/brz/remove-bazaar

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

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Robert Collins
  • Date: 2007-03-06 08:41:40 UTC
  • mto: (2321.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 2322.
  • Revision ID: robertc@robertcollins.net-20070306084140-10rlo4bozpfvuebh
Get _iter_changes on dirstate passing the subtree tests.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007-2010 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""WorkingTree4 format and implementation.
18
18
 
28
28
 
29
29
from bzrlib.lazy_import import lazy_import
30
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
31
34
import errno
 
35
import itertools
 
36
import operator
32
37
import stat
 
38
from time import time
 
39
import warnings
33
40
 
34
41
import bzrlib
35
42
from bzrlib import (
36
43
    bzrdir,
37
44
    cache_utf8,
38
 
    debug,
 
45
    conflicts as _mod_conflicts,
 
46
    delta,
39
47
    dirstate,
40
48
    errors,
41
49
    generate_ids,
 
50
    globbing,
 
51
    hashcache,
 
52
    ignores,
 
53
    merge,
42
54
    osutils,
43
 
    revision as _mod_revision,
44
55
    revisiontree,
45
 
    trace,
 
56
    textui,
46
57
    transform,
47
 
    views,
 
58
    urlutils,
 
59
    xml5,
 
60
    xml6,
48
61
    )
49
62
import bzrlib.branch
 
63
from bzrlib.transport import get_transport
50
64
import bzrlib.ui
51
65
""")
52
66
 
 
67
from bzrlib import symbol_versioning
53
68
from bzrlib.decorators import needs_read_lock, needs_write_lock
54
 
from bzrlib.filters import filtered_input_file, internal_size_sha_file_byname
55
 
from bzrlib.inventory import Inventory, ROOT_ID, entry_factory
 
69
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
70
from bzrlib.lockable_files import LockableFiles, TransportLock
 
71
from bzrlib.lockdir import LockDir
 
72
import bzrlib.mutabletree
56
73
from bzrlib.mutabletree import needs_tree_write_lock
57
74
from bzrlib.osutils import (
58
75
    file_kind,
59
76
    isdir,
 
77
    normpath,
60
78
    pathjoin,
 
79
    rand_chars,
61
80
    realpath,
62
81
    safe_unicode,
 
82
    splitpath,
63
83
    )
64
 
from bzrlib.trace import mutter
 
84
from bzrlib.trace import mutter, note
65
85
from bzrlib.transport.local import LocalTransport
66
86
from bzrlib.tree import InterTree
 
87
from bzrlib.progress import DummyProgress, ProgressPhase
 
88
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
89
from bzrlib.rio import RioReader, rio_file, Stanza
 
90
from bzrlib.symbol_versioning import (deprecated_passed,
 
91
        deprecated_method,
 
92
        deprecated_function,
 
93
        DEPRECATED_PARAMETER,
 
94
        )
67
95
from bzrlib.tree import Tree
68
96
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
69
97
 
70
98
 
71
 
class DirStateWorkingTree(WorkingTree3):
 
99
class WorkingTree4(WorkingTree3):
 
100
    """This is the Format 4 working tree.
 
101
 
 
102
    This differs from WorkingTree3 by:
 
103
     - Having a consolidated internal dirstate, stored in a
 
104
       randomly-accessible sorted file on disk.
 
105
     - Not having a regular inventory attribute.  One can be synthesized 
 
106
       on demand but this is expensive and should be avoided.
 
107
 
 
108
    This is new in bzr 0.15.
 
109
    """
 
110
 
72
111
    def __init__(self, basedir,
73
112
                 branch,
74
113
                 _control_files=None,
83
122
        """
84
123
        self._format = _format
85
124
        self.bzrdir = _bzrdir
 
125
        from bzrlib.trace import note, mutter
 
126
        assert isinstance(basedir, basestring), \
 
127
            "base directory %r is not a string" % basedir
86
128
        basedir = safe_unicode(basedir)
87
129
        mutter("opening working tree %r", basedir)
88
130
        self._branch = branch
 
131
        assert isinstance(self.branch, bzrlib.branch.Branch), \
 
132
            "branch %r is not a Branch" % self.branch
89
133
        self.basedir = realpath(basedir)
90
134
        # if branch is at our basedir and is a format 6 or less
91
135
        # assume all other formats have their own control files.
 
136
        assert isinstance(_control_files, LockableFiles), \
 
137
            "_control_files must be a LockableFiles, not %r" % _control_files
92
138
        self._control_files = _control_files
93
 
        self._transport = self._control_files._transport
94
139
        self._dirty = None
95
140
        #-------------
96
141
        # during a read or write lock these objects are set, and are
98
143
        self._dirstate = None
99
144
        self._inventory = None
100
145
        #-------------
101
 
        self._setup_directory_is_tree_reference()
102
 
        self._detect_case_handling()
103
 
        self._rules_searcher = None
104
 
        self.views = self._make_views()
105
 
        #--- allow tests to select the dirstate iter_changes implementation
106
 
        self._iter_changes = dirstate._process_entry
107
146
 
108
147
    @needs_tree_write_lock
109
148
    def _add(self, files, ids, kinds):
111
150
        state = self.current_dirstate()
112
151
        for f, file_id, kind in zip(files, ids, kinds):
113
152
            f = f.strip('/')
 
153
            assert '//' not in f
 
154
            assert '..' not in f
114
155
            if self.path2id(f):
115
156
                # special case tree root handling.
116
157
                if f == '' and self.path2id(f) == ROOT_ID:
137
178
    @needs_tree_write_lock
138
179
    def add_reference(self, sub_tree):
139
180
        # use standard implementation, which calls back to self._add
140
 
        #
 
181
        # 
141
182
        # So we don't store the reference_revision in the working dirstate,
142
 
        # it's just recorded at the moment of commit.
 
183
        # it's just recorded at the moment of commit. 
143
184
        self._add_reference(sub_tree)
144
185
 
145
186
    def break_lock(self):
184
225
            WorkingTree3._comparison_data(self, entry, path)
185
226
        # it looks like a plain directory, but it's really a reference -- see
186
227
        # also kind()
187
 
        if (self._repo_supports_tree_reference and kind == 'directory'
188
 
            and entry is not None and entry.kind == 'tree-reference'):
 
228
        if kind == 'directory' and self._directory_is_tree_reference(path):
189
229
            kind = 'tree-reference'
190
230
        return kind, executable, stat_value
191
231
 
217
257
            return self._dirstate
218
258
        local_path = self.bzrdir.get_workingtree_transport(None
219
259
            ).local_abspath('dirstate')
220
 
        self._dirstate = dirstate.DirState.on_file(local_path,
221
 
            self._sha1_provider())
 
260
        self._dirstate = dirstate.DirState.on_file(local_path)
222
261
        return self._dirstate
223
262
 
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
 
263
    def _directory_is_tree_reference(self, relpath):
 
264
        # as a special case, if a directory contains control files then 
 
265
        # it's a tree reference, except that the root of the tree is not
 
266
        return relpath and osutils.isdir(self.abspath(relpath) + "/.bzr")
 
267
        # TODO: We could ask all the control formats whether they
 
268
        # recognize this directory, but at the moment there's no cheap api
 
269
        # to do that.  Since we probably can only nest bzr checkouts and
 
270
        # they always use this name it's ok for now.  -- mbp 20060306
 
271
        #
 
272
        # FIXME: There is an unhandled case here of a subdirectory
 
273
        # containing .bzr but not a branch; that will probably blow up
 
274
        # when you try to commit it.  It might happen if there is a
 
275
        # checkout in a subdirectory.  This can be avoided by not adding
 
276
        # it.  mbp 20070306
235
277
 
236
278
    def filter_unversioned_files(self, paths):
237
279
        """Filter out paths that are versioned.
261
303
        self._inventory = None
262
304
        self._dirty = False
263
305
 
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
306
    def _generate_inventory(self):
272
307
        """Create and set self.inventory from the dirstate object.
273
 
 
 
308
        
274
309
        This is relatively expensive: we have to walk the entire dirstate.
275
310
        Ideally we would not, and can deprecate this function.
276
311
        """
280
315
        state._read_dirblocks_if_needed()
281
316
        root_key, current_entry = self._get_entry(path='')
282
317
        current_id = root_key[2]
283
 
        if not (current_entry[0][0] == 'd'): # directory
284
 
            raise AssertionError(current_entry)
 
318
        assert current_entry[0][0] == 'd' # directory
285
319
        inv = Inventory(root_id=current_id)
286
320
        # Turn some things into local variables
287
321
        minikind_to_kind = dirstate.DirState._minikind_to_kind
310
344
                inv_entry = factory[kind](file_id, name_unicode,
311
345
                                          parent_ie.file_id)
312
346
                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
347
                    # not strictly needed: working tree
317
 
                    #inv_entry.text_size = size
318
 
                    #inv_entry.text_sha1 = sha1
 
348
                    #entry.executable = executable
 
349
                    #entry.text_size = size
 
350
                    #entry.text_sha1 = sha1
 
351
                    pass
319
352
                elif kind == 'directory':
320
353
                    # add this entry to the parent map.
321
354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
322
355
                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)
 
356
                    inv_entry.reference_revision = link_or_sha1
 
357
                else:
 
358
                    assert 'unknown kind'
330
359
                # 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,))
 
360
                assert file_id not in inv_byid, ('file_id %s already in'
 
361
                    ' inventory as %s' % (file_id, inv_byid[file_id]))
 
362
                assert name_unicode not in parent_ie.children
337
363
                inv_byid[file_id] = inv_entry
338
364
                parent_ie.children[name_unicode] = inv_entry
339
365
        self._inventory = inv
344
370
        If either file_id or path is supplied, it is used as the key to lookup.
345
371
        If both are supplied, the fastest lookup is used, and an error is
346
372
        raised if they do not both point at the same row.
347
 
 
 
373
        
348
374
        :param file_id: An optional unicode file_id to be looked up.
349
375
        :param path: An optional unicode path to be looked up.
350
376
        :return: The dirstate row tuple for path/file_id, or (None, None)
359
385
    def get_file_sha1(self, file_id, path=None, stat_value=None):
360
386
        # check file id is valid unconditionally.
361
387
        entry = self._get_entry(file_id=file_id, path=path)
362
 
        if entry[0] is None:
363
 
            raise errors.NoSuchId(self, file_id)
 
388
        assert entry[0] is not None, 'what error should this raise'
 
389
        # TODO:
 
390
        # if row stat is valid, use cached sha1, else, get a new sha1.
364
391
        if path is None:
365
392
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
366
393
 
367
394
        file_abspath = self.abspath(path)
368
395
        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)
 
396
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
397
                                          stat_value=stat_value)
379
398
        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
 
399
            return link_or_sha1
390
400
        return None
391
401
 
392
402
    def _get_inventory(self):
393
403
        """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
404
        if self._inventory is not None:
398
405
            return self._inventory
399
406
        self._must_be_locked()
406
413
    @needs_read_lock
407
414
    def get_parent_ids(self):
408
415
        """See Tree.get_parent_ids.
409
 
 
 
416
        
410
417
        This implementation requests the ids list from the dirstate file.
411
418
        """
412
419
        return self.current_dirstate().get_parent_ids()
413
420
 
414
 
    def get_reference_revision(self, file_id, path=None):
 
421
    def get_reference_revision(self, entry, path=None):
415
422
        # referenced tree's revision is whatever's currently there
416
 
        return self.get_nested_tree(file_id, path).last_revision()
 
423
        return self.get_nested_tree(entry, path).last_revision()
417
424
 
418
 
    def get_nested_tree(self, file_id, path=None):
 
425
    def get_nested_tree(self, entry, path=None):
419
426
        if path is None:
420
 
            path = self.id2path(file_id)
421
 
        # else: check file_id is at path?
 
427
            path = self.id2path(entry.file_id)
422
428
        return WorkingTree.open(self.abspath(path))
423
429
 
424
430
    @needs_read_lock
428
434
 
429
435
    def has_id(self, file_id):
430
436
        state = self.current_dirstate()
 
437
        file_id = osutils.safe_file_id(file_id)
431
438
        row, parents = self._get_entry(file_id=file_id)
432
439
        if row is None:
433
440
            return False
434
441
        return osutils.lexists(pathjoin(
435
442
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
436
443
 
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
444
    @needs_read_lock
443
445
    def id2path(self, file_id):
444
 
        "Convert a file-id to a path."
 
446
        file_id = osutils.safe_file_id(file_id)
445
447
        state = self.current_dirstate()
446
448
        entry = self._get_entry(file_id=file_id)
447
449
        if entry == (None, None):
449
451
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
450
452
        return path_utf8.decode('utf8')
451
453
 
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
454
    @needs_read_lock
494
455
    def __iter__(self):
495
456
        """Iterate through file_ids for this tree.
507
468
                result.append(key[2])
508
469
        return iter(result)
509
470
 
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
 
 
 
471
    @needs_read_lock
536
472
    def kind(self, file_id):
537
473
        """Return the kind of a file.
538
474
 
539
475
        This is always the actual kind that's on disk, regardless of what it
540
476
        was added as.
541
 
 
542
 
        Note: The caller is expected to take a read-lock before calling this.
543
477
        """
544
478
        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):
 
479
        assert relpath != None, \
 
480
            "path for id {%s} is None!" % file_id
551
481
        abspath = self.abspath(relpath)
552
482
        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'
 
483
        if kind == 'directory' and self._directory_is_tree_reference(relpath):
 
484
            kind = 'tree-reference'
558
485
        return kind
559
486
 
560
487
    @needs_read_lock
564
491
        if parent_ids:
565
492
            return parent_ids[0]
566
493
        else:
567
 
            return _mod_revision.NULL_REVISION
 
494
            return None
568
495
 
569
496
    def lock_read(self):
570
497
        """See Branch.lock_read, and WorkingTree.unlock."""
575
502
                state = self.current_dirstate()
576
503
                if not state._lock_token:
577
504
                    state.lock_read()
578
 
                # set our support for tree references from the repository in
579
 
                # use.
580
 
                self._repo_supports_tree_reference = getattr(
581
 
                    self.branch.repository._format, "supports_tree_reference",
582
 
                    False)
583
505
            except:
584
506
                self._control_files.unlock()
585
507
                raise
595
517
                state = self.current_dirstate()
596
518
                if not state._lock_token:
597
519
                    state.lock_write()
598
 
                # set our support for tree references from the repository in
599
 
                # use.
600
 
                self._repo_supports_tree_reference = getattr(
601
 
                    self.branch.repository._format, "supports_tree_reference",
602
 
                    False)
603
520
            except:
604
521
                self._control_files.unlock()
605
522
                raise
623
540
        result = []
624
541
        if not from_paths:
625
542
            return result
 
543
 
626
544
        state = self.current_dirstate()
627
 
        if isinstance(from_paths, basestring):
628
 
            raise ValueError()
 
545
 
 
546
        assert not isinstance(from_paths, basestring)
629
547
        to_dir_utf8 = to_dir.encode('utf8')
630
548
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
631
549
        id_index = state._get_id_index()
653
571
        if self._inventory is not None:
654
572
            update_inventory = True
655
573
            inv = self.inventory
 
574
            to_dir_ie = inv[to_dir_id]
656
575
            to_dir_id = to_entry[0][2]
657
 
            to_dir_ie = inv[to_dir_id]
658
576
        else:
659
577
            update_inventory = False
660
578
 
683
601
            new_entry = to_block[1][added_entry_index]
684
602
            rollbacks.append(lambda:state._make_absent(new_entry))
685
603
 
 
604
        # create rename entries and tuples
686
605
        for from_rel in from_paths:
687
606
            # from_rel is 'pathinroot/foo/bar'
688
607
            from_rel_utf8 = from_rel.encode('utf8')
691
610
            from_entry = self._get_entry(path=from_rel)
692
611
            if from_entry == (None, None):
693
612
                raise errors.BzrMoveFailedError(from_rel,to_dir,
694
 
                    errors.NotVersionedError(path=from_rel))
 
613
                    errors.NotVersionedError(path=str(from_rel)))
695
614
 
696
615
            from_id = from_entry[0][2]
697
616
            to_rel = pathjoin(to_dir, from_tail)
724
643
                if from_missing: # implicitly just update our path mapping
725
644
                    move_file = False
726
645
                elif not after:
727
 
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
646
                    raise errors.RenameFailedFilesExist(from_rel, to_rel,
 
647
                        extra="(Use --after to update the Bazaar id)")
728
648
 
729
649
            rollbacks = []
730
650
            def rollback_rename():
731
651
                """A single rename has failed, roll it back."""
732
 
                # roll back everything, even if we encounter trouble doing one
733
 
                # of them.
734
 
                #
735
 
                # TODO: at least log the other exceptions rather than just
736
 
                # losing them mbp 20070307
737
652
                exc_info = None
738
653
                for rollback in reversed(rollbacks):
739
654
                    try:
740
655
                        rollback()
741
656
                    except Exception, e:
 
657
                        import pdb;pdb.set_trace()
742
658
                        exc_info = sys.exc_info()
743
659
                if exc_info:
744
660
                    raise exc_info[0], exc_info[1], exc_info[2]
785
701
 
786
702
                if minikind == 'd':
787
703
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
788
 
                        """Recursively update all entries in this dirblock."""
789
 
                        if from_dir == '':
790
 
                            raise AssertionError("renaming root not supported")
 
704
                        """all entries in this block need updating.
 
705
 
 
706
                        TODO: This is pretty ugly, and doesn't support
 
707
                        reverting, but it works.
 
708
                        """
 
709
                        assert from_dir != '', "renaming root not supported"
791
710
                        from_key = (from_dir, '')
792
711
                        from_block_idx, present = \
793
712
                            state._find_block_index_from_key(from_key)
803
722
                        to_block_index = state._ensure_block(
804
723
                            to_block_index, to_entry_index, to_dir_utf8)
805
724
                        to_block = state._dirblocks[to_block_index]
806
 
 
807
 
                        # Grab a copy since move_one may update the list.
808
 
                        for entry in from_block[1][:]:
809
 
                            if not (entry[0][0] == from_dir):
810
 
                                raise AssertionError()
 
725
                        for entry in from_block[1]:
 
726
                            assert entry[0][0] == from_dir
811
727
                            cur_details = entry[1][0]
812
728
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
813
729
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
814
730
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
815
731
                            minikind = cur_details[0]
816
 
                            if minikind in 'ar':
817
 
                                # Deleted children of a renamed directory
818
 
                                # Do not need to be updated.
819
 
                                # Children that have been renamed out of this
820
 
                                # directory should also not be updated
821
 
                                continue
822
732
                            move_one(entry, from_path_utf8=from_path_utf8,
823
733
                                     minikind=minikind,
824
734
                                     executable=cur_details[3],
827
737
                                     size=cur_details[2],
828
738
                                     to_block=to_block,
829
739
                                     to_key=to_key,
830
 
                                     to_path_utf8=to_path_utf8)
 
740
                                     to_path_utf8=to_rel_utf8)
831
741
                            if minikind == 'd':
832
742
                                # We need to move all the children of this
833
743
                                # entry
872
782
        for tree in trees:
873
783
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
874
784
                parents):
875
 
                return super(DirStateWorkingTree, self).paths2ids(paths,
876
 
                    trees, require_versioned)
 
785
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
877
786
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
878
787
        # -- make all paths utf8 --
879
788
        paths_utf8 = set()
939
848
            if not all_versioned:
940
849
                raise errors.PathsNotVersionedError(paths)
941
850
        # -- remove redundancy in supplied paths to prevent over-scanning --
942
 
        search_paths = osutils.minimum_path_selection(paths)
943
 
        # sketch:
 
851
        search_paths = set()
 
852
        for path in paths:
 
853
            other_paths = paths.difference(set([path]))
 
854
            if not osutils.is_inside_any(other_paths, path):
 
855
                # this is a top level path, we must check it.
 
856
                search_paths.add(path)
 
857
        # sketch: 
944
858
        # for all search_indexs in each path at or under each element of
945
859
        # search_paths, if the detail is relocated: add the id, and add the
946
860
        # relocated path as one to search if its not searched already. If the
1002
916
 
1003
917
    def read_working_inventory(self):
1004
918
        """Read the working inventory.
1005
 
 
 
919
        
1006
920
        This is a meaningless operation for dirstate, but we obey it anyhow.
1007
921
        """
1008
922
        return self.inventory
1013
927
 
1014
928
        WorkingTree4 supplies revision_trees for any basis tree.
1015
929
        """
 
930
        revision_id = osutils.safe_revision_id(revision_id)
1016
931
        dirstate = self.current_dirstate()
1017
932
        parent_ids = dirstate.get_parent_ids()
1018
933
        if revision_id not in parent_ids:
1025
940
    @needs_tree_write_lock
1026
941
    def set_last_revision(self, new_revision):
1027
942
        """Change the last revision in the working tree."""
 
943
        new_revision = osutils.safe_revision_id(new_revision)
1028
944
        parents = self.get_parent_ids()
1029
 
        if new_revision in (_mod_revision.NULL_REVISION, None):
1030
 
            if len(parents) >= 2:
1031
 
                raise AssertionError(
1032
 
                    "setting the last parent to none with a pending merge is "
1033
 
                    "unsupported.")
 
945
        if new_revision in (NULL_REVISION, None):
 
946
            assert len(parents) < 2, (
 
947
                "setting the last parent to none with a pending merge is "
 
948
                "unsupported.")
1034
949
            self.set_parent_ids([])
1035
950
        else:
1036
951
            self.set_parent_ids([new_revision] + parents[1:],
1039
954
    @needs_tree_write_lock
1040
955
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
1041
956
        """Set the parent ids to revision_ids.
1042
 
 
 
957
        
1043
958
        See also set_parent_trees. This api will try to retrieve the tree data
1044
959
        for each element of revision_ids from the trees repository. If you have
1045
960
        tree data already available, it is more efficient to use
1049
964
        :param revision_ids: The revision_ids to set as the parent ids of this
1050
965
            working tree. Any of these may be ghosts.
1051
966
        """
 
967
        revision_ids = [osutils.safe_revision_id(r) for r in revision_ids]
1052
968
        trees = []
1053
969
        for revision_id in revision_ids:
1054
970
            try:
1060
976
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
1061
977
                revtree = None
1062
978
            trees.append((revision_id, revtree))
 
979
        self.current_dirstate()._validate()
1063
980
        self.set_parent_trees(trees,
1064
981
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
982
        self.current_dirstate()._validate()
1065
983
 
1066
984
    @needs_tree_write_lock
1067
985
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
1072
990
            parent tree - i.e. a ghost.
1073
991
        """
1074
992
        dirstate = self.current_dirstate()
 
993
        dirstate._validate()
1075
994
        if len(parents_list) > 0:
1076
995
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
1077
996
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
1078
997
        real_trees = []
1079
998
        ghosts = []
1080
 
 
1081
 
        parent_ids = [rev_id for rev_id, tree in parents_list]
1082
 
        graph = self.branch.repository.get_graph()
1083
 
        heads = graph.heads(parent_ids)
1084
 
        accepted_revisions = set()
1085
 
 
1086
999
        # convert absent trees to the null tree, which we convert back to
1087
1000
        # missing on access.
1088
1001
        for rev_id, tree in parents_list:
1089
 
            if len(accepted_revisions) > 0:
1090
 
                # we always accept the first tree
1091
 
                if rev_id in accepted_revisions or rev_id not in heads:
1092
 
                    # We have already included either this tree, or its
1093
 
                    # descendent, so we skip it.
1094
 
                    continue
1095
 
            _mod_revision.check_not_reserved_id(rev_id)
 
1002
            rev_id = osutils.safe_revision_id(rev_id)
1096
1003
            if tree is not None:
1097
1004
                real_trees.append((rev_id, tree))
1098
1005
            else:
1099
1006
                real_trees.append((rev_id,
1100
 
                    self.branch.repository.revision_tree(
1101
 
                        _mod_revision.NULL_REVISION)))
 
1007
                    self.branch.repository.revision_tree(None)))
1102
1008
                ghosts.append(rev_id)
1103
 
            accepted_revisions.add(rev_id)
 
1009
        dirstate._validate()
1104
1010
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1011
        dirstate._validate()
1105
1012
        self._make_dirty(reset_inventory=False)
 
1013
        dirstate._validate()
1106
1014
 
1107
1015
    def _set_root_id(self, file_id):
1108
1016
        """See WorkingTree.set_root_id."""
1111
1019
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
1112
1020
            self._make_dirty(reset_inventory=True)
1113
1021
 
1114
 
    def _sha_from_stat(self, path, stat_result):
1115
 
        """Get a sha digest from the tree's stat cache.
1116
 
 
1117
 
        The default implementation assumes no stat cache is present.
1118
 
 
1119
 
        :param path: The path.
1120
 
        :param stat_result: The stat result being looked up.
1121
 
        """
1122
 
        return self.current_dirstate().sha1_from_stat(path, stat_result)
1123
 
 
1124
 
    @needs_read_lock
1125
 
    def supports_tree_reference(self):
1126
 
        return self._repo_supports_tree_reference
1127
 
 
1128
1022
    def unlock(self):
1129
1023
        """Unlock in format 4 trees needs to write the entire dirstate."""
1130
 
        # do non-implementation specific cleanup
1131
 
        self._cleanup()
1132
 
 
1133
1024
        if self._control_files._lock_count == 1:
1134
1025
            # eventually we should do signature checking during read locks for
1135
1026
            # dirstate updates.
1166
1057
            return
1167
1058
        state = self.current_dirstate()
1168
1059
        state._read_dirblocks_if_needed()
1169
 
        ids_to_unversion = set(file_ids)
 
1060
        ids_to_unversion = set()
 
1061
        for file_id in file_ids:
 
1062
            ids_to_unversion.add(osutils.safe_file_id(file_id))
1170
1063
        paths_to_unversion = set()
1171
1064
        # sketch:
1172
1065
        # check if the root is to be unversioned, if so, assert for now.
1199
1092
                # just forget the whole block.
1200
1093
                entry_index = 0
1201
1094
                while entry_index < len(block[1]):
1202
 
                    entry = block[1][entry_index]
1203
 
                    if entry[1][0][0] in 'ar':
1204
 
                        # don't remove absent or renamed entries
 
1095
                    # Mark this file id as having been removed
 
1096
                    ids_to_unversion.discard(block[1][entry_index][0][2])
 
1097
                    if not state._make_absent(block[1][entry_index]):
1205
1098
                        entry_index += 1
1206
 
                    else:
1207
 
                        # Mark this file id as having been removed
1208
 
                        ids_to_unversion.discard(entry[0][2])
1209
 
                        if not state._make_absent(entry):
1210
 
                            # The block has not shrunk.
1211
 
                            entry_index += 1
1212
1099
                # go to the next block. (At the moment we dont delete empty
1213
1100
                # dirblocks)
1214
1101
                block_index += 1
1238
1125
                self._inventory.remove_recursive_id(file_id)
1239
1126
 
1240
1127
    @needs_tree_write_lock
1241
 
    def rename_one(self, from_rel, to_rel, after=False):
1242
 
        """See WorkingTree.rename_one"""
1243
 
        self.flush()
1244
 
        WorkingTree.rename_one(self, from_rel, to_rel, after)
1245
 
 
1246
 
    @needs_tree_write_lock
1247
 
    def apply_inventory_delta(self, changes):
1248
 
        """See MutableTree.apply_inventory_delta"""
1249
 
        state = self.current_dirstate()
1250
 
        state.update_by_delta(changes)
1251
 
        self._make_dirty(reset_inventory=True)
1252
 
 
1253
 
    def update_basis_by_delta(self, new_revid, delta):
1254
 
        """See MutableTree.update_basis_by_delta."""
1255
 
        if self.last_revision() == new_revid:
1256
 
            raise AssertionError()
1257
 
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
1258
 
 
1259
 
    @needs_read_lock
1260
 
    def _validate(self):
1261
 
        self._dirstate._validate()
1262
 
 
1263
 
    @needs_tree_write_lock
1264
1128
    def _write_inventory(self, inv):
1265
1129
        """Write inventory as the current inventory."""
1266
 
        if self._dirty:
1267
 
            raise AssertionError("attempting to write an inventory when the "
1268
 
                "dirstate is dirty will lose pending changes")
1269
 
        had_inventory = self._inventory is not None
1270
 
        # Setting self._inventory = None forces the dirstate to regenerate the
1271
 
        # working inventory. We do this because self.inventory may be inv, or
1272
 
        # may have been modified, and either case would prevent a clean delta
1273
 
        # being created.
1274
 
        self._inventory = None
1275
 
        # generate a delta,
1276
 
        delta = inv._make_delta(self.inventory)
1277
 
        # and apply it.
1278
 
        self.apply_inventory_delta(delta)
1279
 
        if had_inventory:
 
1130
        assert not self._dirty, "attempting to write an inventory when the dirstate is dirty will cause data loss"
 
1131
        self.current_dirstate().set_state_from_inventory(inv)
 
1132
        self._make_dirty(reset_inventory=False)
 
1133
        if self._inventory is not None:
1280
1134
            self._inventory = inv
1281
1135
        self.flush()
1282
1136
 
1283
1137
 
1284
 
class ContentFilterAwareSHA1Provider(dirstate.SHA1Provider):
1285
 
 
1286
 
    def __init__(self, tree):
1287
 
        self.tree = tree
1288
 
 
1289
 
    def sha1(self, abspath):
1290
 
        """See dirstate.SHA1Provider.sha1()."""
1291
 
        filters = self.tree._content_filter_stack(
1292
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1293
 
        return internal_size_sha_file_byname(abspath, filters)[1]
1294
 
 
1295
 
    def stat_and_sha1(self, abspath):
1296
 
        """See dirstate.SHA1Provider.stat_and_sha1()."""
1297
 
        filters = self.tree._content_filter_stack(
1298
 
            self.tree.relpath(osutils.safe_unicode(abspath)))
1299
 
        file_obj = file(abspath, 'rb', 65000)
1300
 
        try:
1301
 
            statvalue = os.fstat(file_obj.fileno())
1302
 
            if filters:
1303
 
                file_obj = filtered_input_file(file_obj, filters)
1304
 
            sha1 = osutils.size_sha_file(file_obj)[1]
1305
 
        finally:
1306
 
            file_obj.close()
1307
 
        return statvalue, sha1
1308
 
 
1309
 
 
1310
 
class ContentFilteringDirStateWorkingTree(DirStateWorkingTree):
1311
 
    """Dirstate working tree that supports content filtering.
1312
 
 
1313
 
    The dirstate holds the hash and size of the canonical form of the file, 
1314
 
    and most methods must return that.
1315
 
    """
1316
 
 
1317
 
    def _file_content_summary(self, path, stat_result):
1318
 
        # This is to support the somewhat obsolete path_content_summary method
1319
 
        # with content filtering: see
1320
 
        # <https://bugs.edge.launchpad.net/bzr/+bug/415508>.
1321
 
        #
1322
 
        # If the dirstate cache is up to date and knows the hash and size,
1323
 
        # return that.
1324
 
        # Otherwise if there are no content filters, return the on-disk size
1325
 
        # and leave the hash blank.
1326
 
        # Otherwise, read and filter the on-disk file and use its size and
1327
 
        # hash.
1328
 
        #
1329
 
        # The dirstate doesn't store the size of the canonical form so we
1330
 
        # can't trust it for content-filtered trees.  We just return None.
1331
 
        dirstate_sha1 = self._dirstate.sha1_from_stat(path, stat_result)
1332
 
        executable = self._is_executable_from_path_and_stat(path, stat_result)
1333
 
        return ('file', None, executable, dirstate_sha1)
1334
 
 
1335
 
 
1336
 
class WorkingTree4(DirStateWorkingTree):
1337
 
    """This is the Format 4 working tree.
1338
 
 
1339
 
    This differs from WorkingTree3 by:
1340
 
     - Having a consolidated internal dirstate, stored in a
1341
 
       randomly-accessible sorted file on disk.
1342
 
     - Not having a regular inventory attribute.  One can be synthesized
1343
 
       on demand but this is expensive and should be avoided.
1344
 
 
1345
 
    This is new in bzr 0.15.
1346
 
    """
1347
 
 
1348
 
 
1349
 
class WorkingTree5(ContentFilteringDirStateWorkingTree):
1350
 
    """This is the Format 5 working tree.
1351
 
 
1352
 
    This differs from WorkingTree4 by:
1353
 
     - Supporting content filtering.
1354
 
 
1355
 
    This is new in bzr 1.11.
1356
 
    """
1357
 
 
1358
 
 
1359
 
class WorkingTree6(ContentFilteringDirStateWorkingTree):
1360
 
    """This is the Format 6 working tree.
1361
 
 
1362
 
    This differs from WorkingTree5 by:
1363
 
     - Supporting a current view that may mask the set of files in a tree
1364
 
       impacted by most user operations.
1365
 
 
1366
 
    This is new in bzr 1.14.
1367
 
    """
1368
 
 
1369
 
    def _make_views(self):
1370
 
        return views.PathBasedViews(self)
1371
 
 
1372
 
 
1373
 
class DirStateWorkingTreeFormat(WorkingTreeFormat3):
1374
 
 
1375
 
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
1376
 
                   accelerator_tree=None, hardlink=False):
 
1138
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1139
    """The first consolidated dirstate working tree format.
 
1140
 
 
1141
    This format:
 
1142
        - exists within a metadir controlling .bzr
 
1143
        - includes an explicit version marker for the workingtree control
 
1144
          files, separate from the BzrDir format
 
1145
        - modifies the hash cache format
 
1146
        - is new in bzr TODO FIXME SETBEFOREMERGE
 
1147
        - uses a LockDir to guard access to it.
 
1148
    """
 
1149
 
 
1150
    supports_tree_reference = True
 
1151
 
 
1152
    def get_format_string(self):
 
1153
        """See WorkingTreeFormat.get_format_string()."""
 
1154
        return "Bazaar Working Tree format 4\n"
 
1155
 
 
1156
    def get_format_description(self):
 
1157
        """See WorkingTreeFormat.get_format_description()."""
 
1158
        return "Working tree format 4"
 
1159
 
 
1160
    def initialize(self, a_bzrdir, revision_id=None):
1377
1161
        """See WorkingTreeFormat.initialize().
1378
1162
 
1379
1163
        :param revision_id: allows creating a working tree at a different
1380
1164
        revision than the branch is at.
1381
 
        :param accelerator_tree: A tree which can be used for retrieving file
1382
 
            contents more quickly than the revision tree, i.e. a workingtree.
1383
 
            The revision tree will be used for cases where accelerator_tree's
1384
 
            content is different.
1385
 
        :param hardlink: If true, hard-link files from accelerator_tree,
1386
 
            where possible.
1387
1165
 
1388
 
        These trees get an initial random root id, if their repository supports
1389
 
        rich root data, TREE_ROOT otherwise.
 
1166
        These trees get an initial random root id.
1390
1167
        """
 
1168
        revision_id = osutils.safe_revision_id(revision_id)
1391
1169
        if not isinstance(a_bzrdir.transport, LocalTransport):
1392
1170
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
1393
1171
        transport = a_bzrdir.get_workingtree_transport(self)
1394
1172
        control_files = self._open_control_files(a_bzrdir)
1395
1173
        control_files.create_lock()
1396
1174
        control_files.lock_write()
1397
 
        transport.put_bytes('format', self.get_format_string(),
1398
 
            mode=a_bzrdir._get_file_mode())
1399
 
        if from_branch is not None:
1400
 
            branch = from_branch
1401
 
        else:
1402
 
            branch = a_bzrdir.open_branch()
 
1175
        control_files.put_utf8('format', self.get_format_string())
 
1176
        branch = a_bzrdir.open_branch()
1403
1177
        if revision_id is None:
1404
1178
            revision_id = branch.last_revision()
1405
1179
        local_path = transport.local_abspath('dirstate')
1406
1180
        # write out new dirstate (must exist when we create the tree)
1407
1181
        state = dirstate.DirState.initialize(local_path)
1408
1182
        state.unlock()
1409
 
        del state
1410
 
        wt = self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1183
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1411
1184
                         branch,
1412
1185
                         _format=self,
1413
1186
                         _bzrdir=a_bzrdir,
1414
1187
                         _control_files=control_files)
1415
1188
        wt._new_tree()
1416
1189
        wt.lock_tree_write()
 
1190
        state._validate()
1417
1191
        try:
1418
 
            self._init_custom_control_files(wt)
1419
 
            if revision_id in (None, _mod_revision.NULL_REVISION):
1420
 
                if branch.repository.supports_rich_root():
1421
 
                    wt._set_root_id(generate_ids.gen_root_id())
1422
 
                else:
1423
 
                    wt._set_root_id(ROOT_ID)
 
1192
            if revision_id in (None, NULL_REVISION):
 
1193
                wt._set_root_id(generate_ids.gen_root_id())
1424
1194
                wt.flush()
1425
 
            basis = None
1426
 
            # frequently, we will get here due to branching.  The accelerator
1427
 
            # tree will be the tree from the branch, so the desired basis
1428
 
            # tree will often be a parent of the accelerator tree.
1429
 
            if accelerator_tree is not None:
1430
 
                try:
1431
 
                    basis = accelerator_tree.revision_tree(revision_id)
1432
 
                except errors.NoSuchRevision:
1433
 
                    pass
1434
 
            if basis is None:
1435
 
                basis = branch.repository.revision_tree(revision_id)
1436
 
            if revision_id == _mod_revision.NULL_REVISION:
1437
 
                parents_list = []
1438
 
            else:
1439
 
                parents_list = [(revision_id, basis)]
 
1195
                wt.current_dirstate()._validate()
 
1196
            wt.set_last_revision(revision_id)
 
1197
            wt.flush()
 
1198
            basis = wt.basis_tree()
1440
1199
            basis.lock_read()
1441
 
            try:
1442
 
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1200
            # if the basis has a root id we have to use that; otherwise we use
 
1201
            # a new random one
 
1202
            basis_root_id = basis.get_root_id()
 
1203
            if basis_root_id is not None:
 
1204
                wt._set_root_id(basis_root_id)
1443
1205
                wt.flush()
1444
 
                # if the basis has a root id we have to use that; otherwise we
1445
 
                # use a new random one
1446
 
                basis_root_id = basis.get_root_id()
1447
 
                if basis_root_id is not None:
1448
 
                    wt._set_root_id(basis_root_id)
1449
 
                    wt.flush()
1450
 
                if wt.supports_content_filtering():
1451
 
                    # The original tree may not have the same content filters
1452
 
                    # applied so we can't safely build the inventory delta from
1453
 
                    # the source tree.
1454
 
                    delta_from_tree = False
1455
 
                else:
1456
 
                    delta_from_tree = True
1457
 
                # delta_from_tree is safe even for DirStateRevisionTrees,
1458
 
                # because wt4.apply_inventory_delta does not mutate the input
1459
 
                # inventory entries.
1460
 
                transform.build_tree(basis, wt, accelerator_tree,
1461
 
                                     hardlink=hardlink,
1462
 
                                     delta_from_tree=delta_from_tree)
1463
 
            finally:
1464
 
                basis.unlock()
 
1206
            transform.build_tree(basis, wt)
 
1207
            basis.unlock()
1465
1208
        finally:
1466
1209
            control_files.unlock()
1467
1210
            wt.unlock()
1468
1211
        return wt
1469
1212
 
1470
 
    def _init_custom_control_files(self, wt):
1471
 
        """Subclasses with custom control files should override this method.
1472
 
 
1473
 
        The working tree and control files are locked for writing when this
1474
 
        method is called.
1475
 
 
1476
 
        :param wt: the WorkingTree object
1477
 
        """
1478
 
 
1479
1213
    def _open(self, a_bzrdir, control_files):
1480
1214
        """Open the tree itself.
1481
1215
 
1482
1216
        :param a_bzrdir: the dir for the tree.
1483
1217
        :param control_files: the control files for the tree.
1484
1218
        """
1485
 
        return self._tree_class(a_bzrdir.root_transport.local_abspath('.'),
 
1219
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
1486
1220
                           branch=a_bzrdir.open_branch(),
1487
1221
                           _format=self,
1488
1222
                           _bzrdir=a_bzrdir,
1489
1223
                           _control_files=control_files)
1490
1224
 
1491
1225
    def __get_matchingbzrdir(self):
1492
 
        return self._get_matchingbzrdir()
1493
 
 
1494
 
    def _get_matchingbzrdir(self):
1495
 
        """Overrideable method to get a bzrdir for testing."""
1496
1226
        # please test against something that will let us do tree references
1497
1227
        return bzrdir.format_registry.make_bzrdir(
1498
1228
            'dirstate-with-subtree')
1500
1230
    _matchingbzrdir = property(__get_matchingbzrdir)
1501
1231
 
1502
1232
 
1503
 
class WorkingTreeFormat4(DirStateWorkingTreeFormat):
1504
 
    """The first consolidated dirstate working tree format.
1505
 
 
1506
 
    This format:
1507
 
        - exists within a metadir controlling .bzr
1508
 
        - includes an explicit version marker for the workingtree control
1509
 
          files, separate from the BzrDir format
1510
 
        - modifies the hash cache format
1511
 
        - is new in bzr 0.15
1512
 
        - uses a LockDir to guard access to it.
1513
 
    """
1514
 
 
1515
 
    upgrade_recommended = False
1516
 
 
1517
 
    _tree_class = WorkingTree4
1518
 
 
1519
 
    def get_format_string(self):
1520
 
        """See WorkingTreeFormat.get_format_string()."""
1521
 
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
1522
 
 
1523
 
    def get_format_description(self):
1524
 
        """See WorkingTreeFormat.get_format_description()."""
1525
 
        return "Working tree format 4"
1526
 
 
1527
 
 
1528
 
class WorkingTreeFormat5(DirStateWorkingTreeFormat):
1529
 
    """WorkingTree format supporting content filtering.
1530
 
    """
1531
 
 
1532
 
    upgrade_recommended = False
1533
 
 
1534
 
    _tree_class = WorkingTree5
1535
 
 
1536
 
    def get_format_string(self):
1537
 
        """See WorkingTreeFormat.get_format_string()."""
1538
 
        return "Bazaar Working Tree Format 5 (bzr 1.11)\n"
1539
 
 
1540
 
    def get_format_description(self):
1541
 
        """See WorkingTreeFormat.get_format_description()."""
1542
 
        return "Working tree format 5"
1543
 
 
1544
 
    def supports_content_filtering(self):
1545
 
        return True
1546
 
 
1547
 
 
1548
 
class WorkingTreeFormat6(DirStateWorkingTreeFormat):
1549
 
    """WorkingTree format supporting views.
1550
 
    """
1551
 
 
1552
 
    upgrade_recommended = False
1553
 
 
1554
 
    _tree_class = WorkingTree6
1555
 
 
1556
 
    def get_format_string(self):
1557
 
        """See WorkingTreeFormat.get_format_string()."""
1558
 
        return "Bazaar Working Tree Format 6 (bzr 1.14)\n"
1559
 
 
1560
 
    def get_format_description(self):
1561
 
        """See WorkingTreeFormat.get_format_description()."""
1562
 
        return "Working tree format 6"
1563
 
 
1564
 
    def _init_custom_control_files(self, wt):
1565
 
        """Subclasses with custom control files should override this method."""
1566
 
        wt._transport.put_bytes('views', '', mode=wt.bzrdir._get_file_mode())
1567
 
 
1568
 
    def supports_content_filtering(self):
1569
 
        return True
1570
 
 
1571
 
    def supports_views(self):
1572
 
        return True
1573
 
 
1574
 
 
1575
1233
class DirStateRevisionTree(Tree):
1576
 
    """A revision tree pulling the inventory from a dirstate.
1577
 
    
1578
 
    Note that this is one of the historical (ie revision) trees cached in the
1579
 
    dirstate for easy access, not the workingtree.
1580
 
    """
 
1234
    """A revision tree pulling the inventory from a dirstate."""
1581
1235
 
1582
1236
    def __init__(self, dirstate, revision_id, repository):
1583
1237
        self._dirstate = dirstate
1584
 
        self._revision_id = revision_id
 
1238
        self._revision_id = osutils.safe_revision_id(revision_id)
1585
1239
        self._repository = repository
1586
1240
        self._inventory = None
1587
1241
        self._locked = 0
1588
1242
        self._dirstate_locked = False
1589
 
        self._repo_supports_tree_reference = getattr(
1590
 
            repository._format, "supports_tree_reference",
1591
 
            False)
1592
1243
 
1593
1244
    def __repr__(self):
1594
1245
        return "<%s of %s in %s>" % \
1595
1246
            (self.__class__.__name__, self._revision_id, self._dirstate)
1596
1247
 
1597
 
    def annotate_iter(self, file_id,
1598
 
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1248
    def annotate_iter(self, file_id):
1599
1249
        """See Tree.annotate_iter"""
1600
 
        text_key = (file_id, self.inventory[file_id].revision)
1601
 
        annotations = self._repository.texts.annotate(text_key)
1602
 
        return [(key[-1], line) for (key, line) in annotations]
 
1250
        w = self._repository.weave_store.get_weave(file_id,
 
1251
                           self._repository.get_transaction())
 
1252
        return w.annotate_iter(self.inventory[file_id].revision)
1603
1253
 
1604
 
    def _get_ancestors(self, default_revision):
1605
 
        return set(self._repository.get_ancestry(self._revision_id,
1606
 
                                                 topo_sorted=False))
1607
1254
    def _comparison_data(self, entry, path):
1608
1255
        """See Tree._comparison_data."""
1609
1256
        if entry is None:
1626
1273
    def get_root_id(self):
1627
1274
        return self.path2id('')
1628
1275
 
1629
 
    def id2path(self, file_id):
1630
 
        "Convert a file-id to a path."
1631
 
        entry = self._get_entry(file_id=file_id)
1632
 
        if entry == (None, None):
1633
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1634
 
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
1635
 
        return path_utf8.decode('utf8')
1636
 
 
1637
 
    def iter_references(self):
1638
 
        if not self._repo_supports_tree_reference:
1639
 
            # When the repo doesn't support references, we will have nothing to
1640
 
            # return
1641
 
            return iter([])
1642
 
        # Otherwise, fall back to the default implementation
1643
 
        return super(DirStateRevisionTree, self).iter_references()
1644
 
 
1645
1276
    def _get_parent_index(self):
1646
1277
        """Return the index in the dirstate referenced by this tree."""
1647
1278
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
1652
1283
        If either file_id or path is supplied, it is used as the key to lookup.
1653
1284
        If both are supplied, the fastest lookup is used, and an error is
1654
1285
        raised if they do not both point at the same row.
1655
 
 
 
1286
        
1656
1287
        :param file_id: An optional unicode file_id to be looked up.
1657
1288
        :param path: An optional unicode path to be looked up.
1658
1289
        :return: The dirstate row tuple for path/file_id, or (None, None)
1659
1290
        """
1660
1291
        if file_id is None and path is None:
1661
1292
            raise errors.BzrError('must supply file_id or path')
 
1293
        file_id = osutils.safe_file_id(file_id)
1662
1294
        if path is not None:
1663
1295
            path = path.encode('utf8')
1664
1296
        parent_index = self._get_parent_index()
1672
1304
 
1673
1305
        This is relatively expensive: we have to walk the entire dirstate.
1674
1306
        """
1675
 
        if not self._locked:
1676
 
            raise AssertionError(
1677
 
                'cannot generate inventory of an unlocked '
1678
 
                'dirstate revision tree')
 
1307
        assert self._locked, 'cannot generate inventory of an unlocked '\
 
1308
            'dirstate revision tree'
1679
1309
        # separate call for profiling - makes it clear where the costs are.
1680
1310
        self._dirstate._read_dirblocks_if_needed()
1681
 
        if self._revision_id not in self._dirstate.get_parent_ids():
1682
 
            raise AssertionError(
1683
 
                'parent %s has disappeared from %s' % (
1684
 
                self._revision_id, self._dirstate.get_parent_ids()))
 
1311
        assert self._revision_id in self._dirstate.get_parent_ids(), \
 
1312
            'parent %s has disappeared from %s' % (
 
1313
            self._revision_id, self._dirstate.get_parent_ids())
1685
1314
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
1686
1315
        # This is identical now to the WorkingTree _generate_inventory except
1687
1316
        # for the tree index use.
1688
1317
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
1689
1318
        current_id = root_key[2]
1690
 
        if current_entry[parent_index][0] != 'd':
1691
 
            raise AssertionError()
 
1319
        assert current_entry[parent_index][0] == 'd'
1692
1320
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
1693
1321
        inv.root.revision = current_entry[parent_index][4]
1694
1322
        # Turn some things into local variables
1726
1354
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
1727
1355
                elif kind == 'symlink':
1728
1356
                    inv_entry.executable = False
1729
 
                    inv_entry.text_size = None
 
1357
                    inv_entry.text_size = size
1730
1358
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
1731
1359
                elif kind == 'tree-reference':
1732
 
                    inv_entry.reference_revision = fingerprint or None
 
1360
                    inv_entry.reference_revision = fingerprint
1733
1361
                else:
1734
1362
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
1735
1363
                            % entry)
1736
1364
                # These checks cost us around 40ms on a 55k entry tree
1737
 
                if file_id in inv_byid:
1738
 
                    raise AssertionError('file_id %s already in'
1739
 
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
1740
 
                if name_unicode in parent_ie.children:
1741
 
                    raise AssertionError('name %r already in parent'
1742
 
                        % (name_unicode,))
 
1365
                assert file_id not in inv_byid
 
1366
                assert name_unicode not in parent_ie.children
1743
1367
                inv_byid[file_id] = inv_entry
1744
1368
                parent_ie.children[name_unicode] = inv_entry
1745
1369
        self._inventory = inv
1755
1379
            return None
1756
1380
        parent_index = self._get_parent_index()
1757
1381
        last_changed_revision = entry[1][parent_index][4]
1758
 
        try:
1759
 
            rev = self._repository.get_revision(last_changed_revision)
1760
 
        except errors.NoSuchRevision:
1761
 
            raise errors.FileTimestampUnavailable(self.id2path(file_id))
1762
 
        return rev.timestamp
 
1382
        return self._repository.get_revision(last_changed_revision).timestamp
1763
1383
 
1764
1384
    def get_file_sha1(self, file_id, path=None, stat_value=None):
1765
1385
        entry = self._get_entry(file_id=file_id, path=path)
1769
1389
            return parent_details[1]
1770
1390
        return None
1771
1391
 
1772
 
    def get_file(self, file_id, path=None):
 
1392
    def get_file(self, file_id):
1773
1393
        return StringIO(self.get_file_text(file_id))
1774
1394
 
 
1395
    def get_file_lines(self, file_id):
 
1396
        ie = self.inventory[file_id]
 
1397
        return self._repository.weave_store.get_weave(file_id,
 
1398
                self._repository.get_transaction()).get_lines(ie.revision)
 
1399
 
1775
1400
    def get_file_size(self, file_id):
1776
 
        """See Tree.get_file_size"""
1777
1401
        return self.inventory[file_id].text_size
1778
1402
 
1779
 
    def get_file_text(self, file_id, path=None):
1780
 
        _, content = list(self.iter_files_bytes([(file_id, None)]))[0]
1781
 
        return ''.join(content)
1782
 
 
1783
 
    def get_reference_revision(self, file_id, path=None):
1784
 
        return self.inventory[file_id].reference_revision
1785
 
 
1786
 
    def iter_files_bytes(self, desired_files):
1787
 
        """See Tree.iter_files_bytes.
1788
 
 
1789
 
        This version is implemented on top of Repository.iter_files_bytes"""
1790
 
        parent_index = self._get_parent_index()
1791
 
        repo_desired_files = []
1792
 
        for file_id, identifier in desired_files:
1793
 
            entry = self._get_entry(file_id)
1794
 
            if entry == (None, None):
1795
 
                raise errors.NoSuchId(self, file_id)
1796
 
            repo_desired_files.append((file_id, entry[1][parent_index][4],
1797
 
                                       identifier))
1798
 
        return self._repository.iter_files_bytes(repo_desired_files)
 
1403
    def get_file_text(self, file_id):
 
1404
        return ''.join(self.get_file_lines(file_id))
1799
1405
 
1800
1406
    def get_symlink_target(self, file_id):
1801
1407
        entry = self._get_entry(file_id=file_id)
1803
1409
        if entry[1][parent_index][0] != 'l':
1804
1410
            return None
1805
1411
        else:
1806
 
            target = entry[1][parent_index][1]
1807
 
            target = target.decode('utf8')
1808
 
            return target
 
1412
            # At present, none of the tree implementations supports non-ascii
 
1413
            # symlink targets. So we will just assume that the dirstate path is
 
1414
            # correct.
 
1415
            return entry[1][parent_index][1]
1809
1416
 
1810
1417
    def get_revision_id(self):
1811
1418
        """Return the revision id for this tree."""
1829
1436
        return bool(self.path2id(filename))
1830
1437
 
1831
1438
    def kind(self, file_id):
1832
 
        entry = self._get_entry(file_id=file_id)[1]
1833
 
        if entry is None:
1834
 
            raise errors.NoSuchId(tree=self, file_id=file_id)
1835
 
        parent_index = self._get_parent_index()
1836
 
        return dirstate.DirState._minikind_to_kind[entry[parent_index][0]]
1837
 
 
1838
 
    def stored_kind(self, file_id):
1839
 
        """See Tree.stored_kind"""
1840
 
        return self.kind(file_id)
1841
 
 
1842
 
    def path_content_summary(self, path):
1843
 
        """See Tree.path_content_summary."""
1844
 
        id = self.inventory.path2id(path)
1845
 
        if id is None:
1846
 
            return ('missing', None, None, None)
1847
 
        entry = self._inventory[id]
1848
 
        kind = entry.kind
1849
 
        if kind == 'file':
1850
 
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
1851
 
        elif kind == 'symlink':
1852
 
            return (kind, None, None, entry.symlink_target)
1853
 
        else:
1854
 
            return (kind, None, None, None)
 
1439
        return self.inventory[file_id].kind
1855
1440
 
1856
1441
    def is_executable(self, file_id, path=None):
1857
1442
        ie = self.inventory[file_id]
1859
1444
            return None
1860
1445
        return ie.executable
1861
1446
 
1862
 
    def list_files(self, include_root=False, from_dir=None, recursive=True):
 
1447
    def list_files(self, include_root=False):
1863
1448
        # We use a standard implementation, because DirStateRevisionTree is
1864
1449
        # dealing with one of the parents of the current state
1865
1450
        inv = self._get_inventory()
1866
 
        if from_dir is None:
1867
 
            from_dir_id = None
1868
 
        else:
1869
 
            from_dir_id = inv.path2id(from_dir)
1870
 
            if from_dir_id is None:
1871
 
                # Directory not versioned
1872
 
                return
1873
 
        entries = inv.iter_entries(from_dir=from_dir_id, recursive=recursive)
1874
 
        if inv.root is not None and not include_root and from_dir is None:
 
1451
        entries = inv.iter_entries()
 
1452
        if self.inventory.root is not None and not include_root:
1875
1453
            entries.next()
1876
1454
        for path, entry in entries:
1877
1455
            yield path, 'V', entry.kind, entry.file_id, entry
1910
1488
                self._dirstate_locked = False
1911
1489
            self._repository.unlock()
1912
1490
 
1913
 
    @needs_read_lock
1914
 
    def supports_tree_reference(self):
1915
 
        return self._repo_supports_tree_reference
1916
 
 
1917
1491
    def walkdirs(self, prefix=""):
1918
 
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
1919
 
        # implementation based on an inventory.
1920
 
        # This should be cleaned up to use the much faster Dirstate code
1921
 
        # So for now, we just build up the parent inventory, and extract
1922
 
        # it the same way RevisionTree does.
 
1492
        # TODO: jam 20070215 This is the cheap way by cheating and using the
 
1493
        #       RevisionTree implementation.
 
1494
        #       This should be cleaned up to use the much faster Dirstate code
 
1495
        #       This is a little tricky, though, because the dirstate is
 
1496
        #       indexed by current path, not by parent path.
 
1497
        #       So for now, we just build up the parent inventory, and extract
 
1498
        #       it the same way RevisionTree does.
1923
1499
        _directory = 'directory'
1924
1500
        inv = self._get_inventory()
1925
1501
        top_id = inv.path2id(prefix)
1951
1527
 
1952
1528
class InterDirStateTree(InterTree):
1953
1529
    """Fast path optimiser for changes_from with dirstate trees.
1954
 
 
1955
 
    This is used only when both trees are in the dirstate working file, and
1956
 
    the source is any parent within the dirstate, and the destination is
 
1530
    
 
1531
    This is used only when both trees are in the dirstate working file, and 
 
1532
    the source is any parent within the dirstate, and the destination is 
1957
1533
    the current working tree of the same dirstate.
1958
1534
    """
1959
1535
    # this could be generalized to allow comparisons between any trees in the
1972
1548
        target.set_parent_ids([revid])
1973
1549
        return target.basis_tree(), target
1974
1550
 
1975
 
    @classmethod
1976
 
    def make_source_parent_tree_python_dirstate(klass, test_case, source, target):
1977
 
        result = klass.make_source_parent_tree(source, target)
1978
 
        result[1]._iter_changes = dirstate.ProcessEntryPython
1979
 
        return result
1980
 
 
1981
 
    @classmethod
1982
 
    def make_source_parent_tree_compiled_dirstate(klass, test_case, source,
1983
 
                                                  target):
1984
 
        from bzrlib.tests.test__dirstate_helpers import \
1985
 
            compiled_dirstate_helpers_feature
1986
 
        test_case.requireFeature(compiled_dirstate_helpers_feature)
1987
 
        from bzrlib._dirstate_helpers_pyx import ProcessEntryC
1988
 
        result = klass.make_source_parent_tree(source, target)
1989
 
        result[1]._iter_changes = ProcessEntryC
1990
 
        return result
1991
 
 
1992
1551
    _matching_from_tree_format = WorkingTreeFormat4()
1993
1552
    _matching_to_tree_format = WorkingTreeFormat4()
1994
 
 
1995
 
    @classmethod
1996
 
    def _test_mutable_trees_to_test_trees(klass, test_case, source, target):
1997
 
        # This method shouldn't be called, because we have python and C
1998
 
        # specific flavours.
1999
 
        raise NotImplementedError
2000
 
 
2001
 
    def iter_changes(self, include_unchanged=False,
 
1553
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1554
 
 
1555
    def _iter_changes(self, include_unchanged=False,
2002
1556
                      specific_files=None, pb=None, extra_trees=[],
2003
1557
                      require_versioned=True, want_unversioned=False):
2004
1558
        """Return the changes from source to target.
2005
1559
 
2006
 
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1560
        :return: An iterator that yields tuples. See InterTree._iter_changes
2007
1561
            for details.
2008
1562
        :param specific_files: An optional list of file paths to restrict the
2009
1563
            comparison to. When mapping filenames to ids, all matches in all
2020
1574
            output. An unversioned file is defined as one with (False, False)
2021
1575
            for the versioned pair.
2022
1576
        """
 
1577
        utf8_decode = cache_utf8._utf8_decode_with_None
 
1578
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1579
        # NB: show_status depends on being able to pass in non-versioned files
 
1580
        # and report them as unknown
2023
1581
        # TODO: handle extra trees in the dirstate.
2024
 
        if (extra_trees or specific_files == []):
 
1582
        # TODO: handle comparisons as an empty tree as a different special
 
1583
        # case? mbp 20070226
 
1584
        if extra_trees or (self.source._revision_id == NULL_REVISION):
2025
1585
            # we can't fast-path these cases (yet)
2026
 
            return super(InterDirStateTree, self).iter_changes(
 
1586
            for f in super(InterDirStateTree, self)._iter_changes(
2027
1587
                include_unchanged, specific_files, pb, extra_trees,
2028
 
                require_versioned, want_unversioned=want_unversioned)
 
1588
                require_versioned, want_unversioned=want_unversioned):
 
1589
                yield f
 
1590
            return
2029
1591
        parent_ids = self.target.get_parent_ids()
2030
 
        if not (self.source._revision_id in parent_ids
2031
 
                or self.source._revision_id == _mod_revision.NULL_REVISION):
2032
 
            raise AssertionError(
2033
 
                "revision {%s} is not stored in {%s}, but %s "
2034
 
                "can only be used for trees stored in the dirstate"
2035
 
                % (self.source._revision_id, self.target, self.iter_changes))
 
1592
        assert (self.source._revision_id in parent_ids), \
 
1593
                "revision {%s} is not stored in {%s}, but %s " \
 
1594
                "can only be used for trees stored in the dirstate" \
 
1595
                % (self.source._revision_id, self.target, self._iter_changes)
2036
1596
        target_index = 0
2037
 
        if self.source._revision_id == _mod_revision.NULL_REVISION:
 
1597
        if self.source._revision_id == NULL_REVISION:
2038
1598
            source_index = None
2039
1599
            indices = (target_index,)
2040
1600
        else:
2041
 
            if not (self.source._revision_id in parent_ids):
2042
 
                raise AssertionError(
2043
 
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
2044
 
                    self.source._revision_id, parent_ids))
 
1601
            assert (self.source._revision_id in parent_ids), \
 
1602
                "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1603
                self.source._revision_id, parent_ids)
2045
1604
            source_index = 1 + parent_ids.index(self.source._revision_id)
2046
 
            indices = (source_index, target_index)
 
1605
            indices = (source_index,target_index)
2047
1606
        # -- make all specific_files utf8 --
2048
1607
        if specific_files:
2049
1608
            specific_files_utf8 = set()
2050
1609
            for path in specific_files:
2051
 
                # Note, if there are many specific files, using cache_utf8
2052
 
                # would be good here.
2053
1610
                specific_files_utf8.add(path.encode('utf8'))
2054
1611
            specific_files = specific_files_utf8
2055
1612
        else:
2056
1613
            specific_files = set([''])
2057
1614
        # -- specific_files is now a utf8 path set --
2058
 
 
2059
1615
        # -- get the state object and prepare it.
2060
1616
        state = self.target.current_dirstate()
2061
1617
        state._read_dirblocks_if_needed()
 
1618
        def _entries_for_path(path):
 
1619
            """Return a list with all the entries that match path for all ids.
 
1620
            """
 
1621
            dirname, basename = os.path.split(path)
 
1622
            key = (dirname, basename, '')
 
1623
            block_index, present = state._find_block_index_from_key(key)
 
1624
            if not present:
 
1625
                # the block which should contain path is absent.
 
1626
                return []
 
1627
            result = []
 
1628
            block = state._dirblocks[block_index][1]
 
1629
            entry_index, _ = state._find_entry_index(key, block)
 
1630
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1631
            while (entry_index < len(block) and
 
1632
                block[entry_index][0][0:2] == key[0:2]):
 
1633
                result.append(block[entry_index])
 
1634
                entry_index += 1
 
1635
            return result
2062
1636
        if require_versioned:
2063
1637
            # -- check all supplied paths are versioned in a search tree. --
2064
 
            not_versioned = []
 
1638
            all_versioned = True
2065
1639
            for path in specific_files:
2066
 
                path_entries = state._entries_for_path(path)
 
1640
                path_entries = _entries_for_path(path)
2067
1641
                if not path_entries:
2068
1642
                    # this specified path is not present at all: error
2069
 
                    not_versioned.append(path)
2070
 
                    continue
 
1643
                    all_versioned = False
 
1644
                    break
2071
1645
                found_versioned = False
2072
1646
                # for each id at this path
2073
1647
                for entry in path_entries:
2080
1654
                if not found_versioned:
2081
1655
                    # none of the indexes was not 'absent' at all ids for this
2082
1656
                    # path.
2083
 
                    not_versioned.append(path)
2084
 
            if len(not_versioned) > 0:
2085
 
                raise errors.PathsNotVersionedError(not_versioned)
 
1657
                    all_versioned = False
 
1658
                    break
 
1659
            if not all_versioned:
 
1660
                raise errors.PathsNotVersionedError(specific_files)
2086
1661
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
2087
 
        search_specific_files = osutils.minimum_path_selection(specific_files)
 
1662
        search_specific_files = set()
 
1663
        for path in specific_files:
 
1664
            other_specific_files = specific_files.difference(set([path]))
 
1665
            if not osutils.is_inside_any(other_specific_files, path):
 
1666
                # this is a top level path, we must check it.
 
1667
                search_specific_files.add(path)
 
1668
        # sketch: 
 
1669
        # compare source_index and target_index at or under each element of search_specific_files.
 
1670
        # follow the following comparison table. Note that we only want to do diff operations when
 
1671
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1672
        # for the target.
 
1673
        # cases:
 
1674
        # 
 
1675
        # Source | Target | disk | action
 
1676
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1677
        #        |        |      | diff check on source-target
 
1678
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1679
        #        |        |      | ???
 
1680
        #   r    |  a     |      | add source to search
 
1681
        #   r    |  a     |  a   | 
 
1682
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1683
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1684
        #   a    | fdlt   |      | add new id
 
1685
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1686
        #   a    |  a     |      | not present in either tree, skip
 
1687
        #   a    |  a     |  a   | not present in any tree, skip
 
1688
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1689
        #        |        |      | may not be selected by the users list of paths.
 
1690
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1691
        #        |        |      | may not be selected by the users list of paths.
 
1692
        #  fdlt  | fdlt   |      | content in both: diff them
 
1693
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1694
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1695
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1696
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1697
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1698
        #        |        |      | this id at the other path.
 
1699
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1700
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1701
        #        |        |      | this id at the other path.
 
1702
 
 
1703
        # for all search_indexs in each path at or under each element of
 
1704
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1705
        # relocated path as one to search if its not searched already. If the
 
1706
        # detail is not relocated, add the id.
 
1707
        searched_specific_files = set()
 
1708
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1709
        # Using a list so that we can access the values and change them in
 
1710
        # nested scope. Each one is [path, file_id, entry]
 
1711
        last_source_parent = [None, None, None]
 
1712
        last_target_parent = [None, None, None]
2088
1713
 
2089
1714
        use_filesystem_for_exec = (sys.platform != 'win32')
2090
 
        iter_changes = self.target._iter_changes(include_unchanged,
2091
 
            use_filesystem_for_exec, search_specific_files, state,
2092
 
            source_index, target_index, want_unversioned, self.target)
2093
 
        return iter_changes.iter_changes()
 
1715
 
 
1716
        def _process_entry(entry, path_info):
 
1717
            """Compare an entry and real disk to generate delta information.
 
1718
 
 
1719
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1720
                the path of entry. If None, then the path is considered absent.
 
1721
                (Perhaps we should pass in a concrete entry for this ?)
 
1722
                Basename is returned as a utf8 string because we expect this
 
1723
                tuple will be ignored, and don't want to take the time to
 
1724
                decode.
 
1725
            """
 
1726
            # TODO: when a parent has been renamed, dont emit path renames for children,
 
1727
            ## if path_info[1] == 'sub':
 
1728
            ##     import pdb;pdb.set_trace()
 
1729
            if source_index is None:
 
1730
                source_details = NULL_PARENT_DETAILS
 
1731
            else:
 
1732
                source_details = entry[1][source_index]
 
1733
            target_details = entry[1][target_index]
 
1734
            target_minikind = target_details[0]
 
1735
            if path_info is not None and target_minikind in 'fdlt':
 
1736
                assert target_index == 0
 
1737
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1738
                                                  stat_value=path_info[3])
 
1739
                # The entry may have been modified by update_entry
 
1740
                target_details = entry[1][target_index]
 
1741
                target_minikind = target_details[0]
 
1742
            else:
 
1743
                link_or_sha1 = None
 
1744
            source_minikind = source_details[0]
 
1745
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1746
                # claimed content in both: diff
 
1747
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1748
                #        |        |      | diff check on source-target
 
1749
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1750
                #        |        |      | ???
 
1751
                if source_minikind in 'r':
 
1752
                    # add the source to the search path to find any children it
 
1753
                    # has.  TODO ? : only add if it is a container ?
 
1754
                    if not osutils.is_inside_any(searched_specific_files,
 
1755
                                                 source_details[1]):
 
1756
                        search_specific_files.add(source_details[1])
 
1757
                    # generate the old path; this is needed for stating later
 
1758
                    # as well.
 
1759
                    old_path = source_details[1]
 
1760
                    old_dirname, old_basename = os.path.split(old_path)
 
1761
                    path = pathjoin(entry[0][0], entry[0][1])
 
1762
                    old_entry = state._get_entry(source_index,
 
1763
                                                 path_utf8=old_path)
 
1764
                    # update the source details variable to be the real
 
1765
                    # location.
 
1766
                    source_details = old_entry[1][source_index]
 
1767
                    source_minikind = source_details[0]
 
1768
                else:
 
1769
                    old_dirname = entry[0][0]
 
1770
                    old_basename = entry[0][1]
 
1771
                    old_path = path = pathjoin(old_dirname, old_basename)
 
1772
                if path_info is None:
 
1773
                    # the file is missing on disk, show as removed.
 
1774
                    content_change = True
 
1775
                    target_kind = None
 
1776
                    target_exec = False
 
1777
                else:
 
1778
                    # source and target are both versioned and disk file is present.
 
1779
                    target_kind = path_info[2]
 
1780
                    if target_kind == 'directory':
 
1781
                        if source_minikind != 'd':
 
1782
                            content_change = True
 
1783
                        else:
 
1784
                            # directories have no fingerprint
 
1785
                            content_change = False
 
1786
                        target_exec = False
 
1787
                    elif target_kind == 'file':
 
1788
                        if source_minikind != 'f':
 
1789
                            content_change = True
 
1790
                        else:
 
1791
                            # We could check the size, but we already have the
 
1792
                            # sha1 hash.
 
1793
                            content_change = (link_or_sha1 != source_details[1])
 
1794
                        # Target details is updated at update_entry time
 
1795
                        if use_filesystem_for_exec:
 
1796
                            # We don't need S_ISREG here, because we are sure
 
1797
                            # we are dealing with a file.
 
1798
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
1799
                        else:
 
1800
                            target_exec = target_details[3]
 
1801
                    elif target_kind == 'symlink':
 
1802
                        if source_minikind != 'l':
 
1803
                            content_change = True
 
1804
                        else:
 
1805
                            content_change = (link_or_sha1 != source_details[1])
 
1806
                        target_exec = False
 
1807
                    elif target_kind == 'tree-reference':
 
1808
                        if source_minikind != 't':
 
1809
                            content_change = True
 
1810
                        else:
 
1811
                            content_change = False
 
1812
                        target_exec = False
 
1813
                    else:
 
1814
                        raise Exception, "unknown kind %s" % path_info[2]
 
1815
                # parent id is the entry for the path in the target tree
 
1816
                if old_dirname == last_source_parent[0]:
 
1817
                    source_parent_id = last_source_parent[1]
 
1818
                else:
 
1819
                    source_parent_entry = state._get_entry(source_index,
 
1820
                                                           path_utf8=old_dirname)
 
1821
                    source_parent_id = source_parent_entry[0][2]
 
1822
                    if source_parent_id == entry[0][2]:
 
1823
                        # This is the root, so the parent is None
 
1824
                        source_parent_id = None
 
1825
                    else:
 
1826
                        last_source_parent[0] = old_dirname
 
1827
                        last_source_parent[1] = source_parent_id
 
1828
                        last_source_parent[2] = source_parent_entry
 
1829
 
 
1830
                new_dirname = entry[0][0]
 
1831
                if new_dirname == last_target_parent[0]:
 
1832
                    target_parent_id = last_target_parent[1]
 
1833
                else:
 
1834
                    # TODO: We don't always need to do the lookup, because the
 
1835
                    #       parent entry will be the same as the source entry.
 
1836
                    target_parent_entry = state._get_entry(target_index,
 
1837
                                                           path_utf8=new_dirname)
 
1838
                    target_parent_id = target_parent_entry[0][2]
 
1839
                    if target_parent_id == entry[0][2]:
 
1840
                        # This is the root, so the parent is None
 
1841
                        target_parent_id = None
 
1842
                    else:
 
1843
                        last_target_parent[0] = new_dirname
 
1844
                        last_target_parent[1] = target_parent_id
 
1845
                        last_target_parent[2] = target_parent_entry
 
1846
 
 
1847
                source_exec = source_details[3]
 
1848
                return ((entry[0][2], (old_path, path), content_change,
 
1849
                        (True, True),
 
1850
                        (source_parent_id, target_parent_id),
 
1851
                        (old_basename, entry[0][1]),
 
1852
                        (_minikind_to_kind[source_minikind], target_kind),
 
1853
                        (source_exec, target_exec)),)
 
1854
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
1855
                # looks like a new file
 
1856
                if path_info is not None:
 
1857
                    path = pathjoin(entry[0][0], entry[0][1])
 
1858
                    # parent id is the entry for the path in the target tree
 
1859
                    # TODO: these are the same for an entire directory: cache em.
 
1860
                    parent_id = state._get_entry(target_index,
 
1861
                                                 path_utf8=entry[0][0])[0][2]
 
1862
                    if parent_id == entry[0][2]:
 
1863
                        parent_id = None
 
1864
                    if use_filesystem_for_exec:
 
1865
                        # We need S_ISREG here, because we aren't sure if this
 
1866
                        # is a file or not.
 
1867
                        target_exec = bool(
 
1868
                            stat.S_ISREG(path_info[3].st_mode)
 
1869
                            and stat.S_IEXEC & path_info[3].st_mode)
 
1870
                    else:
 
1871
                        target_exec = target_details[3]
 
1872
                    return ((entry[0][2], (None, path), True,
 
1873
                            (False, True),
 
1874
                            (None, parent_id),
 
1875
                            (None, entry[0][1]),
 
1876
                            (None, path_info[2]),
 
1877
                            (None, target_exec)),)
 
1878
                else:
 
1879
                    # but its not on disk: we deliberately treat this as just
 
1880
                    # never-present. (Why ?! - RBC 20070224)
 
1881
                    pass
 
1882
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
1883
                # unversioned, possibly, or possibly not deleted: we dont care.
 
1884
                # if its still on disk, *and* theres no other entry at this
 
1885
                # path [we dont know this in this routine at the moment -
 
1886
                # perhaps we should change this - then it would be an unknown.
 
1887
                old_path = pathjoin(entry[0][0], entry[0][1])
 
1888
                # parent id is the entry for the path in the target tree
 
1889
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
1890
                if parent_id == entry[0][2]:
 
1891
                    parent_id = None
 
1892
                return ((entry[0][2], (old_path, None), True,
 
1893
                        (True, False),
 
1894
                        (parent_id, None),
 
1895
                        (entry[0][1], None),
 
1896
                        (_minikind_to_kind[source_minikind], None),
 
1897
                        (source_details[3], None)),)
 
1898
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
1899
                # a rename; could be a true rename, or a rename inherited from
 
1900
                # a renamed parent. TODO: handle this efficiently. Its not
 
1901
                # common case to rename dirs though, so a correct but slow
 
1902
                # implementation will do.
 
1903
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
1904
                    search_specific_files.add(target_details[1])
 
1905
            elif source_minikind in 'r' and target_minikind in 'r':
 
1906
                # neither of the selected trees contain this file,
 
1907
                # so skip over it. This is not currently directly tested, but
 
1908
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
1909
                pass
 
1910
            else:
 
1911
                raise AssertionError("don't know how to compare "
 
1912
                    "source_minikind=%r, target_minikind=%r"
 
1913
                    % (source_minikind, target_minikind))
 
1914
                ## import pdb;pdb.set_trace()
 
1915
            return ()
 
1916
        while search_specific_files:
 
1917
            # TODO: the pending list should be lexically sorted?
 
1918
            current_root = search_specific_files.pop()
 
1919
            searched_specific_files.add(current_root)
 
1920
            # process the entries for this containing directory: the rest will be
 
1921
            # found by their parents recursively.
 
1922
            root_entries = _entries_for_path(current_root)
 
1923
            root_abspath = self.target.abspath(current_root)
 
1924
            try:
 
1925
                root_stat = os.lstat(root_abspath)
 
1926
            except OSError, e:
 
1927
                if e.errno == errno.ENOENT:
 
1928
                    # the path does not exist: let _process_entry know that.
 
1929
                    root_dir_info = None
 
1930
                else:
 
1931
                    # some other random error: hand it up.
 
1932
                    raise
 
1933
            else:
 
1934
                root_dir_info = ('', current_root,
 
1935
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
1936
                    root_abspath)
 
1937
                if root_dir_info[2] == 'directory':
 
1938
                    if self.target._directory_is_tree_reference(current_root):
 
1939
                        root_dir_info = root_dir_info[:2] + \
 
1940
                            ('tree-reference',) + root_dir_info[3:]
 
1941
 
 
1942
            if not root_entries and not root_dir_info:
 
1943
                # this specified path is not present at all, skip it.
 
1944
                continue
 
1945
            path_handled = False
 
1946
            for entry in root_entries:
 
1947
                for result in _process_entry(entry, root_dir_info):
 
1948
                    # this check should probably be outside the loop: one
 
1949
                    # 'iterate two trees' api, and then _iter_changes filters
 
1950
                    # unchanged pairs. - RBC 20070226
 
1951
                    path_handled = True
 
1952
                    if (include_unchanged
 
1953
                        or result[2]                    # content change
 
1954
                        or result[3][0] != result[3][1] # versioned status
 
1955
                        or result[4][0] != result[4][1] # parent id
 
1956
                        or result[5][0] != result[5][1] # name
 
1957
                        or result[6][0] != result[6][1] # kind
 
1958
                        or result[7][0] != result[7][1] # executable
 
1959
                        ):
 
1960
                        result = (result[0],
 
1961
                            ((utf8_decode(result[1][0])[0]),
 
1962
                             utf8_decode(result[1][1])[0]),) + result[2:]
 
1963
                        yield result
 
1964
            if want_unversioned and not path_handled:
 
1965
                new_executable = bool(
 
1966
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
1967
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
1968
                yield (None, (None, current_root), True, (False, False),
 
1969
                    (None, None),
 
1970
                    (None, splitpath(current_root)[-1]),
 
1971
                    (None, root_dir_info[2]), (None, new_executable))
 
1972
            dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
1973
            initial_key = (current_root, '', '')
 
1974
            block_index, _ = state._find_block_index_from_key(initial_key)
 
1975
            if block_index == 0:
 
1976
                # we have processed the total root already, but because the
 
1977
                # initial key matched it we should skip it here.
 
1978
                block_index +=1
 
1979
            try:
 
1980
                current_dir_info = dir_iterator.next()
 
1981
            except OSError, e:
 
1982
                if e.errno in (errno.ENOENT, errno.ENOTDIR):
 
1983
                    # there may be directories in the inventory even though
 
1984
                    # this path is not a file on disk: so mark it as end of
 
1985
                    # iterator
 
1986
                    current_dir_info = None
 
1987
                else:
 
1988
                    raise
 
1989
            else:
 
1990
                if current_dir_info[0][0] == '':
 
1991
                    # remove .bzr from iteration
 
1992
                    bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
1993
                    assert current_dir_info[1][bzr_index][0] == '.bzr'
 
1994
                    del current_dir_info[1][bzr_index]
 
1995
            # walk until both the directory listing and the versioned metadata
 
1996
            # are exhausted. TODO: reevaluate this, perhaps we should stop when
 
1997
            # the versioned data runs out.
 
1998
            if (block_index < len(state._dirblocks) and
 
1999
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2000
                current_block = state._dirblocks[block_index]
 
2001
            else:
 
2002
                current_block = None
 
2003
            while (current_dir_info is not None or
 
2004
                   current_block is not None):
 
2005
                if (current_dir_info and current_block
 
2006
                    and current_dir_info[0][0] != current_block[0]):
 
2007
                    if current_dir_info[0][0] < current_block[0] :
 
2008
                        # filesystem data refers to paths not covered by the dirblock.
 
2009
                        # this has two possibilities:
 
2010
                        # A) it is versioned but empty, so there is no block for it
 
2011
                        # B) it is not versioned.
 
2012
                        # in either case it was processed by the containing directories walk:
 
2013
                        # if it is root/foo, when we walked root we emitted it,
 
2014
                        # or if we ere given root/foo to walk specifically, we
 
2015
                        # emitted it when checking the walk-root entries
 
2016
                        # advance the iterator and loop - we dont need to emit it.
 
2017
                        try:
 
2018
                            current_dir_info = dir_iterator.next()
 
2019
                        except StopIteration:
 
2020
                            current_dir_info = None
 
2021
                    else:
 
2022
                        # We have a dirblock entry for this location, but there
 
2023
                        # is no filesystem path for this. This is most likely
 
2024
                        # because a directory was removed from the disk.
 
2025
                        # We don't have to report the missing directory,
 
2026
                        # because that should have already been handled, but we
 
2027
                        # need to handle all of the files that are contained
 
2028
                        # within.
 
2029
                        for current_entry in current_block[1]:
 
2030
                            # entry referring to file not present on disk.
 
2031
                            # advance the entry only, after processing.
 
2032
                            for result in _process_entry(current_entry, None):
 
2033
                                # this check should probably be outside the loop: one
 
2034
                                # 'iterate two trees' api, and then _iter_changes filters
 
2035
                                # unchanged pairs. - RBC 20070226
 
2036
                                if (include_unchanged
 
2037
                                    or result[2]                    # content change
 
2038
                                    or result[3][0] != result[3][1] # versioned status
 
2039
                                    or result[4][0] != result[4][1] # parent id
 
2040
                                    or result[5][0] != result[5][1] # name
 
2041
                                    or result[6][0] != result[6][1] # kind
 
2042
                                    or result[7][0] != result[7][1] # executable
 
2043
                                    ):
 
2044
                                    result = (result[0],
 
2045
                                        ((utf8_decode(result[1][0])[0]),
 
2046
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2047
                                    yield result
 
2048
                        block_index +=1
 
2049
                        if (block_index < len(state._dirblocks) and
 
2050
                            osutils.is_inside(current_root,
 
2051
                                              state._dirblocks[block_index][0])):
 
2052
                            current_block = state._dirblocks[block_index]
 
2053
                        else:
 
2054
                            current_block = None
 
2055
                    continue
 
2056
                entry_index = 0
 
2057
                if current_block and entry_index < len(current_block[1]):
 
2058
                    current_entry = current_block[1][entry_index]
 
2059
                else:
 
2060
                    current_entry = None
 
2061
                advance_entry = True
 
2062
                path_index = 0
 
2063
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2064
                    current_path_info = current_dir_info[1][path_index]
 
2065
                    if current_path_info[2] == 'directory':
 
2066
                        if self.target._directory_is_tree_reference(
 
2067
                            current_path_info[0]):
 
2068
                            current_path_info = current_path_info[:2] + \
 
2069
                                ('tree-reference',) + current_path_info[3:]
 
2070
                else:
 
2071
                    current_path_info = None
 
2072
                advance_path = True
 
2073
                path_handled = False
 
2074
                while (current_entry is not None or
 
2075
                    current_path_info is not None):
 
2076
                    if current_entry is None:
 
2077
                        # the check for path_handled when the path is adnvaced
 
2078
                        # will yield this path if needed.
 
2079
                        pass
 
2080
                    elif current_path_info is None:
 
2081
                        # no path is fine: the per entry code will handle it.
 
2082
                        for result in _process_entry(current_entry, current_path_info):
 
2083
                            # this check should probably be outside the loop: one
 
2084
                            # 'iterate two trees' api, and then _iter_changes filters
 
2085
                            # unchanged pairs. - RBC 20070226
 
2086
                            if (include_unchanged
 
2087
                                or result[2]                    # content change
 
2088
                                or result[3][0] != result[3][1] # versioned status
 
2089
                                or result[4][0] != result[4][1] # parent id
 
2090
                                or result[5][0] != result[5][1] # name
 
2091
                                or result[6][0] != result[6][1] # kind
 
2092
                                or result[7][0] != result[7][1] # executable
 
2093
                                ):
 
2094
                                result = (result[0],
 
2095
                                    ((utf8_decode(result[1][0])[0]),
 
2096
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2097
                                yield result
 
2098
                    elif current_entry[0][1] != current_path_info[1]:
 
2099
                        if current_path_info[1] < current_entry[0][1]:
 
2100
                            # extra file on disk: pass for now, but only
 
2101
                            # increment the path, not the entry
 
2102
                            # import pdb; pdb.set_trace()
 
2103
                            # print 'unversioned file'
 
2104
                            advance_entry = False
 
2105
                        else:
 
2106
                            # entry referring to file not present on disk.
 
2107
                            # advance the entry only, after processing.
 
2108
                            for result in _process_entry(current_entry, None):
 
2109
                                # this check should probably be outside the loop: one
 
2110
                                # 'iterate two trees' api, and then _iter_changes filters
 
2111
                                # unchanged pairs. - RBC 20070226
 
2112
                                path_handled = True
 
2113
                                if (include_unchanged
 
2114
                                    or result[2]                    # content change
 
2115
                                    or result[3][0] != result[3][1] # versioned status
 
2116
                                    or result[4][0] != result[4][1] # parent id
 
2117
                                    or result[5][0] != result[5][1] # name
 
2118
                                    or result[6][0] != result[6][1] # kind
 
2119
                                    or result[7][0] != result[7][1] # executable
 
2120
                                    ):
 
2121
                                    result = (result[0],
 
2122
                                        ((utf8_decode(result[1][0])[0]),
 
2123
                                         utf8_decode(result[1][1])[0]),) + result[2:]
 
2124
                                    yield result
 
2125
                            advance_path = False
 
2126
                    else:
 
2127
                        for result in _process_entry(current_entry, current_path_info):
 
2128
                            # this check should probably be outside the loop: one
 
2129
                            # 'iterate two trees' api, and then _iter_changes filters
 
2130
                            # unchanged pairs. - RBC 20070226
 
2131
                            path_handled = True
 
2132
                            if (include_unchanged
 
2133
                                or result[2]                    # content change
 
2134
                                or result[3][0] != result[3][1] # versioned status
 
2135
                                or result[4][0] != result[4][1] # parent id
 
2136
                                or result[5][0] != result[5][1] # name
 
2137
                                or result[6][0] != result[6][1] # kind
 
2138
                                or result[7][0] != result[7][1] # executable
 
2139
                                ):
 
2140
                                result = (result[0],
 
2141
                                    ((utf8_decode(result[1][0])[0]),
 
2142
                                     utf8_decode(result[1][1])[0]),) + result[2:]
 
2143
                                yield result
 
2144
                    if advance_entry and current_entry is not None:
 
2145
                        entry_index += 1
 
2146
                        if entry_index < len(current_block[1]):
 
2147
                            current_entry = current_block[1][entry_index]
 
2148
                        else:
 
2149
                            current_entry = None
 
2150
                    else:
 
2151
                        advance_entry = True # reset the advance flaga
 
2152
                    if advance_path and current_path_info is not None:
 
2153
                        if not path_handled:
 
2154
                            # unversioned in all regards
 
2155
                            if want_unversioned:
 
2156
                                new_executable = bool(
 
2157
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2158
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2159
                                if want_unversioned:
 
2160
                                    yield (None, (None, current_path_info[0]),
 
2161
                                        True,
 
2162
                                        (False, False),
 
2163
                                        (None, None),
 
2164
                                        (None, current_path_info[1]),
 
2165
                                        (None, current_path_info[2]),
 
2166
                                        (None, new_executable))
 
2167
                            # dont descend into this unversioned path if it is
 
2168
                            # a dir
 
2169
                            if current_path_info[2] in (
 
2170
                                'directory', 'tree-referene'):
 
2171
                                del current_dir_info[1][path_index]
 
2172
                                path_index -= 1
 
2173
                        path_index += 1
 
2174
                        if path_index < len(current_dir_info[1]):
 
2175
                            current_path_info = current_dir_info[1][path_index]
 
2176
                            if current_path_info[2] == 'directory':
 
2177
                                if self.target._directory_is_tree_reference(
 
2178
                                    current_path_info[0]):
 
2179
                                    current_path_info = current_path_info[:2] + \
 
2180
                                        ('tree-reference',) + current_path_info[3:]
 
2181
                        else:
 
2182
                            current_path_info = None
 
2183
                        path_handled = False
 
2184
                    else:
 
2185
                        advance_path = True # reset the advance flagg.
 
2186
                if current_block is not None:
 
2187
                    block_index += 1
 
2188
                    if (block_index < len(state._dirblocks) and
 
2189
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2190
                        current_block = state._dirblocks[block_index]
 
2191
                    else:
 
2192
                        current_block = None
 
2193
                if current_dir_info is not None:
 
2194
                    try:
 
2195
                        current_dir_info = dir_iterator.next()
 
2196
                    except StopIteration:
 
2197
                        current_dir_info = None
 
2198
 
2094
2199
 
2095
2200
    @staticmethod
2096
2201
    def is_compatible(source, target):
2097
2202
        # the target must be a dirstate working tree
2098
 
        if not isinstance(target, DirStateWorkingTree):
 
2203
        if not isinstance(target, WorkingTree4):
2099
2204
            return False
2100
 
        # the source must be a revtree or dirstate rev tree.
 
2205
        # the source must be a revtreee or dirstate rev tree.
2101
2206
        if not isinstance(source,
2102
2207
            (revisiontree.RevisionTree, DirStateRevisionTree)):
2103
2208
            return False
2104
2209
        # the source revid must be in the target dirstate
2105
 
        if not (source._revision_id == _mod_revision.NULL_REVISION or
 
2210
        if not (source._revision_id == NULL_REVISION or
2106
2211
            source._revision_id in target.get_parent_ids()):
2107
 
            # TODO: what about ghosts? it may well need to
 
2212
            # TODO: what about ghosts? it may well need to 
2108
2213
            # check for them explicitly.
2109
2214
            return False
2110
2215
        return True
2120
2225
 
2121
2226
    def convert(self, tree):
2122
2227
        # lock the control files not the tree, so that we dont get tree
2123
 
        # on-unlock behaviours, and so that noone else diddles with the
 
2228
        # on-unlock behaviours, and so that noone else diddles with the 
2124
2229
        # tree during upgrade.
2125
2230
        tree._control_files.lock_write()
2126
2231
        try:
2127
 
            tree.read_working_inventory()
2128
2232
            self.create_dirstate_data(tree)
2129
2233
            self.update_format(tree)
2130
2234
            self.remove_xml_files(tree)
2152
2256
 
2153
2257
    def update_format(self, tree):
2154
2258
        """Change the format marker."""
2155
 
        tree._transport.put_bytes('format',
2156
 
            self.target_format.get_format_string(),
2157
 
            mode=tree.bzrdir._get_file_mode())
2158
 
 
2159
 
 
2160
 
class Converter4to5(object):
2161
 
    """Perform an in-place upgrade of format 4 to format 5 trees."""
2162
 
 
2163
 
    def __init__(self):
2164
 
        self.target_format = WorkingTreeFormat5()
2165
 
 
2166
 
    def convert(self, tree):
2167
 
        # lock the control files not the tree, so that we don't get tree
2168
 
        # on-unlock behaviours, and so that no-one else diddles with the
2169
 
        # tree during upgrade.
2170
 
        tree._control_files.lock_write()
2171
 
        try:
2172
 
            self.update_format(tree)
2173
 
        finally:
2174
 
            tree._control_files.unlock()
2175
 
 
2176
 
    def update_format(self, tree):
2177
 
        """Change the format marker."""
2178
 
        tree._transport.put_bytes('format',
2179
 
            self.target_format.get_format_string(),
2180
 
            mode=tree.bzrdir._get_file_mode())
2181
 
 
2182
 
 
2183
 
class Converter4or5to6(object):
2184
 
    """Perform an in-place upgrade of format 4 or 5 to format 6 trees."""
2185
 
 
2186
 
    def __init__(self):
2187
 
        self.target_format = WorkingTreeFormat6()
2188
 
 
2189
 
    def convert(self, tree):
2190
 
        # lock the control files not the tree, so that we don't get tree
2191
 
        # on-unlock behaviours, and so that no-one else diddles with the
2192
 
        # tree during upgrade.
2193
 
        tree._control_files.lock_write()
2194
 
        try:
2195
 
            self.init_custom_control_files(tree)
2196
 
            self.update_format(tree)
2197
 
        finally:
2198
 
            tree._control_files.unlock()
2199
 
 
2200
 
    def init_custom_control_files(self, tree):
2201
 
        """Initialize custom control files."""
2202
 
        tree._transport.put_bytes('views', '',
2203
 
            mode=tree.bzrdir._get_file_mode())
2204
 
 
2205
 
    def update_format(self, tree):
2206
 
        """Change the format marker."""
2207
 
        tree._transport.put_bytes('format',
2208
 
            self.target_format.get_format_string(),
2209
 
            mode=tree.bzrdir._get_file_mode())
 
2259
        tree._control_files.put_utf8('format',
 
2260
            self.target_format.get_format_string())