target_parent[2] = target_parent_entry
 
1937
 
 
1938
                source_exec = source_details[3]
 
1939
                return ((entry[0][2], (old_path, path), content_change,
 
1940
                        (True, True),
 
1941
                        (source_parent_id, target_parent_id),
 
1942
                        (old_basename, entry[0][1]),
 
1943
                        (_minikind_to_kind[source_minikind], target_kind),
 
1944
                        (source_exec, target_exec)),)
 
1945
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1946
                # looks like a new file
 
1947
                if path_info is not None:
 
1948
                    path = pathjoin(entry[0][0], entry[0][1])
 
1949
                    # parent id is the entry for the path in the target tree
 
1950
                    # TODO: these are the same for an entire directory: cache em.
 
1951
                    parent_id = state._get_entry(target_index,
 
1952
                                                 path_utf8=entry[0][0])[0][2]
 
1953
                    if parent_id == entry[0][2]:
 
1954
                        parent_id = None
 
1955
                    if use_filesystem_for_exec:
 
1956
                        # We need S_ISREG here, because we aren't sure if this
 
1957
                        # is a file or not.
 
1958
                        target_exec = bool(
 
1959
                            stat.S_ISREG(path_info[3].st_mode)
 
1960
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1961
                    else:
 
1962
                        target_exec = target_details[3]
 
1963
                    return ((entry[0][2], (None, path), True,
 
1964
                            (False, True),
 
1965
                            (None, parent_id),
 
1966
                            (None, entry[0][1]),
 
1967
                            (None, path_info[2]),
 
1968
                            (None, target_exec)),)
 
1969
                else:
 
1970
                    # but its not on disk: we deliberately treat this as just
 
1971
                    # never-present. (Why ?! - RBC 20070224)
 
1972
                    pass
 
1973
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1974
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1975
                # if its still on disk, *and* theres no other entry at this
 
1976
                # path [we dont know this in this routine at the moment -
 
1977
                # perhaps we should change this - then it would be an unknown.
 
1978
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1979
                # parent id is the entry for the path in the target tree
 
1980
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1981
                if parent_id == entry[0][2]:
 
1982
                    parent_id = None
 
1983
                return ((entry[0][2], (old_path, None), True,
 
1984
                        (True, False),
 
1985
                        (parent_id, None),
 
1986
                        (entry[0][1], None),
 
1987
                        (_minikind_to_kind[source_minikind], None),
 
1988
                        (source_details[3], None)),)
 
1989
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1990
                # a rename; could be a true rename, or a rename inherited from
 
1991
                # a renamed parent. TODO: handle this efficiently. Its not
 
1992
                # common case to rename dirs though, so a correct but slow
 
1993
                # implementation will do.
 
1994
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1995
                    search_specific_files.add(target_details[1])
 
1996
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1997
                # neither of the selected trees contain this file,
 
1998
                # so skip over it. This is not currently directly tested, but
 
1999
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2000
                pass
 
2001
            else:
 
2002
                raise AssertionError("don't know how to compare "
 
2003
                    "source_minikind=%r, target_minikind=%r"
 
2004
                    % (source_minikind, target_minikind))
 
2005
                ## import pdb;pdb.set_trace()
 
2006
            return ()
 
2007
 
 
2008
        while search_specific_files:
 
2009
            # TODO: the pending list should be lexically sorted?  the
 
2010
            # interface doesn't require it.
 
2011
            current_root = search_specific_files.pop()
 
2012
            current_root_unicode = current_root.decode('utf8')
 
2013
            searched_specific_files.add(current_root)
 
2014
            # process the entries for this containing directory: the rest will be
 
2015
            # found by their parents recursively.
 
2016
            root_entries = _entries_for_path(current_root)
 
2017
            root_abspath = self.target.abspath(current_root_unicode)
 
2018
            try:
 
2019
                root_stat = os.lstat(root_abspath)
 
2020
            except OSError, e:
 
2021
                if e.errno == errno.ENOENT:
 
2022
                    # the path does not exist: let _process_entry know that.
 
2023
                    root_dir_info = None
 
2024
                else:
 
2025
                    # some other random error: hand it up.
 
2026
                    raise
 
2027
            else:
 
2028
                root_dir_info = ('', current_root,
 
2029
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2030
                    root_abspath)
 
2031
                if root_dir_info[2] == 'directory':
 
2032
                    if self.target._directory_is_tree_reference(
 
2033
                        current_root.decode('utf8')):
 
2034
                        root_dir_info = root_dir_info[:2] + \
 
2035
                            ('tree-reference',) + root_dir_info[3:]
 
2036
 
 
2037
            if not root_entries and not root_dir_info:
 
2038
                # this specified path is not present at all, skip it.
 
2039
                continue
 
2040
            path_handled = False
 
2041
            for entry in root_entries:
 
2042
                for result in _process_entry(entry, root_dir_info):
 
2043
                    # this check should probably be outside the loop: one
 
2044
                    # 'iterate two trees' api, and then _iter_changes filters
 
2045
                    # unchanged pairs. - RBC 20070226
 
2046
                    path_handled = True
 
2047
                    if (include_unchanged
 
2048
                        or result[2]                    # content change
 
2049
                        or result[3][0] != result[3][1] # versioned status
 
2050
                        or result[4][0] != result[4][1] # parent id
 
2051
                        or result[5][0] != result[5][1] # name
 
2052
                        or result[6][0] != result[6][1] # kind
 
2053
                        or result[7][0] != result[7][1] # executable
 
2054
                        ):
 
2055
                        yield (result[0],
 
2056
                               (utf8_decode_or_none(result[1][0]),
 
2057
                                utf8_decode_or_none(result[1][1])),
 
2058
                               result[2],
 
2059
                               result[3],
 
2060
                               result[4],
 
2061
                               (utf8_decode_or_none(result[5][0]),
 
2062
                                utf8_decode_or_none(result[5][1])),
 
2063
                               result[6],
 
2064
                               result[7],
 
2065
                              )
 
2066
            if want_unversioned and not path_handled and root_dir_info:
 
2067
                new_executable = bool(
 
2068
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2069
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2070
                yield (None,
 
2071
                       (None, current_root_unicode),
 
2072
                       True,
 
2073
                       (False, False),
 
2074
                       (None, None),
 
2075
                       (None, splitpath(current_root_unicode)[-1]),
 
2076
                       (None, root_dir_info[2]),
 
2077
                       (None, new_executable)
 
2078
                      )
 
2079
            initial_key = (current_root, '', '')
 
2080
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2081
            if block_index == 0:
 
2082
                # we have processed the total root already, but because the
 
2083
                # initial key matched it we should skip it here.
 
2084
                block_index +=1
 
2085
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2086
                current_dir_info = None
 
2087
            else:
 
2088
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2089
                try:
 
2090
                    current_dir_info = dir_iterator.next()
 
2091
                except OSError, e:
 
2092
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2093
                    # python 2.5 has e.errno == EINVAL,
 
2094
                    #            and e.winerror == ERROR_DIRECTORY
 
2095
                    e_winerror = getattr(e, 'winerror', None)
 
2096
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2097
                    # there may be directories in the inventory even though
 
2098
                    # this path is not a file on disk: so mark it as end of
 
2099
                    # iterator
 
2100
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2101
                        current_dir_info = None
 
2102
                    elif (sys.platform == 'win32'
 
2103
                          and (e.errno in win_errors
 
2104
                               or e_winerror in win_errors)):
 
2105
                        current_dir_info = None
 
2106
                    else:
 
2107
                        raise
 
2108
                else:
 
2109
                    if current_dir_info[0][0] == '':
 
2110
                        # remove .bzr from iteration
 
2111
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2112
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2113
                        del current_dir_info[1][bzr_index]
 
2114
            # walk until both the directory listing and the versioned metadata
 
2115
            # are exhausted. 
 
2116
            if (block_index < len(state._dirblocks) and
 
2117
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2118
                current_block = state._dirblocks[block_index]
 
2119
            else:
 
2120
                current_block = None
 
2121
            while (current_dir_info is not None or
 
2122
                   current_block is not None):
 
2123
                if (current_dir_info and current_block
 
2124
                    and current_dir_info[0][0] != current_block[0]):
 
2125
                    if current_dir_info[0][0].split('/') < current_block[0].split('/'):
 
2126
                        # filesystem data refers to paths not covered by the dirblock.
 
2127
                        # this has two possibilities:
 
2128
                        # A) it is versioned but empty, so there is no block for it
 
2129
                        # B) it is not versioned.
 
2130
 
 
2131
                        # if (A) then we need to recurse into it to check for
 
2132
                        # new unknown files or directories.
 
2133
                        # if (B) then we should ignore it, because we don't
 
2134
                        # recurse into unknown directories.
 
2135
                        path_index = 0
 
2136
                        while path_index < len(current_dir_info[1]):
 
2137
                                current_path_info = current_dir_info[1][path_index]
 
2138
                                if want_unversioned:
 
2139
                                    if current_path_info[2] == 'directory':
 
2140
                                        if self.target._directory_is_tree_reference(
 
2141
                                            current_path_info[0].decode('utf8')):
 
2142
                                            current_path_info = current_path_info[:2] + \
 
2143
                                                ('tree-reference',) + current_path_info[3:]
 
2144
                                    new_executable = bool(
 
2145
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2146
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2147
                                    yield (None,
 
2148
                                        (None, utf8_decode_or_none(current_path_info[0])),
 
2149
                                        True,
 
2150
                                        (False, False),
 
2151
                                        (None, None),
 
2152
                                        (None, utf8_decode_or_none(current_path_info[1])),
 
2153
                                        (None, current_path_info[2]),
 
2154
                                        (None, new_executable))
 
2155
                                # dont descend into this unversioned path if it is
 
2156
                                # a dir
 
2157
                                if current_path_info[2] in ('directory',
 
2158
                                                            'tree-reference'):
 
2159
                                    del current_dir_info[1][path_index]
 
2160
                                    path_index -= 1
 
2161
                                path_index += 1
 
2162
 
 
2163
                        # This dir info has been handled, go to the next
 
2164
                        try:
 
2165
                            current_dir_info = dir_iterator.next()
 
2166
                        except StopIteration:
 
2167
                            current_dir_info = None
 
2168
                    else:
 
2169
                        # We have a dirblock entry for this location, but there
 
2170
                        # is no filesystem path for this. This is most likely
 
2171
                        # because a directory was removed from the disk.
 
2172
                        # We don't have to report the missing directory,
 
2173
                        # because that should have already been handled, but we
 
2174
                        # need to handle all of the files that are contained
 
2175
                        # within.
 
2176
                        for current_entry in current_block[1]:
 
2177
                            # entry referring to file not present on disk.
 
2178
                            # advance the entry only, after processing.
 
2179
                            for result in _process_entry(current_entry, None):
 
2180
                                # this check should probably be outside the loop: one
 
2181
                                # 'iterate two trees' api, and then _iter_changes filters
 
2182
                                # unchanged pairs. - RBC 20070226
 
2183
                                if (include_unchanged
 
2184
                                    or result[2]                    # content change
 
2185
                                    or result[3][0] != result[3][1] # versioned status
 
2186
                                    or result[4][0] != result[4][1] # parent id
 
2187
                                    or result[5][0] != result[5][1] # name
 
2188
                                    or result[6][0] != result[6][1] # kind
 
2189
                                    or result[7][0] != result[7][1] # executable
 
2190
                                    ):
 
2191
                                    yield (result[0],
 
2192
                                           (utf8_decode_or_none(result[1][0]),
 
2193
                                            utf8_decode_or_none(result[1][1])),
 
2194
                                           result[2],
 
2195
                                           result[3],
 
2196
                                           result[4],
 
2197
                                           (utf8_decode_or_none(result[5][0]),
 
2198
                                            utf8_decode_or_none(result[5][1])),
 
2199
                                           result[6],
 
2200
                                           result[7],
 
2201
                                          )
 
2202
                        block_index +=1
 
2203
                        if (block_index < len(state._dirblocks) and
 
2204
                            osutils.is_inside(current_root,
 
2205
                                              state._dirblocks[block_index][0])):
 
2206
                            current_block = state._dirblocks[block_index]
 
2207
                        else:
 
2208
                            current_block = None
 
2209
                    continue
 
2210
                entry_index = 0
 
2211
                if current_block and entry_index < len(current_block[1]):
 
2212
                    current_entry = current_block[1][entry_index]
 
2213
                else:
 
2214
                    current_entry = None
 
2215
                advance_entry = True
 
2216
                path_index = 0
 
2217
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2218
                    current_path_info = current_dir_info[1][path_index]
 
2219
                    if current_path_info[2] == 'directory':
 
2220
                        if self.target._directory_is_tree_reference(
 
2221
                            current_path_info[0].decode('utf8')):
 
2222
                            current_path_info = current_path_info[:2] + \
 
2223
                                ('tree-reference',) + current_path_info[3:]
 
2224
                else:
 
2225
                    current_path_info = None
 
2226
                advance_path = True
 
2227
                path_handled = False
 
2228
                while (current_entry is not None or
 
2229
                    current_path_info is not None):
 
2230
                    if current_entry is None:
 
2231
                        # the check for path_handled when the path is adnvaced
 
2232
                        # will yield this path if needed.
 
2233
                        pass
 
2234
                    elif current_path_info is None:
 
2235
                        # no path is fine: the per entry code will handle it.
 
2236
                        for result in _process_entry(current_entry, current_path_info):
 
2237
                            # this check should probably be outside the loop: one
 
2238
                            # 'iterate two trees' api, and then _iter_changes filters
 
2239
                            # unchanged pairs. - RBC 20070226
 
2240
                            if (include_unchanged
 
2241
                                or result[2]                    # content change
 
2242
                                or result[3][0] != result[3][1] # versioned status
 
2243
                                or result[4][0] != result[4][1] # parent id
 
2244
                                or result[5][0] != result[5][1] # name
 
2245
                                or result[6][0] != result[6][1] # kind
 
2246
                                or result[7][0] != result[7][1] # executable
 
2247
                                ):
 
2248
                                yield (result[0],
 
2249
                                       (utf8_decode_or_none(result[1][0]),
 
2250
                                        utf8_decode_or_none(result[1][1])),
 
2251
                                       result[2],
 
2252
                                       result[3],
 
2253
                                       result[4],
 
2254
                                       (utf8_decode_or_none(result[5][0]),
 
2255
                                        utf8_decode_or_none(result[5][1])),
 
2256
                                       result[6],
 
2257
                                       result[7],
 
2258
                                      )
 
2259
                    elif (current_entry[0][1] != current_path_info[1]
 
2260
                          or current_entry[1][target_index][0] in 'ar'):
 
2261
                        # The current path on disk doesn't match the dirblock
 
2262
                        # record. Either the dirblock is marked as absent, or
 
2263
                        # the file on disk is not present at all in the
 
2264
                        # dirblock. Either way, report about the dirblock
 
2265
                        # entry, and let other code handle the filesystem one.
 
2266
                        if current_path_info[1].split('/') < current_entry[0][1].split('/'):
 
2267
                            # extra file on disk: pass for now, but only
 
2268
                            # increment the path, not the entry
 
2269
                            advance_entry = False
 
2270
                        else:
 
2271
                            # entry referring to file not present on disk.
 
2272
                            # advance the entry only, after processing.
 
2273
                            for result in _process_entry(current_entry, None):
 
2274
                                # this check should probably be outside the loop: one
 
2275
                                # 'iterate two trees' api, and then _iter_changes filters
 
2276
                                # unchanged pairs. - RBC 20070226
 
2277
                                if (include_unchanged
 
2278
                                    or result[2]                    # content change
 
2279
                                    or result[3][0] != result[3][1] # versioned status
 
2280
                                    or result[4][0] != result[4][1] # parent id
 
2281
                                    or result[5][0] != result[5][1] # name
 
2282
                                    or result[6][0] != result[6][1] # kind
 
2283
                                    or result[7][0] != result[7][1] # executable
 
2284
                                    ):
 
2285
                                    yield (result[0],
 
2286
                                           (utf8_decode_or_none(result[1][0]),
 
2287
                                            utf8_decode_or_none(result[1][1])),
 
2288
                                           result[2],
 
2289
                                           result[3],
 
2290
                                           result[4],
 
2291
                                           (utf8_decode_or_none(result[5][0]),
 
2292
                                            utf8_decode_or_none(result[5][1])),
 
2293
                                           result[6],
 
2294
                                           result[7],
 
2295
                                          )
 
2296
                            advance_path = False
 
2297
                    else:
 
2298
                        for result in _process_entry(current_entry, current_path_info):
 
2299
                            # this check should probably be outside the loop: one
 
2300
                            # 'iterate two trees' api, and then _iter_changes filters
 
2301
                            # unchanged pairs. - RBC 20070226
 
2302
                            path_handled = True
 
2303
                            if (include_unchanged
 
2304
                                or result[2]                    # content change
 
2305
                                or result[3][0] != result[3][1] # versioned status
 
2306
                                or result[4][0] != result[4][1] # parent id
 
2307
                                or result[5][0] != result[5][1] # name
 
2308
                                or result[6][0] != result[6][1] # kind
 
2309
                                or result[7][0] != result[7][1] # executable
 
2310
                                ):
 
2311
                                yield (result[0],
 
2312
                                       (utf8_decode_or_none(result[1][0]),
 
2313
                                        utf8_decode_or_none(result[1][1])),
 
2314
                                       result[2],
 
2315
                                       result[3],
 
2316
                                       result[4],
 
2317
                                       (utf8_decode_or_none(result[5][0]),
 
2318
                                        utf8_decode_or_none(result[5][1])),
 
2319
                                       result[6],
 
2320
                                       result[7],
 
2321
                                      )
 
2322
                    if advance_entry and current_entry is not None:
 
2323
                        entry_index += 1
 
2324
                        if entry_index < len(current_block[1]):
 
2325
                            current_entry = current_block[1][entry_index]
 
2326
                        else:
 
2327
                            current_entry = None
 
2328
                    else:
 
2329
                        advance_entry = True # reset the advance flaga
 
2330
                    if advance_path and current_path_info is not None:
 
2331
                        if not path_handled:
 
2332
                            # unversioned in all regards
 
2333
                            if want_unversioned:
 
2334
                                new_executable = bool(
 
2335
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2336
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2337
                                yield (None,
 
2338
                                    (None, utf8_decode_or_none(current_path_info[0])),
 
2339
                                    True,
 
2340
                                    (False, False),
 
2341
                                    (None, None),
 
2342
                                    (None, utf8_decode_or_none(current_path_info[1])),
 
2343
                                    (None, current_path_info[2]),
 
2344
                                    (None, new_executable))
 
2345
                            # dont descend into this unversioned path if it is
 
2346
                            # a dir
 
2347
                            if current_path_info[2] in ('directory'):
 
2348
                                del current_dir_info[1][path_index]
 
2349
                                path_index -= 1
 
2350
                        # dont descend the disk iterator into any tree 
 
2351
                        # paths.
 
2352
                        if current_path_info[2] == 'tree-reference':
 
2353
                            del current_dir_info[1][path_index]
 
2354
                            path_index -= 1
 
2355
                        path_index += 1
 
2356
                        if path_index < len(current_dir_info[1]):
 
2357
                            current_path_info = current_dir_info[1][path_index]
 
2358
                            if current_path_info[2] == 'directory':
 
2359
                                if self.target._directory_is_tree_reference(
 
2360
                                    current_path_info[0].decode('utf8')):
 
2361
                                    current_path_info = current_path_info[:2] + \
 
2362
                                        ('tree-reference',) + current_path_info[3:]
 
2363
                        else:
 
2364
                            current_path_info = None
 
2365
                        path_handled = False
 
2366
                    else:
 
2367
                        advance_path = True # reset the advance flagg.
 
2368
                if current_block is not None:
 
2369
                    block_index += 1
 
2370
                    if (block_index < len(state._dirblocks) and
 
2371
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2372
                        current_block = state._dirblocks[block_index]
 
2373
                    else:
 
2374
                        current_block = None
 
2375
                if current_dir_info is not None:
 
2376
                    try:
 
2377
                        current_dir_info = dir_iterator.next()
 
2378
                    except StopIteration:
 
2379
                        current_dir_info = None
 
2380
 
2113
2381
 
2114
2382
    @staticmethod
2115
2383
    def is_compatible(source, target):
2116
2384
        # the target must be a dirstate working tree
2117
 
        if not isinstance(target, DirStateWorkingTree):
 
2385
        if not isinstance(target, WorkingTree4):
2118
2386
            return False
2119
 
        # the source must be a revtree or dirstate rev tree.
 
2387
        # the source must be a revtreee or dirstate rev tree.
2120
2388
        if not isinstance(source,
2121
2389
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2122
2390
            return False
2123
2391
        # the source revid must be in the target dirstate
2124
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2392
        if not (source._revision_id == NULL_REVISION or
2125
2393
            source._revision_id in target.get_parent_ids()):
2126
 
            # TODO: what about ghosts? it may well need to
 
2394
            # TODO: what about ghosts? it may well need to 
2127
2395
            # check for them explicitly.
2128
2396
            return False
2129
2397
        return True
2139
2407
 
2140
2408
    def convert(self, tree):
2141
2409
        # lock the control files not the tree, so that we dont get tree
2142
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2410
        # on-unlock behaviours, and so that noone else diddles with the 
2143
2411
        # tree during upgrade.
2144
2412
        tree._control_files.lock_write()
2145
2413
        try:
2171
2439
 
2172
2440
    def update_format(self, tree):
2173
2441
        """Change the format marker."""
2174
 
        tree._transport.put_bytes('format',
2175
 
            self.target_format.get_format_string(),
2176
 
            mode=tree.bzrdir._get_file_mode())
2177
 
 
2178
 
 
2179
 
class Converter4to5(object):
2180
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2181
 
 
2182
 
    def __init__(self):
2183
 
        self.target_format = WorkingTreeFormat5()
2184
 
 
2185
 
    def convert(self, tree):
2186
 
        # lock the control files not the tree, so that we don't get tree
2187
 
        # on-unlock behaviours, and so that no-one else diddles with the
2188
 
        # tree during upgrade.
2189
 
        tree._control_files.lock_write()
2190
 
        try:
2191
 
            self.update_format(tree)
2192
 
        finally:
2193
 
            tree._control_files.unlock()
2194
 
 
2195
 
    def update_format(self, tree):
2196
 
        """Change the format marker."""
2197
 
        tree._transport.put_bytes('format',
2198
 
            self.target_format.get_format_string(),
2199
 
            mode=tree.bzrdir._get_file_mode())
2200
 
 
2201
 
 
2202
 
class Converter4or5to6(object):
2203
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2204
 
 
2205
 
    def __init__(self):
2206
 
        self.target_format = WorkingTreeFormat6()
2207
 
 
2208
 
    def convert(self, tree):
2209
 
        # lock the control files not the tree, so that we don't get tree
2210
 
        # on-unlock behaviours, and so that no-one else diddles with the
2211
 
        # tree during upgrade.
2212
 
        tree._control_files.lock_write()
2213
 
        try:
2214
 
            self.init_custom_control_files(tree)
2215
 
            self.update_format(tree)
2216
 
        finally:
2217
 
            tree._control_files.unlock()
2218
 
 
2219
 
    def init_custom_control_files(self, tree):
2220
 
        """Initialize custom control files."""
2221
 
        tree._transport.put_bytes('views', '',
2222
 
            mode=tree.bzrdir._get_file_mode())
2223
 
 
2224
 
    def update_format(self, tree):
2225
 
        """Change the format marker."""
2226
 
        tree._transport.put_bytes('format',
2227
 
            self.target_format.get_format_string(),
2228
 
            mode=tree.bzrdir._get_file_mode())
 
2442
        tree._control_files.put_utf8('format',
 
2443
            self.target_format.get_format_string())