~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

IntroduceĀ dirstate-tagsĀ format

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
 
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
56
73
from bzrlib.mutabletree import needs_tree_write_lock
57
74
from bzrlib.osutils import (
58
75
    file_kind,
59
76
    isdir,
 
77
    normpath,
60
78
    pathjoin,
 
79
    rand_chars,
61
80
    realpath,
62
81
    safe_unicode,
 
82
    splitpath,
63
83
    )
64
 
from bzrlib.trace import mutter
 
84
from bzrlib.trace import mutter, note
65
85
from bzrlib.transport.local import LocalTransport
66
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
        )
67
95
from bzrlib.tree import Tree
68
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
97
 
70
98
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
99
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
 
110
 
72
111
    def __init__(self, basedir,
73
112
                 branch,
74
113
                 _control_files=None,
83
122
        """
84
123
        self._format = _format
85
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
86
128
        basedir = safe_unicode(basedir)
87
129
        mutter("opening working tree %r", basedir)
88
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
89
133
        self.basedir = realpath(basedir)
90
134
        # if branch is at our basedir and is a format 6 or less
91
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
92
138
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
139
        self._dirty = None
95
140
        #-------------
96
141
        # during a read or write lock these objects are set, and are
98
143
        self._dirstate = None
99
144
        self._inventory = None
100
145
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
146
 
108
147
    @needs_tree_write_lock
109
148
    def _add(self, files, ids, kinds):
111
150
        state = self.current_dirstate()
112
151
        for f, file_id, kind in zip(files, ids, kinds):
113
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
114
155
            if self.path2id(f):
115
156
                # special case tree root handling.
116
157
                if f == '' and self.path2id(f) == ROOT_ID:
137
178
    @needs_tree_write_lock
138
179
    def add_reference(self, sub_tree):
139
180
        # use standard implementation, which calls back to self._add
140
 
        #
 
181
        # 
141
182
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
183
        # it's just recorded at the moment of commit. 
143
184
        self._add_reference(sub_tree)
144
185
 
145
186
    def break_lock(self):
184
225
            WorkingTree3._comparison_data(self, entry, path)
185
226
        # it looks like a plain directory, but it's really a reference -- see
186
227
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
228
        if (self._repo_supports_tree_reference and
 
229
            kind == 'directory' and
 
230
            self._directory_is_tree_reference(path)):
189
231
            kind = 'tree-reference'
190
232
        return kind, executable, stat_value
191
233
 
217
259
            return self._dirstate
218
260
        local_path = self.bzrdir.get_workingtree_transport(None
219
261
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
262
        self._dirstate = dirstate.DirState.on_file(local_path)
222
263
        return self._dirstate
223
264
 
224
 
    def _sha1_provider(self):
225
 
        """A function that returns a SHA1Provider suitable for this tree.
226
 
 
227
 
        :return: None if content filtering is not supported by this tree.
228
 
          Otherwise, a SHA1Provider is returned that sha's the canonical
229
 
          form of files, i.e. after read filters are applied.
230
 
        """
231
 
        if self.supports_content_filtering():
232
 
            return ContentFilterAwareSHA1Provider(self)
233
 
        else:
234
 
            return None
 
265
    def _directory_is_tree_reference(self, relpath):
 
266
        # as a special case, if a directory contains control files then 
 
267
        # it's a tree reference, except that the root of the tree is not
 
268
        return relpath and osutils.isdir(self.abspath(relpath) + u"/.bzr")
 
269
        # TODO: We could ask all the control formats whether they
 
270
        # recognize this directory, but at the moment there's no cheap api
 
271
        # to do that.  Since we probably can only nest bzr checkouts and
 
272
        # they always use this name it's ok for now.  -- mbp 20060306
 
273
        #
 
274
        # FIXME: There is an unhandled case here of a subdirectory
 
275
        # containing .bzr but not a branch; that will probably blow up
 
276
        # when you try to commit it.  It might happen if there is a
 
277
        # checkout in a subdirectory.  This can be avoided by not adding
 
278
        # it.  mbp 20070306
235
279
 
236
280
    def filter_unversioned_files(self, paths):
237
281
        """Filter out paths that are versioned.
270
314
 
271
315
    def _generate_inventory(self):
272
316
        """Create and set self.inventory from the dirstate object.
273
 
 
 
317
        
274
318
        This is relatively expensive: we have to walk the entire dirstate.
275
319
        Ideally we would not, and can deprecate this function.
276
320
        """
280
324
        state._read_dirblocks_if_needed()
281
325
        root_key, current_entry = self._get_entry(path='')
282
326
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
327
        assert current_entry[0][0] == 'd' # directory
285
328
        inv = Inventory(root_id=current_id)
286
329
        # Turn some things into local variables
287
330
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
353
                inv_entry = factory[kind](file_id, name_unicode,
311
354
                                          parent_ie.file_id)
312
355
                if kind == 'file':
313
 
                    # This is only needed on win32, where this is the only way
314
 
                    # we know the executable bit.
315
 
                    inv_entry.executable = executable
316
356
                    # not strictly needed: working tree
317
 
                    #inv_entry.text_size = size
318
 
                    #inv_entry.text_sha1 = sha1
 
357
                    #entry.executable = executable
 
358
                    #entry.text_size = size
 
359
                    #entry.text_sha1 = sha1
 
360
                    pass
319
361
                elif kind == 'directory':
320
362
                    # add this entry to the parent map.
321
363
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
364
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
 
365
                    assert self._repo_supports_tree_reference
327
366
                    inv_entry.reference_revision = link_or_sha1 or None
328
 
                elif kind != 'symlink':
329
 
                    raise AssertionError("unknown kind %r" % kind)
 
367
                else:
 
368
                    assert 'unknown kind'
330
369
                # These checks cost us around 40ms on a 55k entry tree
331
 
                if file_id in inv_byid:
332
 
                    raise AssertionError('file_id %s already in'
333
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
334
 
                if name_unicode in parent_ie.children:
335
 
                    raise AssertionError('name %r already in parent'
336
 
                        % (name_unicode,))
 
370
                assert file_id not in inv_byid, ('file_id %s already in'
 
371
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
372
                assert name_unicode not in parent_ie.children
337
373
                inv_byid[file_id] = inv_entry
338
374
                parent_ie.children[name_unicode] = inv_entry
339
375
        self._inventory = inv
344
380
        If either file_id or path is supplied, it is used as the key to lookup.
345
381
        If both are supplied, the fastest lookup is used, and an error is
346
382
        raised if they do not both point at the same row.
347
 
 
 
383
        
348
384
        :param file_id: An optional unicode file_id to be looked up.
349
385
        :param path: An optional unicode path to be looked up.
350
386
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
395
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
396
        # check file id is valid unconditionally.
361
397
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
398
        assert entry[0] is not None, 'what error should this raise'
364
399
        if path is None:
365
400
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
401
 
367
402
        file_abspath = self.abspath(path)
368
403
        state = self.current_dirstate()
369
 
        if stat_value is None:
370
 
            try:
371
 
                stat_value = os.lstat(file_abspath)
372
 
            except OSError, e:
373
 
                if e.errno == errno.ENOENT:
374
 
                    return None
375
 
                else:
376
 
                    raise
377
 
        link_or_sha1 = dirstate.update_entry(state, entry, file_abspath,
378
 
            stat_value=stat_value)
 
404
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
405
                                          stat_value=stat_value)
379
406
        if entry[1][0][0] == 'f':
380
 
            if link_or_sha1 is None:
381
 
                file_obj, statvalue = self.get_file_with_stat(file_id, path)
382
 
                try:
383
 
                    sha1 = osutils.sha_file(file_obj)
384
 
                finally:
385
 
                    file_obj.close()
386
 
                self._observed_sha1(file_id, path, (sha1, statvalue))
387
 
                return sha1
388
 
            else:
389
 
                return link_or_sha1
 
407
            return link_or_sha1
390
408
        return None
391
409
 
392
410
    def _get_inventory(self):
393
411
        """Get the inventory for the tree. This is only valid within a lock."""
394
 
        if 'evil' in debug.debug_flags:
395
 
            trace.mutter_callsite(2,
396
 
                "accessing .inventory forces a size of tree translation.")
397
412
        if self._inventory is not None:
398
413
            return self._inventory
399
414
        self._must_be_locked()
406
421
    @needs_read_lock
407
422
    def get_parent_ids(self):
408
423
        """See Tree.get_parent_ids.
409
 
 
 
424
        
410
425
        This implementation requests the ids list from the dirstate file.
411
426
        """
412
427
        return self.current_dirstate().get_parent_ids()
428
443
 
429
444
    def has_id(self, file_id):
430
445
        state = self.current_dirstate()
 
446
        file_id = osutils.safe_file_id(file_id)
431
447
        row, parents = self._get_entry(file_id=file_id)
432
448
        if row is None:
433
449
            return False
434
450
        return osutils.lexists(pathjoin(
435
451
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
452
 
437
 
    def has_or_had_id(self, file_id):
438
 
        state = self.current_dirstate()
439
 
        row, parents = self._get_entry(file_id=file_id)
440
 
        return row is not None
441
 
 
442
453
    @needs_read_lock
443
454
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
455
        file_id = osutils.safe_file_id(file_id)
445
456
        state = self.current_dirstate()
446
457
        entry = self._get_entry(file_id=file_id)
447
458
        if entry == (None, None):
449
460
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
461
        return path_utf8.decode('utf8')
451
462
 
452
 
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
453
 
        entry = self._get_entry(path=path)
454
 
        if entry == (None, None):
455
 
            return False # Missing entries are not executable
456
 
        return entry[1][0][3] # Executable?
457
 
 
458
 
    if not osutils.supports_executable():
459
 
        def is_executable(self, file_id, path=None):
460
 
            """Test if a file is executable or not.
461
 
 
462
 
            Note: The caller is expected to take a read-lock before calling this.
463
 
            """
464
 
            entry = self._get_entry(file_id=file_id, path=path)
465
 
            if entry == (None, None):
466
 
                return False
467
 
            return entry[1][0][3]
468
 
 
469
 
        _is_executable_from_path_and_stat = \
470
 
            _is_executable_from_path_and_stat_from_basis
471
 
    else:
472
 
        def is_executable(self, file_id, path=None):
473
 
            """Test if a file is executable or not.
474
 
 
475
 
            Note: The caller is expected to take a read-lock before calling this.
476
 
            """
477
 
            self._must_be_locked()
478
 
            if not path:
479
 
                path = self.id2path(file_id)
480
 
            mode = os.lstat(self.abspath(path)).st_mode
481
 
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
482
 
 
483
 
    def all_file_ids(self):
484
 
        """See Tree.iter_all_file_ids"""
485
 
        self._must_be_locked()
486
 
        result = set()
487
 
        for key, tree_details in self.current_dirstate()._iter_entries():
488
 
            if tree_details[0][0] in ('a', 'r'): # relocated
489
 
                continue
490
 
            result.add(key[2])
491
 
        return result
492
 
 
493
463
    @needs_read_lock
494
464
    def __iter__(self):
495
465
        """Iterate through file_ids for this tree.
508
478
        return iter(result)
509
479
 
510
480
    def iter_references(self):
511
 
        if not self._repo_supports_tree_reference:
512
 
            # When the repo doesn't support references, we will have nothing to
513
 
            # return
514
 
            return
515
481
        for key, tree_details in self.current_dirstate()._iter_entries():
516
482
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
483
                # not relevant to the working tree
519
485
            if not key[1]:
520
486
                # the root is not a reference.
521
487
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
 
488
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
523
489
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
 
490
                if self._kind(path) == 'tree-reference':
 
491
                    yield path, key[2]
526
492
            except errors.NoSuchFile:
527
493
                # path is missing on disk.
528
494
                continue
529
495
 
530
 
    def _observed_sha1(self, file_id, path, (sha1, statvalue)):
531
 
        """See MutableTree._observed_sha1."""
532
 
        state = self.current_dirstate()
533
 
        entry = self._get_entry(file_id=file_id, path=path)
534
 
        state._observed_sha1(entry, sha1, statvalue)
535
 
 
 
496
    @needs_read_lock
536
497
    def kind(self, file_id):
537
498
        """Return the kind of a file.
538
499
 
539
500
        This is always the actual kind that's on disk, regardless of what it
540
501
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
502
        """
544
503
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
 
504
        assert relpath != None, \
 
505
            "path for id {%s} is None!" % file_id
548
506
        return self._kind(relpath)
549
507
 
550
508
    def _kind(self, relpath):
551
509
        abspath = self.abspath(relpath)
552
510
        kind = file_kind(abspath)
553
 
        if (self._repo_supports_tree_reference and kind == 'directory'):
554
 
            entry = self._get_entry(path=relpath)
555
 
            if entry[1] is not None:
556
 
                if entry[1][0][0] == 't':
557
 
                    kind = 'tree-reference'
 
511
        if (self._repo_supports_tree_reference and
 
512
            kind == 'directory' and
 
513
            self._directory_is_tree_reference(relpath)):
 
514
            kind = 'tree-reference'
558
515
        return kind
559
516
 
560
517
    @needs_read_lock
564
521
        if parent_ids:
565
522
            return parent_ids[0]
566
523
        else:
567
 
            return _mod_revision.NULL_REVISION
 
524
            return None
568
525
 
569
526
    def lock_read(self):
570
 
        """See Branch.lock_read, and WorkingTree.unlock.
571
 
 
572
 
        :return: An object with an unlock method which will release the lock
573
 
            obtained.
574
 
        """
 
527
        """See Branch.lock_read, and WorkingTree.unlock."""
575
528
        self.branch.lock_read()
576
529
        try:
577
530
            self._control_files.lock_read()
582
535
                # set our support for tree references from the repository in
583
536
                # use.
584
537
                self._repo_supports_tree_reference = getattr(
585
 
                    self.branch.repository._format, "supports_tree_reference",
 
538
                    self.branch.repository._format, "support_tree_reference",
586
539
                    False)
587
540
            except:
588
541
                self._control_files.unlock()
590
543
        except:
591
544
            self.branch.unlock()
592
545
            raise
593
 
        return self
594
546
 
595
547
    def _lock_self_write(self):
596
548
        """This should be called after the branch is locked."""
603
555
                # set our support for tree references from the repository in
604
556
                # use.
605
557
                self._repo_supports_tree_reference = getattr(
606
 
                    self.branch.repository._format, "supports_tree_reference",
 
558
                    self.branch.repository._format, "support_tree_reference",
607
559
                    False)
608
560
            except:
609
561
                self._control_files.unlock()
611
563
        except:
612
564
            self.branch.unlock()
613
565
            raise
614
 
        return self
615
566
 
616
567
    def lock_tree_write(self):
617
 
        """See MutableTree.lock_tree_write, and WorkingTree.unlock.
618
 
 
619
 
        :return: An object with an unlock method which will release the lock
620
 
            obtained.
621
 
        """
 
568
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
622
569
        self.branch.lock_read()
623
 
        return self._lock_self_write()
 
570
        self._lock_self_write()
624
571
 
625
572
    def lock_write(self):
626
 
        """See MutableTree.lock_write, and WorkingTree.unlock.
627
 
 
628
 
        :return: An object with an unlock method which will release the lock
629
 
            obtained.
630
 
        """
 
573
        """See MutableTree.lock_write, and WorkingTree.unlock."""
631
574
        self.branch.lock_write()
632
 
        return self._lock_self_write()
 
575
        self._lock_self_write()
633
576
 
634
577
    @needs_tree_write_lock
635
578
    def move(self, from_paths, to_dir, after=False):
637
580
        result = []
638
581
        if not from_paths:
639
582
            return result
 
583
 
640
584
        state = self.current_dirstate()
641
 
        if isinstance(from_paths, basestring):
642
 
            raise ValueError()
 
585
 
 
586
        assert not isinstance(from_paths, basestring)
643
587
        to_dir_utf8 = to_dir.encode('utf8')
644
588
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
645
589
        id_index = state._get_id_index()
667
611
        if self._inventory is not None:
668
612
            update_inventory = True
669
613
            inv = self.inventory
 
614
            to_dir_ie = inv[to_dir_id]
670
615
            to_dir_id = to_entry[0][2]
671
 
            to_dir_ie = inv[to_dir_id]
672
616
        else:
673
617
            update_inventory = False
674
618
 
697
641
            new_entry = to_block[1][added_entry_index]
698
642
            rollbacks.append(lambda:state._make_absent(new_entry))
699
643
 
 
644
        # create rename entries and tuples
700
645
        for from_rel in from_paths:
701
646
            # from_rel is 'pathinroot/foo/bar'
702
647
            from_rel_utf8 = from_rel.encode('utf8')
705
650
            from_entry = self._get_entry(path=from_rel)
706
651
            if from_entry == (None, None):
707
652
                raise errors.BzrMoveFailedError(from_rel,to_dir,
708
 
                    errors.NotVersionedError(path=from_rel))
 
653
                    errors.NotVersionedError(path=str(from_rel)))
709
654
 
710
655
            from_id = from_entry[0][2]
711
656
            to_rel = pathjoin(to_dir, from_tail)
738
683
                if from_missing: # implicitly just update our path mapping
739
684
                    move_file = False
740
685
                elif not after:
741
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
686
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
687
                        extra="(Use --after to update the Bazaar id)")
742
688
 
743
689
            rollbacks = []
744
690
            def rollback_rename():
799
745
 
800
746
                if minikind == 'd':
801
747
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
802
 
                        """Recursively update all entries in this dirblock."""
803
 
                        if from_dir == '':
804
 
                            raise AssertionError("renaming root not supported")
 
748
                        """all entries in this block need updating.
 
749
 
 
750
                        TODO: This is pretty ugly, and doesn't support
 
751
                        reverting, but it works.
 
752
                        """
 
753
                        assert from_dir != '', "renaming root not supported"
805
754
                        from_key = (from_dir, '')
806
755
                        from_block_idx, present = \
807
756
                            state._find_block_index_from_key(from_key)
817
766
                        to_block_index = state._ensure_block(
818
767
                            to_block_index, to_entry_index, to_dir_utf8)
819
768
                        to_block = state._dirblocks[to_block_index]
820
 
 
821
 
                        # Grab a copy since move_one may update the list.
822
 
                        for entry in from_block[1][:]:
823
 
                            if not (entry[0][0] == from_dir):
824
 
                                raise AssertionError()
 
769
                        for entry in from_block[1]:
 
770
                            assert entry[0][0] == from_dir
825
771
                            cur_details = entry[1][0]
826
772
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
827
773
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
828
774
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
829
775
                            minikind = cur_details[0]
830
 
                            if minikind in 'ar':
831
 
                                # Deleted children of a renamed directory
832
 
                                # Do not need to be updated.
833
 
                                # Children that have been renamed out of this
834
 
                                # directory should also not be updated
835
 
                                continue
836
776
                            move_one(entry, from_path_utf8=from_path_utf8,
837
777
                                     minikind=minikind,
838
778
                                     executable=cur_details[3],
841
781
                                     size=cur_details[2],
842
782
                                     to_block=to_block,
843
783
                                     to_key=to_key,
844
 
                                     to_path_utf8=to_path_utf8)
 
784
                                     to_path_utf8=to_rel_utf8)
845
785
                            if minikind == 'd':
846
786
                                # We need to move all the children of this
847
787
                                # entry
886
826
        for tree in trees:
887
827
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
888
828
                parents):
889
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
890
 
                    trees, require_versioned)
 
829
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
891
830
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
892
831
        # -- make all paths utf8 --
893
832
        paths_utf8 = set()
953
892
            if not all_versioned:
954
893
                raise errors.PathsNotVersionedError(paths)
955
894
        # -- remove redundancy in supplied paths to prevent over-scanning --
956
 
        search_paths = osutils.minimum_path_selection(paths)
957
 
        # sketch:
 
895
        search_paths = set()
 
896
        for path in paths:
 
897
            other_paths = paths.difference(set([path]))
 
898
            if not osutils.is_inside_any(other_paths, path):
 
899
                # this is a top level path, we must check it.
 
900
                search_paths.add(path)
 
901
        # sketch: 
958
902
        # for all search_indexs in each path at or under each element of
959
903
        # search_paths, if the detail is relocated: add the id, and add the
960
904
        # relocated path as one to search if its not searched already. If the
1016
960
 
1017
961
    def read_working_inventory(self):
1018
962
        """Read the working inventory.
1019
 
 
 
963
        
1020
964
        This is a meaningless operation for dirstate, but we obey it anyhow.
1021
965
        """
1022
966
        return self.inventory
1027
971
 
1028
972
        WorkingTree4 supplies revision_trees for any basis tree.
1029
973
        """
 
974
        revision_id = osutils.safe_revision_id(revision_id)
1030
975
        dirstate = self.current_dirstate()
1031
976
        parent_ids = dirstate.get_parent_ids()
1032
977
        if revision_id not in parent_ids:
1039
984
    @needs_tree_write_lock
1040
985
    def set_last_revision(self, new_revision):
1041
986
        """Change the last revision in the working tree."""
 
987
        new_revision = osutils.safe_revision_id(new_revision)
1042
988
        parents = self.get_parent_ids()
1043
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1044
 
            if len(parents) >= 2:
1045
 
                raise AssertionError(
1046
 
                    "setting the last parent to none with a pending merge is "
1047
 
                    "unsupported.")
 
989
        if new_revision in (NULL_REVISION, None):
 
990
            assert len(parents) < 2, (
 
991
                "setting the last parent to none with a pending merge is "
 
992
                "unsupported.")
1048
993
            self.set_parent_ids([])
1049
994
        else:
1050
995
            self.set_parent_ids([new_revision] + parents[1:],
1053
998
    @needs_tree_write_lock
1054
999
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
1000
        """Set the parent ids to revision_ids.
1056
 
 
 
1001
        
1057
1002
        See also set_parent_trees. This api will try to retrieve the tree data
1058
1003
        for each element of revision_ids from the trees repository. If you have
1059
1004
        tree data already available, it is more efficient to use
1063
1008
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
1009
            working tree. Any of these may be ghosts.
1065
1010
        """
 
1011
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1066
1012
        trees = []
1067
1013
        for revision_id in revision_ids:
1068
1014
            try:
1074
1020
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1075
1021
                revtree = None
1076
1022
            trees.append((revision_id, revtree))
 
1023
        self.current_dirstate()._validate()
1077
1024
        self.set_parent_trees(trees,
1078
1025
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1026
        self.current_dirstate()._validate()
1079
1027
 
1080
1028
    @needs_tree_write_lock
1081
1029
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1086
1034
            parent tree - i.e. a ghost.
1087
1035
        """
1088
1036
        dirstate = self.current_dirstate()
 
1037
        dirstate._validate()
1089
1038
        if len(parents_list) > 0:
1090
1039
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1091
1040
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1092
1041
        real_trees = []
1093
1042
        ghosts = []
1094
 
 
1095
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1096
 
        graph = self.branch.repository.get_graph()
1097
 
        heads = graph.heads(parent_ids)
1098
 
        accepted_revisions = set()
1099
 
 
1100
1043
        # convert absent trees to the null tree, which we convert back to
1101
1044
        # missing on access.
1102
1045
        for rev_id, tree in parents_list:
1103
 
            if len(accepted_revisions) > 0:
1104
 
                # we always accept the first tree
1105
 
                if rev_id in accepted_revisions or rev_id not in heads:
1106
 
                    # We have already included either this tree, or its
1107
 
                    # descendent, so we skip it.
1108
 
                    continue
1109
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1046
            rev_id = osutils.safe_revision_id(rev_id)
1110
1047
            if tree is not None:
1111
1048
                real_trees.append((rev_id, tree))
1112
1049
            else:
1113
1050
                real_trees.append((rev_id,
1114
 
                    self.branch.repository.revision_tree(
1115
 
                        _mod_revision.NULL_REVISION)))
 
1051
                    self.branch.repository.revision_tree(None)))
1116
1052
                ghosts.append(rev_id)
1117
 
            accepted_revisions.add(rev_id)
 
1053
        dirstate._validate()
1118
1054
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1055
        dirstate._validate()
1119
1056
        self._make_dirty(reset_inventory=False)
 
1057
        dirstate._validate()
1120
1058
 
1121
1059
    def _set_root_id(self, file_id):
1122
1060
        """See WorkingTree.set_root_id."""
1125
1063
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1126
1064
            self._make_dirty(reset_inventory=True)
1127
1065
 
1128
 
    def _sha_from_stat(self, path, stat_result):
1129
 
        """Get a sha digest from the tree's stat cache.
1130
 
 
1131
 
        The default implementation assumes no stat cache is present.
1132
 
 
1133
 
        :param path: The path.
1134
 
        :param stat_result: The stat result being looked up.
1135
 
        """
1136
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1137
 
 
1138
1066
    @needs_read_lock
1139
1067
    def supports_tree_reference(self):
1140
1068
        return self._repo_supports_tree_reference
1141
1069
 
1142
1070
    def unlock(self):
1143
1071
        """Unlock in format 4 trees needs to write the entire dirstate."""
1144
 
        # do non-implementation specific cleanup
1145
 
        self._cleanup()
1146
 
 
1147
1072
        if self._control_files._lock_count == 1:
1148
1073
            # eventually we should do signature checking during read locks for
1149
1074
            # dirstate updates.
1180
1105
            return
1181
1106
        state = self.current_dirstate()
1182
1107
        state._read_dirblocks_if_needed()
1183
 
        ids_to_unversion = set(file_ids)
 
1108
        ids_to_unversion = set()
 
1109
        for file_id in file_ids:
 
1110
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1184
1111
        paths_to_unversion = set()
1185
1112
        # sketch:
1186
1113
        # check if the root is to be unversioned, if so, assert for now.
1213
1140
                # just forget the whole block.
1214
1141
                entry_index = 0
1215
1142
                while entry_index < len(block[1]):
1216
 
                    entry = block[1][entry_index]
1217
 
                    if entry[1][0][0] in 'ar':
1218
 
                        # don't remove absent or renamed entries
 
1143
                    # Mark this file id as having been removed
 
1144
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1145
                    if not state._make_absent(block[1][entry_index]):
1219
1146
                        entry_index += 1
1220
 
                    else:
1221
 
                        # Mark this file id as having been removed
1222
 
                        ids_to_unversion.discard(entry[0][2])
1223
 
                        if not state._make_absent(entry):
1224
 
                            # The block has not shrunk.
1225
 
                            entry_index += 1
1226
1147
                # go to the next block. (At the moment we dont delete empty
1227
1148
                # dirblocks)
1228
1149
                block_index += 1
1252
1173
                self._inventory.remove_recursive_id(file_id)
1253
1174
 
1254
1175
    @needs_tree_write_lock
1255
 
    def rename_one(self, from_rel, to_rel, after=False):
1256
 
        """See WorkingTree.rename_one"""
1257
 
        self.flush()
1258
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1259
 
 
1260
 
    @needs_tree_write_lock
1261
 
    def apply_inventory_delta(self, changes):
1262
 
        """See MutableTree.apply_inventory_delta"""
1263
 
        state = self.current_dirstate()
1264
 
        state.update_by_delta(changes)
1265
 
        self._make_dirty(reset_inventory=True)
1266
 
 
1267
 
    def update_basis_by_delta(self, new_revid, delta):
1268
 
        """See MutableTree.update_basis_by_delta."""
1269
 
        if self.last_revision() == new_revid:
1270
 
            raise AssertionError()
1271
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1272
 
 
1273
 
    @needs_read_lock
1274
 
    def _validate(self):
1275
 
        self._dirstate._validate()
1276
 
 
1277
 
    @needs_tree_write_lock
1278
1176
    def _write_inventory(self, inv):
1279
1177
        """Write inventory as the current inventory."""
1280
 
        if self._dirty:
1281
 
            raise AssertionError("attempting to write an inventory when the "
1282
 
                "dirstate is dirty will lose pending changes")
1283
 
        had_inventory = self._inventory is not None
1284
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1285
 
        # working inventory. We do this because self.inventory may be inv, or
1286
 
        # may have been modified, and either case would prevent a clean delta
1287
 
        # being created.
1288
 
        self._inventory = None
1289
 
        # generate a delta,
1290
 
        delta = inv._make_delta(self.inventory)
1291
 
        # and apply it.
1292
 
        self.apply_inventory_delta(delta)
1293
 
        if had_inventory:
 
1178
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1179
        self.current_dirstate().set_state_from_inventory(inv)
 
1180
        self._make_dirty(reset_inventory=False)
 
1181
        if self._inventory is not None:
1294
1182
            self._inventory = inv
1295
1183
        self.flush()
1296
1184
 
1297
1185
 
1298
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1299
 
 
1300
 
    def __init__(self, tree):
1301
 
        self.tree = tree
1302
 
 
1303
 
    def sha1(self, abspath):
1304
 
        """See dirstate.SHA1Provider.sha1()."""
1305
 
        filters = self.tree._content_filter_stack(
1306
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1307
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1308
 
 
1309
 
    def stat_and_sha1(self, abspath):
1310
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1311
 
        filters = self.tree._content_filter_stack(
1312
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1313
 
        file_obj = file(abspath, 'rb', 65000)
1314
 
        try:
1315
 
            statvalue = os.fstat(file_obj.fileno())
1316
 
            if filters:
1317
 
                file_obj = filtered_input_file(file_obj, filters)
1318
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1319
 
        finally:
1320
 
            file_obj.close()
1321
 
        return statvalue, sha1
1322
 
 
1323
 
 
1324
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1325
 
    """Dirstate working tree that supports content filtering.
1326
 
 
1327
 
    The dirstate holds the hash and size of the canonical form of the file, 
1328
 
    and most methods must return that.
1329
 
    """
1330
 
 
1331
 
    def _file_content_summary(self, path, stat_result):
1332
 
        # This is to support the somewhat obsolete path_content_summary method
1333
 
        # with content filtering: see
1334
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1335
 
        #
1336
 
        # If the dirstate cache is up to date and knows the hash and size,
1337
 
        # return that.
1338
 
        # Otherwise if there are no content filters, return the on-disk size
1339
 
        # and leave the hash blank.
1340
 
        # Otherwise, read and filter the on-disk file and use its size and
1341
 
        # hash.
1342
 
        #
1343
 
        # The dirstate doesn't store the size of the canonical form so we
1344
 
        # can't trust it for content-filtered trees.  We just return None.
1345
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1346
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1347
 
        return ('file', None, executable, dirstate_sha1)
1348
 
 
1349
 
 
1350
 
class WorkingTree4(DirStateWorkingTree):
1351
 
    """This is the Format 4 working tree.
1352
 
 
1353
 
    This differs from WorkingTree3 by:
1354
 
     - Having a consolidated internal dirstate, stored in a
1355
 
       randomly-accessible sorted file on disk.
1356
 
     - Not having a regular inventory attribute.  One can be synthesized
1357
 
       on demand but this is expensive and should be avoided.
1358
 
 
1359
 
    This is new in bzr 0.15.
1360
 
    """
1361
 
 
1362
 
 
1363
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1364
 
    """This is the Format 5 working tree.
1365
 
 
1366
 
    This differs from WorkingTree4 by:
1367
 
     - Supporting content filtering.
1368
 
 
1369
 
    This is new in bzr 1.11.
1370
 
    """
1371
 
 
1372
 
 
1373
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1374
 
    """This is the Format 6 working tree.
1375
 
 
1376
 
    This differs from WorkingTree5 by:
1377
 
     - Supporting a current view that may mask the set of files in a tree
1378
 
       impacted by most user operations.
1379
 
 
1380
 
    This is new in bzr 1.14.
1381
 
    """
1382
 
 
1383
 
    def _make_views(self):
1384
 
        return views.PathBasedViews(self)
1385
 
 
1386
 
 
1387
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1388
 
 
1389
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1390
 
                   accelerator_tree=None, hardlink=False):
 
1186
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1187
    """The first consolidated dirstate working tree format.
 
1188
 
 
1189
    This format:
 
1190
        - exists within a metadir controlling .bzr
 
1191
        - includes an explicit version marker for the workingtree control
 
1192
          files, separate from the BzrDir format
 
1193
        - modifies the hash cache format
 
1194
        - is new in bzr 0.15
 
1195
        - uses a LockDir to guard access to it.
 
1196
    """
 
1197
 
 
1198
    def get_format_string(self):
 
1199
        """See WorkingTreeFormat.get_format_string()."""
 
1200
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1201
 
 
1202
    def get_format_description(self):
 
1203
        """See WorkingTreeFormat.get_format_description()."""
 
1204
        return "Working tree format 4"
 
1205
 
 
1206
    def initialize(self, a_bzrdir, revision_id=None):
1391
1207
        """See WorkingTreeFormat.initialize().
1392
1208
 
1393
1209
        :param revision_id: allows creating a working tree at a different
1394
1210
        revision than the branch is at.
1395
 
        :param accelerator_tree: A tree which can be used for retrieving file
1396
 
            contents more quickly than the revision tree, i.e. a workingtree.
1397
 
            The revision tree will be used for cases where accelerator_tree's
1398
 
            content is different.
1399
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1400
 
            where possible.
1401
1211
 
1402
 
        These trees get an initial random root id, if their repository supports
1403
 
        rich root data, TREE_ROOT otherwise.
 
1212
        These trees get an initial random root id.
1404
1213
        """
 
1214
        revision_id = osutils.safe_revision_id(revision_id)
1405
1215
        if not isinstance(a_bzrdir.transport, LocalTransport):
1406
1216
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1407
1217
        transport = a_bzrdir.get_workingtree_transport(self)
1408
1218
        control_files = self._open_control_files(a_bzrdir)
1409
1219
        control_files.create_lock()
1410
1220
        control_files.lock_write()
1411
 
        transport.put_bytes('format', self.get_format_string(),
1412
 
            mode=a_bzrdir._get_file_mode())
1413
 
        if from_branch is not None:
1414
 
            branch = from_branch
1415
 
        else:
1416
 
            branch = a_bzrdir.open_branch()
 
1221
        control_files.put_utf8('format', self.get_format_string())
 
1222
        branch = a_bzrdir.open_branch()
1417
1223
        if revision_id is None:
1418
1224
            revision_id = branch.last_revision()
1419
1225
        local_path = transport.local_abspath('dirstate')
1420
1226
        # write out new dirstate (must exist when we create the tree)
1421
1227
        state = dirstate.DirState.initialize(local_path)
1422
1228
        state.unlock()
1423
 
        del state
1424
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1229
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1425
1230
                         branch,
1426
1231
                         _format=self,
1427
1232
                         _bzrdir=a_bzrdir,
1428
1233
                         _control_files=control_files)
1429
1234
        wt._new_tree()
1430
1235
        wt.lock_tree_write()
 
1236
        state._validate()
1431
1237
        try:
1432
 
            self._init_custom_control_files(wt)
1433
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1434
 
                if branch.repository.supports_rich_root():
1435
 
                    wt._set_root_id(generate_ids.gen_root_id())
1436
 
                else:
1437
 
                    wt._set_root_id(ROOT_ID)
 
1238
            if revision_id in (None, NULL_REVISION):
 
1239
                wt._set_root_id(generate_ids.gen_root_id())
1438
1240
                wt.flush()
1439
 
            basis = None
1440
 
            # frequently, we will get here due to branching.  The accelerator
1441
 
            # tree will be the tree from the branch, so the desired basis
1442
 
            # tree will often be a parent of the accelerator tree.
1443
 
            if accelerator_tree is not None:
1444
 
                try:
1445
 
                    basis = accelerator_tree.revision_tree(revision_id)
1446
 
                except errors.NoSuchRevision:
1447
 
                    pass
1448
 
            if basis is None:
1449
 
                basis = branch.repository.revision_tree(revision_id)
1450
 
            if revision_id == _mod_revision.NULL_REVISION:
1451
 
                parents_list = []
1452
 
            else:
1453
 
                parents_list = [(revision_id, basis)]
 
1241
                wt.current_dirstate()._validate()
 
1242
            wt.set_last_revision(revision_id)
 
1243
            wt.flush()
 
1244
            basis = wt.basis_tree()
1454
1245
            basis.lock_read()
1455
 
            try:
1456
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1246
            # if the basis has a root id we have to use that; otherwise we use
 
1247
            # a new random one
 
1248
            basis_root_id = basis.get_root_id()
 
1249
            if basis_root_id is not None:
 
1250
                wt._set_root_id(basis_root_id)
1457
1251
                wt.flush()
1458
 
                # if the basis has a root id we have to use that; otherwise we
1459
 
                # use a new random one
1460
 
                basis_root_id = basis.get_root_id()
1461
 
                if basis_root_id is not None:
1462
 
                    wt._set_root_id(basis_root_id)
1463
 
                    wt.flush()
1464
 
                if wt.supports_content_filtering():
1465
 
                    # The original tree may not have the same content filters
1466
 
                    # applied so we can't safely build the inventory delta from
1467
 
                    # the source tree.
1468
 
                    delta_from_tree = False
1469
 
                else:
1470
 
                    delta_from_tree = True
1471
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1472
 
                # because wt4.apply_inventory_delta does not mutate the input
1473
 
                # inventory entries.
1474
 
                transform.build_tree(basis, wt, accelerator_tree,
1475
 
                                     hardlink=hardlink,
1476
 
                                     delta_from_tree=delta_from_tree)
1477
 
            finally:
1478
 
                basis.unlock()
 
1252
            transform.build_tree(basis, wt)
 
1253
            basis.unlock()
1479
1254
        finally:
1480
1255
            control_files.unlock()
1481
1256
            wt.unlock()
1482
1257
        return wt
1483
1258
 
1484
 
    def _init_custom_control_files(self, wt):
1485
 
        """Subclasses with custom control files should override this method.
1486
 
 
1487
 
        The working tree and control files are locked for writing when this
1488
 
        method is called.
1489
 
 
1490
 
        :param wt: the WorkingTree object
1491
 
        """
1492
 
 
1493
1259
    def _open(self, a_bzrdir, control_files):
1494
1260
        """Open the tree itself.
1495
1261
 
1496
1262
        :param a_bzrdir: the dir for the tree.
1497
1263
        :param control_files: the control files for the tree.
1498
1264
        """
1499
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1265
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1500
1266
                           branch=a_bzrdir.open_branch(),
1501
1267
                           _format=self,
1502
1268
                           _bzrdir=a_bzrdir,
1503
1269
                           _control_files=control_files)
1504
1270
 
1505
1271
    def __get_matchingbzrdir(self):
1506
 
        return self._get_matchingbzrdir()
1507
 
 
1508
 
    def _get_matchingbzrdir(self):
1509
 
        """Overrideable method to get a bzrdir for testing."""
1510
1272
        # please test against something that will let us do tree references
1511
1273
        return bzrdir.format_registry.make_bzrdir(
1512
1274
            'dirstate-with-subtree')
1514
1276
    _matchingbzrdir = property(__get_matchingbzrdir)
1515
1277
 
1516
1278
 
1517
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1518
 
    """The first consolidated dirstate working tree format.
1519
 
 
1520
 
    This format:
1521
 
        - exists within a metadir controlling .bzr
1522
 
        - includes an explicit version marker for the workingtree control
1523
 
          files, separate from the BzrDir format
1524
 
        - modifies the hash cache format
1525
 
        - is new in bzr 0.15
1526
 
        - uses a LockDir to guard access to it.
1527
 
    """
1528
 
 
1529
 
    upgrade_recommended = False
1530
 
 
1531
 
    _tree_class = WorkingTree4
1532
 
 
1533
 
    def get_format_string(self):
1534
 
        """See WorkingTreeFormat.get_format_string()."""
1535
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1536
 
 
1537
 
    def get_format_description(self):
1538
 
        """See WorkingTreeFormat.get_format_description()."""
1539
 
        return "Working tree format 4"
1540
 
 
1541
 
 
1542
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1543
 
    """WorkingTree format supporting content filtering.
1544
 
    """
1545
 
 
1546
 
    upgrade_recommended = False
1547
 
 
1548
 
    _tree_class = WorkingTree5
1549
 
 
1550
 
    def get_format_string(self):
1551
 
        """See WorkingTreeFormat.get_format_string()."""
1552
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1553
 
 
1554
 
    def get_format_description(self):
1555
 
        """See WorkingTreeFormat.get_format_description()."""
1556
 
        return "Working tree format 5"
1557
 
 
1558
 
    def supports_content_filtering(self):
1559
 
        return True
1560
 
 
1561
 
 
1562
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1563
 
    """WorkingTree format supporting views.
1564
 
    """
1565
 
 
1566
 
    upgrade_recommended = False
1567
 
 
1568
 
    _tree_class = WorkingTree6
1569
 
 
1570
 
    def get_format_string(self):
1571
 
        """See WorkingTreeFormat.get_format_string()."""
1572
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1573
 
 
1574
 
    def get_format_description(self):
1575
 
        """See WorkingTreeFormat.get_format_description()."""
1576
 
        return "Working tree format 6"
1577
 
 
1578
 
    def _init_custom_control_files(self, wt):
1579
 
        """Subclasses with custom control files should override this method."""
1580
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1581
 
 
1582
 
    def supports_content_filtering(self):
1583
 
        return True
1584
 
 
1585
 
    def supports_views(self):
1586
 
        return True
1587
 
 
1588
 
 
1589
1279
class DirStateRevisionTree(Tree):
1590
 
    """A revision tree pulling the inventory from a dirstate.
1591
 
    
1592
 
    Note that this is one of the historical (ie revision) trees cached in the
1593
 
    dirstate for easy access, not the workingtree.
1594
 
    """
 
1280
    """A revision tree pulling the inventory from a dirstate."""
1595
1281
 
1596
1282
    def __init__(self, dirstate, revision_id, repository):
1597
1283
        self._dirstate = dirstate
1598
 
        self._revision_id = revision_id
 
1284
        self._revision_id = osutils.safe_revision_id(revision_id)
1599
1285
        self._repository = repository
1600
1286
        self._inventory = None
1601
1287
        self._locked = 0
1602
1288
        self._dirstate_locked = False
1603
 
        self._repo_supports_tree_reference = getattr(
1604
 
            repository._format, "supports_tree_reference",
1605
 
            False)
1606
1289
 
1607
1290
    def __repr__(self):
1608
1291
        return "<%s of %s in %s>" % \
1609
1292
            (self.__class__.__name__, self._revision_id, self._dirstate)
1610
1293
 
1611
 
    def annotate_iter(self, file_id,
1612
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1294
    def annotate_iter(self, file_id):
1613
1295
        """See Tree.annotate_iter"""
1614
 
        text_key = (file_id, self.inventory[file_id].revision)
1615
 
        annotations = self._repository.texts.annotate(text_key)
1616
 
        return [(key[-1], line) for (key, line) in annotations]
 
1296
        w = self._repository.weave_store.get_weave(file_id,
 
1297
                           self._repository.get_transaction())
 
1298
        return w.annotate_iter(self.inventory[file_id].revision)
1617
1299
 
1618
 
    def _get_ancestors(self, default_revision):
1619
 
        return set(self._repository.get_ancestry(self._revision_id,
1620
 
                                                 topo_sorted=False))
1621
1300
    def _comparison_data(self, entry, path):
1622
1301
        """See Tree._comparison_data."""
1623
1302
        if entry is None:
1640
1319
    def get_root_id(self):
1641
1320
        return self.path2id('')
1642
1321
 
1643
 
    def id2path(self, file_id):
1644
 
        "Convert a file-id to a path."
1645
 
        entry = self._get_entry(file_id=file_id)
1646
 
        if entry == (None, None):
1647
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1648
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1649
 
        return path_utf8.decode('utf8')
1650
 
 
1651
 
    def iter_references(self):
1652
 
        if not self._repo_supports_tree_reference:
1653
 
            # When the repo doesn't support references, we will have nothing to
1654
 
            # return
1655
 
            return iter([])
1656
 
        # Otherwise, fall back to the default implementation
1657
 
        return super(DirStateRevisionTree, self).iter_references()
1658
 
 
1659
1322
    def _get_parent_index(self):
1660
1323
        """Return the index in the dirstate referenced by this tree."""
1661
1324
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1666
1329
        If either file_id or path is supplied, it is used as the key to lookup.
1667
1330
        If both are supplied, the fastest lookup is used, and an error is
1668
1331
        raised if they do not both point at the same row.
1669
 
 
 
1332
        
1670
1333
        :param file_id: An optional unicode file_id to be looked up.
1671
1334
        :param path: An optional unicode path to be looked up.
1672
1335
        :return: The dirstate row tuple for path/file_id, or (None, None)
1673
1336
        """
1674
1337
        if file_id is None and path is None:
1675
1338
            raise errors.BzrError('must supply file_id or path')
 
1339
        file_id = osutils.safe_file_id(file_id)
1676
1340
        if path is not None:
1677
1341
            path = path.encode('utf8')
1678
1342
        parent_index = self._get_parent_index()
1686
1350
 
1687
1351
        This is relatively expensive: we have to walk the entire dirstate.
1688
1352
        """
1689
 
        if not self._locked:
1690
 
            raise AssertionError(
1691
 
                'cannot generate inventory of an unlocked '
1692
 
                'dirstate revision tree')
 
1353
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1354
            'dirstate revision tree'
1693
1355
        # separate call for profiling - makes it clear where the costs are.
1694
1356
        self._dirstate._read_dirblocks_if_needed()
1695
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1696
 
            raise AssertionError(
1697
 
                'parent %s has disappeared from %s' % (
1698
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1357
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1358
            'parent %s has disappeared from %s' % (
 
1359
            self._revision_id, self._dirstate.get_parent_ids())
1699
1360
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1700
1361
        # This is identical now to the WorkingTree _generate_inventory except
1701
1362
        # for the tree index use.
1702
1363
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1703
1364
        current_id = root_key[2]
1704
 
        if current_entry[parent_index][0] != 'd':
1705
 
            raise AssertionError()
 
1365
        assert current_entry[parent_index][0] == 'd'
1706
1366
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1707
1367
        inv.root.revision = current_entry[parent_index][4]
1708
1368
        # Turn some things into local variables
1740
1400
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1741
1401
                elif kind == 'symlink':
1742
1402
                    inv_entry.executable = False
1743
 
                    inv_entry.text_size = None
 
1403
                    inv_entry.text_size = size
1744
1404
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1745
1405
                elif kind == 'tree-reference':
1746
1406
                    inv_entry.reference_revision = fingerprint or None
1748
1408
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1749
1409
                            % entry)
1750
1410
                # These checks cost us around 40ms on a 55k entry tree
1751
 
                if file_id in inv_byid:
1752
 
                    raise AssertionError('file_id %s already in'
1753
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1754
 
                if name_unicode in parent_ie.children:
1755
 
                    raise AssertionError('name %r already in parent'
1756
 
                        % (name_unicode,))
 
1411
                assert file_id not in inv_byid
 
1412
                assert name_unicode not in parent_ie.children
1757
1413
                inv_byid[file_id] = inv_entry
1758
1414
                parent_ie.children[name_unicode] = inv_entry
1759
1415
        self._inventory = inv
1769
1425
            return None
1770
1426
        parent_index = self._get_parent_index()
1771
1427
        last_changed_revision = entry[1][parent_index][4]
1772
 
        try:
1773
 
            rev = self._repository.get_revision(last_changed_revision)
1774
 
        except errors.NoSuchRevision:
1775
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1776
 
        return rev.timestamp
 
1428
        return self._repository.get_revision(last_changed_revision).timestamp
1777
1429
 
1778
1430
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1779
1431
        entry = self._get_entry(file_id=file_id, path=path)
1783
1435
            return parent_details[1]
1784
1436
        return None
1785
1437
 
1786
 
    def get_file(self, file_id, path=None):
 
1438
    def get_file(self, file_id):
1787
1439
        return StringIO(self.get_file_text(file_id))
1788
1440
 
 
1441
    def get_file_lines(self, file_id):
 
1442
        ie = self.inventory[file_id]
 
1443
        return self._repository.weave_store.get_weave(file_id,
 
1444
                self._repository.get_transaction()).get_lines(ie.revision)
 
1445
 
1789
1446
    def get_file_size(self, file_id):
1790
 
        """See Tree.get_file_size"""
1791
1447
        return self.inventory[file_id].text_size
1792
1448
 
1793
 
    def get_file_text(self, file_id, path=None):
1794
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1795
 
        return ''.join(content)
 
1449
    def get_file_text(self, file_id):
 
1450
        return ''.join(self.get_file_lines(file_id))
1796
1451
 
1797
1452
    def get_reference_revision(self, file_id, path=None):
1798
1453
        return self.inventory[file_id].reference_revision
1799
1454
 
1800
 
    def iter_files_bytes(self, desired_files):
1801
 
        """See Tree.iter_files_bytes.
1802
 
 
1803
 
        This version is implemented on top of Repository.iter_files_bytes"""
1804
 
        parent_index = self._get_parent_index()
1805
 
        repo_desired_files = []
1806
 
        for file_id, identifier in desired_files:
1807
 
            entry = self._get_entry(file_id)
1808
 
            if entry == (None, None):
1809
 
                raise errors.NoSuchId(self, file_id)
1810
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1811
 
                                       identifier))
1812
 
        return self._repository.iter_files_bytes(repo_desired_files)
1813
 
 
1814
1455
    def get_symlink_target(self, file_id):
1815
1456
        entry = self._get_entry(file_id=file_id)
1816
1457
        parent_index = self._get_parent_index()
1817
1458
        if entry[1][parent_index][0] != 'l':
1818
1459
            return None
1819
1460
        else:
1820
 
            target = entry[1][parent_index][1]
1821
 
            target = target.decode('utf8')
1822
 
            return target
 
1461
            # At present, none of the tree implementations supports non-ascii
 
1462
            # symlink targets. So we will just assume that the dirstate path is
 
1463
            # correct.
 
1464
            return entry[1][parent_index][1]
1823
1465
 
1824
1466
    def get_revision_id(self):
1825
1467
        """Return the revision id for this tree."""
1843
1485
        return bool(self.path2id(filename))
1844
1486
 
1845
1487
    def kind(self, file_id):
1846
 
        entry = self._get_entry(file_id=file_id)[1]
1847
 
        if entry is None:
1848
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1849
 
        parent_index = self._get_parent_index()
1850
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1851
 
 
1852
 
    def stored_kind(self, file_id):
1853
 
        """See Tree.stored_kind"""
1854
 
        return self.kind(file_id)
1855
 
 
1856
 
    def path_content_summary(self, path):
1857
 
        """See Tree.path_content_summary."""
1858
 
        id = self.inventory.path2id(path)
1859
 
        if id is None:
1860
 
            return ('missing', None, None, None)
1861
 
        entry = self._inventory[id]
1862
 
        kind = entry.kind
1863
 
        if kind == 'file':
1864
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1865
 
        elif kind == 'symlink':
1866
 
            return (kind, None, None, entry.symlink_target)
1867
 
        else:
1868
 
            return (kind, None, None, None)
 
1488
        return self.inventory[file_id].kind
1869
1489
 
1870
1490
    def is_executable(self, file_id, path=None):
1871
1491
        ie = self.inventory[file_id]
1873
1493
            return None
1874
1494
        return ie.executable
1875
1495
 
1876
 
    def is_locked(self):
1877
 
        return self._locked
1878
 
 
1879
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1496
    def list_files(self, include_root=False):
1880
1497
        # We use a standard implementation, because DirStateRevisionTree is
1881
1498
        # dealing with one of the parents of the current state
1882
1499
        inv = self._get_inventory()
1883
 
        if from_dir is None:
1884
 
            from_dir_id = None
1885
 
        else:
1886
 
            from_dir_id = inv.path2id(from_dir)
1887
 
            if from_dir_id is None:
1888
 
                # Directory not versioned
1889
 
                return
1890
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1891
 
        if inv.root is not None and not include_root and from_dir is None:
 
1500
        entries = inv.iter_entries()
 
1501
        if self.inventory.root is not None and not include_root:
1892
1502
            entries.next()
1893
1503
        for path, entry in entries:
1894
1504
            yield path, 'V', entry.kind, entry.file_id, entry
1895
1505
 
1896
1506
    def lock_read(self):
1897
 
        """Lock the tree for a set of operations.
1898
 
 
1899
 
        :return: An object with an unlock method which will release the lock
1900
 
            obtained.
1901
 
        """
 
1507
        """Lock the tree for a set of operations."""
1902
1508
        if not self._locked:
1903
1509
            self._repository.lock_read()
1904
1510
            if self._dirstate._lock_token is None:
1905
1511
                self._dirstate.lock_read()
1906
1512
                self._dirstate_locked = True
1907
1513
        self._locked += 1
1908
 
        return self
1909
1514
 
1910
1515
    def _must_be_locked(self):
1911
1516
        if not self._locked:
1932
1537
                self._dirstate_locked = False
1933
1538
            self._repository.unlock()
1934
1539
 
1935
 
    @needs_read_lock
1936
 
    def supports_tree_reference(self):
1937
 
        return self._repo_supports_tree_reference
1938
 
 
1939
1540
    def walkdirs(self, prefix=""):
1940
1541
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1941
 
        # implementation based on an inventory.
 
1542
        # implementation based on an inventory.  
1942
1543
        # This should be cleaned up to use the much faster Dirstate code
1943
1544
        # So for now, we just build up the parent inventory, and extract
1944
1545
        # it the same way RevisionTree does.
1973
1574
 
1974
1575
class InterDirStateTree(InterTree):
1975
1576
    """Fast path optimiser for changes_from with dirstate trees.
1976
 
 
1977
 
    This is used only when both trees are in the dirstate working file, and
1978
 
    the source is any parent within the dirstate, and the destination is
 
1577
    
 
1578
    This is used only when both trees are in the dirstate working file, and 
 
1579
    the source is any parent within the dirstate, and the destination is 
1979
1580
    the current working tree of the same dirstate.
1980
1581
    """
1981
1582
    # this could be generalized to allow comparisons between any trees in the
1994
1595
        target.set_parent_ids([revid])
1995
1596
        return target.basis_tree(), target
1996
1597
 
1997
 
    @classmethod
1998
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1999
 
        result = klass.make_source_parent_tree(source, target)
2000
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
2001
 
        return result
2002
 
 
2003
 
    @classmethod
2004
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
2005
 
                                                  target):
2006
 
        from bzrlib.tests.test__dirstate_helpers import \
2007
 
            compiled_dirstate_helpers_feature
2008
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
2009
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
2010
 
        result = klass.make_source_parent_tree(source, target)
2011
 
        result[1]._iter_changes = ProcessEntryC
2012
 
        return result
2013
 
 
2014
1598
    _matching_from_tree_format = WorkingTreeFormat4()
2015
1599
    _matching_to_tree_format = WorkingTreeFormat4()
2016
 
 
2017
 
    @classmethod
2018
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
2019
 
        # This method shouldn't be called, because we have python and C
2020
 
        # specific flavours.
2021
 
        raise NotImplementedError
2022
 
 
2023
 
    def iter_changes(self, include_unchanged=False,
 
1600
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1601
 
 
1602
    def _iter_changes(self, include_unchanged=False,
2024
1603
                      specific_files=None, pb=None, extra_trees=[],
2025
1604
                      require_versioned=True, want_unversioned=False):
2026
1605
        """Return the changes from source to target.
2027
1606
 
2028
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1607
        :return: An iterator that yields tuples. See InterTree._iter_changes
2029
1608
            for details.
2030
1609
        :param specific_files: An optional list of file paths to restrict the
2031
1610
            comparison to. When mapping filenames to ids, all matches in all
2042
1621
            output. An unversioned file is defined as one with (False, False)
2043
1622
            for the versioned pair.
2044
1623
        """
 
1624
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1625
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1626
        # NB: show_status depends on being able to pass in non-versioned files
 
1627
        # and report them as unknown
2045
1628
        # TODO: handle extra trees in the dirstate.
2046
 
        if (extra_trees or specific_files == []):
 
1629
        # TODO: handle comparisons as an empty tree as a different special
 
1630
        # case? mbp 20070226
 
1631
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2047
1632
            # we can't fast-path these cases (yet)
2048
 
            return super(InterDirStateTree, self).iter_changes(
 
1633
            for f in super(InterDirStateTree, self)._iter_changes(
2049
1634
                include_unchanged, specific_files, pb, extra_trees,
2050
 
                require_versioned, want_unversioned=want_unversioned)
 
1635
                require_versioned, want_unversioned=want_unversioned):
 
1636
                yield f
 
1637
            return
2051
1638
        parent_ids = self.target.get_parent_ids()
2052
 
        if not (self.source._revision_id in parent_ids
2053
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2054
 
            raise AssertionError(
2055
 
                "revision {%s} is not stored in {%s}, but %s "
2056
 
                "can only be used for trees stored in the dirstate"
2057
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1639
        assert (self.source._revision_id in parent_ids), \
 
1640
                "revision {%s} is not stored in {%s}, but %s " \
 
1641
                "can only be used for trees stored in the dirstate" \
 
1642
                % (self.source._revision_id, self.target, self._iter_changes)
2058
1643
        target_index = 0
2059
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1644
        if self.source._revision_id == NULL_REVISION:
2060
1645
            source_index = None
2061
1646
            indices = (target_index,)
2062
1647
        else:
2063
 
            if not (self.source._revision_id in parent_ids):
2064
 
                raise AssertionError(
2065
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2066
 
                    self.source._revision_id, parent_ids))
 
1648
            assert (self.source._revision_id in parent_ids), \
 
1649
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1650
                self.source._revision_id, parent_ids)
2067
1651
            source_index = 1 + parent_ids.index(self.source._revision_id)
2068
 
            indices = (source_index, target_index)
 
1652
            indices = (source_index,target_index)
2069
1653
        # -- make all specific_files utf8 --
2070
1654
        if specific_files:
2071
1655
            specific_files_utf8 = set()
2072
1656
            for path in specific_files:
2073
 
                # Note, if there are many specific files, using cache_utf8
2074
 
                # would be good here.
2075
1657
                specific_files_utf8.add(path.encode('utf8'))
2076
1658
            specific_files = specific_files_utf8
2077
1659
        else:
2078
1660
            specific_files = set([''])
2079
1661
        # -- specific_files is now a utf8 path set --
2080
 
 
2081
1662
        # -- get the state object and prepare it.
2082
1663
        state = self.target.current_dirstate()
2083
1664
        state._read_dirblocks_if_needed()
 
1665
        def _entries_for_path(path):
 
1666
            """Return a list with all the entries that match path for all ids.
 
1667
            """
 
1668
            dirname, basename = os.path.split(path)
 
1669
            key = (dirname, basename, '')
 
1670
            block_index, present = state._find_block_index_from_key(key)
 
1671
            if not present:
 
1672
                # the block which should contain path is absent.
 
1673
                return []
 
1674
            result = []
 
1675
            block = state._dirblocks[block_index][1]
 
1676
            entry_index, _ = state._find_entry_index(key, block)
 
1677
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1678
            while (entry_index < len(block) and
 
1679
                block[entry_index][0][0:2] == key[0:2]):
 
1680
                result.append(block[entry_index])
 
1681
                entry_index += 1
 
1682
            return result
2084
1683
        if require_versioned:
2085
1684
            # -- check all supplied paths are versioned in a search tree. --
2086
 
            not_versioned = []
 
1685
            all_versioned = True
2087
1686
            for path in specific_files:
2088
 
                path_entries = state._entries_for_path(path)
 
1687
                path_entries = _entries_for_path(path)
2089
1688
                if not path_entries:
2090
1689
                    # this specified path is not present at all: error
2091
 
                    not_versioned.append(path)
2092
 
                    continue
 
1690
                    all_versioned = False
 
1691
                    break
2093
1692
                found_versioned = False
2094
1693
                # for each id at this path
2095
1694
                for entry in path_entries:
2102
1701
                if not found_versioned:
2103
1702
                    # none of the indexes was not 'absent' at all ids for this
2104
1703
                    # path.
2105
 
                    not_versioned.append(path)
2106
 
            if len(not_versioned) > 0:
2107
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1704
                    all_versioned = False
 
1705
                    break
 
1706
            if not all_versioned:
 
1707
                raise errors.PathsNotVersionedError(specific_files)
2108
1708
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2109
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1709
        search_specific_files = set()
 
1710
        for path in specific_files:
 
1711
            other_specific_files = specific_files.difference(set([path]))
 
1712
            if not osutils.is_inside_any(other_specific_files, path):
 
1713
                # this is a top level path, we must check it.
 
1714
                search_specific_files.add(path)
 
1715
        # sketch: 
 
1716
        # compare source_index and target_index at or under each element of search_specific_files.
 
1717
        # follow the following comparison table. Note that we only want to do diff operations when
 
1718
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1719
        # for the target.
 
1720
        # cases:
 
1721
        # 
 
1722
        # Source | Target | disk | action
 
1723
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1724
        #        |        |      | diff check on source-target
 
1725
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1726
        #        |        |      | ???
 
1727
        #   r    |  a     |      | add source to search
 
1728
        #   r    |  a     |  a   | 
 
1729
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1730
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1731
        #   a    | fdlt   |      | add new id
 
1732
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1733
        #   a    |  a     |      | not present in either tree, skip
 
1734
        #   a    |  a     |  a   | not present in any tree, skip
 
1735
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1736
        #        |        |      | may not be selected by the users list of paths.
 
1737
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1738
        #        |        |      | may not be selected by the users list of paths.
 
1739
        #  fdlt  | fdlt   |      | content in both: diff them
 
1740
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1741
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1742
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1743
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1744
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1745
        #        |        |      | this id at the other path.
 
1746
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1747
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1748
        #        |        |      | this id at the other path.
 
1749
 
 
1750
        # for all search_indexs in each path at or under each element of
 
1751
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1752
        # relocated path as one to search if its not searched already. If the
 
1753
        # detail is not relocated, add the id.
 
1754
        searched_specific_files = set()
 
1755
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1756
        # Using a list so that we can access the values and change them in
 
1757
        # nested scope. Each one is [path, file_id, entry]
 
1758
        last_source_parent = [None, None, None]
 
1759
        last_target_parent = [None, None, None]
2110
1760
 
2111
1761
        use_filesystem_for_exec = (sys.platform != 'win32')
2112
 
        iter_changes = self.target._iter_changes(include_unchanged,
2113
 
            use_filesystem_for_exec, search_specific_files, state,
2114
 
            source_index, target_index, want_unversioned, self.target)
2115
 
        return iter_changes.iter_changes()
 
1762
 
 
1763
        def _process_entry(entry, path_info):
 
1764
            """Compare an entry and real disk to generate delta information.
 
1765
 
 
1766
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1767
                the path of entry. If None, then the path is considered absent.
 
1768
                (Perhaps we should pass in a concrete entry for this ?)
 
1769
                Basename is returned as a utf8 string because we expect this
 
1770
                tuple will be ignored, and don't want to take the time to
 
1771
                decode.
 
1772
            """
 
1773
            if source_index is None:
 
1774
                source_details = NULL_PARENT_DETAILS
 
1775
            else:
 
1776
                source_details = entry[1][source_index]
 
1777
            target_details = entry[1][target_index]
 
1778
            target_minikind = target_details[0]
 
1779
            if path_info is not None and target_minikind in 'fdlt':
 
1780
                assert target_index == 0
 
1781
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1782
                                                  stat_value=path_info[3])
 
1783
                # The entry may have been modified by update_entry
 
1784
                target_details = entry[1][target_index]
 
1785
                target_minikind = target_details[0]
 
1786
            else:
 
1787
                link_or_sha1 = None
 
1788
            source_minikind = source_details[0]
 
1789
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1790
                # claimed content in both: diff
 
1791
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1792
                #        |        |      | diff check on source-target
 
1793
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1794
                #        |        |      | ???
 
1795
                if source_minikind in 'r':
 
1796
                    # add the source to the search path to find any children it
 
1797
                    # has.  TODO ? : only add if it is a container ?
 
1798
                    if not osutils.is_inside_any(searched_specific_files,
 
1799
                                                 source_details[1]):
 
1800
                        search_specific_files.add(source_details[1])
 
1801
                    # generate the old path; this is needed for stating later
 
1802
                    # as well.
 
1803
                    old_path = source_details[1]
 
1804
                    old_dirname, old_basename = os.path.split(old_path)
 
1805
                    path = pathjoin(entry[0][0], entry[0][1])
 
1806
                    old_entry = state._get_entry(source_index,
 
1807
                                                 path_utf8=old_path)
 
1808
                    # update the source details variable to be the real
 
1809
                    # location.
 
1810
                    source_details = old_entry[1][source_index]
 
1811
                    source_minikind = source_details[0]
 
1812
                else:
 
1813
                    old_dirname = entry[0][0]
 
1814
                    old_basename = entry[0][1]
 
1815
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1816
                if path_info is None:
 
1817
                    # the file is missing on disk, show as removed.
 
1818
                    content_change = True
 
1819
                    target_kind = None
 
1820
                    target_exec = False
 
1821
                else:
 
1822
                    # source and target are both versioned and disk file is present.
 
1823
                    target_kind = path_info[2]
 
1824
                    if target_kind == 'directory':
 
1825
                        if source_minikind != 'd':
 
1826
                            content_change = True
 
1827
                        else:
 
1828
                            # directories have no fingerprint
 
1829
                            content_change = False
 
1830
                        target_exec = False
 
1831
                    elif target_kind == 'file':
 
1832
                        if source_minikind != 'f':
 
1833
                            content_change = True
 
1834
                        else:
 
1835
                            # We could check the size, but we already have the
 
1836
                            # sha1 hash.
 
1837
                            content_change = (link_or_sha1 != source_details[1])
 
1838
                        # Target details is updated at update_entry time
 
1839
                        if use_filesystem_for_exec:
 
1840
                            # We don't need S_ISREG here, because we are sure
 
1841
                            # we are dealing with a file.
 
1842
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1843
                        else:
 
1844
                            target_exec = target_details[3]
 
1845
                    elif target_kind == 'symlink':
 
1846
                        if source_minikind != 'l':
 
1847
                            content_change = True
 
1848
                        else:
 
1849
                            content_change = (link_or_sha1 != source_details[1])
 
1850
                        target_exec = False
 
1851
                    elif target_kind == 'tree-reference':
 
1852
                        if source_minikind != 't':
 
1853
                            content_change = True
 
1854
                        else:
 
1855
                            content_change = False
 
1856
                        target_exec = False
 
1857
                    else:
 
1858
                        raise Exception, "unknown kind %s" % path_info[2]
 
1859
                # parent id is the entry for the path in the target tree
 
1860
                if old_dirname == last_source_parent[0]:
 
1861
                    source_parent_id = last_source_parent[1]
 
1862
                else:
 
1863
                    source_parent_entry = state._get_entry(source_index,
 
1864
                                                           path_utf8=old_dirname)
 
1865
                    source_parent_id = source_parent_entry[0][2]
 
1866
                    if source_parent_id == entry[0][2]:
 
1867
                        # This is the root, so the parent is None
 
1868
                        source_parent_id = None
 
1869
                    else:
 
1870
                        last_source_parent[0] = old_dirname
 
1871
                        last_source_parent[1] = source_parent_id
 
1872
                        last_source_parent[2] = source_parent_entry
 
1873
                new_dirname = entry[0][0]
 
1874
                if new_dirname == last_target_parent[0]:
 
1875
                    target_parent_id = last_target_parent[1]
 
1876
                else:
 
1877
                    # TODO: We don't always need to do the lookup, because the
 
1878
                    #       parent entry will be the same as the source entry.
 
1879
                    target_parent_entry = state._get_entry(target_index,
 
1880
                                                           path_utf8=new_dirname)
 
1881
                    target_parent_id = target_parent_entry[0][2]
 
1882
                    if target_parent_id == entry[0][2]:
 
1883
                        # This is the root, so the parent is None
 
1884
                        target_parent_id = None
 
1885
                    else:
 
1886
                        last_target_parent[0] = new_dirname
 
1887
                        last_target_parent[1] = target_parent_id
 
1888
                        last_target_parent[2] = target_parent_entry
 
1889
 
 
1890
                source_exec = source_details[3]
 
1891
                return ((entry[0][2], (old_path, path), content_change,
 
1892
                        (True, True),
 
1893
                        (source_parent_id, target_parent_id),
 
1894
                        (old_basename, entry[0][1]),
 
1895
                        (_minikind_to_kind[source_minikind], target_kind),
 
1896
                        (source_exec, target_exec)),)
 
1897
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1898
                # looks like a new file
 
1899
                if path_info is not None:
 
1900
                    path = pathjoin(entry[0][0], entry[0][1])
 
1901
                    # parent id is the entry for the path in the target tree
 
1902
                    # TODO: these are the same for an entire directory: cache em.
 
1903
                    parent_id = state._get_entry(target_index,
 
1904
                                                 path_utf8=entry[0][0])[0][2]
 
1905
                    if parent_id == entry[0][2]:
 
1906
                        parent_id = None
 
1907
                    if use_filesystem_for_exec:
 
1908
                        # We need S_ISREG here, because we aren't sure if this
 
1909
                        # is a file or not.
 
1910
                        target_exec = bool(
 
1911
                            stat.S_ISREG(path_info[3].st_mode)
 
1912
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1913
                    else:
 
1914
                        target_exec = target_details[3]
 
1915
                    return ((entry[0][2], (None, path), True,
 
1916
                            (False, True),
 
1917
                            (None, parent_id),
 
1918
                            (None, entry[0][1]),
 
1919
                            (None, path_info[2]),
 
1920
                            (None, target_exec)),)
 
1921
                else:
 
1922
                    # but its not on disk: we deliberately treat this as just
 
1923
                    # never-present. (Why ?! - RBC 20070224)
 
1924
                    pass
 
1925
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1926
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1927
                # if its still on disk, *and* theres no other entry at this
 
1928
                # path [we dont know this in this routine at the moment -
 
1929
                # perhaps we should change this - then it would be an unknown.
 
1930
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1931
                # parent id is the entry for the path in the target tree
 
1932
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1933
                if parent_id == entry[0][2]:
 
1934
                    parent_id = None
 
1935
                return ((entry[0][2], (old_path, None), True,
 
1936
                        (True, False),
 
1937
                        (parent_id, None),
 
1938
                        (entry[0][1], None),
 
1939
                        (_minikind_to_kind[source_minikind], None),
 
1940
                        (source_details[3], None)),)
 
1941
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1942
                # a rename; could be a true rename, or a rename inherited from
 
1943
                # a renamed parent. TODO: handle this efficiently. Its not
 
1944
                # common case to rename dirs though, so a correct but slow
 
1945
                # implementation will do.
 
1946
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1947
                    search_specific_files.add(target_details[1])
 
1948
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
1949
                # neither of the selected trees contain this file,
 
1950
                # so skip over it. This is not currently directly tested, but
 
1951
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1952
                pass
 
1953
            else:
 
1954
                raise AssertionError("don't know how to compare "
 
1955
                    "source_minikind=%r, target_minikind=%r"
 
1956
                    % (source_minikind, target_minikind))
 
1957
                ## import pdb;pdb.set_trace()
 
1958
            return ()
 
1959
 
 
1960
        while search_specific_files:
 
1961
            # TODO: the pending list should be lexically sorted?  the
 
1962
            # interface doesn't require it.
 
1963
            current_root = search_specific_files.pop()
 
1964
            searched_specific_files.add(current_root)
 
1965
            # process the entries for this containing directory: the rest will be
 
1966
            # found by their parents recursively.
 
1967
            root_entries = _entries_for_path(current_root)
 
1968
            root_abspath = self.target.abspath(current_root)
 
1969
            try:
 
1970
                root_stat = os.lstat(root_abspath)
 
1971
            except OSError, e:
 
1972
                if e.errno == errno.ENOENT:
 
1973
                    # the path does not exist: let _process_entry know that.
 
1974
                    root_dir_info = None
 
1975
                else:
 
1976
                    # some other random error: hand it up.
 
1977
                    raise
 
1978
            else:
 
1979
                root_dir_info = ('', current_root,
 
1980
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1981
                    root_abspath)
 
1982
                if root_dir_info[2] == 'directory':
 
1983
                    if self.target._directory_is_tree_reference(
 
1984
                        current_root.decode('utf8')):
 
1985
                        root_dir_info = root_dir_info[:2] + \
 
1986
                            ('tree-reference',) + root_dir_info[3:]
 
1987
 
 
1988
            if not root_entries and not root_dir_info:
 
1989
                # this specified path is not present at all, skip it.
 
1990
                continue
 
1991
            path_handled = False
 
1992
            for entry in root_entries:
 
1993
                for result in _process_entry(entry, root_dir_info):
 
1994
                    # this check should probably be outside the loop: one
 
1995
                    # 'iterate two trees' api, and then _iter_changes filters
 
1996
                    # unchanged pairs. - RBC 20070226
 
1997
                    path_handled = True
 
1998
                    if (include_unchanged
 
1999
                        or result[2]                    # content change
 
2000
                        or result[3][0] != result[3][1] # versioned status
 
2001
                        or result[4][0] != result[4][1] # parent id
 
2002
                        or result[5][0] != result[5][1] # name
 
2003
                        or result[6][0] != result[6][1] # kind
 
2004
                        or result[7][0] != result[7][1] # executable
 
2005
                        ):
 
2006
                        result = (result[0],
 
2007
                            ((utf8_decode(result[1][0])[0]),
 
2008
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
2009
                        yield result
 
2010
            if want_unversioned and not path_handled:
 
2011
                new_executable = bool(
 
2012
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2013
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2014
                yield (None, (None, current_root), True, (False, False),
 
2015
                    (None, None),
 
2016
                    (None, splitpath(current_root)[-1]),
 
2017
                    (None, root_dir_info[2]), (None, new_executable))
 
2018
            initial_key = (current_root, '', '')
 
2019
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2020
            if block_index == 0:
 
2021
                # we have processed the total root already, but because the
 
2022
                # initial key matched it we should skip it here.
 
2023
                block_index +=1
 
2024
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2025
                current_dir_info = None
 
2026
            else:
 
2027
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2028
                try:
 
2029
                    current_dir_info = dir_iterator.next()
 
2030
                except OSError, e:
 
2031
                    if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
2032
                        # there may be directories in the inventory even though
 
2033
                        # this path is not a file on disk: so mark it as end of
 
2034
                        # iterator
 
2035
                        current_dir_info = None
 
2036
                    else:
 
2037
                        raise
 
2038
                else:
 
2039
                    if current_dir_info[0][0] == '':
 
2040
                        # remove .bzr from iteration
 
2041
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2042
                        assert current_dir_info[1][bzr_index][0] == '.bzr'
 
2043
                        del current_dir_info[1][bzr_index]
 
2044
            # walk until both the directory listing and the versioned metadata
 
2045
            # are exhausted. 
 
2046
            if (block_index < len(state._dirblocks) and
 
2047
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2048
                current_block = state._dirblocks[block_index]
 
2049
            else:
 
2050
                current_block = None
 
2051
            while (current_dir_info is not None or
 
2052
                   current_block is not None):
 
2053
                if (current_dir_info and current_block
 
2054
                    and current_dir_info[0][0] != current_block[0]):
 
2055
                    if current_dir_info[0][0] < current_block[0] :
 
2056
                        # filesystem data refers to paths not covered by the dirblock.
 
2057
                        # this has two possibilities:
 
2058
                        # A) it is versioned but empty, so there is no block for it
 
2059
                        # B) it is not versioned.
 
2060
                        # in either case it was processed by the containing directories walk:
 
2061
                        # if it is root/foo, when we walked root we emitted it,
 
2062
                        # or if we ere given root/foo to walk specifically, we
 
2063
                        # emitted it when checking the walk-root entries
 
2064
                        # advance the iterator and loop - we dont need to emit it.
 
2065
                        try:
 
2066
                            current_dir_info = dir_iterator.next()
 
2067
                        except StopIteration:
 
2068
                            current_dir_info = None
 
2069
                    else:
 
2070
                        # We have a dirblock entry for this location, but there
 
2071
                        # is no filesystem path for this. This is most likely
 
2072
                        # because a directory was removed from the disk.
 
2073
                        # We don't have to report the missing directory,
 
2074
                        # because that should have already been handled, but we
 
2075
                        # need to handle all of the files that are contained
 
2076
                        # within.
 
2077
                        for current_entry in current_block[1]:
 
2078
                            # entry referring to file not present on disk.
 
2079
                            # advance the entry only, after processing.
 
2080
                            for result in _process_entry(current_entry, None):
 
2081
                                # this check should probably be outside the loop: one
 
2082
                                # 'iterate two trees' api, and then _iter_changes filters
 
2083
                                # unchanged pairs. - RBC 20070226
 
2084
                                if (include_unchanged
 
2085
                                    or result[2]                    # content change
 
2086
                                    or result[3][0] != result[3][1] # versioned status
 
2087
                                    or result[4][0] != result[4][1] # parent id
 
2088
                                    or result[5][0] != result[5][1] # name
 
2089
                                    or result[6][0] != result[6][1] # kind
 
2090
                                    or result[7][0] != result[7][1] # executable
 
2091
                                    ):
 
2092
                                    result = (result[0],
 
2093
                                        ((utf8_decode(result[1][0])[0]),
 
2094
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2095
                                    yield result
 
2096
                        block_index +=1
 
2097
                        if (block_index < len(state._dirblocks) and
 
2098
                            osutils.is_inside(current_root,
 
2099
                                              state._dirblocks[block_index][0])):
 
2100
                            current_block = state._dirblocks[block_index]
 
2101
                        else:
 
2102
                            current_block = None
 
2103
                    continue
 
2104
                entry_index = 0
 
2105
                if current_block and entry_index < len(current_block[1]):
 
2106
                    current_entry = current_block[1][entry_index]
 
2107
                else:
 
2108
                    current_entry = None
 
2109
                advance_entry = True
 
2110
                path_index = 0
 
2111
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2112
                    current_path_info = current_dir_info[1][path_index]
 
2113
                    if current_path_info[2] == 'directory':
 
2114
                        if self.target._directory_is_tree_reference(
 
2115
                            current_path_info[0].decode('utf8')):
 
2116
                            current_path_info = current_path_info[:2] + \
 
2117
                                ('tree-reference',) + current_path_info[3:]
 
2118
                else:
 
2119
                    current_path_info = None
 
2120
                advance_path = True
 
2121
                path_handled = False
 
2122
                while (current_entry is not None or
 
2123
                    current_path_info is not None):
 
2124
                    if current_entry is None:
 
2125
                        # the check for path_handled when the path is adnvaced
 
2126
                        # will yield this path if needed.
 
2127
                        pass
 
2128
                    elif current_path_info is None:
 
2129
                        # no path is fine: the per entry code will handle it.
 
2130
                        for result in _process_entry(current_entry, current_path_info):
 
2131
                            # this check should probably be outside the loop: one
 
2132
                            # 'iterate two trees' api, and then _iter_changes filters
 
2133
                            # unchanged pairs. - RBC 20070226
 
2134
                            if (include_unchanged
 
2135
                                or result[2]                    # content change
 
2136
                                or result[3][0] != result[3][1] # versioned status
 
2137
                                or result[4][0] != result[4][1] # parent id
 
2138
                                or result[5][0] != result[5][1] # name
 
2139
                                or result[6][0] != result[6][1] # kind
 
2140
                                or result[7][0] != result[7][1] # executable
 
2141
                                ):
 
2142
                                result = (result[0],
 
2143
                                    ((utf8_decode(result[1][0])[0]),
 
2144
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2145
                                yield result
 
2146
                    elif current_entry[0][1] != current_path_info[1]:
 
2147
                        if current_path_info[1] < current_entry[0][1]:
 
2148
                            # extra file on disk: pass for now, but only
 
2149
                            # increment the path, not the entry
 
2150
                            advance_entry = False
 
2151
                        else:
 
2152
                            # entry referring to file not present on disk.
 
2153
                            # advance the entry only, after processing.
 
2154
                            for result in _process_entry(current_entry, None):
 
2155
                                # this check should probably be outside the loop: one
 
2156
                                # 'iterate two trees' api, and then _iter_changes filters
 
2157
                                # unchanged pairs. - RBC 20070226
 
2158
                                path_handled = True
 
2159
                                if (include_unchanged
 
2160
                                    or result[2]                    # content change
 
2161
                                    or result[3][0] != result[3][1] # versioned status
 
2162
                                    or result[4][0] != result[4][1] # parent id
 
2163
                                    or result[5][0] != result[5][1] # name
 
2164
                                    or result[6][0] != result[6][1] # kind
 
2165
                                    or result[7][0] != result[7][1] # executable
 
2166
                                    ):
 
2167
                                    result = (result[0],
 
2168
                                        ((utf8_decode(result[1][0])[0]),
 
2169
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2170
                                    yield result
 
2171
                            advance_path = False
 
2172
                    else:
 
2173
                        for result in _process_entry(current_entry, current_path_info):
 
2174
                            # this check should probably be outside the loop: one
 
2175
                            # 'iterate two trees' api, and then _iter_changes filters
 
2176
                            # unchanged pairs. - RBC 20070226
 
2177
                            path_handled = True
 
2178
                            if (include_unchanged
 
2179
                                or result[2]                    # content change
 
2180
                                or result[3][0] != result[3][1] # versioned status
 
2181
                                or result[4][0] != result[4][1] # parent id
 
2182
                                or result[5][0] != result[5][1] # name
 
2183
                                or result[6][0] != result[6][1] # kind
 
2184
                                or result[7][0] != result[7][1] # executable
 
2185
                                ):
 
2186
                                result = (result[0],
 
2187
                                    ((utf8_decode(result[1][0])[0]),
 
2188
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2189
                                yield result
 
2190
                    if advance_entry and current_entry is not None:
 
2191
                        entry_index += 1
 
2192
                        if entry_index < len(current_block[1]):
 
2193
                            current_entry = current_block[1][entry_index]
 
2194
                        else:
 
2195
                            current_entry = None
 
2196
                    else:
 
2197
                        advance_entry = True # reset the advance flaga
 
2198
                    if advance_path and current_path_info is not None:
 
2199
                        if not path_handled:
 
2200
                            # unversioned in all regards
 
2201
                            if want_unversioned:
 
2202
                                new_executable = bool(
 
2203
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2204
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2205
                                if want_unversioned:
 
2206
                                    yield (None, (None, current_path_info[0]),
 
2207
                                        True,
 
2208
                                        (False, False),
 
2209
                                        (None, None),
 
2210
                                        (None, current_path_info[1]),
 
2211
                                        (None, current_path_info[2]),
 
2212
                                        (None, new_executable))
 
2213
                            # dont descend into this unversioned path if it is
 
2214
                            # a dir
 
2215
                            if current_path_info[2] in ('directory'):
 
2216
                                del current_dir_info[1][path_index]
 
2217
                                path_index -= 1
 
2218
                        # dont descend the disk iterator into any tree 
 
2219
                        # paths.
 
2220
                        if current_path_info[2] == 'tree-reference':
 
2221
                            del current_dir_info[1][path_index]
 
2222
                            path_index -= 1
 
2223
                        path_index += 1
 
2224
                        if path_index < len(current_dir_info[1]):
 
2225
                            current_path_info = current_dir_info[1][path_index]
 
2226
                            if current_path_info[2] == 'directory':
 
2227
                                if self.target._directory_is_tree_reference(
 
2228
                                    current_path_info[0].decode('utf8')):
 
2229
                                    current_path_info = current_path_info[:2] + \
 
2230
                                        ('tree-reference',) + current_path_info[3:]
 
2231
                        else:
 
2232
                            current_path_info = None
 
2233
                        path_handled = False
 
2234
                    else:
 
2235
                        advance_path = True # reset the advance flagg.
 
2236
                if current_block is not None:
 
2237
                    block_index += 1
 
2238
                    if (block_index < len(state._dirblocks) and
 
2239
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2240
                        current_block = state._dirblocks[block_index]
 
2241
                    else:
 
2242
                        current_block = None
 
2243
                if current_dir_info is not None:
 
2244
                    try:
 
2245
                        current_dir_info = dir_iterator.next()
 
2246
                    except StopIteration:
 
2247
                        current_dir_info = None
 
2248
 
2116
2249
 
2117
2250
    @staticmethod
2118
2251
    def is_compatible(source, target):
2119
2252
        # the target must be a dirstate working tree
2120
 
        if not isinstance(target, DirStateWorkingTree):
 
2253
        if not isinstance(target, WorkingTree4):
2121
2254
            return False
2122
 
        # the source must be a revtree or dirstate rev tree.
 
2255
        # the source must be a revtreee or dirstate rev tree.
2123
2256
        if not isinstance(source,
2124
2257
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2125
2258
            return False
2126
2259
        # the source revid must be in the target dirstate
2127
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2260
        if not (source._revision_id == NULL_REVISION or
2128
2261
            source._revision_id in target.get_parent_ids()):
2129
 
            # TODO: what about ghosts? it may well need to
 
2262
            # TODO: what about ghosts? it may well need to 
2130
2263
            # check for them explicitly.
2131
2264
            return False
2132
2265
        return True
2142
2275
 
2143
2276
    def convert(self, tree):
2144
2277
        # lock the control files not the tree, so that we dont get tree
2145
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2278
        # on-unlock behaviours, and so that noone else diddles with the 
2146
2279
        # tree during upgrade.
2147
2280
        tree._control_files.lock_write()
2148
2281
        try:
2149
 
            tree.read_working_inventory()
2150
2282
            self.create_dirstate_data(tree)
2151
2283
            self.update_format(tree)
2152
2284
            self.remove_xml_files(tree)
2174
2306
 
2175
2307
    def update_format(self, tree):
2176
2308
        """Change the format marker."""
2177
 
        tree._transport.put_bytes('format',
2178
 
            self.target_format.get_format_string(),
2179
 
            mode=tree.bzrdir._get_file_mode())
2180
 
 
2181
 
 
2182
 
class Converter4to5(object):
2183
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2184
 
 
2185
 
    def __init__(self):
2186
 
        self.target_format = WorkingTreeFormat5()
2187
 
 
2188
 
    def convert(self, tree):
2189
 
        # lock the control files not the tree, so that we don't get tree
2190
 
        # on-unlock behaviours, and so that no-one else diddles with the
2191
 
        # tree during upgrade.
2192
 
        tree._control_files.lock_write()
2193
 
        try:
2194
 
            self.update_format(tree)
2195
 
        finally:
2196
 
            tree._control_files.unlock()
2197
 
 
2198
 
    def update_format(self, tree):
2199
 
        """Change the format marker."""
2200
 
        tree._transport.put_bytes('format',
2201
 
            self.target_format.get_format_string(),
2202
 
            mode=tree.bzrdir._get_file_mode())
2203
 
 
2204
 
 
2205
 
class Converter4or5to6(object):
2206
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2207
 
 
2208
 
    def __init__(self):
2209
 
        self.target_format = WorkingTreeFormat6()
2210
 
 
2211
 
    def convert(self, tree):
2212
 
        # lock the control files not the tree, so that we don't get tree
2213
 
        # on-unlock behaviours, and so that no-one else diddles with the
2214
 
        # tree during upgrade.
2215
 
        tree._control_files.lock_write()
2216
 
        try:
2217
 
            self.init_custom_control_files(tree)
2218
 
            self.update_format(tree)
2219
 
        finally:
2220
 
            tree._control_files.unlock()
2221
 
 
2222
 
    def init_custom_control_files(self, tree):
2223
 
        """Initialize custom control files."""
2224
 
        tree._transport.put_bytes('views', '',
2225
 
            mode=tree.bzrdir._get_file_mode())
2226
 
 
2227
 
    def update_format(self, tree):
2228
 
        """Change the format marker."""
2229
 
        tree._transport.put_bytes('format',
2230
 
            self.target_format.get_format_string(),
2231
 
            mode=tree.bzrdir._get_file_mode())
 
2309
        tree._control_files.put_utf8('format',
 
2310
            self.target_format.get_format_string())