~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

Fix test_upgrade defects related to non local or absent working trees.

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
 
    file_kind,
59
75
    isdir,
 
76
    normpath,
60
77
    pathjoin,
 
78
    rand_chars,
61
79
    realpath,
62
80
    safe_unicode,
 
81
    splitpath,
63
82
    )
64
 
from bzrlib.trace import mutter
 
83
from bzrlib.trace import mutter, note
65
84
from bzrlib.transport.local import LocalTransport
66
85
from bzrlib.tree import InterTree
 
86
from bzrlib.progress import DummyProgress, ProgressPhase
 
87
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
88
from bzrlib.rio import RioReader, rio_file, Stanza
 
89
from bzrlib.symbol_versioning import (deprecated_passed,
 
90
        deprecated_method,
 
91
        deprecated_function,
 
92
        DEPRECATED_PARAMETER,
 
93
        )
67
94
from bzrlib.tree import Tree
68
95
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
96
 
70
97
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
98
class WorkingTree4(WorkingTree3):
 
99
    """This is the Format 4 working tree.
 
100
 
 
101
    This differs from WorkingTree3 by:
 
102
     - Having a consolidated internal dirstate, stored in a
 
103
       randomly-accessible sorted file on disk.
 
104
     - Not having a regular inventory attribute.  One can be synthesized 
 
105
       on demand but this is expensive and should be avoided.
 
106
 
 
107
    This is new in bzr 0.15.
 
108
    """
 
109
 
72
110
    def __init__(self, basedir,
73
111
                 branch,
74
112
                 _control_files=None,
83
121
        """
84
122
        self._format = _format
85
123
        self.bzrdir = _bzrdir
 
124
        from bzrlib.trace import note, mutter
 
125
        assert isinstance(basedir, basestring), \
 
126
            "base directory %r is not a string" % basedir
86
127
        basedir = safe_unicode(basedir)
87
128
        mutter("opening working tree %r", basedir)
88
129
        self._branch = branch
 
130
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
131
            "branch %r is not a Branch" % self.branch
89
132
        self.basedir = realpath(basedir)
90
133
        # if branch is at our basedir and is a format 6 or less
91
134
        # assume all other formats have their own control files.
 
135
        assert isinstance(_control_files, LockableFiles), \
 
136
            "_control_files must be a LockableFiles, not %r" % _control_files
92
137
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
138
        self._dirty = None
95
139
        #-------------
96
140
        # during a read or write lock these objects are set, and are
98
142
        self._dirstate = None
99
143
        self._inventory = None
100
144
        #-------------
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
145
 
108
146
    @needs_tree_write_lock
109
147
    def _add(self, files, ids, kinds):
111
149
        state = self.current_dirstate()
112
150
        for f, file_id, kind in zip(files, ids, kinds):
113
151
            f = f.strip('/')
 
152
            assert '//' not in f
 
153
            assert '..' not in f
114
154
            if self.path2id(f):
115
155
                # special case tree root handling.
116
156
                if f == '' and self.path2id(f) == ROOT_ID:
137
177
    @needs_tree_write_lock
138
178
    def add_reference(self, sub_tree):
139
179
        # use standard implementation, which calls back to self._add
140
 
        #
 
180
        # 
141
181
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
182
        # it's just recorded at the moment of commit. 
143
183
        self._add_reference(sub_tree)
144
184
 
145
185
    def break_lock(self):
182
222
    def _comparison_data(self, entry, path):
183
223
        kind, executable, stat_value = \
184
224
            WorkingTree3._comparison_data(self, entry, path)
185
 
        # it looks like a plain directory, but it's really a reference -- see
186
 
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
225
        # it looks like a plain directory, but it's really a reference
 
226
        if kind == 'directory' and entry.kind == 'tree-reference':
189
227
            kind = 'tree-reference'
190
228
        return kind, executable, stat_value
191
229
 
217
255
            return self._dirstate
218
256
        local_path = self.bzrdir.get_workingtree_transport(None
219
257
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
258
        self._dirstate = dirstate.DirState.on_file(local_path)
222
259
        return self._dirstate
223
260
 
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
235
 
 
236
261
    def filter_unversioned_files(self, paths):
237
262
        """Filter out paths that are versioned.
238
263
 
261
286
        self._inventory = None
262
287
        self._dirty = False
263
288
 
264
 
    @needs_tree_write_lock
265
 
    def _gather_kinds(self, files, kinds):
266
 
        """See MutableTree._gather_kinds."""
267
 
        for pos, f in enumerate(files):
268
 
            if kinds[pos] is None:
269
 
                kinds[pos] = self._kind(f)
270
 
 
271
289
    def _generate_inventory(self):
272
290
        """Create and set self.inventory from the dirstate object.
273
 
 
 
291
        
274
292
        This is relatively expensive: we have to walk the entire dirstate.
275
293
        Ideally we would not, and can deprecate this function.
276
294
        """
280
298
        state._read_dirblocks_if_needed()
281
299
        root_key, current_entry = self._get_entry(path='')
282
300
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
301
        assert current_entry[0][0] == 'd' # directory
285
302
        inv = Inventory(root_id=current_id)
286
303
        # Turn some things into local variables
287
304
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
327
                inv_entry = factory[kind](file_id, name_unicode,
311
328
                                          parent_ie.file_id)
312
329
                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
330
                    # not strictly needed: working tree
317
 
                    #inv_entry.text_size = size
318
 
                    #inv_entry.text_sha1 = sha1
 
331
                    #entry.executable = executable
 
332
                    #entry.text_size = size
 
333
                    #entry.text_sha1 = sha1
 
334
                    pass
319
335
                elif kind == 'directory':
320
336
                    # add this entry to the parent map.
321
337
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
338
                elif kind == 'tree-reference':
323
 
                    if not self._repo_supports_tree_reference:
324
 
                        raise errors.UnsupportedOperation(
325
 
                            self._generate_inventory,
326
 
                            self.branch.repository)
327
 
                    inv_entry.reference_revision = link_or_sha1 or None
328
 
                elif kind != 'symlink':
329
 
                    raise AssertionError("unknown kind %r" % kind)
 
339
                    inv_entry.reference_revision = link_or_sha1
 
340
                else:
 
341
                    assert 'unknown kind'
330
342
                # 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,))
 
343
                assert file_id not in inv_byid, ('file_id %s already in'
 
344
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
345
                assert name_unicode not in parent_ie.children
337
346
                inv_byid[file_id] = inv_entry
338
347
                parent_ie.children[name_unicode] = inv_entry
339
348
        self._inventory = inv
344
353
        If either file_id or path is supplied, it is used as the key to lookup.
345
354
        If both are supplied, the fastest lookup is used, and an error is
346
355
        raised if they do not both point at the same row.
347
 
 
 
356
        
348
357
        :param file_id: An optional unicode file_id to be looked up.
349
358
        :param path: An optional unicode path to be looked up.
350
359
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
368
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
369
        # check file id is valid unconditionally.
361
370
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
371
        assert entry[0] is not None, 'what error should this raise'
 
372
        # TODO:
 
373
        # if row stat is valid, use cached sha1, else, get a new sha1.
364
374
        if path is None:
365
375
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
376
 
367
377
        file_abspath = self.abspath(path)
368
378
        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)
 
379
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
380
                                          stat_value=stat_value)
379
381
        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
 
382
            return link_or_sha1
390
383
        return None
391
384
 
392
385
    def _get_inventory(self):
393
386
        """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
387
        if self._inventory is not None:
398
388
            return self._inventory
399
389
        self._must_be_locked()
406
396
    @needs_read_lock
407
397
    def get_parent_ids(self):
408
398
        """See Tree.get_parent_ids.
409
 
 
 
399
        
410
400
        This implementation requests the ids list from the dirstate file.
411
401
        """
412
402
        return self.current_dirstate().get_parent_ids()
413
403
 
414
 
    def get_reference_revision(self, file_id, path=None):
 
404
    def get_reference_revision(self, entry, path=None):
415
405
        # referenced tree's revision is whatever's currently there
416
 
        return self.get_nested_tree(file_id, path).last_revision()
 
406
        return self.get_nested_tree(entry, path).last_revision()
417
407
 
418
 
    def get_nested_tree(self, file_id, path=None):
 
408
    def get_nested_tree(self, entry, path=None):
419
409
        if path is None:
420
 
            path = self.id2path(file_id)
421
 
        # else: check file_id is at path?
 
410
            path = self.id2path(entry.file_id)
422
411
        return WorkingTree.open(self.abspath(path))
423
412
 
424
413
    @needs_read_lock
428
417
 
429
418
    def has_id(self, file_id):
430
419
        state = self.current_dirstate()
 
420
        file_id = osutils.safe_file_id(file_id)
431
421
        row, parents = self._get_entry(file_id=file_id)
432
422
        if row is None:
433
423
            return False
434
424
        return osutils.lexists(pathjoin(
435
425
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
426
 
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
427
    @needs_read_lock
443
428
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
429
        file_id = osutils.safe_file_id(file_id)
445
430
        state = self.current_dirstate()
446
431
        entry = self._get_entry(file_id=file_id)
447
432
        if entry == (None, None):
449
434
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
435
        return path_utf8.decode('utf8')
451
436
 
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
437
    @needs_read_lock
494
438
    def __iter__(self):
495
439
        """Iterate through file_ids for this tree.
507
451
                result.append(key[2])
508
452
        return iter(result)
509
453
 
510
 
    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
 
        for key, tree_details in self.current_dirstate()._iter_entries():
516
 
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
517
 
                # not relevant to the working tree
518
 
                continue
519
 
            if not key[1]:
520
 
                # the root is not a reference.
521
 
                continue
522
 
            relpath = pathjoin(key[0].decode('utf8'), key[1].decode('utf8'))
523
 
            try:
524
 
                if self._kind(relpath) == 'tree-reference':
525
 
                    yield relpath, key[2]
526
 
            except errors.NoSuchFile:
527
 
                # path is missing on disk.
528
 
                continue
529
 
 
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
 
 
 
454
    @needs_read_lock
536
455
    def kind(self, file_id):
537
 
        """Return the kind of a file.
538
 
 
539
 
        This is always the actual kind that's on disk, regardless of what it
540
 
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
 
        """
544
 
        relpath = self.id2path(file_id)
545
 
        if relpath is None:
546
 
            raise AssertionError(
547
 
                "path for id {%s} is None!" % file_id)
548
 
        return self._kind(relpath)
549
 
 
550
 
    def _kind(self, relpath):
551
 
        abspath = self.abspath(relpath)
552
 
        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'
 
456
        # The kind of a file is whatever it actually is on disk, except that 
 
457
        # tree-references need to be reported as such rather than as the
 
458
        # directiories
 
459
        #
 
460
        # TODO: Possibly we should check that the directory still really
 
461
        # contains a subtree, at least during commit? mbp 20070227
 
462
        kind = WorkingTree3.kind(self, file_id)
 
463
        if kind == 'directory':
 
464
            # TODO: ask the dirstate not the inventory -- mbp 20060227
 
465
            entry = self.inventory[file_id]
 
466
            if entry.kind == 'tree-reference':
 
467
                kind = 'tree-reference'
558
468
        return kind
559
469
 
560
470
    @needs_read_lock
564
474
        if parent_ids:
565
475
            return parent_ids[0]
566
476
        else:
567
 
            return _mod_revision.NULL_REVISION
 
477
            return None
568
478
 
569
479
    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
 
        """
 
480
        """See Branch.lock_read, and WorkingTree.unlock."""
575
481
        self.branch.lock_read()
576
482
        try:
577
483
            self._control_files.lock_read()
579
485
                state = self.current_dirstate()
580
486
                if not state._lock_token:
581
487
                    state.lock_read()
582
 
                # set our support for tree references from the repository in
583
 
                # use.
584
 
                self._repo_supports_tree_reference = getattr(
585
 
                    self.branch.repository._format, "supports_tree_reference",
586
 
                    False)
587
488
            except:
588
489
                self._control_files.unlock()
589
490
                raise
590
491
        except:
591
492
            self.branch.unlock()
592
493
            raise
593
 
        return self
594
494
 
595
495
    def _lock_self_write(self):
596
496
        """This should be called after the branch is locked."""
600
500
                state = self.current_dirstate()
601
501
                if not state._lock_token:
602
502
                    state.lock_write()
603
 
                # set our support for tree references from the repository in
604
 
                # use.
605
 
                self._repo_supports_tree_reference = getattr(
606
 
                    self.branch.repository._format, "supports_tree_reference",
607
 
                    False)
608
503
            except:
609
504
                self._control_files.unlock()
610
505
                raise
611
506
        except:
612
507
            self.branch.unlock()
613
508
            raise
614
 
        return self
615
509
 
616
510
    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
 
        """
 
511
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
622
512
        self.branch.lock_read()
623
 
        return self._lock_self_write()
 
513
        self._lock_self_write()
624
514
 
625
515
    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
 
        """
 
516
        """See MutableTree.lock_write, and WorkingTree.unlock."""
631
517
        self.branch.lock_write()
632
 
        return self._lock_self_write()
 
518
        self._lock_self_write()
633
519
 
634
520
    @needs_tree_write_lock
635
521
    def move(self, from_paths, to_dir, after=False):
637
523
        result = []
638
524
        if not from_paths:
639
525
            return result
 
526
 
640
527
        state = self.current_dirstate()
641
 
        if isinstance(from_paths, basestring):
642
 
            raise ValueError()
 
528
 
 
529
        assert not isinstance(from_paths, basestring)
643
530
        to_dir_utf8 = to_dir.encode('utf8')
644
531
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
645
532
        id_index = state._get_id_index()
667
554
        if self._inventory is not None:
668
555
            update_inventory = True
669
556
            inv = self.inventory
 
557
            to_dir_ie = inv[to_dir_id]
670
558
            to_dir_id = to_entry[0][2]
671
 
            to_dir_ie = inv[to_dir_id]
672
559
        else:
673
560
            update_inventory = False
674
561
 
697
584
            new_entry = to_block[1][added_entry_index]
698
585
            rollbacks.append(lambda:state._make_absent(new_entry))
699
586
 
 
587
        # create rename entries and tuples
700
588
        for from_rel in from_paths:
701
589
            # from_rel is 'pathinroot/foo/bar'
702
590
            from_rel_utf8 = from_rel.encode('utf8')
705
593
            from_entry = self._get_entry(path=from_rel)
706
594
            if from_entry == (None, None):
707
595
                raise errors.BzrMoveFailedError(from_rel,to_dir,
708
 
                    errors.NotVersionedError(path=from_rel))
 
596
                    errors.NotVersionedError(path=str(from_rel)))
709
597
 
710
598
            from_id = from_entry[0][2]
711
599
            to_rel = pathjoin(to_dir, from_tail)
738
626
                if from_missing: # implicitly just update our path mapping
739
627
                    move_file = False
740
628
                elif not after:
741
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
629
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
630
                        extra="(Use --after to update the Bazaar id)")
742
631
 
743
632
            rollbacks = []
744
633
            def rollback_rename():
745
634
                """A single rename has failed, roll it back."""
746
 
                # roll back everything, even if we encounter trouble doing one
747
 
                # of them.
748
 
                #
749
 
                # TODO: at least log the other exceptions rather than just
750
 
                # losing them mbp 20070307
751
635
                exc_info = None
752
636
                for rollback in reversed(rollbacks):
753
637
                    try:
754
638
                        rollback()
755
639
                    except Exception, e:
 
640
                        import pdb;pdb.set_trace()
756
641
                        exc_info = sys.exc_info()
757
642
                if exc_info:
758
643
                    raise exc_info[0], exc_info[1], exc_info[2]
799
684
 
800
685
                if minikind == 'd':
801
686
                    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")
 
687
                        """all entries in this block need updating.
 
688
 
 
689
                        TODO: This is pretty ugly, and doesn't support
 
690
                        reverting, but it works.
 
691
                        """
 
692
                        assert from_dir != '', "renaming root not supported"
805
693
                        from_key = (from_dir, '')
806
694
                        from_block_idx, present = \
807
695
                            state._find_block_index_from_key(from_key)
817
705
                        to_block_index = state._ensure_block(
818
706
                            to_block_index, to_entry_index, to_dir_utf8)
819
707
                        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()
 
708
                        for entry in from_block[1]:
 
709
                            assert entry[0][0] == from_dir
825
710
                            cur_details = entry[1][0]
826
711
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
827
712
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
828
713
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
829
714
                            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
715
                            move_one(entry, from_path_utf8=from_path_utf8,
837
716
                                     minikind=minikind,
838
717
                                     executable=cur_details[3],
841
720
                                     size=cur_details[2],
842
721
                                     to_block=to_block,
843
722
                                     to_key=to_key,
844
 
                                     to_path_utf8=to_path_utf8)
 
723
                                     to_path_utf8=to_rel_utf8)
845
724
                            if minikind == 'd':
846
725
                                # We need to move all the children of this
847
726
                                # entry
886
765
        for tree in trees:
887
766
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
888
767
                parents):
889
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
890
 
                    trees, require_versioned)
 
768
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
891
769
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
892
770
        # -- make all paths utf8 --
893
771
        paths_utf8 = set()
953
831
            if not all_versioned:
954
832
                raise errors.PathsNotVersionedError(paths)
955
833
        # -- remove redundancy in supplied paths to prevent over-scanning --
956
 
        search_paths = osutils.minimum_path_selection(paths)
957
 
        # sketch:
 
834
        search_paths = set()
 
835
        for path in paths:
 
836
            other_paths = paths.difference(set([path]))
 
837
            if not osutils.is_inside_any(other_paths, path):
 
838
                # this is a top level path, we must check it.
 
839
                search_paths.add(path)
 
840
        # sketch: 
958
841
        # for all search_indexs in each path at or under each element of
959
842
        # search_paths, if the detail is relocated: add the id, and add the
960
843
        # relocated path as one to search if its not searched already. If the
1016
899
 
1017
900
    def read_working_inventory(self):
1018
901
        """Read the working inventory.
1019
 
 
 
902
        
1020
903
        This is a meaningless operation for dirstate, but we obey it anyhow.
1021
904
        """
1022
905
        return self.inventory
1027
910
 
1028
911
        WorkingTree4 supplies revision_trees for any basis tree.
1029
912
        """
 
913
        revision_id = osutils.safe_revision_id(revision_id)
1030
914
        dirstate = self.current_dirstate()
1031
915
        parent_ids = dirstate.get_parent_ids()
1032
916
        if revision_id not in parent_ids:
1039
923
    @needs_tree_write_lock
1040
924
    def set_last_revision(self, new_revision):
1041
925
        """Change the last revision in the working tree."""
 
926
        new_revision = osutils.safe_revision_id(new_revision)
1042
927
        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.")
 
928
        if new_revision in (NULL_REVISION, None):
 
929
            assert len(parents) < 2, (
 
930
                "setting the last parent to none with a pending merge is "
 
931
                "unsupported.")
1048
932
            self.set_parent_ids([])
1049
933
        else:
1050
934
            self.set_parent_ids([new_revision] + parents[1:],
1053
937
    @needs_tree_write_lock
1054
938
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1055
939
        """Set the parent ids to revision_ids.
1056
 
 
 
940
        
1057
941
        See also set_parent_trees. This api will try to retrieve the tree data
1058
942
        for each element of revision_ids from the trees repository. If you have
1059
943
        tree data already available, it is more efficient to use
1063
947
        :param revision_ids: The revision_ids to set as the parent ids of this
1064
948
            working tree. Any of these may be ghosts.
1065
949
        """
 
950
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1066
951
        trees = []
1067
952
        for revision_id in revision_ids:
1068
953
            try:
1074
959
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1075
960
                revtree = None
1076
961
            trees.append((revision_id, revtree))
 
962
        self.current_dirstate()._validate()
1077
963
        self.set_parent_trees(trees,
1078
964
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
965
        self.current_dirstate()._validate()
1079
966
 
1080
967
    @needs_tree_write_lock
1081
968
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1086
973
            parent tree - i.e. a ghost.
1087
974
        """
1088
975
        dirstate = self.current_dirstate()
 
976
        dirstate._validate()
1089
977
        if len(parents_list) > 0:
1090
978
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1091
979
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1092
980
        real_trees = []
1093
981
        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
982
        # convert absent trees to the null tree, which we convert back to
1101
983
        # missing on access.
1102
984
        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)
 
985
            rev_id = osutils.safe_revision_id(rev_id)
1110
986
            if tree is not None:
1111
987
                real_trees.append((rev_id, tree))
1112
988
            else:
1113
989
                real_trees.append((rev_id,
1114
 
                    self.branch.repository.revision_tree(
1115
 
                        _mod_revision.NULL_REVISION)))
 
990
                    self.branch.repository.revision_tree(None)))
1116
991
                ghosts.append(rev_id)
1117
 
            accepted_revisions.add(rev_id)
 
992
        dirstate._validate()
1118
993
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
994
        dirstate._validate()
1119
995
        self._make_dirty(reset_inventory=False)
 
996
        dirstate._validate()
1120
997
 
1121
998
    def _set_root_id(self, file_id):
1122
999
        """See WorkingTree.set_root_id."""
1125
1002
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1126
1003
            self._make_dirty(reset_inventory=True)
1127
1004
 
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
 
    @needs_read_lock
1139
 
    def supports_tree_reference(self):
1140
 
        return self._repo_supports_tree_reference
1141
 
 
1142
1005
    def unlock(self):
1143
1006
        """Unlock in format 4 trees needs to write the entire dirstate."""
1144
 
        # do non-implementation specific cleanup
1145
 
        self._cleanup()
1146
 
 
1147
1007
        if self._control_files._lock_count == 1:
1148
1008
            # eventually we should do signature checking during read locks for
1149
1009
            # dirstate updates.
1180
1040
            return
1181
1041
        state = self.current_dirstate()
1182
1042
        state._read_dirblocks_if_needed()
1183
 
        ids_to_unversion = set(file_ids)
 
1043
        ids_to_unversion = set()
 
1044
        for file_id in file_ids:
 
1045
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1184
1046
        paths_to_unversion = set()
1185
1047
        # sketch:
1186
1048
        # check if the root is to be unversioned, if so, assert for now.
1213
1075
                # just forget the whole block.
1214
1076
                entry_index = 0
1215
1077
                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
 
1078
                    # Mark this file id as having been removed
 
1079
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1080
                    if not state._make_absent(block[1][entry_index]):
1219
1081
                        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
1082
                # go to the next block. (At the moment we dont delete empty
1227
1083
                # dirblocks)
1228
1084
                block_index += 1
1252
1108
                self._inventory.remove_recursive_id(file_id)
1253
1109
 
1254
1110
    @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
1111
    def _write_inventory(self, inv):
1279
1112
        """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:
 
1113
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1114
        self.current_dirstate().set_state_from_inventory(inv)
 
1115
        self._make_dirty(reset_inventory=False)
 
1116
        if self._inventory is not None:
1294
1117
            self._inventory = inv
1295
1118
        self.flush()
1296
1119
 
1297
1120
 
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):
 
1121
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1122
    """The first consolidated dirstate working tree format.
 
1123
 
 
1124
    This format:
 
1125
        - exists within a metadir controlling .bzr
 
1126
        - includes an explicit version marker for the workingtree control
 
1127
          files, separate from the BzrDir format
 
1128
        - modifies the hash cache format
 
1129
        - is new in bzr TODO FIXME SETBEFOREMERGE
 
1130
        - uses a LockDir to guard access to it.
 
1131
    """
 
1132
 
 
1133
    supports_tree_reference = True
 
1134
 
 
1135
    def get_format_string(self):
 
1136
        """See WorkingTreeFormat.get_format_string()."""
 
1137
        return "Bazaar Working Tree format 4\n"
 
1138
 
 
1139
    def get_format_description(self):
 
1140
        """See WorkingTreeFormat.get_format_description()."""
 
1141
        return "Working tree format 4"
 
1142
 
 
1143
    def initialize(self, a_bzrdir, revision_id=None):
1391
1144
        """See WorkingTreeFormat.initialize().
1392
1145
 
1393
1146
        :param revision_id: allows creating a working tree at a different
1394
1147
        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
1148
 
1402
 
        These trees get an initial random root id, if their repository supports
1403
 
        rich root data, TREE_ROOT otherwise.
 
1149
        These trees get an initial random root id.
1404
1150
        """
 
1151
        revision_id = osutils.safe_revision_id(revision_id)
1405
1152
        if not isinstance(a_bzrdir.transport, LocalTransport):
1406
1153
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1407
1154
        transport = a_bzrdir.get_workingtree_transport(self)
1408
1155
        control_files = self._open_control_files(a_bzrdir)
1409
1156
        control_files.create_lock()
1410
1157
        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()
 
1158
        control_files.put_utf8('format', self.get_format_string())
 
1159
        branch = a_bzrdir.open_branch()
1417
1160
        if revision_id is None:
1418
1161
            revision_id = branch.last_revision()
1419
1162
        local_path = transport.local_abspath('dirstate')
1420
1163
        # write out new dirstate (must exist when we create the tree)
1421
1164
        state = dirstate.DirState.initialize(local_path)
1422
1165
        state.unlock()
1423
 
        del state
1424
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1166
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1425
1167
                         branch,
1426
1168
                         _format=self,
1427
1169
                         _bzrdir=a_bzrdir,
1428
1170
                         _control_files=control_files)
1429
1171
        wt._new_tree()
1430
1172
        wt.lock_tree_write()
 
1173
        state._validate()
1431
1174
        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)
 
1175
            if revision_id in (None, NULL_REVISION):
 
1176
                wt._set_root_id(generate_ids.gen_root_id())
1438
1177
                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)]
 
1178
                wt.current_dirstate()._validate()
 
1179
            wt.set_last_revision(revision_id)
 
1180
            wt.flush()
 
1181
            basis = wt.basis_tree()
1454
1182
            basis.lock_read()
1455
 
            try:
1456
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1183
            # if the basis has a root id we have to use that; otherwise we use
 
1184
            # a new random one
 
1185
            basis_root_id = basis.get_root_id()
 
1186
            if basis_root_id is not None:
 
1187
                wt._set_root_id(basis_root_id)
1457
1188
                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()
 
1189
            transform.build_tree(basis, wt)
 
1190
            basis.unlock()
1479
1191
        finally:
1480
1192
            control_files.unlock()
1481
1193
            wt.unlock()
1482
1194
        return wt
1483
1195
 
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
1196
    def _open(self, a_bzrdir, control_files):
1494
1197
        """Open the tree itself.
1495
1198
 
1496
1199
        :param a_bzrdir: the dir for the tree.
1497
1200
        :param control_files: the control files for the tree.
1498
1201
        """
1499
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1202
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1500
1203
                           branch=a_bzrdir.open_branch(),
1501
1204
                           _format=self,
1502
1205
                           _bzrdir=a_bzrdir,
1503
1206
                           _control_files=control_files)
1504
1207
 
1505
1208
    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
1209
        # please test against something that will let us do tree references
1511
1210
        return bzrdir.format_registry.make_bzrdir(
1512
1211
            'dirstate-with-subtree')
1514
1213
    _matchingbzrdir = property(__get_matchingbzrdir)
1515
1214
 
1516
1215
 
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
1216
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
 
    """
 
1217
    """A revision tree pulling the inventory from a dirstate."""
1595
1218
 
1596
1219
    def __init__(self, dirstate, revision_id, repository):
1597
1220
        self._dirstate = dirstate
1598
 
        self._revision_id = revision_id
 
1221
        self._revision_id = osutils.safe_revision_id(revision_id)
1599
1222
        self._repository = repository
1600
1223
        self._inventory = None
1601
1224
        self._locked = 0
1602
1225
        self._dirstate_locked = False
1603
 
        self._repo_supports_tree_reference = getattr(
1604
 
            repository._format, "supports_tree_reference",
1605
 
            False)
1606
1226
 
1607
1227
    def __repr__(self):
1608
1228
        return "<%s of %s in %s>" % \
1609
1229
            (self.__class__.__name__, self._revision_id, self._dirstate)
1610
1230
 
1611
 
    def annotate_iter(self, file_id,
1612
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1231
    def annotate_iter(self, file_id):
1613
1232
        """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]
 
1233
        w = self._repository.weave_store.get_weave(file_id,
 
1234
                           self._repository.get_transaction())
 
1235
        return w.annotate_iter(self.inventory[file_id].revision)
1617
1236
 
1618
 
    def _get_ancestors(self, default_revision):
1619
 
        return set(self._repository.get_ancestry(self._revision_id,
1620
 
                                                 topo_sorted=False))
1621
1237
    def _comparison_data(self, entry, path):
1622
1238
        """See Tree._comparison_data."""
1623
1239
        if entry is None:
1640
1256
    def get_root_id(self):
1641
1257
        return self.path2id('')
1642
1258
 
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
1259
    def _get_parent_index(self):
1660
1260
        """Return the index in the dirstate referenced by this tree."""
1661
1261
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1666
1266
        If either file_id or path is supplied, it is used as the key to lookup.
1667
1267
        If both are supplied, the fastest lookup is used, and an error is
1668
1268
        raised if they do not both point at the same row.
1669
 
 
 
1269
        
1670
1270
        :param file_id: An optional unicode file_id to be looked up.
1671
1271
        :param path: An optional unicode path to be looked up.
1672
1272
        :return: The dirstate row tuple for path/file_id, or (None, None)
1673
1273
        """
1674
1274
        if file_id is None and path is None:
1675
1275
            raise errors.BzrError('must supply file_id or path')
 
1276
        file_id = osutils.safe_file_id(file_id)
1676
1277
        if path is not None:
1677
1278
            path = path.encode('utf8')
1678
1279
        parent_index = self._get_parent_index()
1686
1287
 
1687
1288
        This is relatively expensive: we have to walk the entire dirstate.
1688
1289
        """
1689
 
        if not self._locked:
1690
 
            raise AssertionError(
1691
 
                'cannot generate inventory of an unlocked '
1692
 
                'dirstate revision tree')
 
1290
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1291
            'dirstate revision tree'
1693
1292
        # separate call for profiling - makes it clear where the costs are.
1694
1293
        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()))
 
1294
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1295
            'parent %s has disappeared from %s' % (
 
1296
            self._revision_id, self._dirstate.get_parent_ids())
1699
1297
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1700
1298
        # This is identical now to the WorkingTree _generate_inventory except
1701
1299
        # for the tree index use.
1702
1300
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1703
1301
        current_id = root_key[2]
1704
 
        if current_entry[parent_index][0] != 'd':
1705
 
            raise AssertionError()
 
1302
        assert current_entry[parent_index][0] == 'd'
1706
1303
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1707
1304
        inv.root.revision = current_entry[parent_index][4]
1708
1305
        # Turn some things into local variables
1740
1337
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1741
1338
                elif kind == 'symlink':
1742
1339
                    inv_entry.executable = False
1743
 
                    inv_entry.text_size = None
 
1340
                    inv_entry.text_size = size
1744
1341
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1745
1342
                elif kind == 'tree-reference':
1746
 
                    inv_entry.reference_revision = fingerprint or None
 
1343
                    inv_entry.reference_revision = fingerprint
1747
1344
                else:
1748
1345
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1749
1346
                            % entry)
1750
1347
                # 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,))
 
1348
                assert file_id not in inv_byid
 
1349
                assert name_unicode not in parent_ie.children
1757
1350
                inv_byid[file_id] = inv_entry
1758
1351
                parent_ie.children[name_unicode] = inv_entry
1759
1352
        self._inventory = inv
1769
1362
            return None
1770
1363
        parent_index = self._get_parent_index()
1771
1364
        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
 
1365
        return self._repository.get_revision(last_changed_revision).timestamp
1777
1366
 
1778
1367
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1779
1368
        entry = self._get_entry(file_id=file_id, path=path)
1783
1372
            return parent_details[1]
1784
1373
        return None
1785
1374
 
1786
 
    def get_file(self, file_id, path=None):
 
1375
    def get_file(self, file_id):
1787
1376
        return StringIO(self.get_file_text(file_id))
1788
1377
 
 
1378
    def get_file_lines(self, file_id):
 
1379
        ie = self.inventory[file_id]
 
1380
        return self._repository.weave_store.get_weave(file_id,
 
1381
                self._repository.get_transaction()).get_lines(ie.revision)
 
1382
 
1789
1383
    def get_file_size(self, file_id):
1790
 
        """See Tree.get_file_size"""
1791
1384
        return self.inventory[file_id].text_size
1792
1385
 
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)
1796
 
 
1797
 
    def get_reference_revision(self, file_id, path=None):
1798
 
        return self.inventory[file_id].reference_revision
1799
 
 
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)
 
1386
    def get_file_text(self, file_id):
 
1387
        return ''.join(self.get_file_lines(file_id))
1813
1388
 
1814
1389
    def get_symlink_target(self, file_id):
1815
1390
        entry = self._get_entry(file_id=file_id)
1817
1392
        if entry[1][parent_index][0] != 'l':
1818
1393
            return None
1819
1394
        else:
1820
 
            target = entry[1][parent_index][1]
1821
 
            target = target.decode('utf8')
1822
 
            return target
 
1395
            # At present, none of the tree implementations supports non-ascii
 
1396
            # symlink targets. So we will just assume that the dirstate path is
 
1397
            # correct.
 
1398
            return entry[1][parent_index][1]
1823
1399
 
1824
1400
    def get_revision_id(self):
1825
1401
        """Return the revision id for this tree."""
1843
1419
        return bool(self.path2id(filename))
1844
1420
 
1845
1421
    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)
 
1422
        return self.inventory[file_id].kind
1869
1423
 
1870
1424
    def is_executable(self, file_id, path=None):
1871
1425
        ie = self.inventory[file_id]
1873
1427
            return None
1874
1428
        return ie.executable
1875
1429
 
1876
 
    def is_locked(self):
1877
 
        return self._locked
1878
 
 
1879
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1430
    def list_files(self, include_root=False):
1880
1431
        # We use a standard implementation, because DirStateRevisionTree is
1881
1432
        # dealing with one of the parents of the current state
1882
1433
        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:
 
1434
        entries = inv.iter_entries()
 
1435
        if self.inventory.root is not None and not include_root:
1892
1436
            entries.next()
1893
1437
        for path, entry in entries:
1894
1438
            yield path, 'V', entry.kind, entry.file_id, entry
1895
1439
 
1896
1440
    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
 
        """
 
1441
        """Lock the tree for a set of operations."""
1902
1442
        if not self._locked:
1903
1443
            self._repository.lock_read()
1904
1444
            if self._dirstate._lock_token is None:
1905
1445
                self._dirstate.lock_read()
1906
1446
                self._dirstate_locked = True
1907
1447
        self._locked += 1
1908
 
        return self
1909
1448
 
1910
1449
    def _must_be_locked(self):
1911
1450
        if not self._locked:
1932
1471
                self._dirstate_locked = False
1933
1472
            self._repository.unlock()
1934
1473
 
1935
 
    @needs_read_lock
1936
 
    def supports_tree_reference(self):
1937
 
        return self._repo_supports_tree_reference
1938
 
 
1939
1474
    def walkdirs(self, prefix=""):
1940
 
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1941
 
        # implementation based on an inventory.
1942
 
        # This should be cleaned up to use the much faster Dirstate code
1943
 
        # So for now, we just build up the parent inventory, and extract
1944
 
        # it the same way RevisionTree does.
 
1475
        # TODO: jam 20070215 This is the cheap way by cheating and using the
 
1476
        #       RevisionTree implementation.
 
1477
        #       This should be cleaned up to use the much faster Dirstate code
 
1478
        #       This is a little tricky, though, because the dirstate is
 
1479
        #       indexed by current path, not by parent path.
 
1480
        #       So for now, we just build up the parent inventory, and extract
 
1481
        #       it the same way RevisionTree does.
1945
1482
        _directory = 'directory'
1946
1483
        inv = self._get_inventory()
1947
1484
        top_id = inv.path2id(prefix)
1972
1509
 
1973
1510
 
1974
1511
class InterDirStateTree(InterTree):
1975
 
    """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
1979
 
    the current working tree of the same dirstate.
1980
 
    """
1981
 
    # this could be generalized to allow comparisons between any trees in the
1982
 
    # dirstate, and possibly between trees stored in different dirstates.
 
1512
    """Fast path optimiser for changes_from with dirstate trees."""
1983
1513
 
1984
1514
    def __init__(self, source, target):
1985
1515
        super(InterDirStateTree, self).__init__(source, target)
1994
1524
        target.set_parent_ids([revid])
1995
1525
        return target.basis_tree(), target
1996
1526
 
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
1527
    _matching_from_tree_format = WorkingTreeFormat4()
2015
1528
    _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,
 
1529
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1530
 
 
1531
    def _iter_changes(self, include_unchanged=False,
2024
1532
                      specific_files=None, pb=None, extra_trees=[],
2025
1533
                      require_versioned=True, want_unversioned=False):
2026
1534
        """Return the changes from source to target.
2027
1535
 
2028
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1536
        :return: An iterator that yields tuples. See InterTree._iter_changes
2029
1537
            for details.
2030
1538
        :param specific_files: An optional list of file paths to restrict the
2031
1539
            comparison to. When mapping filenames to ids, all matches in all
2042
1550
            output. An unversioned file is defined as one with (False, False)
2043
1551
            for the versioned pair.
2044
1552
        """
 
1553
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1554
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1555
        # NB: show_status depends on being able to pass in non-versioned files
 
1556
        # and report them as unknown
2045
1557
        # TODO: handle extra trees in the dirstate.
2046
 
        if (extra_trees or specific_files == []):
 
1558
        # TODO: handle comparisons as an empty tree as a different special
 
1559
        # case? mbp 20070226
 
1560
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2047
1561
            # we can't fast-path these cases (yet)
2048
 
            return super(InterDirStateTree, self).iter_changes(
 
1562
            for f in super(InterDirStateTree, self)._iter_changes(
2049
1563
                include_unchanged, specific_files, pb, extra_trees,
2050
 
                require_versioned, want_unversioned=want_unversioned)
 
1564
                require_versioned):
 
1565
                yield f
 
1566
            return
2051
1567
        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))
 
1568
        assert (self.source._revision_id in parent_ids), \
 
1569
                "revision {%s} is not stored in {%s}, but %s " \
 
1570
                "can only be used for trees stored in the dirstate" \
 
1571
                % (self.source._revision_id, self.target, self._iter_changes)
2058
1572
        target_index = 0
2059
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1573
        if self.source._revision_id == NULL_REVISION:
2060
1574
            source_index = None
2061
1575
            indices = (target_index,)
2062
1576
        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))
 
1577
            assert (self.source._revision_id in parent_ids), \
 
1578
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1579
                self.source._revision_id, parent_ids)
2067
1580
            source_index = 1 + parent_ids.index(self.source._revision_id)
2068
 
            indices = (source_index, target_index)
 
1581
            indices = (source_index,target_index)
2069
1582
        # -- make all specific_files utf8 --
2070
1583
        if specific_files:
2071
1584
            specific_files_utf8 = set()
2072
1585
            for path in specific_files:
2073
 
                # Note, if there are many specific files, using cache_utf8
2074
 
                # would be good here.
2075
1586
                specific_files_utf8.add(path.encode('utf8'))
2076
1587
            specific_files = specific_files_utf8
2077
1588
        else:
2078
1589
            specific_files = set([''])
2079
1590
        # -- specific_files is now a utf8 path set --
2080
 
 
2081
1591
        # -- get the state object and prepare it.
2082
1592
        state = self.target.current_dirstate()
2083
1593
        state._read_dirblocks_if_needed()
 
1594
        def _entries_for_path(path):
 
1595
            """Return a list with all the entries that match path for all ids.
 
1596
            """
 
1597
            dirname, basename = os.path.split(path)
 
1598
            key = (dirname, basename, '')
 
1599
            block_index, present = state._find_block_index_from_key(key)
 
1600
            if not present:
 
1601
                # the block which should contain path is absent.
 
1602
                return []
 
1603
            result = []
 
1604
            block = state._dirblocks[block_index][1]
 
1605
            entry_index, _ = state._find_entry_index(key, block)
 
1606
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1607
            while (entry_index < len(block) and
 
1608
                block[entry_index][0][0:2] == key[0:2]):
 
1609
                result.append(block[entry_index])
 
1610
                entry_index += 1
 
1611
            return result
2084
1612
        if require_versioned:
2085
1613
            # -- check all supplied paths are versioned in a search tree. --
2086
 
            not_versioned = []
 
1614
            all_versioned = True
2087
1615
            for path in specific_files:
2088
 
                path_entries = state._entries_for_path(path)
 
1616
                path_entries = _entries_for_path(path)
2089
1617
                if not path_entries:
2090
1618
                    # this specified path is not present at all: error
2091
 
                    not_versioned.append(path)
2092
 
                    continue
 
1619
                    all_versioned = False
 
1620
                    break
2093
1621
                found_versioned = False
2094
1622
                # for each id at this path
2095
1623
                for entry in path_entries:
2102
1630
                if not found_versioned:
2103
1631
                    # none of the indexes was not 'absent' at all ids for this
2104
1632
                    # path.
2105
 
                    not_versioned.append(path)
2106
 
            if len(not_versioned) > 0:
2107
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1633
                    all_versioned = False
 
1634
                    break
 
1635
            if not all_versioned:
 
1636
                raise errors.PathsNotVersionedError(specific_files)
2108
1637
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2109
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1638
        search_specific_files = set()
 
1639
        for path in specific_files:
 
1640
            other_specific_files = specific_files.difference(set([path]))
 
1641
            if not osutils.is_inside_any(other_specific_files, path):
 
1642
                # this is a top level path, we must check it.
 
1643
                search_specific_files.add(path)
 
1644
        # sketch: 
 
1645
        # compare source_index and target_index at or under each element of search_specific_files.
 
1646
        # follow the following comparison table. Note that we only want to do diff operations when
 
1647
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1648
        # for the target.
 
1649
        # cases:
 
1650
        # 
 
1651
        # Source | Target | disk | action
 
1652
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1653
        #        |        |      | diff check on source-target
 
1654
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1655
        #        |        |      | ???
 
1656
        #   r    |  a     |      | add source to search
 
1657
        #   r    |  a     |  a   | 
 
1658
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1659
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1660
        #   a    | fdlt   |      | add new id
 
1661
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1662
        #   a    |  a     |      | not present in either tree, skip
 
1663
        #   a    |  a     |  a   | not present in any tree, skip
 
1664
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1665
        #        |        |      | may not be selected by the users list of paths.
 
1666
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1667
        #        |        |      | may not be selected by the users list of paths.
 
1668
        #  fdlt  | fdlt   |      | content in both: diff them
 
1669
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1670
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1671
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1672
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1673
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1674
        #        |        |      | this id at the other path.
 
1675
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1676
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1677
        #        |        |      | this id at the other path.
 
1678
 
 
1679
        # for all search_indexs in each path at or under each element of
 
1680
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1681
        # relocated path as one to search if its not searched already. If the
 
1682
        # detail is not relocated, add the id.
 
1683
        searched_specific_files = set()
 
1684
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1685
        # Using a list so that we can access the values and change them in
 
1686
        # nested scope. Each one is [path, file_id, entry]
 
1687
        last_source_parent = [None, None, None]
 
1688
        last_target_parent = [None, None, None]
2110
1689
 
2111
1690
        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()
 
1691
 
 
1692
        def _process_entry(entry, path_info):
 
1693
            """Compare an entry and real disk to generate delta information.
 
1694
 
 
1695
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1696
                the path of entry. If None, then the path is considered absent.
 
1697
                (Perhaps we should pass in a concrete entry for this ?)
 
1698
                Basename is returned as a utf8 string because we expect this
 
1699
                tuple will be ignored, and don't want to take the time to
 
1700
                decode.
 
1701
            """
 
1702
            # TODO: when a parent has been renamed, dont emit path renames for children,
 
1703
            if source_index is None:
 
1704
                source_details = NULL_PARENT_DETAILS
 
1705
            else:
 
1706
                source_details = entry[1][source_index]
 
1707
            target_details = entry[1][target_index]
 
1708
            target_minikind = target_details[0]
 
1709
            if path_info is not None and target_minikind in 'fdl':
 
1710
                assert target_index == 0
 
1711
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1712
                                                  stat_value=path_info[3])
 
1713
                # The entry may have been modified by update_entry
 
1714
                target_details = entry[1][target_index]
 
1715
                target_minikind = target_details[0]
 
1716
            else:
 
1717
                link_or_sha1 = None
 
1718
            source_minikind = source_details[0]
 
1719
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1720
                # claimed content in both: diff
 
1721
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1722
                #        |        |      | diff check on source-target
 
1723
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1724
                #        |        |      | ???
 
1725
                if source_minikind in 'r':
 
1726
                    # add the source to the search path to find any children it
 
1727
                    # has.  TODO ? : only add if it is a container ?
 
1728
                    if not osutils.is_inside_any(searched_specific_files,
 
1729
                                                 source_details[1]):
 
1730
                        search_specific_files.add(source_details[1])
 
1731
                    # generate the old path; this is needed for stating later
 
1732
                    # as well.
 
1733
                    old_path = source_details[1]
 
1734
                    old_dirname, old_basename = os.path.split(old_path)
 
1735
                    path = pathjoin(entry[0][0], entry[0][1])
 
1736
                    old_entry = state._get_entry(source_index,
 
1737
                                                 path_utf8=old_path)
 
1738
                    # update the source details variable to be the real
 
1739
                    # location.
 
1740
                    source_details = old_entry[1][source_index]
 
1741
                    source_minikind = source_details[0]
 
1742
                else:
 
1743
                    old_dirname = entry[0][0]
 
1744
                    old_basename = entry[0][1]
 
1745
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1746
                if path_info is None:
 
1747
                    # the file is missing on disk, show as removed.
 
1748
                    content_change = True
 
1749
                    target_kind = None
 
1750
                    target_exec = False
 
1751
                else:
 
1752
                    # source and target are both versioned and disk file is present.
 
1753
                    target_kind = path_info[2]
 
1754
                    if target_kind == 'directory':
 
1755
                        if source_minikind != 'd':
 
1756
                            content_change = True
 
1757
                        else:
 
1758
                            # directories have no fingerprint
 
1759
                            content_change = False
 
1760
                        target_exec = False
 
1761
                    elif target_kind == 'file':
 
1762
                        if source_minikind != 'f':
 
1763
                            content_change = True
 
1764
                        else:
 
1765
                            # We could check the size, but we already have the
 
1766
                            # sha1 hash.
 
1767
                            content_change = (link_or_sha1 != source_details[1])
 
1768
                        # Target details is updated at update_entry time
 
1769
                        if use_filesystem_for_exec:
 
1770
                            # We don't need S_ISREG here, because we are sure
 
1771
                            # we are dealing with a file.
 
1772
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1773
                        else:
 
1774
                            target_exec = target_details[3]
 
1775
                    elif target_kind == 'symlink':
 
1776
                        if source_minikind != 'l':
 
1777
                            content_change = True
 
1778
                        else:
 
1779
                            content_change = (link_or_sha1 != source_details[1])
 
1780
                        target_exec = False
 
1781
                    elif target_kind == 'tree-reference':
 
1782
                        if source_minikind != 't':
 
1783
                            content_change = True
 
1784
                        else:
 
1785
                            content_change = False
 
1786
                    else:
 
1787
                        raise Exception, "unknown kind %s" % path_info[2]
 
1788
                # parent id is the entry for the path in the target tree
 
1789
                if old_dirname == last_source_parent[0]:
 
1790
                    source_parent_id = last_source_parent[1]
 
1791
                else:
 
1792
                    source_parent_entry = state._get_entry(source_index,
 
1793
                                                           path_utf8=old_dirname)
 
1794
                    source_parent_id = source_parent_entry[0][2]
 
1795
                    if source_parent_id == entry[0][2]:
 
1796
                        # This is the root, so the parent is None
 
1797
                        source_parent_id = None
 
1798
                    else:
 
1799
                        last_source_parent[0] = old_dirname
 
1800
                        last_source_parent[1] = source_parent_id
 
1801
                        last_source_parent[2] = source_parent_entry
 
1802
 
 
1803
                new_dirname = entry[0][0]
 
1804
                if new_dirname == last_target_parent[0]:
 
1805
                    target_parent_id = last_target_parent[1]
 
1806
                else:
 
1807
                    # TODO: We don't always need to do the lookup, because the
 
1808
                    #       parent entry will be the same as the source entry.
 
1809
                    target_parent_entry = state._get_entry(target_index,
 
1810
                                                           path_utf8=new_dirname)
 
1811
                    target_parent_id = target_parent_entry[0][2]
 
1812
                    if target_parent_id == entry[0][2]:
 
1813
                        # This is the root, so the parent is None
 
1814
                        target_parent_id = None
 
1815
                    else:
 
1816
                        last_target_parent[0] = new_dirname
 
1817
                        last_target_parent[1] = target_parent_id
 
1818
                        last_target_parent[2] = target_parent_entry
 
1819
 
 
1820
                source_exec = source_details[3]
 
1821
                return ((entry[0][2], (old_path, path), content_change,
 
1822
                        (True, True),
 
1823
                        (source_parent_id, target_parent_id),
 
1824
                        (old_basename, entry[0][1]),
 
1825
                        (_minikind_to_kind[source_minikind], target_kind),
 
1826
                        (source_exec, target_exec)),)
 
1827
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1828
                # looks like a new file
 
1829
                if path_info is not None:
 
1830
                    path = pathjoin(entry[0][0], entry[0][1])
 
1831
                    # parent id is the entry for the path in the target tree
 
1832
                    # TODO: these are the same for an entire directory: cache em.
 
1833
                    parent_id = state._get_entry(target_index,
 
1834
                                                 path_utf8=entry[0][0])[0][2]
 
1835
                    if parent_id == entry[0][2]:
 
1836
                        parent_id = None
 
1837
                    if use_filesystem_for_exec:
 
1838
                        # We need S_ISREG here, because we aren't sure if this
 
1839
                        # is a file or not.
 
1840
                        target_exec = bool(
 
1841
                            stat.S_ISREG(path_info[3].st_mode)
 
1842
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1843
                    else:
 
1844
                        target_exec = target_details[3]
 
1845
                    return ((entry[0][2], (None, path), True,
 
1846
                            (False, True),
 
1847
                            (None, parent_id),
 
1848
                            (None, entry[0][1]),
 
1849
                            (None, path_info[2]),
 
1850
                            (None, target_exec)),)
 
1851
                else:
 
1852
                    # but its not on disk: we deliberately treat this as just
 
1853
                    # never-present. (Why ?! - RBC 20070224)
 
1854
                    pass
 
1855
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1856
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1857
                # if its still on disk, *and* theres no other entry at this
 
1858
                # path [we dont know this in this routine at the moment -
 
1859
                # perhaps we should change this - then it would be an unknown.
 
1860
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1861
                # parent id is the entry for the path in the target tree
 
1862
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1863
                if parent_id == entry[0][2]:
 
1864
                    parent_id = None
 
1865
                return ((entry[0][2], (old_path, None), True,
 
1866
                        (True, False),
 
1867
                        (parent_id, None),
 
1868
                        (entry[0][1], None),
 
1869
                        (_minikind_to_kind[source_minikind], None),
 
1870
                        (source_details[3], None)),)
 
1871
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1872
                # a rename; could be a true rename, or a rename inherited from
 
1873
                # a renamed parent. TODO: handle this efficiently. Its not
 
1874
                # common case to rename dirs though, so a correct but slow
 
1875
                # implementation will do.
 
1876
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1877
                    search_specific_files.add(target_details[1])
 
1878
            elif source_minikind in 'r' and target_minikind in 'r':
 
1879
                # neither of the selected trees contain this file,
 
1880
                # so skip over it. This is not currently directly tested, but
 
1881
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1882
                pass
 
1883
            else:
 
1884
                print "*******", source_minikind, target_minikind
 
1885
                import pdb;pdb.set_trace()
 
1886
            return ()
 
1887
        while search_specific_files:
 
1888
            # TODO: the pending list should be lexically sorted?
 
1889
            current_root = search_specific_files.pop()
 
1890
            searched_specific_files.add(current_root)
 
1891
            # process the entries for this containing directory: the rest will be
 
1892
            # found by their parents recursively.
 
1893
            root_entries = _entries_for_path(current_root)
 
1894
            root_abspath = self.target.abspath(current_root)
 
1895
            try:
 
1896
                root_stat = os.lstat(root_abspath)
 
1897
            except OSError, e:
 
1898
                if e.errno == errno.ENOENT:
 
1899
                    # the path does not exist: let _process_entry know that.
 
1900
                    root_dir_info = None
 
1901
                else:
 
1902
                    # some other random error: hand it up.
 
1903
                    raise
 
1904
            else:
 
1905
                root_dir_info = ('', current_root,
 
1906
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1907
                    root_abspath)
 
1908
            if not root_entries and not root_dir_info:
 
1909
                # this specified path is not present at all, skip it.
 
1910
                continue
 
1911
            path_handled = False
 
1912
            for entry in root_entries:
 
1913
                for result in _process_entry(entry, root_dir_info):
 
1914
                    # this check should probably be outside the loop: one
 
1915
                    # 'iterate two trees' api, and then _iter_changes filters
 
1916
                    # unchanged pairs. - RBC 20070226
 
1917
                    path_handled = True
 
1918
                    if (include_unchanged
 
1919
                        or result[2]                    # content change
 
1920
                        or result[3][0] != result[3][1] # versioned status
 
1921
                        or result[4][0] != result[4][1] # parent id
 
1922
                        or result[5][0] != result[5][1] # name
 
1923
                        or result[6][0] != result[6][1] # kind
 
1924
                        or result[7][0] != result[7][1] # executable
 
1925
                        ):
 
1926
                        result = (result[0],
 
1927
                            ((utf8_decode(result[1][0])[0]),
 
1928
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
1929
                        yield result
 
1930
            if want_unversioned and not path_handled:
 
1931
                new_executable = bool(
 
1932
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
1933
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
1934
                yield (None, (None, current_root), True, (False, False),
 
1935
                    (None, None),
 
1936
                    (None, splitpath(current_root)[-1]),
 
1937
                    (None, root_dir_info[2]), (None, new_executable))
 
1938
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
1939
            initial_key = (current_root, '', '')
 
1940
            block_index, _ = state._find_block_index_from_key(initial_key)
 
1941
            if block_index == 0:
 
1942
                # we have processed the total root already, but because the
 
1943
                # initial key matched it we should skip it here.
 
1944
                block_index +=1
 
1945
            try:
 
1946
                current_dir_info = dir_iterator.next()
 
1947
            except OSError, e:
 
1948
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
1949
                    # there may be directories in the inventory even though
 
1950
                    # this path is not a file on disk: so mark it as end of
 
1951
                    # iterator
 
1952
                    current_dir_info = None
 
1953
                else:
 
1954
                    raise
 
1955
            else:
 
1956
                if current_dir_info[0][0] == '':
 
1957
                    # remove .bzr from iteration
 
1958
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
1959
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
1960
                    del current_dir_info[1][bzr_index]
 
1961
            # walk until both the directory listing and the versioned metadata
 
1962
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
 
1963
            # the versioned data runs out.
 
1964
            if (block_index < len(state._dirblocks) and
 
1965
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
1966
                current_block = state._dirblocks[block_index]
 
1967
            else:
 
1968
                current_block = None
 
1969
            while (current_dir_info is not None or
 
1970
                   current_block is not None):
 
1971
                if (current_dir_info and current_block
 
1972
                    and current_dir_info[0][0] != current_block[0]):
 
1973
                    if current_dir_info[0][0] < current_block[0] :
 
1974
                        # import pdb; pdb.set_trace()
 
1975
                        # print 'unversioned dir'
 
1976
                        # filesystem data refers to paths not covered by the dirblock.
 
1977
                        # this has two possibilities:
 
1978
                        # A) it is versioned but empty, so there is no block for it
 
1979
                        # B) it is not versioned.
 
1980
                        # in either case it was processed by the containing directories walk:
 
1981
                        # if it is root/foo, when we walked root we emitted it,
 
1982
                        # or if we ere given root/foo to walk specifically, we
 
1983
                        # emitted it when checking the walk-root entries
 
1984
                        # advance the iterator and loop - we dont need to emit it.
 
1985
                        try:
 
1986
                            current_dir_info = dir_iterator.next()
 
1987
                        except StopIteration:
 
1988
                            current_dir_info = None
 
1989
                    else:
 
1990
                        # We have a dirblock entry for this location, but there
 
1991
                        # is no filesystem path for this. This is most likely
 
1992
                        # because a directory was removed from the disk.
 
1993
                        # We don't have to report the missing directory,
 
1994
                        # because that should have already been handled, but we
 
1995
                        # need to handle all of the files that are contained
 
1996
                        # within.
 
1997
                        for current_entry in current_block[1]:
 
1998
                            # entry referring to file not present on disk.
 
1999
                            # advance the entry only, after processing.
 
2000
                            for result in _process_entry(current_entry, None):
 
2001
                                # this check should probably be outside the loop: one
 
2002
                                # 'iterate two trees' api, and then _iter_changes filters
 
2003
                                # unchanged pairs. - RBC 20070226
 
2004
                                if (include_unchanged
 
2005
                                    or result[2]                    # content change
 
2006
                                    or result[3][0] != result[3][1] # versioned status
 
2007
                                    or result[4][0] != result[4][1] # parent id
 
2008
                                    or result[5][0] != result[5][1] # name
 
2009
                                    or result[6][0] != result[6][1] # kind
 
2010
                                    or result[7][0] != result[7][1] # executable
 
2011
                                    ):
 
2012
                                    result = (result[0],
 
2013
                                        ((utf8_decode(result[1][0])[0]),
 
2014
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2015
                                    yield result
 
2016
                        block_index +=1
 
2017
                        if (block_index < len(state._dirblocks) and
 
2018
                            osutils.is_inside(current_root,
 
2019
                                              state._dirblocks[block_index][0])):
 
2020
                            current_block = state._dirblocks[block_index]
 
2021
                        else:
 
2022
                            current_block = None
 
2023
                    continue
 
2024
                entry_index = 0
 
2025
                if current_block and entry_index < len(current_block[1]):
 
2026
                    current_entry = current_block[1][entry_index]
 
2027
                else:
 
2028
                    current_entry = None
 
2029
                advance_entry = True
 
2030
                path_index = 0
 
2031
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2032
                    current_path_info = current_dir_info[1][path_index]
 
2033
                else:
 
2034
                    current_path_info = None
 
2035
                advance_path = True
 
2036
                path_handled = False
 
2037
                while (current_entry is not None or
 
2038
                    current_path_info is not None):
 
2039
                    if current_entry is None:
 
2040
                        # the check for path_handled when the path is adnvaced
 
2041
                        # will yield this path if needed.
 
2042
                        pass
 
2043
                    elif current_path_info is None:
 
2044
                        # no path is fine: the per entry code will handle it.
 
2045
                        for result in _process_entry(current_entry, current_path_info):
 
2046
                            # this check should probably be outside the loop: one
 
2047
                            # 'iterate two trees' api, and then _iter_changes filters
 
2048
                            # unchanged pairs. - RBC 20070226
 
2049
                            if (include_unchanged
 
2050
                                or result[2]                    # content change
 
2051
                                or result[3][0] != result[3][1] # versioned status
 
2052
                                or result[4][0] != result[4][1] # parent id
 
2053
                                or result[5][0] != result[5][1] # name
 
2054
                                or result[6][0] != result[6][1] # kind
 
2055
                                or result[7][0] != result[7][1] # executable
 
2056
                                ):
 
2057
                                result = (result[0],
 
2058
                                    ((utf8_decode(result[1][0])[0]),
 
2059
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2060
                                yield result
 
2061
                    elif current_entry[0][1] != current_path_info[1]:
 
2062
                        if current_path_info[1] < current_entry[0][1]:
 
2063
                            # extra file on disk: pass for now, but only
 
2064
                            # increment the path, not the entry
 
2065
                            # import pdb; pdb.set_trace()
 
2066
                            # print 'unversioned file'
 
2067
                            advance_entry = False
 
2068
                        else:
 
2069
                            # entry referring to file not present on disk.
 
2070
                            # advance the entry only, after processing.
 
2071
                            for result in _process_entry(current_entry, None):
 
2072
                                # this check should probably be outside the loop: one
 
2073
                                # 'iterate two trees' api, and then _iter_changes filters
 
2074
                                # unchanged pairs. - RBC 20070226
 
2075
                                path_handled = True
 
2076
                                if (include_unchanged
 
2077
                                    or result[2]                    # content change
 
2078
                                    or result[3][0] != result[3][1] # versioned status
 
2079
                                    or result[4][0] != result[4][1] # parent id
 
2080
                                    or result[5][0] != result[5][1] # name
 
2081
                                    or result[6][0] != result[6][1] # kind
 
2082
                                    or result[7][0] != result[7][1] # executable
 
2083
                                    ):
 
2084
                                    result = (result[0],
 
2085
                                        ((utf8_decode(result[1][0])[0]),
 
2086
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2087
                                    yield result
 
2088
                            advance_path = False
 
2089
                    else:
 
2090
                        for result in _process_entry(current_entry, current_path_info):
 
2091
                            # this check should probably be outside the loop: one
 
2092
                            # 'iterate two trees' api, and then _iter_changes filters
 
2093
                            # unchanged pairs. - RBC 20070226
 
2094
                            path_handled = True
 
2095
                            if (include_unchanged
 
2096
                                or result[2]                    # content change
 
2097
                                or result[3][0] != result[3][1] # versioned status
 
2098
                                or result[4][0] != result[4][1] # parent id
 
2099
                                or result[5][0] != result[5][1] # name
 
2100
                                or result[6][0] != result[6][1] # kind
 
2101
                                or result[7][0] != result[7][1] # executable
 
2102
                                ):
 
2103
                                result = (result[0],
 
2104
                                    ((utf8_decode(result[1][0])[0]),
 
2105
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2106
                                yield result
 
2107
                    if advance_entry and current_entry is not None:
 
2108
                        entry_index += 1
 
2109
                        if entry_index < len(current_block[1]):
 
2110
                            current_entry = current_block[1][entry_index]
 
2111
                        else:
 
2112
                            current_entry = None
 
2113
                    else:
 
2114
                        advance_entry = True # reset the advance flaga
 
2115
                    if advance_path and current_path_info is not None:
 
2116
                        if not path_handled:
 
2117
                            # unversioned in all regards
 
2118
                            if want_unversioned:
 
2119
                                new_executable = bool(
 
2120
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2121
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2122
                                if want_unversioned:
 
2123
                                    yield (None, (None, current_path_info[0]),
 
2124
                                        True,
 
2125
                                        (False, False),
 
2126
                                        (None, None),
 
2127
                                        (None, current_path_info[1]),
 
2128
                                        (None, current_path_info[2]),
 
2129
                                        (None, new_executable))
 
2130
                            # dont descend into this unversioned path if it is
 
2131
                            # a dir
 
2132
                            if current_path_info[2] == 'directory':
 
2133
                                del current_dir_info[1][path_index]
 
2134
                                path_index -= 1
 
2135
                        path_index += 1
 
2136
                        if path_index < len(current_dir_info[1]):
 
2137
                            current_path_info = current_dir_info[1][path_index]
 
2138
                        else:
 
2139
                            current_path_info = None
 
2140
                        path_handled = False
 
2141
                    else:
 
2142
                        advance_path = True # reset the advance flagg.
 
2143
                if current_block is not None:
 
2144
                    block_index += 1
 
2145
                    if (block_index < len(state._dirblocks) and
 
2146
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2147
                        current_block = state._dirblocks[block_index]
 
2148
                    else:
 
2149
                        current_block = None
 
2150
                if current_dir_info is not None:
 
2151
                    try:
 
2152
                        current_dir_info = dir_iterator.next()
 
2153
                    except StopIteration:
 
2154
                        current_dir_info = None
 
2155
 
2116
2156
 
2117
2157
    @staticmethod
2118
2158
    def is_compatible(source, target):
2119
2159
        # the target must be a dirstate working tree
2120
 
        if not isinstance(target, DirStateWorkingTree):
 
2160
        if not isinstance(target, WorkingTree4):
2121
2161
            return False
2122
 
        # the source must be a revtree or dirstate rev tree.
 
2162
        # the source must be a revtreee or dirstate rev tree.
2123
2163
        if not isinstance(source,
2124
2164
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2125
2165
            return False
2126
2166
        # the source revid must be in the target dirstate
2127
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2167
        if not (source._revision_id == NULL_REVISION or
2128
2168
            source._revision_id in target.get_parent_ids()):
2129
 
            # TODO: what about ghosts? it may well need to
 
2169
            # TODO: what about ghosts? it may well need to 
2130
2170
            # check for them explicitly.
2131
2171
            return False
2132
2172
        return True
2142
2182
 
2143
2183
    def convert(self, tree):
2144
2184
        # 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
 
2185
        # on-unlock behaviours, and so that noone else diddles with the 
2146
2186
        # tree during upgrade.
2147
2187
        tree._control_files.lock_write()
2148
2188
        try:
2149
 
            tree.read_working_inventory()
2150
2189
            self.create_dirstate_data(tree)
2151
2190
            self.update_format(tree)
2152
2191
            self.remove_xml_files(tree)
2174
2213
 
2175
2214
    def update_format(self, tree):
2176
2215
        """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())
 
2216
        tree._control_files.put_utf8('format',
 
2217
            self.target_format.get_format_string())