~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/workingtree_4.py

  • Committer: Canonical.com Patch Queue Manager
  • Date: 2008-05-29 15:57:16 UTC
  • mfrom: (3427.5.9 dep_warnings)
  • Revision ID: pqm@pqm.ubuntu.com-20080529155716-0w3kic8lioa63231
(jam) Enable Deprecation Warnings when running -Werror and when
        running selftest

Show diffs side-by-side

added added

removed removed

Lines of Context:
 
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
 
2
#
 
3
# This program is free software; you can redistribute it and/or modify
 
4
# it under the terms of the GNU General Public License as published by
 
5
# the Free Software Foundation; either version 2 of the License, or
 
6
# (at your option) any later version.
 
7
#
 
8
# This program is distributed in the hope that it will be useful,
 
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
 
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 
11
# GNU General Public License for more details.
 
12
#
 
13
# You should have received a copy of the GNU General Public License
 
14
# along with this program; if not, write to the Free Software
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
"""WorkingTree4 format and implementation.
 
18
 
 
19
WorkingTree4 provides the dirstate based working tree logic.
 
20
 
 
21
To get a WorkingTree, call bzrdir.open_workingtree() or
 
22
WorkingTree.open(dir).
 
23
"""
 
24
 
 
25
from cStringIO import StringIO
 
26
import os
 
27
import sys
 
28
 
 
29
from bzrlib.lazy_import import lazy_import
 
30
lazy_import(globals(), """
 
31
from bisect import bisect_left
 
32
import collections
 
33
from copy import deepcopy
 
34
import errno
 
35
import itertools
 
36
import operator
 
37
import stat
 
38
from time import time
 
39
import warnings
 
40
 
 
41
import bzrlib
 
42
from bzrlib import (
 
43
    bzrdir,
 
44
    cache_utf8,
 
45
    conflicts as _mod_conflicts,
 
46
    debug,
 
47
    delta,
 
48
    dirstate,
 
49
    errors,
 
50
    generate_ids,
 
51
    globbing,
 
52
    ignores,
 
53
    merge,
 
54
    osutils,
 
55
    revision as _mod_revision,
 
56
    revisiontree,
 
57
    textui,
 
58
    trace,
 
59
    transform,
 
60
    urlutils,
 
61
    xml5,
 
62
    xml6,
 
63
    )
 
64
import bzrlib.branch
 
65
from bzrlib.transport import get_transport
 
66
import bzrlib.ui
 
67
""")
 
68
 
 
69
from bzrlib import symbol_versioning
 
70
from bzrlib.decorators import needs_read_lock, needs_write_lock
 
71
from bzrlib.inventory import InventoryEntry, Inventory, ROOT_ID, entry_factory
 
72
from bzrlib.lockable_files import LockableFiles, TransportLock
 
73
from bzrlib.lockdir import LockDir
 
74
import bzrlib.mutabletree
 
75
from bzrlib.mutabletree import needs_tree_write_lock
 
76
from bzrlib.osutils import (
 
77
    file_kind,
 
78
    isdir,
 
79
    normpath,
 
80
    pathjoin,
 
81
    rand_chars,
 
82
    realpath,
 
83
    safe_unicode,
 
84
    splitpath,
 
85
    )
 
86
from bzrlib.trace import mutter, note
 
87
from bzrlib.transport.local import LocalTransport
 
88
from bzrlib.tree import InterTree
 
89
from bzrlib.progress import DummyProgress, ProgressPhase
 
90
from bzrlib.revision import NULL_REVISION, CURRENT_REVISION
 
91
from bzrlib.rio import RioReader, rio_file, Stanza
 
92
from bzrlib.symbol_versioning import (deprecated_passed,
 
93
        deprecated_method,
 
94
        deprecated_function,
 
95
        DEPRECATED_PARAMETER,
 
96
        )
 
97
from bzrlib.tree import Tree
 
98
from bzrlib.workingtree import WorkingTree, WorkingTree3, WorkingTreeFormat3
 
99
 
 
100
 
 
101
# This is the Windows equivalent of ENOTDIR
 
102
# It is defined in pywin32.winerror, but we don't want a strong dependency for
 
103
# just an error code.
 
104
ERROR_PATH_NOT_FOUND = 3
 
105
ERROR_DIRECTORY = 267
 
106
 
 
107
 
 
108
class WorkingTree4(WorkingTree3):
 
109
    """This is the Format 4 working tree.
 
110
 
 
111
    This differs from WorkingTree3 by:
 
112
     - Having a consolidated internal dirstate, stored in a
 
113
       randomly-accessible sorted file on disk.
 
114
     - Not having a regular inventory attribute.  One can be synthesized 
 
115
       on demand but this is expensive and should be avoided.
 
116
 
 
117
    This is new in bzr 0.15.
 
118
    """
 
119
 
 
120
    def __init__(self, basedir,
 
121
                 branch,
 
122
                 _control_files=None,
 
123
                 _format=None,
 
124
                 _bzrdir=None):
 
125
        """Construct a WorkingTree for basedir.
 
126
 
 
127
        If the branch is not supplied, it is opened automatically.
 
128
        If the branch is supplied, it must be the branch for this basedir.
 
129
        (branch.base is not cross checked, because for remote branches that
 
130
        would be meaningless).
 
131
        """
 
132
        self._format = _format
 
133
        self.bzrdir = _bzrdir
 
134
        basedir = safe_unicode(basedir)
 
135
        mutter("opening working tree %r", basedir)
 
136
        self._branch = branch
 
137
        self.basedir = realpath(basedir)
 
138
        # if branch is at our basedir and is a format 6 or less
 
139
        # assume all other formats have their own control files.
 
140
        self._control_files = _control_files
 
141
        self._transport = self._control_files._transport
 
142
        self._dirty = None
 
143
        #-------------
 
144
        # during a read or write lock these objects are set, and are
 
145
        # None the rest of the time.
 
146
        self._dirstate = None
 
147
        self._inventory = None
 
148
        #-------------
 
149
        self._setup_directory_is_tree_reference()
 
150
        self._detect_case_handling()
 
151
 
 
152
    @needs_tree_write_lock
 
153
    def _add(self, files, ids, kinds):
 
154
        """See MutableTree._add."""
 
155
        state = self.current_dirstate()
 
156
        for f, file_id, kind in zip(files, ids, kinds):
 
157
            f = f.strip('/')
 
158
            if self.path2id(f):
 
159
                # special case tree root handling.
 
160
                if f == '' and self.path2id(f) == ROOT_ID:
 
161
                    state.set_path_id('', generate_ids.gen_file_id(f))
 
162
                continue
 
163
            if file_id is None:
 
164
                file_id = generate_ids.gen_file_id(f)
 
165
            # deliberately add the file with no cached stat or sha1
 
166
            # - on the first access it will be gathered, and we can
 
167
            # always change this once tests are all passing.
 
168
            state.add(f, file_id, kind, None, '')
 
169
        self._make_dirty(reset_inventory=True)
 
170
 
 
171
    def _make_dirty(self, reset_inventory):
 
172
        """Make the tree state dirty.
 
173
 
 
174
        :param reset_inventory: True if the cached inventory should be removed
 
175
            (presuming there is one).
 
176
        """
 
177
        self._dirty = True
 
178
        if reset_inventory and self._inventory is not None:
 
179
            self._inventory = None
 
180
 
 
181
    @needs_tree_write_lock
 
182
    def add_reference(self, sub_tree):
 
183
        # use standard implementation, which calls back to self._add
 
184
        # 
 
185
        # So we don't store the reference_revision in the working dirstate,
 
186
        # it's just recorded at the moment of commit. 
 
187
        self._add_reference(sub_tree)
 
188
 
 
189
    def break_lock(self):
 
190
        """Break a lock if one is present from another instance.
 
191
 
 
192
        Uses the ui factory to ask for confirmation if the lock may be from
 
193
        an active process.
 
194
 
 
195
        This will probe the repository for its lock as well.
 
196
        """
 
197
        # if the dirstate is locked by an active process, reject the break lock
 
198
        # call.
 
199
        try:
 
200
            if self._dirstate is None:
 
201
                clear = True
 
202
            else:
 
203
                clear = False
 
204
            state = self._current_dirstate()
 
205
            if state._lock_token is not None:
 
206
                # we already have it locked. sheese, cant break our own lock.
 
207
                raise errors.LockActive(self.basedir)
 
208
            else:
 
209
                try:
 
210
                    # try for a write lock - need permission to get one anyhow
 
211
                    # to break locks.
 
212
                    state.lock_write()
 
213
                except errors.LockContention:
 
214
                    # oslocks fail when a process is still live: fail.
 
215
                    # TODO: get the locked lockdir info and give to the user to
 
216
                    # assist in debugging.
 
217
                    raise errors.LockActive(self.basedir)
 
218
                else:
 
219
                    state.unlock()
 
220
        finally:
 
221
            if clear:
 
222
                self._dirstate = None
 
223
        self._control_files.break_lock()
 
224
        self.branch.break_lock()
 
225
 
 
226
    def _comparison_data(self, entry, path):
 
227
        kind, executable, stat_value = \
 
228
            WorkingTree3._comparison_data(self, entry, path)
 
229
        # it looks like a plain directory, but it's really a reference -- see
 
230
        # also kind()
 
231
        if (self._repo_supports_tree_reference and
 
232
            kind == 'directory' and
 
233
            self._directory_is_tree_reference(path)):
 
234
            kind = 'tree-reference'
 
235
        return kind, executable, stat_value
 
236
 
 
237
    @needs_write_lock
 
238
    def commit(self, message=None, revprops=None, *args, **kwargs):
 
239
        # mark the tree as dirty post commit - commit
 
240
        # can change the current versioned list by doing deletes.
 
241
        result = WorkingTree3.commit(self, message, revprops, *args, **kwargs)
 
242
        self._make_dirty(reset_inventory=True)
 
243
        return result
 
244
 
 
245
    def current_dirstate(self):
 
246
        """Return the current dirstate object.
 
247
 
 
248
        This is not part of the tree interface and only exposed for ease of
 
249
        testing.
 
250
 
 
251
        :raises errors.NotWriteLocked: when not in a lock.
 
252
        """
 
253
        self._must_be_locked()
 
254
        return self._current_dirstate()
 
255
 
 
256
    def _current_dirstate(self):
 
257
        """Internal function that does not check lock status.
 
258
 
 
259
        This is needed for break_lock which also needs the dirstate.
 
260
        """
 
261
        if self._dirstate is not None:
 
262
            return self._dirstate
 
263
        local_path = self.bzrdir.get_workingtree_transport(None
 
264
            ).local_abspath('dirstate')
 
265
        self._dirstate = dirstate.DirState.on_file(local_path)
 
266
        return self._dirstate
 
267
 
 
268
    def filter_unversioned_files(self, paths):
 
269
        """Filter out paths that are versioned.
 
270
 
 
271
        :return: set of paths.
 
272
        """
 
273
        # TODO: make a generic multi-bisect routine roughly that should list
 
274
        # the paths, then process one half at a time recursively, and feed the
 
275
        # results of each bisect in further still
 
276
        paths = sorted(paths)
 
277
        result = set()
 
278
        state = self.current_dirstate()
 
279
        # TODO we want a paths_to_dirblocks helper I think
 
280
        for path in paths:
 
281
            dirname, basename = os.path.split(path.encode('utf8'))
 
282
            _, _, _, path_is_versioned = state._get_block_entry_index(
 
283
                dirname, basename, 0)
 
284
            if not path_is_versioned:
 
285
                result.add(path)
 
286
        return result
 
287
 
 
288
    def flush(self):
 
289
        """Write all cached data to disk."""
 
290
        if self._control_files._lock_mode != 'w':
 
291
            raise errors.NotWriteLocked(self)
 
292
        self.current_dirstate().save()
 
293
        self._inventory = None
 
294
        self._dirty = False
 
295
 
 
296
    @needs_tree_write_lock
 
297
    def _gather_kinds(self, files, kinds):
 
298
        """See MutableTree._gather_kinds."""
 
299
        for pos, f in enumerate(files):
 
300
            if kinds[pos] is None:
 
301
                kinds[pos] = self._kind(f)
 
302
 
 
303
    def _generate_inventory(self):
 
304
        """Create and set self.inventory from the dirstate object.
 
305
        
 
306
        This is relatively expensive: we have to walk the entire dirstate.
 
307
        Ideally we would not, and can deprecate this function.
 
308
        """
 
309
        #: uncomment to trap on inventory requests.
 
310
        # import pdb;pdb.set_trace()
 
311
        state = self.current_dirstate()
 
312
        state._read_dirblocks_if_needed()
 
313
        root_key, current_entry = self._get_entry(path='')
 
314
        current_id = root_key[2]
 
315
        if not (current_entry[0][0] == 'd'): # directory
 
316
            raise AssertionError(current_entry)
 
317
        inv = Inventory(root_id=current_id)
 
318
        # Turn some things into local variables
 
319
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
320
        factory = entry_factory
 
321
        utf8_decode = cache_utf8._utf8_decode
 
322
        inv_byid = inv._byid
 
323
        # we could do this straight out of the dirstate; it might be fast
 
324
        # and should be profiled - RBC 20070216
 
325
        parent_ies = {'' : inv.root}
 
326
        for block in state._dirblocks[1:]: # skip the root
 
327
            dirname = block[0]
 
328
            try:
 
329
                parent_ie = parent_ies[dirname]
 
330
            except KeyError:
 
331
                # all the paths in this block are not versioned in this tree
 
332
                continue
 
333
            for key, entry in block[1]:
 
334
                minikind, link_or_sha1, size, executable, stat = entry[0]
 
335
                if minikind in ('a', 'r'): # absent, relocated
 
336
                    # a parent tree only entry
 
337
                    continue
 
338
                name = key[1]
 
339
                name_unicode = utf8_decode(name)[0]
 
340
                file_id = key[2]
 
341
                kind = minikind_to_kind[minikind]
 
342
                inv_entry = factory[kind](file_id, name_unicode,
 
343
                                          parent_ie.file_id)
 
344
                if kind == 'file':
 
345
                    # This is only needed on win32, where this is the only way
 
346
                    # we know the executable bit.
 
347
                    inv_entry.executable = executable
 
348
                    # not strictly needed: working tree
 
349
                    #inv_entry.text_size = size
 
350
                    #inv_entry.text_sha1 = sha1
 
351
                elif kind == 'directory':
 
352
                    # add this entry to the parent map.
 
353
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
354
                elif kind == 'tree-reference':
 
355
                    if not self._repo_supports_tree_reference:
 
356
                        raise AssertionError(
 
357
                            "repository of %r "
 
358
                            "doesn't support tree references "
 
359
                            "required by entry %r"
 
360
                            % (self, name))
 
361
                    inv_entry.reference_revision = link_or_sha1 or None
 
362
                elif kind != 'symlink':
 
363
                    raise AssertionError("unknown kind %r" % kind)
 
364
                # These checks cost us around 40ms on a 55k entry tree
 
365
                if file_id in inv_byid:
 
366
                    raise AssertionError('file_id %s already in'
 
367
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
368
                if name_unicode in parent_ie.children:
 
369
                    raise AssertionError('name %r already in parent'
 
370
                        % (name_unicode,))
 
371
                inv_byid[file_id] = inv_entry
 
372
                parent_ie.children[name_unicode] = inv_entry
 
373
        self._inventory = inv
 
374
 
 
375
    def _get_entry(self, file_id=None, path=None):
 
376
        """Get the dirstate row for file_id or path.
 
377
 
 
378
        If either file_id or path is supplied, it is used as the key to lookup.
 
379
        If both are supplied, the fastest lookup is used, and an error is
 
380
        raised if they do not both point at the same row.
 
381
        
 
382
        :param file_id: An optional unicode file_id to be looked up.
 
383
        :param path: An optional unicode path to be looked up.
 
384
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
385
        """
 
386
        if file_id is None and path is None:
 
387
            raise errors.BzrError('must supply file_id or path')
 
388
        state = self.current_dirstate()
 
389
        if path is not None:
 
390
            path = path.encode('utf8')
 
391
        return state._get_entry(0, fileid_utf8=file_id, path_utf8=path)
 
392
 
 
393
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
394
        # check file id is valid unconditionally.
 
395
        entry = self._get_entry(file_id=file_id, path=path)
 
396
        if entry[0] is None:
 
397
            raise errors.NoSuchId(self, file_id)
 
398
        if path is None:
 
399
            path = pathjoin(entry[0][0], entry[0][1]).decode('utf8')
 
400
 
 
401
        file_abspath = self.abspath(path)
 
402
        state = self.current_dirstate()
 
403
        if stat_value is None:
 
404
            try:
 
405
                stat_value = os.lstat(file_abspath)
 
406
            except OSError, e:
 
407
                if e.errno == errno.ENOENT:
 
408
                    return None
 
409
                else:
 
410
                    raise
 
411
        link_or_sha1 = state.update_entry(entry, file_abspath,
 
412
                                          stat_value=stat_value)
 
413
        if entry[1][0][0] == 'f':
 
414
            return link_or_sha1
 
415
        return None
 
416
 
 
417
    def _get_inventory(self):
 
418
        """Get the inventory for the tree. This is only valid within a lock."""
 
419
        if 'evil' in debug.debug_flags:
 
420
            trace.mutter_callsite(2,
 
421
                "accessing .inventory forces a size of tree translation.")
 
422
        if self._inventory is not None:
 
423
            return self._inventory
 
424
        self._must_be_locked()
 
425
        self._generate_inventory()
 
426
        return self._inventory
 
427
 
 
428
    inventory = property(_get_inventory,
 
429
                         doc="Inventory of this Tree")
 
430
 
 
431
    @needs_read_lock
 
432
    def get_parent_ids(self):
 
433
        """See Tree.get_parent_ids.
 
434
        
 
435
        This implementation requests the ids list from the dirstate file.
 
436
        """
 
437
        return self.current_dirstate().get_parent_ids()
 
438
 
 
439
    def get_reference_revision(self, file_id, path=None):
 
440
        # referenced tree's revision is whatever's currently there
 
441
        return self.get_nested_tree(file_id, path).last_revision()
 
442
 
 
443
    def get_nested_tree(self, file_id, path=None):
 
444
        if path is None:
 
445
            path = self.id2path(file_id)
 
446
        # else: check file_id is at path?
 
447
        return WorkingTree.open(self.abspath(path))
 
448
 
 
449
    @needs_read_lock
 
450
    def get_root_id(self):
 
451
        """Return the id of this trees root"""
 
452
        return self._get_entry(path='')[0][2]
 
453
 
 
454
    def has_id(self, file_id):
 
455
        state = self.current_dirstate()
 
456
        row, parents = self._get_entry(file_id=file_id)
 
457
        if row is None:
 
458
            return False
 
459
        return osutils.lexists(pathjoin(
 
460
                    self.basedir, row[0].decode('utf8'), row[1].decode('utf8')))
 
461
 
 
462
    @needs_read_lock
 
463
    def id2path(self, file_id):
 
464
        "Convert a file-id to a path."
 
465
        state = self.current_dirstate()
 
466
        entry = self._get_entry(file_id=file_id)
 
467
        if entry == (None, None):
 
468
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
469
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
470
        return path_utf8.decode('utf8')
 
471
 
 
472
    def _is_executable_from_path_and_stat_from_basis(self, path, stat_result):
 
473
        entry = self._get_entry(path=path)
 
474
        if entry == (None, None):
 
475
            return False # Missing entries are not executable
 
476
        return entry[1][0][3] # Executable?
 
477
 
 
478
    if not osutils.supports_executable():
 
479
        def is_executable(self, file_id, path=None):
 
480
            """Test if a file is executable or not.
 
481
 
 
482
            Note: The caller is expected to take a read-lock before calling this.
 
483
            """
 
484
            entry = self._get_entry(file_id=file_id, path=path)
 
485
            if entry == (None, None):
 
486
                return False
 
487
            return entry[1][0][3]
 
488
 
 
489
        _is_executable_from_path_and_stat = \
 
490
            _is_executable_from_path_and_stat_from_basis
 
491
    else:
 
492
        def is_executable(self, file_id, path=None):
 
493
            """Test if a file is executable or not.
 
494
 
 
495
            Note: The caller is expected to take a read-lock before calling this.
 
496
            """
 
497
            self._must_be_locked()
 
498
            if not path:
 
499
                path = self.id2path(file_id)
 
500
            mode = os.lstat(self.abspath(path)).st_mode
 
501
            return bool(stat.S_ISREG(mode) and stat.S_IEXEC & mode)
 
502
 
 
503
    def all_file_ids(self):
 
504
        """See Tree.iter_all_file_ids"""
 
505
        self._must_be_locked()
 
506
        result = set()
 
507
        for key, tree_details in self.current_dirstate()._iter_entries():
 
508
            if tree_details[0][0] in ('a', 'r'): # relocated
 
509
                continue
 
510
            result.add(key[2])
 
511
        return result
 
512
 
 
513
    @needs_read_lock
 
514
    def __iter__(self):
 
515
        """Iterate through file_ids for this tree.
 
516
 
 
517
        file_ids are in a WorkingTree if they are in the working inventory
 
518
        and the working file exists.
 
519
        """
 
520
        result = []
 
521
        for key, tree_details in self.current_dirstate()._iter_entries():
 
522
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
523
                # not relevant to the working tree
 
524
                continue
 
525
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
526
            if osutils.lexists(path):
 
527
                result.append(key[2])
 
528
        return iter(result)
 
529
 
 
530
    def iter_references(self):
 
531
        for key, tree_details in self.current_dirstate()._iter_entries():
 
532
            if tree_details[0][0] in ('a', 'r'): # absent, relocated
 
533
                # not relevant to the working tree
 
534
                continue
 
535
            if not key[1]:
 
536
                # the root is not a reference.
 
537
                continue
 
538
            path = pathjoin(self.basedir, key[0].decode('utf8'), key[1].decode('utf8'))
 
539
            try:
 
540
                if self._kind(path) == 'tree-reference':
 
541
                    yield path, key[2]
 
542
            except errors.NoSuchFile:
 
543
                # path is missing on disk.
 
544
                continue
 
545
 
 
546
    def kind(self, file_id):
 
547
        """Return the kind of a file.
 
548
 
 
549
        This is always the actual kind that's on disk, regardless of what it
 
550
        was added as.
 
551
 
 
552
        Note: The caller is expected to take a read-lock before calling this.
 
553
        """
 
554
        relpath = self.id2path(file_id)
 
555
        if relpath is None:
 
556
            raise AssertionError(
 
557
                "path for id {%s} is None!" % file_id)
 
558
        return self._kind(relpath)
 
559
 
 
560
    def _kind(self, relpath):
 
561
        abspath = self.abspath(relpath)
 
562
        kind = file_kind(abspath)
 
563
        if (self._repo_supports_tree_reference and
 
564
            kind == 'directory' and
 
565
            self._directory_is_tree_reference(relpath)):
 
566
            kind = 'tree-reference'
 
567
        return kind
 
568
 
 
569
    @needs_read_lock
 
570
    def _last_revision(self):
 
571
        """See Mutable.last_revision."""
 
572
        parent_ids = self.current_dirstate().get_parent_ids()
 
573
        if parent_ids:
 
574
            return parent_ids[0]
 
575
        else:
 
576
            return _mod_revision.NULL_REVISION
 
577
 
 
578
    def lock_read(self):
 
579
        """See Branch.lock_read, and WorkingTree.unlock."""
 
580
        self.branch.lock_read()
 
581
        try:
 
582
            self._control_files.lock_read()
 
583
            try:
 
584
                state = self.current_dirstate()
 
585
                if not state._lock_token:
 
586
                    state.lock_read()
 
587
                # set our support for tree references from the repository in
 
588
                # use.
 
589
                self._repo_supports_tree_reference = getattr(
 
590
                    self.branch.repository._format, "supports_tree_reference",
 
591
                    False)
 
592
            except:
 
593
                self._control_files.unlock()
 
594
                raise
 
595
        except:
 
596
            self.branch.unlock()
 
597
            raise
 
598
 
 
599
    def _lock_self_write(self):
 
600
        """This should be called after the branch is locked."""
 
601
        try:
 
602
            self._control_files.lock_write()
 
603
            try:
 
604
                state = self.current_dirstate()
 
605
                if not state._lock_token:
 
606
                    state.lock_write()
 
607
                # set our support for tree references from the repository in
 
608
                # use.
 
609
                self._repo_supports_tree_reference = getattr(
 
610
                    self.branch.repository._format, "supports_tree_reference",
 
611
                    False)
 
612
            except:
 
613
                self._control_files.unlock()
 
614
                raise
 
615
        except:
 
616
            self.branch.unlock()
 
617
            raise
 
618
 
 
619
    def lock_tree_write(self):
 
620
        """See MutableTree.lock_tree_write, and WorkingTree.unlock."""
 
621
        self.branch.lock_read()
 
622
        self._lock_self_write()
 
623
 
 
624
    def lock_write(self):
 
625
        """See MutableTree.lock_write, and WorkingTree.unlock."""
 
626
        self.branch.lock_write()
 
627
        self._lock_self_write()
 
628
 
 
629
    @needs_tree_write_lock
 
630
    def move(self, from_paths, to_dir, after=False):
 
631
        """See WorkingTree.move()."""
 
632
        result = []
 
633
        if not from_paths:
 
634
            return result
 
635
        state = self.current_dirstate()
 
636
        if isinstance(from_paths, basestring):
 
637
            raise ValueError()
 
638
        to_dir_utf8 = to_dir.encode('utf8')
 
639
        to_entry_dirname, to_basename = os.path.split(to_dir_utf8)
 
640
        id_index = state._get_id_index()
 
641
        # check destination directory
 
642
        # get the details for it
 
643
        to_entry_block_index, to_entry_entry_index, dir_present, entry_present = \
 
644
            state._get_block_entry_index(to_entry_dirname, to_basename, 0)
 
645
        if not entry_present:
 
646
            raise errors.BzrMoveFailedError('', to_dir,
 
647
                errors.NotVersionedError(to_dir))
 
648
        to_entry = state._dirblocks[to_entry_block_index][1][to_entry_entry_index]
 
649
        # get a handle on the block itself.
 
650
        to_block_index = state._ensure_block(
 
651
            to_entry_block_index, to_entry_entry_index, to_dir_utf8)
 
652
        to_block = state._dirblocks[to_block_index]
 
653
        to_abs = self.abspath(to_dir)
 
654
        if not isdir(to_abs):
 
655
            raise errors.BzrMoveFailedError('',to_dir,
 
656
                errors.NotADirectory(to_abs))
 
657
 
 
658
        if to_entry[1][0][0] != 'd':
 
659
            raise errors.BzrMoveFailedError('',to_dir,
 
660
                errors.NotADirectory(to_abs))
 
661
 
 
662
        if self._inventory is not None:
 
663
            update_inventory = True
 
664
            inv = self.inventory
 
665
            to_dir_id = to_entry[0][2]
 
666
            to_dir_ie = inv[to_dir_id]
 
667
        else:
 
668
            update_inventory = False
 
669
 
 
670
        rollbacks = []
 
671
        def move_one(old_entry, from_path_utf8, minikind, executable,
 
672
                     fingerprint, packed_stat, size,
 
673
                     to_block, to_key, to_path_utf8):
 
674
            state._make_absent(old_entry)
 
675
            from_key = old_entry[0]
 
676
            rollbacks.append(
 
677
                lambda:state.update_minimal(from_key,
 
678
                    minikind,
 
679
                    executable=executable,
 
680
                    fingerprint=fingerprint,
 
681
                    packed_stat=packed_stat,
 
682
                    size=size,
 
683
                    path_utf8=from_path_utf8))
 
684
            state.update_minimal(to_key,
 
685
                    minikind,
 
686
                    executable=executable,
 
687
                    fingerprint=fingerprint,
 
688
                    packed_stat=packed_stat,
 
689
                    size=size,
 
690
                    path_utf8=to_path_utf8)
 
691
            added_entry_index, _ = state._find_entry_index(to_key, to_block[1])
 
692
            new_entry = to_block[1][added_entry_index]
 
693
            rollbacks.append(lambda:state._make_absent(new_entry))
 
694
 
 
695
        for from_rel in from_paths:
 
696
            # from_rel is 'pathinroot/foo/bar'
 
697
            from_rel_utf8 = from_rel.encode('utf8')
 
698
            from_dirname, from_tail = osutils.split(from_rel)
 
699
            from_dirname, from_tail_utf8 = osutils.split(from_rel_utf8)
 
700
            from_entry = self._get_entry(path=from_rel)
 
701
            if from_entry == (None, None):
 
702
                raise errors.BzrMoveFailedError(from_rel,to_dir,
 
703
                    errors.NotVersionedError(path=str(from_rel)))
 
704
 
 
705
            from_id = from_entry[0][2]
 
706
            to_rel = pathjoin(to_dir, from_tail)
 
707
            to_rel_utf8 = pathjoin(to_dir_utf8, from_tail_utf8)
 
708
            item_to_entry = self._get_entry(path=to_rel)
 
709
            if item_to_entry != (None, None):
 
710
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
711
                    "Target is already versioned.")
 
712
 
 
713
            if from_rel == to_rel:
 
714
                raise errors.BzrMoveFailedError(from_rel, to_rel,
 
715
                    "Source and target are identical.")
 
716
 
 
717
            from_missing = not self.has_filename(from_rel)
 
718
            to_missing = not self.has_filename(to_rel)
 
719
            if after:
 
720
                move_file = False
 
721
            else:
 
722
                move_file = True
 
723
            if to_missing:
 
724
                if not move_file:
 
725
                    raise errors.BzrMoveFailedError(from_rel, to_rel,
 
726
                        errors.NoSuchFile(path=to_rel,
 
727
                        extra="New file has not been created yet"))
 
728
                elif from_missing:
 
729
                    # neither path exists
 
730
                    raise errors.BzrRenameFailedError(from_rel, to_rel,
 
731
                        errors.PathsDoNotExist(paths=(from_rel, to_rel)))
 
732
            else:
 
733
                if from_missing: # implicitly just update our path mapping
 
734
                    move_file = False
 
735
                elif not after:
 
736
                    raise errors.RenameFailedFilesExist(from_rel, to_rel)
 
737
 
 
738
            rollbacks = []
 
739
            def rollback_rename():
 
740
                """A single rename has failed, roll it back."""
 
741
                # roll back everything, even if we encounter trouble doing one
 
742
                # of them.
 
743
                #
 
744
                # TODO: at least log the other exceptions rather than just
 
745
                # losing them mbp 20070307
 
746
                exc_info = None
 
747
                for rollback in reversed(rollbacks):
 
748
                    try:
 
749
                        rollback()
 
750
                    except Exception, e:
 
751
                        exc_info = sys.exc_info()
 
752
                if exc_info:
 
753
                    raise exc_info[0], exc_info[1], exc_info[2]
 
754
 
 
755
            # perform the disk move first - its the most likely failure point.
 
756
            if move_file:
 
757
                from_rel_abs = self.abspath(from_rel)
 
758
                to_rel_abs = self.abspath(to_rel)
 
759
                try:
 
760
                    osutils.rename(from_rel_abs, to_rel_abs)
 
761
                except OSError, e:
 
762
                    raise errors.BzrMoveFailedError(from_rel, to_rel, e[1])
 
763
                rollbacks.append(lambda: osutils.rename(to_rel_abs, from_rel_abs))
 
764
            try:
 
765
                # perform the rename in the inventory next if needed: its easy
 
766
                # to rollback
 
767
                if update_inventory:
 
768
                    # rename the entry
 
769
                    from_entry = inv[from_id]
 
770
                    current_parent = from_entry.parent_id
 
771
                    inv.rename(from_id, to_dir_id, from_tail)
 
772
                    rollbacks.append(
 
773
                        lambda: inv.rename(from_id, current_parent, from_tail))
 
774
                # finally do the rename in the dirstate, which is a little
 
775
                # tricky to rollback, but least likely to need it.
 
776
                old_block_index, old_entry_index, dir_present, file_present = \
 
777
                    state._get_block_entry_index(from_dirname, from_tail_utf8, 0)
 
778
                old_block = state._dirblocks[old_block_index][1]
 
779
                old_entry = old_block[old_entry_index]
 
780
                from_key, old_entry_details = old_entry
 
781
                cur_details = old_entry_details[0]
 
782
                # remove the old row
 
783
                to_key = ((to_block[0],) + from_key[1:3])
 
784
                minikind = cur_details[0]
 
785
                move_one(old_entry, from_path_utf8=from_rel_utf8,
 
786
                         minikind=minikind,
 
787
                         executable=cur_details[3],
 
788
                         fingerprint=cur_details[1],
 
789
                         packed_stat=cur_details[4],
 
790
                         size=cur_details[2],
 
791
                         to_block=to_block,
 
792
                         to_key=to_key,
 
793
                         to_path_utf8=to_rel_utf8)
 
794
 
 
795
                if minikind == 'd':
 
796
                    def update_dirblock(from_dir, to_key, to_dir_utf8):
 
797
                        """Recursively update all entries in this dirblock."""
 
798
                        if from_dir == '':
 
799
                            raise AssertionError("renaming root not supported")
 
800
                        from_key = (from_dir, '')
 
801
                        from_block_idx, present = \
 
802
                            state._find_block_index_from_key(from_key)
 
803
                        if not present:
 
804
                            # This is the old record, if it isn't present, then
 
805
                            # there is theoretically nothing to update.
 
806
                            # (Unless it isn't present because of lazy loading,
 
807
                            # but we don't do that yet)
 
808
                            return
 
809
                        from_block = state._dirblocks[from_block_idx]
 
810
                        to_block_index, to_entry_index, _, _ = \
 
811
                            state._get_block_entry_index(to_key[0], to_key[1], 0)
 
812
                        to_block_index = state._ensure_block(
 
813
                            to_block_index, to_entry_index, to_dir_utf8)
 
814
                        to_block = state._dirblocks[to_block_index]
 
815
 
 
816
                        # Grab a copy since move_one may update the list.
 
817
                        for entry in from_block[1][:]:
 
818
                            if not (entry[0][0] == from_dir):
 
819
                                raise AssertionError()
 
820
                            cur_details = entry[1][0]
 
821
                            to_key = (to_dir_utf8, entry[0][1], entry[0][2])
 
822
                            from_path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
823
                            to_path_utf8 = osutils.pathjoin(to_dir_utf8, entry[0][1])
 
824
                            minikind = cur_details[0]
 
825
                            if minikind in 'ar':
 
826
                                # Deleted children of a renamed directory
 
827
                                # Do not need to be updated.
 
828
                                # Children that have been renamed out of this
 
829
                                # directory should also not be updated
 
830
                                continue
 
831
                            move_one(entry, from_path_utf8=from_path_utf8,
 
832
                                     minikind=minikind,
 
833
                                     executable=cur_details[3],
 
834
                                     fingerprint=cur_details[1],
 
835
                                     packed_stat=cur_details[4],
 
836
                                     size=cur_details[2],
 
837
                                     to_block=to_block,
 
838
                                     to_key=to_key,
 
839
                                     to_path_utf8=to_path_utf8)
 
840
                            if minikind == 'd':
 
841
                                # We need to move all the children of this
 
842
                                # entry
 
843
                                update_dirblock(from_path_utf8, to_key,
 
844
                                                to_path_utf8)
 
845
                    update_dirblock(from_rel_utf8, to_key, to_rel_utf8)
 
846
            except:
 
847
                rollback_rename()
 
848
                raise
 
849
            result.append((from_rel, to_rel))
 
850
            state._dirblock_state = dirstate.DirState.IN_MEMORY_MODIFIED
 
851
            self._make_dirty(reset_inventory=False)
 
852
 
 
853
        return result
 
854
 
 
855
    def _must_be_locked(self):
 
856
        if not self._control_files._lock_count:
 
857
            raise errors.ObjectNotLocked(self)
 
858
 
 
859
    def _new_tree(self):
 
860
        """Initialize the state in this tree to be a new tree."""
 
861
        self._dirty = True
 
862
 
 
863
    @needs_read_lock
 
864
    def path2id(self, path):
 
865
        """Return the id for path in this tree."""
 
866
        path = path.strip('/')
 
867
        entry = self._get_entry(path=path)
 
868
        if entry == (None, None):
 
869
            return None
 
870
        return entry[0][2]
 
871
 
 
872
    def paths2ids(self, paths, trees=[], require_versioned=True):
 
873
        """See Tree.paths2ids().
 
874
 
 
875
        This specialisation fast-paths the case where all the trees are in the
 
876
        dirstate.
 
877
        """
 
878
        if paths is None:
 
879
            return None
 
880
        parents = self.get_parent_ids()
 
881
        for tree in trees:
 
882
            if not (isinstance(tree, DirStateRevisionTree) and tree._revision_id in
 
883
                parents):
 
884
                return super(WorkingTree4, self).paths2ids(paths, trees, require_versioned)
 
885
        search_indexes = [0] + [1 + parents.index(tree._revision_id) for tree in trees]
 
886
        # -- make all paths utf8 --
 
887
        paths_utf8 = set()
 
888
        for path in paths:
 
889
            paths_utf8.add(path.encode('utf8'))
 
890
        paths = paths_utf8
 
891
        # -- paths is now a utf8 path set --
 
892
        # -- get the state object and prepare it.
 
893
        state = self.current_dirstate()
 
894
        if False and (state._dirblock_state == dirstate.DirState.NOT_IN_MEMORY
 
895
            and '' not in paths):
 
896
            paths2ids = self._paths2ids_using_bisect
 
897
        else:
 
898
            paths2ids = self._paths2ids_in_memory
 
899
        return paths2ids(paths, search_indexes,
 
900
                         require_versioned=require_versioned)
 
901
 
 
902
    def _paths2ids_in_memory(self, paths, search_indexes,
 
903
                             require_versioned=True):
 
904
        state = self.current_dirstate()
 
905
        state._read_dirblocks_if_needed()
 
906
        def _entries_for_path(path):
 
907
            """Return a list with all the entries that match path for all ids.
 
908
            """
 
909
            dirname, basename = os.path.split(path)
 
910
            key = (dirname, basename, '')
 
911
            block_index, present = state._find_block_index_from_key(key)
 
912
            if not present:
 
913
                # the block which should contain path is absent.
 
914
                return []
 
915
            result = []
 
916
            block = state._dirblocks[block_index][1]
 
917
            entry_index, _ = state._find_entry_index(key, block)
 
918
            # we may need to look at multiple entries at this path: walk while the paths match.
 
919
            while (entry_index < len(block) and
 
920
                block[entry_index][0][0:2] == key[0:2]):
 
921
                result.append(block[entry_index])
 
922
                entry_index += 1
 
923
            return result
 
924
        if require_versioned:
 
925
            # -- check all supplied paths are versioned in a search tree. --
 
926
            all_versioned = True
 
927
            for path in paths:
 
928
                path_entries = _entries_for_path(path)
 
929
                if not path_entries:
 
930
                    # this specified path is not present at all: error
 
931
                    all_versioned = False
 
932
                    break
 
933
                found_versioned = False
 
934
                # for each id at this path
 
935
                for entry in path_entries:
 
936
                    # for each tree.
 
937
                    for index in search_indexes:
 
938
                        if entry[1][index][0] != 'a': # absent
 
939
                            found_versioned = True
 
940
                            # all good: found a versioned cell
 
941
                            break
 
942
                if not found_versioned:
 
943
                    # none of the indexes was not 'absent' at all ids for this
 
944
                    # path.
 
945
                    all_versioned = False
 
946
                    break
 
947
            if not all_versioned:
 
948
                raise errors.PathsNotVersionedError(paths)
 
949
        # -- remove redundancy in supplied paths to prevent over-scanning --
 
950
        search_paths = osutils.minimum_path_selection(paths)
 
951
        # sketch: 
 
952
        # for all search_indexs in each path at or under each element of
 
953
        # search_paths, if the detail is relocated: add the id, and add the
 
954
        # relocated path as one to search if its not searched already. If the
 
955
        # detail is not relocated, add the id.
 
956
        searched_paths = set()
 
957
        found_ids = set()
 
958
        def _process_entry(entry):
 
959
            """Look at search_indexes within entry.
 
960
 
 
961
            If a specific tree's details are relocated, add the relocation
 
962
            target to search_paths if not searched already. If it is absent, do
 
963
            nothing. Otherwise add the id to found_ids.
 
964
            """
 
965
            for index in search_indexes:
 
966
                if entry[1][index][0] == 'r': # relocated
 
967
                    if not osutils.is_inside_any(searched_paths, entry[1][index][1]):
 
968
                        search_paths.add(entry[1][index][1])
 
969
                elif entry[1][index][0] != 'a': # absent
 
970
                    found_ids.add(entry[0][2])
 
971
        while search_paths:
 
972
            current_root = search_paths.pop()
 
973
            searched_paths.add(current_root)
 
974
            # process the entries for this containing directory: the rest will be
 
975
            # found by their parents recursively.
 
976
            root_entries = _entries_for_path(current_root)
 
977
            if not root_entries:
 
978
                # this specified path is not present at all, skip it.
 
979
                continue
 
980
            for entry in root_entries:
 
981
                _process_entry(entry)
 
982
            initial_key = (current_root, '', '')
 
983
            block_index, _ = state._find_block_index_from_key(initial_key)
 
984
            while (block_index < len(state._dirblocks) and
 
985
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
986
                for entry in state._dirblocks[block_index][1]:
 
987
                    _process_entry(entry)
 
988
                block_index += 1
 
989
        return found_ids
 
990
 
 
991
    def _paths2ids_using_bisect(self, paths, search_indexes,
 
992
                                require_versioned=True):
 
993
        state = self.current_dirstate()
 
994
        found_ids = set()
 
995
 
 
996
        split_paths = sorted(osutils.split(p) for p in paths)
 
997
        found = state._bisect_recursive(split_paths)
 
998
 
 
999
        if require_versioned:
 
1000
            found_dir_names = set(dir_name_id[:2] for dir_name_id in found)
 
1001
            for dir_name in split_paths:
 
1002
                if dir_name not in found_dir_names:
 
1003
                    raise errors.PathsNotVersionedError(paths)
 
1004
 
 
1005
        for dir_name_id, trees_info in found.iteritems():
 
1006
            for index in search_indexes:
 
1007
                if trees_info[index][0] not in ('r', 'a'):
 
1008
                    found_ids.add(dir_name_id[2])
 
1009
        return found_ids
 
1010
 
 
1011
    def read_working_inventory(self):
 
1012
        """Read the working inventory.
 
1013
        
 
1014
        This is a meaningless operation for dirstate, but we obey it anyhow.
 
1015
        """
 
1016
        return self.inventory
 
1017
 
 
1018
    @needs_read_lock
 
1019
    def revision_tree(self, revision_id):
 
1020
        """See Tree.revision_tree.
 
1021
 
 
1022
        WorkingTree4 supplies revision_trees for any basis tree.
 
1023
        """
 
1024
        dirstate = self.current_dirstate()
 
1025
        parent_ids = dirstate.get_parent_ids()
 
1026
        if revision_id not in parent_ids:
 
1027
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1028
        if revision_id in dirstate.get_ghosts():
 
1029
            raise errors.NoSuchRevisionInTree(self, revision_id)
 
1030
        return DirStateRevisionTree(dirstate, revision_id,
 
1031
            self.branch.repository)
 
1032
 
 
1033
    @needs_tree_write_lock
 
1034
    def set_last_revision(self, new_revision):
 
1035
        """Change the last revision in the working tree."""
 
1036
        parents = self.get_parent_ids()
 
1037
        if new_revision in (NULL_REVISION, None):
 
1038
            if len(parents) >= 2:
 
1039
                raise AssertionError(
 
1040
                    "setting the last parent to none with a pending merge is "
 
1041
                    "unsupported.")
 
1042
            self.set_parent_ids([])
 
1043
        else:
 
1044
            self.set_parent_ids([new_revision] + parents[1:],
 
1045
                allow_leftmost_as_ghost=True)
 
1046
 
 
1047
    @needs_tree_write_lock
 
1048
    def set_parent_ids(self, revision_ids, allow_leftmost_as_ghost=False):
 
1049
        """Set the parent ids to revision_ids.
 
1050
        
 
1051
        See also set_parent_trees. This api will try to retrieve the tree data
 
1052
        for each element of revision_ids from the trees repository. If you have
 
1053
        tree data already available, it is more efficient to use
 
1054
        set_parent_trees rather than set_parent_ids. set_parent_ids is however
 
1055
        an easier API to use.
 
1056
 
 
1057
        :param revision_ids: The revision_ids to set as the parent ids of this
 
1058
            working tree. Any of these may be ghosts.
 
1059
        """
 
1060
        trees = []
 
1061
        for revision_id in revision_ids:
 
1062
            try:
 
1063
                revtree = self.branch.repository.revision_tree(revision_id)
 
1064
                # TODO: jam 20070213 KnitVersionedFile raises
 
1065
                #       RevisionNotPresent rather than NoSuchRevision if a
 
1066
                #       given revision_id is not present. Should Repository be
 
1067
                #       catching it and re-raising NoSuchRevision?
 
1068
            except (errors.NoSuchRevision, errors.RevisionNotPresent):
 
1069
                revtree = None
 
1070
            trees.append((revision_id, revtree))
 
1071
        self.set_parent_trees(trees,
 
1072
            allow_leftmost_as_ghost=allow_leftmost_as_ghost)
 
1073
 
 
1074
    @needs_tree_write_lock
 
1075
    def set_parent_trees(self, parents_list, allow_leftmost_as_ghost=False):
 
1076
        """Set the parents of the working tree.
 
1077
 
 
1078
        :param parents_list: A list of (revision_id, tree) tuples.
 
1079
            If tree is None, then that element is treated as an unreachable
 
1080
            parent tree - i.e. a ghost.
 
1081
        """
 
1082
        dirstate = self.current_dirstate()
 
1083
        if len(parents_list) > 0:
 
1084
            if not allow_leftmost_as_ghost and parents_list[0][1] is None:
 
1085
                raise errors.GhostRevisionUnusableHere(parents_list[0][0])
 
1086
        real_trees = []
 
1087
        ghosts = []
 
1088
        # convert absent trees to the null tree, which we convert back to
 
1089
        # missing on access.
 
1090
        for rev_id, tree in parents_list:
 
1091
            _mod_revision.check_not_reserved_id(rev_id)
 
1092
            if tree is not None:
 
1093
                real_trees.append((rev_id, tree))
 
1094
            else:
 
1095
                real_trees.append((rev_id,
 
1096
                    self.branch.repository.revision_tree(None)))
 
1097
                ghosts.append(rev_id)
 
1098
        dirstate.set_parent_trees(real_trees, ghosts=ghosts)
 
1099
        self._make_dirty(reset_inventory=False)
 
1100
 
 
1101
    def _set_root_id(self, file_id):
 
1102
        """See WorkingTree.set_root_id."""
 
1103
        state = self.current_dirstate()
 
1104
        state.set_path_id('', file_id)
 
1105
        if state._dirblock_state == dirstate.DirState.IN_MEMORY_MODIFIED:
 
1106
            self._make_dirty(reset_inventory=True)
 
1107
 
 
1108
    def _sha_from_stat(self, path, stat_result):
 
1109
        """Get a sha digest from the tree's stat cache.
 
1110
 
 
1111
        The default implementation assumes no stat cache is present.
 
1112
 
 
1113
        :param path: The path.
 
1114
        :param stat_result: The stat result being looked up.
 
1115
        """
 
1116
        return self.current_dirstate().sha1_from_stat(path, stat_result)
 
1117
 
 
1118
    @needs_read_lock
 
1119
    def supports_tree_reference(self):
 
1120
        return self._repo_supports_tree_reference
 
1121
 
 
1122
    def unlock(self):
 
1123
        """Unlock in format 4 trees needs to write the entire dirstate."""
 
1124
        # do non-implementation specific cleanup
 
1125
        self._cleanup()
 
1126
 
 
1127
        if self._control_files._lock_count == 1:
 
1128
            # eventually we should do signature checking during read locks for
 
1129
            # dirstate updates.
 
1130
            if self._control_files._lock_mode == 'w':
 
1131
                if self._dirty:
 
1132
                    self.flush()
 
1133
            if self._dirstate is not None:
 
1134
                # This is a no-op if there are no modifications.
 
1135
                self._dirstate.save()
 
1136
                self._dirstate.unlock()
 
1137
            # TODO: jam 20070301 We shouldn't have to wipe the dirstate at this
 
1138
            #       point. Instead, it could check if the header has been
 
1139
            #       modified when it is locked, and if not, it can hang on to
 
1140
            #       the data it has in memory.
 
1141
            self._dirstate = None
 
1142
            self._inventory = None
 
1143
        # reverse order of locking.
 
1144
        try:
 
1145
            return self._control_files.unlock()
 
1146
        finally:
 
1147
            self.branch.unlock()
 
1148
 
 
1149
    @needs_tree_write_lock
 
1150
    def unversion(self, file_ids):
 
1151
        """Remove the file ids in file_ids from the current versioned set.
 
1152
 
 
1153
        When a file_id is unversioned, all of its children are automatically
 
1154
        unversioned.
 
1155
 
 
1156
        :param file_ids: The file ids to stop versioning.
 
1157
        :raises: NoSuchId if any fileid is not currently versioned.
 
1158
        """
 
1159
        if not file_ids:
 
1160
            return
 
1161
        state = self.current_dirstate()
 
1162
        state._read_dirblocks_if_needed()
 
1163
        ids_to_unversion = set(file_ids)
 
1164
        paths_to_unversion = set()
 
1165
        # sketch:
 
1166
        # check if the root is to be unversioned, if so, assert for now.
 
1167
        # walk the state marking unversioned things as absent.
 
1168
        # if there are any un-unversioned ids at the end, raise
 
1169
        for key, details in state._dirblocks[0][1]:
 
1170
            if (details[0][0] not in ('a', 'r') and # absent or relocated
 
1171
                key[2] in ids_to_unversion):
 
1172
                # I haven't written the code to unversion / yet - it should be
 
1173
                # supported.
 
1174
                raise errors.BzrError('Unversioning the / is not currently supported')
 
1175
        block_index = 0
 
1176
        while block_index < len(state._dirblocks):
 
1177
            # process one directory at a time.
 
1178
            block = state._dirblocks[block_index]
 
1179
            # first check: is the path one to remove - it or its children
 
1180
            delete_block = False
 
1181
            for path in paths_to_unversion:
 
1182
                if (block[0].startswith(path) and
 
1183
                    (len(block[0]) == len(path) or
 
1184
                     block[0][len(path)] == '/')):
 
1185
                    # this entire block should be deleted - its the block for a
 
1186
                    # path to unversion; or the child of one
 
1187
                    delete_block = True
 
1188
                    break
 
1189
            # TODO: trim paths_to_unversion as we pass by paths
 
1190
            if delete_block:
 
1191
                # this block is to be deleted: process it.
 
1192
                # TODO: we can special case the no-parents case and
 
1193
                # just forget the whole block.
 
1194
                entry_index = 0
 
1195
                while entry_index < len(block[1]):
 
1196
                    # Mark this file id as having been removed
 
1197
                    entry = block[1][entry_index]
 
1198
                    ids_to_unversion.discard(entry[0][2])
 
1199
                    if (entry[1][0][0] in 'ar' # don't remove absent or renamed
 
1200
                                               # entries
 
1201
                        or not state._make_absent(entry)):
 
1202
                        entry_index += 1
 
1203
                # go to the next block. (At the moment we dont delete empty
 
1204
                # dirblocks)
 
1205
                block_index += 1
 
1206
                continue
 
1207
            entry_index = 0
 
1208
            while entry_index < len(block[1]):
 
1209
                entry = block[1][entry_index]
 
1210
                if (entry[1][0][0] in ('a', 'r') or # absent, relocated
 
1211
                    # ^ some parent row.
 
1212
                    entry[0][2] not in ids_to_unversion):
 
1213
                    # ^ not an id to unversion
 
1214
                    entry_index += 1
 
1215
                    continue
 
1216
                if entry[1][0][0] == 'd':
 
1217
                    paths_to_unversion.add(pathjoin(entry[0][0], entry[0][1]))
 
1218
                if not state._make_absent(entry):
 
1219
                    entry_index += 1
 
1220
                # we have unversioned this id
 
1221
                ids_to_unversion.remove(entry[0][2])
 
1222
            block_index += 1
 
1223
        if ids_to_unversion:
 
1224
            raise errors.NoSuchId(self, iter(ids_to_unversion).next())
 
1225
        self._make_dirty(reset_inventory=False)
 
1226
        # have to change the legacy inventory too.
 
1227
        if self._inventory is not None:
 
1228
            for file_id in file_ids:
 
1229
                self._inventory.remove_recursive_id(file_id)
 
1230
 
 
1231
    @needs_tree_write_lock
 
1232
    def rename_one(self, from_rel, to_rel, after=False):
 
1233
        """See WorkingTree.rename_one"""
 
1234
        self.flush()
 
1235
        WorkingTree.rename_one(self, from_rel, to_rel, after)
 
1236
 
 
1237
    @needs_tree_write_lock
 
1238
    def apply_inventory_delta(self, changes):
 
1239
        """See MutableTree.apply_inventory_delta"""
 
1240
        state = self.current_dirstate()
 
1241
        state.update_by_delta(changes)
 
1242
        self._make_dirty(reset_inventory=True)
 
1243
 
 
1244
    def update_basis_by_delta(self, new_revid, delta):
 
1245
        """See MutableTree.update_basis_by_delta."""
 
1246
        if self.last_revision() == new_revid:
 
1247
            raise AssertionError()
 
1248
        self.current_dirstate().update_basis_by_delta(delta, new_revid)
 
1249
 
 
1250
    @needs_read_lock
 
1251
    def _validate(self):
 
1252
        self._dirstate._validate()
 
1253
 
 
1254
    @needs_tree_write_lock
 
1255
    def _write_inventory(self, inv):
 
1256
        """Write inventory as the current inventory."""
 
1257
        if self._dirty:
 
1258
            raise AssertionError("attempting to write an inventory when the "
 
1259
                "dirstate is dirty will lose pending changes")
 
1260
        self.current_dirstate().set_state_from_inventory(inv)
 
1261
        self._make_dirty(reset_inventory=False)
 
1262
        if self._inventory is not None:
 
1263
            self._inventory = inv
 
1264
        self.flush()
 
1265
 
 
1266
 
 
1267
class WorkingTreeFormat4(WorkingTreeFormat3):
 
1268
    """The first consolidated dirstate working tree format.
 
1269
 
 
1270
    This format:
 
1271
        - exists within a metadir controlling .bzr
 
1272
        - includes an explicit version marker for the workingtree control
 
1273
          files, separate from the BzrDir format
 
1274
        - modifies the hash cache format
 
1275
        - is new in bzr 0.15
 
1276
        - uses a LockDir to guard access to it.
 
1277
    """
 
1278
 
 
1279
    upgrade_recommended = False
 
1280
 
 
1281
    def get_format_string(self):
 
1282
        """See WorkingTreeFormat.get_format_string()."""
 
1283
        return "Bazaar Working Tree Format 4 (bzr 0.15)\n"
 
1284
 
 
1285
    def get_format_description(self):
 
1286
        """See WorkingTreeFormat.get_format_description()."""
 
1287
        return "Working tree format 4"
 
1288
 
 
1289
    def initialize(self, a_bzrdir, revision_id=None, from_branch=None,
 
1290
                   accelerator_tree=None, hardlink=False):
 
1291
        """See WorkingTreeFormat.initialize().
 
1292
 
 
1293
        :param revision_id: allows creating a working tree at a different
 
1294
        revision than the branch is at.
 
1295
        :param accelerator_tree: A tree which can be used for retrieving file
 
1296
            contents more quickly than the revision tree, i.e. a workingtree.
 
1297
            The revision tree will be used for cases where accelerator_tree's
 
1298
            content is different.
 
1299
        :param hardlink: If true, hard-link files from accelerator_tree,
 
1300
            where possible.
 
1301
 
 
1302
        These trees get an initial random root id, if their repository supports
 
1303
        rich root data, TREE_ROOT otherwise.
 
1304
        """
 
1305
        if not isinstance(a_bzrdir.transport, LocalTransport):
 
1306
            raise errors.NotLocalUrl(a_bzrdir.transport.base)
 
1307
        transport = a_bzrdir.get_workingtree_transport(self)
 
1308
        control_files = self._open_control_files(a_bzrdir)
 
1309
        control_files.create_lock()
 
1310
        control_files.lock_write()
 
1311
        transport.put_bytes('format', self.get_format_string(),
 
1312
            mode=control_files._file_mode)
 
1313
        if from_branch is not None:
 
1314
            branch = from_branch
 
1315
        else:
 
1316
            branch = a_bzrdir.open_branch()
 
1317
        if revision_id is None:
 
1318
            revision_id = branch.last_revision()
 
1319
        local_path = transport.local_abspath('dirstate')
 
1320
        # write out new dirstate (must exist when we create the tree)
 
1321
        state = dirstate.DirState.initialize(local_path)
 
1322
        state.unlock()
 
1323
        del state
 
1324
        wt = WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1325
                         branch,
 
1326
                         _format=self,
 
1327
                         _bzrdir=a_bzrdir,
 
1328
                         _control_files=control_files)
 
1329
        wt._new_tree()
 
1330
        wt.lock_tree_write()
 
1331
        try:
 
1332
            if revision_id in (None, NULL_REVISION):
 
1333
                if branch.repository.supports_rich_root():
 
1334
                    wt._set_root_id(generate_ids.gen_root_id())
 
1335
                else:
 
1336
                    wt._set_root_id(ROOT_ID)
 
1337
                wt.flush()
 
1338
            basis = None
 
1339
            # frequently, we will get here due to branching.  The accelerator
 
1340
            # tree will be the tree from the branch, so the desired basis
 
1341
            # tree will often be a parent of the accelerator tree.
 
1342
            if accelerator_tree is not None:
 
1343
                try:
 
1344
                    basis = accelerator_tree.revision_tree(revision_id)
 
1345
                except errors.NoSuchRevision:
 
1346
                    pass
 
1347
            if basis is None:
 
1348
                basis = branch.repository.revision_tree(revision_id)
 
1349
            if revision_id == NULL_REVISION:
 
1350
                parents_list = []
 
1351
            else:
 
1352
                parents_list = [(revision_id, basis)]
 
1353
            basis.lock_read()
 
1354
            try:
 
1355
                wt.set_parent_trees(parents_list, allow_leftmost_as_ghost=True)
 
1356
                wt.flush()
 
1357
                # if the basis has a root id we have to use that; otherwise we
 
1358
                # use a new random one
 
1359
                basis_root_id = basis.get_root_id()
 
1360
                if basis_root_id is not None:
 
1361
                    wt._set_root_id(basis_root_id)
 
1362
                    wt.flush()
 
1363
                transform.build_tree(basis, wt, accelerator_tree,
 
1364
                                     hardlink=hardlink)
 
1365
            finally:
 
1366
                basis.unlock()
 
1367
        finally:
 
1368
            control_files.unlock()
 
1369
            wt.unlock()
 
1370
        return wt
 
1371
 
 
1372
    def _open(self, a_bzrdir, control_files):
 
1373
        """Open the tree itself.
 
1374
 
 
1375
        :param a_bzrdir: the dir for the tree.
 
1376
        :param control_files: the control files for the tree.
 
1377
        """
 
1378
        return WorkingTree4(a_bzrdir.root_transport.local_abspath('.'),
 
1379
                           branch=a_bzrdir.open_branch(),
 
1380
                           _format=self,
 
1381
                           _bzrdir=a_bzrdir,
 
1382
                           _control_files=control_files)
 
1383
 
 
1384
    def __get_matchingbzrdir(self):
 
1385
        # please test against something that will let us do tree references
 
1386
        return bzrdir.format_registry.make_bzrdir(
 
1387
            'dirstate-with-subtree')
 
1388
 
 
1389
    _matchingbzrdir = property(__get_matchingbzrdir)
 
1390
 
 
1391
 
 
1392
class DirStateRevisionTree(Tree):
 
1393
    """A revision tree pulling the inventory from a dirstate."""
 
1394
 
 
1395
    def __init__(self, dirstate, revision_id, repository):
 
1396
        self._dirstate = dirstate
 
1397
        self._revision_id = revision_id
 
1398
        self._repository = repository
 
1399
        self._inventory = None
 
1400
        self._locked = 0
 
1401
        self._dirstate_locked = False
 
1402
 
 
1403
    def __repr__(self):
 
1404
        return "<%s of %s in %s>" % \
 
1405
            (self.__class__.__name__, self._revision_id, self._dirstate)
 
1406
 
 
1407
    def annotate_iter(self, file_id,
 
1408
                      default_revision=_mod_revision.CURRENT_REVISION):
 
1409
        """See Tree.annotate_iter"""
 
1410
        w = self._get_weave(file_id)
 
1411
        return w.annotate(self.inventory[file_id].revision)
 
1412
 
 
1413
    def _get_ancestors(self, default_revision):
 
1414
        return set(self._repository.get_ancestry(self._revision_id,
 
1415
                                                 topo_sorted=False))
 
1416
    def _comparison_data(self, entry, path):
 
1417
        """See Tree._comparison_data."""
 
1418
        if entry is None:
 
1419
            return None, False, None
 
1420
        # trust the entry as RevisionTree does, but this may not be
 
1421
        # sensible: the entry might not have come from us?
 
1422
        return entry.kind, entry.executable, None
 
1423
 
 
1424
    def _file_size(self, entry, stat_value):
 
1425
        return entry.text_size
 
1426
 
 
1427
    def filter_unversioned_files(self, paths):
 
1428
        """Filter out paths that are not versioned.
 
1429
 
 
1430
        :return: set of paths.
 
1431
        """
 
1432
        pred = self.has_filename
 
1433
        return set((p for p in paths if not pred(p)))
 
1434
 
 
1435
    def get_root_id(self):
 
1436
        return self.path2id('')
 
1437
 
 
1438
    def id2path(self, file_id):
 
1439
        "Convert a file-id to a path."
 
1440
        entry = self._get_entry(file_id=file_id)
 
1441
        if entry == (None, None):
 
1442
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1443
        path_utf8 = osutils.pathjoin(entry[0][0], entry[0][1])
 
1444
        return path_utf8.decode('utf8')
 
1445
 
 
1446
    def _get_parent_index(self):
 
1447
        """Return the index in the dirstate referenced by this tree."""
 
1448
        return self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1449
 
 
1450
    def _get_entry(self, file_id=None, path=None):
 
1451
        """Get the dirstate row for file_id or path.
 
1452
 
 
1453
        If either file_id or path is supplied, it is used as the key to lookup.
 
1454
        If both are supplied, the fastest lookup is used, and an error is
 
1455
        raised if they do not both point at the same row.
 
1456
        
 
1457
        :param file_id: An optional unicode file_id to be looked up.
 
1458
        :param path: An optional unicode path to be looked up.
 
1459
        :return: The dirstate row tuple for path/file_id, or (None, None)
 
1460
        """
 
1461
        if file_id is None and path is None:
 
1462
            raise errors.BzrError('must supply file_id or path')
 
1463
        if path is not None:
 
1464
            path = path.encode('utf8')
 
1465
        parent_index = self._get_parent_index()
 
1466
        return self._dirstate._get_entry(parent_index, fileid_utf8=file_id, path_utf8=path)
 
1467
 
 
1468
    def _generate_inventory(self):
 
1469
        """Create and set self.inventory from the dirstate object.
 
1470
 
 
1471
        (So this is only called the first time the inventory is requested for
 
1472
        this tree; it then remains in memory until it's out of date.)
 
1473
 
 
1474
        This is relatively expensive: we have to walk the entire dirstate.
 
1475
        """
 
1476
        if not self._locked:
 
1477
            raise AssertionError(
 
1478
                'cannot generate inventory of an unlocked '
 
1479
                'dirstate revision tree')
 
1480
        # separate call for profiling - makes it clear where the costs are.
 
1481
        self._dirstate._read_dirblocks_if_needed()
 
1482
        if self._revision_id not in self._dirstate.get_parent_ids():
 
1483
            raise AssertionError(
 
1484
                'parent %s has disappeared from %s' % (
 
1485
                self._revision_id, self._dirstate.get_parent_ids()))
 
1486
        parent_index = self._dirstate.get_parent_ids().index(self._revision_id) + 1
 
1487
        # This is identical now to the WorkingTree _generate_inventory except
 
1488
        # for the tree index use.
 
1489
        root_key, current_entry = self._dirstate._get_entry(parent_index, path_utf8='')
 
1490
        current_id = root_key[2]
 
1491
        if current_entry[parent_index][0] != 'd':
 
1492
            raise AssertionError()
 
1493
        inv = Inventory(root_id=current_id, revision_id=self._revision_id)
 
1494
        inv.root.revision = current_entry[parent_index][4]
 
1495
        # Turn some things into local variables
 
1496
        minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1497
        factory = entry_factory
 
1498
        utf8_decode = cache_utf8._utf8_decode
 
1499
        inv_byid = inv._byid
 
1500
        # we could do this straight out of the dirstate; it might be fast
 
1501
        # and should be profiled - RBC 20070216
 
1502
        parent_ies = {'' : inv.root}
 
1503
        for block in self._dirstate._dirblocks[1:]: #skip root
 
1504
            dirname = block[0]
 
1505
            try:
 
1506
                parent_ie = parent_ies[dirname]
 
1507
            except KeyError:
 
1508
                # all the paths in this block are not versioned in this tree
 
1509
                continue
 
1510
            for key, entry in block[1]:
 
1511
                minikind, fingerprint, size, executable, revid = entry[parent_index]
 
1512
                if minikind in ('a', 'r'): # absent, relocated
 
1513
                    # not this tree
 
1514
                    continue
 
1515
                name = key[1]
 
1516
                name_unicode = utf8_decode(name)[0]
 
1517
                file_id = key[2]
 
1518
                kind = minikind_to_kind[minikind]
 
1519
                inv_entry = factory[kind](file_id, name_unicode,
 
1520
                                          parent_ie.file_id)
 
1521
                inv_entry.revision = revid
 
1522
                if kind == 'file':
 
1523
                    inv_entry.executable = executable
 
1524
                    inv_entry.text_size = size
 
1525
                    inv_entry.text_sha1 = fingerprint
 
1526
                elif kind == 'directory':
 
1527
                    parent_ies[(dirname + '/' + name).strip('/')] = inv_entry
 
1528
                elif kind == 'symlink':
 
1529
                    inv_entry.executable = False
 
1530
                    inv_entry.text_size = None
 
1531
                    inv_entry.symlink_target = utf8_decode(fingerprint)[0]
 
1532
                elif kind == 'tree-reference':
 
1533
                    inv_entry.reference_revision = fingerprint or None
 
1534
                else:
 
1535
                    raise AssertionError("cannot convert entry %r into an InventoryEntry"
 
1536
                            % entry)
 
1537
                # These checks cost us around 40ms on a 55k entry tree
 
1538
                if file_id in inv_byid:
 
1539
                    raise AssertionError('file_id %s already in'
 
1540
                        ' inventory as %s' % (file_id, inv_byid[file_id]))
 
1541
                if name_unicode in parent_ie.children:
 
1542
                    raise AssertionError('name %r already in parent'
 
1543
                        % (name_unicode,))
 
1544
                inv_byid[file_id] = inv_entry
 
1545
                parent_ie.children[name_unicode] = inv_entry
 
1546
        self._inventory = inv
 
1547
 
 
1548
    def get_file_mtime(self, file_id, path=None):
 
1549
        """Return the modification time for this record.
 
1550
 
 
1551
        We return the timestamp of the last-changed revision.
 
1552
        """
 
1553
        # Make sure the file exists
 
1554
        entry = self._get_entry(file_id, path=path)
 
1555
        if entry == (None, None): # do we raise?
 
1556
            return None
 
1557
        parent_index = self._get_parent_index()
 
1558
        last_changed_revision = entry[1][parent_index][4]
 
1559
        return self._repository.get_revision(last_changed_revision).timestamp
 
1560
 
 
1561
    def get_file_sha1(self, file_id, path=None, stat_value=None):
 
1562
        entry = self._get_entry(file_id=file_id, path=path)
 
1563
        parent_index = self._get_parent_index()
 
1564
        parent_details = entry[1][parent_index]
 
1565
        if parent_details[0] == 'f':
 
1566
            return parent_details[1]
 
1567
        return None
 
1568
 
 
1569
    def _get_weave(self, file_id):
 
1570
        return self._repository.weave_store.get_weave(file_id,
 
1571
                self._repository.get_transaction())
 
1572
 
 
1573
    def get_file(self, file_id, path=None):
 
1574
        return StringIO(self.get_file_text(file_id))
 
1575
 
 
1576
    def get_file_lines(self, file_id):
 
1577
        entry = self._get_entry(file_id=file_id)[1]
 
1578
        if entry is None:
 
1579
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1580
        return self._get_weave(file_id).get_lines(entry[1][4])
 
1581
 
 
1582
    def get_file_size(self, file_id):
 
1583
        """See Tree.get_file_size"""
 
1584
        return self.inventory[file_id].text_size
 
1585
 
 
1586
    def get_file_text(self, file_id):
 
1587
        return ''.join(self.get_file_lines(file_id))
 
1588
 
 
1589
    def get_reference_revision(self, file_id, path=None):
 
1590
        return self.inventory[file_id].reference_revision
 
1591
 
 
1592
    def iter_files_bytes(self, desired_files):
 
1593
        """See Tree.iter_files_bytes.
 
1594
 
 
1595
        This version is implemented on top of Repository.iter_files_bytes"""
 
1596
        parent_index = self._get_parent_index()
 
1597
        repo_desired_files = []
 
1598
        for file_id, identifier in desired_files:
 
1599
            entry = self._get_entry(file_id)
 
1600
            if entry == (None, None):
 
1601
                raise errors.NoSuchId(self, file_id)
 
1602
            repo_desired_files.append((file_id, entry[1][parent_index][4],
 
1603
                                       identifier))
 
1604
        return self._repository.iter_files_bytes(repo_desired_files)
 
1605
 
 
1606
    def get_symlink_target(self, file_id):
 
1607
        entry = self._get_entry(file_id=file_id)
 
1608
        parent_index = self._get_parent_index()
 
1609
        if entry[1][parent_index][0] != 'l':
 
1610
            return None
 
1611
        else:
 
1612
            # At present, none of the tree implementations supports non-ascii
 
1613
            # symlink targets. So we will just assume that the dirstate path is
 
1614
            # correct.
 
1615
            return entry[1][parent_index][1]
 
1616
 
 
1617
    def get_revision_id(self):
 
1618
        """Return the revision id for this tree."""
 
1619
        return self._revision_id
 
1620
 
 
1621
    def _get_inventory(self):
 
1622
        if self._inventory is not None:
 
1623
            return self._inventory
 
1624
        self._must_be_locked()
 
1625
        self._generate_inventory()
 
1626
        return self._inventory
 
1627
 
 
1628
    inventory = property(_get_inventory,
 
1629
                         doc="Inventory of this Tree")
 
1630
 
 
1631
    def get_parent_ids(self):
 
1632
        """The parents of a tree in the dirstate are not cached."""
 
1633
        return self._repository.get_revision(self._revision_id).parent_ids
 
1634
 
 
1635
    def has_filename(self, filename):
 
1636
        return bool(self.path2id(filename))
 
1637
 
 
1638
    def kind(self, file_id):
 
1639
        entry = self._get_entry(file_id=file_id)[1]
 
1640
        if entry is None:
 
1641
            raise errors.NoSuchId(tree=self, file_id=file_id)
 
1642
        return dirstate.DirState._minikind_to_kind[entry[1][0]]
 
1643
 
 
1644
    def stored_kind(self, file_id):
 
1645
        """See Tree.stored_kind"""
 
1646
        return self.kind(file_id)
 
1647
 
 
1648
    def path_content_summary(self, path):
 
1649
        """See Tree.path_content_summary."""
 
1650
        id = self.inventory.path2id(path)
 
1651
        if id is None:
 
1652
            return ('missing', None, None, None)
 
1653
        entry = self._inventory[id]
 
1654
        kind = entry.kind
 
1655
        if kind == 'file':
 
1656
            return (kind, entry.text_size, entry.executable, entry.text_sha1)
 
1657
        elif kind == 'symlink':
 
1658
            return (kind, None, None, entry.symlink_target)
 
1659
        else:
 
1660
            return (kind, None, None, None)
 
1661
 
 
1662
    def is_executable(self, file_id, path=None):
 
1663
        ie = self.inventory[file_id]
 
1664
        if ie.kind != "file":
 
1665
            return None
 
1666
        return ie.executable
 
1667
 
 
1668
    def list_files(self, include_root=False):
 
1669
        # We use a standard implementation, because DirStateRevisionTree is
 
1670
        # dealing with one of the parents of the current state
 
1671
        inv = self._get_inventory()
 
1672
        entries = inv.iter_entries()
 
1673
        if self.inventory.root is not None and not include_root:
 
1674
            entries.next()
 
1675
        for path, entry in entries:
 
1676
            yield path, 'V', entry.kind, entry.file_id, entry
 
1677
 
 
1678
    def lock_read(self):
 
1679
        """Lock the tree for a set of operations."""
 
1680
        if not self._locked:
 
1681
            self._repository.lock_read()
 
1682
            if self._dirstate._lock_token is None:
 
1683
                self._dirstate.lock_read()
 
1684
                self._dirstate_locked = True
 
1685
        self._locked += 1
 
1686
 
 
1687
    def _must_be_locked(self):
 
1688
        if not self._locked:
 
1689
            raise errors.ObjectNotLocked(self)
 
1690
 
 
1691
    @needs_read_lock
 
1692
    def path2id(self, path):
 
1693
        """Return the id for path in this tree."""
 
1694
        # lookup by path: faster than splitting and walking the ivnentory.
 
1695
        entry = self._get_entry(path=path)
 
1696
        if entry == (None, None):
 
1697
            return None
 
1698
        return entry[0][2]
 
1699
 
 
1700
    def unlock(self):
 
1701
        """Unlock, freeing any cache memory used during the lock."""
 
1702
        # outside of a lock, the inventory is suspect: release it.
 
1703
        self._locked -=1
 
1704
        if not self._locked:
 
1705
            self._inventory = None
 
1706
            self._locked = 0
 
1707
            if self._dirstate_locked:
 
1708
                self._dirstate.unlock()
 
1709
                self._dirstate_locked = False
 
1710
            self._repository.unlock()
 
1711
 
 
1712
    def walkdirs(self, prefix=""):
 
1713
        # TODO: jam 20070215 This is the lazy way by using the RevisionTree
 
1714
        # implementation based on an inventory.  
 
1715
        # This should be cleaned up to use the much faster Dirstate code
 
1716
        # So for now, we just build up the parent inventory, and extract
 
1717
        # it the same way RevisionTree does.
 
1718
        _directory = 'directory'
 
1719
        inv = self._get_inventory()
 
1720
        top_id = inv.path2id(prefix)
 
1721
        if top_id is None:
 
1722
            pending = []
 
1723
        else:
 
1724
            pending = [(prefix, top_id)]
 
1725
        while pending:
 
1726
            dirblock = []
 
1727
            relpath, file_id = pending.pop()
 
1728
            # 0 - relpath, 1- file-id
 
1729
            if relpath:
 
1730
                relroot = relpath + '/'
 
1731
            else:
 
1732
                relroot = ""
 
1733
            # FIXME: stash the node in pending
 
1734
            entry = inv[file_id]
 
1735
            for name, child in entry.sorted_children():
 
1736
                toppath = relroot + name
 
1737
                dirblock.append((toppath, name, child.kind, None,
 
1738
                    child.file_id, child.kind
 
1739
                    ))
 
1740
            yield (relpath, entry.file_id), dirblock
 
1741
            # push the user specified dirs from dirblock
 
1742
            for dir in reversed(dirblock):
 
1743
                if dir[2] == _directory:
 
1744
                    pending.append((dir[0], dir[4]))
 
1745
 
 
1746
 
 
1747
class InterDirStateTree(InterTree):
 
1748
    """Fast path optimiser for changes_from with dirstate trees.
 
1749
    
 
1750
    This is used only when both trees are in the dirstate working file, and 
 
1751
    the source is any parent within the dirstate, and the destination is 
 
1752
    the current working tree of the same dirstate.
 
1753
    """
 
1754
    # this could be generalized to allow comparisons between any trees in the
 
1755
    # dirstate, and possibly between trees stored in different dirstates.
 
1756
 
 
1757
    def __init__(self, source, target):
 
1758
        super(InterDirStateTree, self).__init__(source, target)
 
1759
        if not InterDirStateTree.is_compatible(source, target):
 
1760
            raise Exception, "invalid source %r and target %r" % (source, target)
 
1761
 
 
1762
    @staticmethod
 
1763
    def make_source_parent_tree(source, target):
 
1764
        """Change the source tree into a parent of the target."""
 
1765
        revid = source.commit('record tree')
 
1766
        target.branch.repository.fetch(source.branch.repository, revid)
 
1767
        target.set_parent_ids([revid])
 
1768
        return target.basis_tree(), target
 
1769
 
 
1770
    _matching_from_tree_format = WorkingTreeFormat4()
 
1771
    _matching_to_tree_format = WorkingTreeFormat4()
 
1772
    _test_mutable_trees_to_test_trees = make_source_parent_tree
 
1773
 
 
1774
    def iter_changes(self, include_unchanged=False,
 
1775
                      specific_files=None, pb=None, extra_trees=[],
 
1776
                      require_versioned=True, want_unversioned=False):
 
1777
        """Return the changes from source to target.
 
1778
 
 
1779
        :return: An iterator that yields tuples. See InterTree.iter_changes
 
1780
            for details.
 
1781
        :param specific_files: An optional list of file paths to restrict the
 
1782
            comparison to. When mapping filenames to ids, all matches in all
 
1783
            trees (including optional extra_trees) are used, and all children of
 
1784
            matched directories are included.
 
1785
        :param include_unchanged: An optional boolean requesting the inclusion of
 
1786
            unchanged entries in the result.
 
1787
        :param extra_trees: An optional list of additional trees to use when
 
1788
            mapping the contents of specific_files (paths) to file_ids.
 
1789
        :param require_versioned: If True, all files in specific_files must be
 
1790
            versioned in one of source, target, extra_trees or
 
1791
            PathsNotVersionedError is raised.
 
1792
        :param want_unversioned: Should unversioned files be returned in the
 
1793
            output. An unversioned file is defined as one with (False, False)
 
1794
            for the versioned pair.
 
1795
        """
 
1796
        utf8_decode = cache_utf8._utf8_decode
 
1797
        _minikind_to_kind = dirstate.DirState._minikind_to_kind
 
1798
        cmp_by_dirs = dirstate.cmp_by_dirs
 
1799
        # NB: show_status depends on being able to pass in non-versioned files
 
1800
        # and report them as unknown
 
1801
        # TODO: handle extra trees in the dirstate.
 
1802
        if (extra_trees or specific_files == []):
 
1803
            # we can't fast-path these cases (yet)
 
1804
            for f in super(InterDirStateTree, self).iter_changes(
 
1805
                include_unchanged, specific_files, pb, extra_trees,
 
1806
                require_versioned, want_unversioned=want_unversioned):
 
1807
                yield f
 
1808
            return
 
1809
        parent_ids = self.target.get_parent_ids()
 
1810
        if not (self.source._revision_id in parent_ids
 
1811
                or self.source._revision_id == NULL_REVISION):
 
1812
            raise AssertionError(
 
1813
                "revision {%s} is not stored in {%s}, but %s "
 
1814
                "can only be used for trees stored in the dirstate"
 
1815
                % (self.source._revision_id, self.target, self.iter_changes))
 
1816
        target_index = 0
 
1817
        if self.source._revision_id == NULL_REVISION:
 
1818
            source_index = None
 
1819
            indices = (target_index,)
 
1820
        else:
 
1821
            if not (self.source._revision_id in parent_ids):
 
1822
                raise AssertionError(
 
1823
                    "Failure: source._revision_id: %s not in target.parent_ids(%s)" % (
 
1824
                    self.source._revision_id, parent_ids))
 
1825
            source_index = 1 + parent_ids.index(self.source._revision_id)
 
1826
            indices = (source_index, target_index)
 
1827
        # -- make all specific_files utf8 --
 
1828
        if specific_files:
 
1829
            specific_files_utf8 = set()
 
1830
            for path in specific_files:
 
1831
                specific_files_utf8.add(path.encode('utf8'))
 
1832
            specific_files = specific_files_utf8
 
1833
        else:
 
1834
            specific_files = set([''])
 
1835
        # -- specific_files is now a utf8 path set --
 
1836
        # -- get the state object and prepare it.
 
1837
        state = self.target.current_dirstate()
 
1838
        state._read_dirblocks_if_needed()
 
1839
        def _entries_for_path(path):
 
1840
            """Return a list with all the entries that match path for all ids.
 
1841
            """
 
1842
            dirname, basename = os.path.split(path)
 
1843
            key = (dirname, basename, '')
 
1844
            block_index, present = state._find_block_index_from_key(key)
 
1845
            if not present:
 
1846
                # the block which should contain path is absent.
 
1847
                return []
 
1848
            result = []
 
1849
            block = state._dirblocks[block_index][1]
 
1850
            entry_index, _ = state._find_entry_index(key, block)
 
1851
            # we may need to look at multiple entries at this path: walk while the specific_files match.
 
1852
            while (entry_index < len(block) and
 
1853
                block[entry_index][0][0:2] == key[0:2]):
 
1854
                result.append(block[entry_index])
 
1855
                entry_index += 1
 
1856
            return result
 
1857
        if require_versioned:
 
1858
            # -- check all supplied paths are versioned in a search tree. --
 
1859
            all_versioned = True
 
1860
            for path in specific_files:
 
1861
                path_entries = _entries_for_path(path)
 
1862
                if not path_entries:
 
1863
                    # this specified path is not present at all: error
 
1864
                    all_versioned = False
 
1865
                    break
 
1866
                found_versioned = False
 
1867
                # for each id at this path
 
1868
                for entry in path_entries:
 
1869
                    # for each tree.
 
1870
                    for index in indices:
 
1871
                        if entry[1][index][0] != 'a': # absent
 
1872
                            found_versioned = True
 
1873
                            # all good: found a versioned cell
 
1874
                            break
 
1875
                if not found_versioned:
 
1876
                    # none of the indexes was not 'absent' at all ids for this
 
1877
                    # path.
 
1878
                    all_versioned = False
 
1879
                    break
 
1880
            if not all_versioned:
 
1881
                raise errors.PathsNotVersionedError(specific_files)
 
1882
        # -- remove redundancy in supplied specific_files to prevent over-scanning --
 
1883
        search_specific_files = set()
 
1884
        for path in specific_files:
 
1885
            other_specific_files = specific_files.difference(set([path]))
 
1886
            if not osutils.is_inside_any(other_specific_files, path):
 
1887
                # this is a top level path, we must check it.
 
1888
                search_specific_files.add(path)
 
1889
        # sketch: 
 
1890
        # compare source_index and target_index at or under each element of search_specific_files.
 
1891
        # follow the following comparison table. Note that we only want to do diff operations when
 
1892
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
 
1893
        # for the target.
 
1894
        # cases:
 
1895
        # 
 
1896
        # Source | Target | disk | action
 
1897
        #   r    | fdlt   |      | add source to search, add id path move and perform
 
1898
        #        |        |      | diff check on source-target
 
1899
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
 
1900
        #        |        |      | ???
 
1901
        #   r    |  a     |      | add source to search
 
1902
        #   r    |  a     |  a   | 
 
1903
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
 
1904
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
 
1905
        #   a    | fdlt   |      | add new id
 
1906
        #   a    | fdlt   |  a   | dangling locally added file, skip
 
1907
        #   a    |  a     |      | not present in either tree, skip
 
1908
        #   a    |  a     |  a   | not present in any tree, skip
 
1909
        #   a    |  r     |      | not present in either tree at this path, skip as it
 
1910
        #        |        |      | may not be selected by the users list of paths.
 
1911
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
 
1912
        #        |        |      | may not be selected by the users list of paths.
 
1913
        #  fdlt  | fdlt   |      | content in both: diff them
 
1914
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
 
1915
        #  fdlt  |  a     |      | unversioned: output deleted id for now
 
1916
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
 
1917
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
 
1918
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1919
        #        |        |      | this id at the other path.
 
1920
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
 
1921
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
 
1922
        #        |        |      | this id at the other path.
 
1923
 
 
1924
        # for all search_indexs in each path at or under each element of
 
1925
        # search_specific_files, if the detail is relocated: add the id, and add the
 
1926
        # relocated path as one to search if its not searched already. If the
 
1927
        # detail is not relocated, add the id.
 
1928
        searched_specific_files = set()
 
1929
        NULL_PARENT_DETAILS = dirstate.DirState.NULL_PARENT_DETAILS
 
1930
        # Using a list so that we can access the values and change them in
 
1931
        # nested scope. Each one is [path, file_id, entry]
 
1932
        last_source_parent = [None, None]
 
1933
        last_target_parent = [None, None]
 
1934
 
 
1935
        use_filesystem_for_exec = (sys.platform != 'win32')
 
1936
 
 
1937
        # Just a sentry, so that _process_entry can say that this
 
1938
        # record is handled, but isn't interesting to process (unchanged)
 
1939
        uninteresting = object()
 
1940
 
 
1941
 
 
1942
        old_dirname_to_file_id = {}
 
1943
        new_dirname_to_file_id = {}
 
1944
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
 
1945
        #       keeping a cache of directories that we have seen.
 
1946
 
 
1947
        def _process_entry(entry, path_info):
 
1948
            """Compare an entry and real disk to generate delta information.
 
1949
 
 
1950
            :param path_info: top_relpath, basename, kind, lstat, abspath for
 
1951
                the path of entry. If None, then the path is considered absent.
 
1952
                (Perhaps we should pass in a concrete entry for this ?)
 
1953
                Basename is returned as a utf8 string because we expect this
 
1954
                tuple will be ignored, and don't want to take the time to
 
1955
                decode.
 
1956
            :return: None if these don't match
 
1957
                     A tuple of information about the change, or
 
1958
                     the object 'uninteresting' if these match, but are
 
1959
                     basically identical.
 
1960
            """
 
1961
            if source_index is None:
 
1962
                source_details = NULL_PARENT_DETAILS
 
1963
            else:
 
1964
                source_details = entry[1][source_index]
 
1965
            target_details = entry[1][target_index]
 
1966
            target_minikind = target_details[0]
 
1967
            if path_info is not None and target_minikind in 'fdlt':
 
1968
                if not (target_index == 0):
 
1969
                    raise AssertionError()
 
1970
                link_or_sha1 = state.update_entry(entry, abspath=path_info[4],
 
1971
                                                  stat_value=path_info[3])
 
1972
                # The entry may have been modified by update_entry
 
1973
                target_details = entry[1][target_index]
 
1974
                target_minikind = target_details[0]
 
1975
            else:
 
1976
                link_or_sha1 = None
 
1977
            file_id = entry[0][2]
 
1978
            source_minikind = source_details[0]
 
1979
            if source_minikind in 'fdltr' and target_minikind in 'fdlt':
 
1980
                # claimed content in both: diff
 
1981
                #   r    | fdlt   |      | add source to search, add id path move and perform
 
1982
                #        |        |      | diff check on source-target
 
1983
                #   r    | fdlt   |  a   | dangling file that was present in the basis.
 
1984
                #        |        |      | ???
 
1985
                if source_minikind in 'r':
 
1986
                    # add the source to the search path to find any children it
 
1987
                    # has.  TODO ? : only add if it is a container ?
 
1988
                    if not osutils.is_inside_any(searched_specific_files,
 
1989
                                                 source_details[1]):
 
1990
                        search_specific_files.add(source_details[1])
 
1991
                    # generate the old path; this is needed for stating later
 
1992
                    # as well.
 
1993
                    old_path = source_details[1]
 
1994
                    old_dirname, old_basename = os.path.split(old_path)
 
1995
                    path = pathjoin(entry[0][0], entry[0][1])
 
1996
                    old_entry = state._get_entry(source_index,
 
1997
                                                 path_utf8=old_path)
 
1998
                    # update the source details variable to be the real
 
1999
                    # location.
 
2000
                    if old_entry == (None, None):
 
2001
                        raise errors.CorruptDirstate(state._filename,
 
2002
                            "entry '%s/%s' is considered renamed from %r"
 
2003
                            " but source does not exist\n"
 
2004
                            "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
 
2005
                    source_details = old_entry[1][source_index]
 
2006
                    source_minikind = source_details[0]
 
2007
                else:
 
2008
                    old_dirname = entry[0][0]
 
2009
                    old_basename = entry[0][1]
 
2010
                    old_path = path = None
 
2011
                if path_info is None:
 
2012
                    # the file is missing on disk, show as removed.
 
2013
                    content_change = True
 
2014
                    target_kind = None
 
2015
                    target_exec = False
 
2016
                else:
 
2017
                    # source and target are both versioned and disk file is present.
 
2018
                    target_kind = path_info[2]
 
2019
                    if target_kind == 'directory':
 
2020
                        if path is None:
 
2021
                            old_path = path = pathjoin(old_dirname, old_basename)
 
2022
                        new_dirname_to_file_id[path] = file_id
 
2023
                        if source_minikind != 'd':
 
2024
                            content_change = True
 
2025
                        else:
 
2026
                            # directories have no fingerprint
 
2027
                            content_change = False
 
2028
                        target_exec = False
 
2029
                    elif target_kind == 'file':
 
2030
                        if source_minikind != 'f':
 
2031
                            content_change = True
 
2032
                        else:
 
2033
                            # We could check the size, but we already have the
 
2034
                            # sha1 hash.
 
2035
                            content_change = (link_or_sha1 != source_details[1])
 
2036
                        # Target details is updated at update_entry time
 
2037
                        if use_filesystem_for_exec:
 
2038
                            # We don't need S_ISREG here, because we are sure
 
2039
                            # we are dealing with a file.
 
2040
                            target_exec = bool(stat.S_IEXEC & path_info[3].st_mode)
 
2041
                        else:
 
2042
                            target_exec = target_details[3]
 
2043
                    elif target_kind == 'symlink':
 
2044
                        if source_minikind != 'l':
 
2045
                            content_change = True
 
2046
                        else:
 
2047
                            content_change = (link_or_sha1 != source_details[1])
 
2048
                        target_exec = False
 
2049
                    elif target_kind == 'tree-reference':
 
2050
                        if source_minikind != 't':
 
2051
                            content_change = True
 
2052
                        else:
 
2053
                            content_change = False
 
2054
                        target_exec = False
 
2055
                    else:
 
2056
                        raise Exception, "unknown kind %s" % path_info[2]
 
2057
                if source_minikind == 'd':
 
2058
                    if path is None:
 
2059
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2060
                    old_dirname_to_file_id[old_path] = file_id
 
2061
                # parent id is the entry for the path in the target tree
 
2062
                if old_dirname == last_source_parent[0]:
 
2063
                    source_parent_id = last_source_parent[1]
 
2064
                else:
 
2065
                    try:
 
2066
                        source_parent_id = old_dirname_to_file_id[old_dirname]
 
2067
                    except KeyError:
 
2068
                        source_parent_entry = state._get_entry(source_index,
 
2069
                                                               path_utf8=old_dirname)
 
2070
                        source_parent_id = source_parent_entry[0][2]
 
2071
                    if source_parent_id == entry[0][2]:
 
2072
                        # This is the root, so the parent is None
 
2073
                        source_parent_id = None
 
2074
                    else:
 
2075
                        last_source_parent[0] = old_dirname
 
2076
                        last_source_parent[1] = source_parent_id
 
2077
                new_dirname = entry[0][0]
 
2078
                if new_dirname == last_target_parent[0]:
 
2079
                    target_parent_id = last_target_parent[1]
 
2080
                else:
 
2081
                    try:
 
2082
                        target_parent_id = new_dirname_to_file_id[new_dirname]
 
2083
                    except KeyError:
 
2084
                        # TODO: We don't always need to do the lookup, because the
 
2085
                        #       parent entry will be the same as the source entry.
 
2086
                        target_parent_entry = state._get_entry(target_index,
 
2087
                                                               path_utf8=new_dirname)
 
2088
                        if target_parent_entry == (None, None):
 
2089
                            raise AssertionError(
 
2090
                                "Could not find target parent in wt: %s\nparent of: %s"
 
2091
                                % (new_dirname, entry))
 
2092
                        target_parent_id = target_parent_entry[0][2]
 
2093
                    if target_parent_id == entry[0][2]:
 
2094
                        # This is the root, so the parent is None
 
2095
                        target_parent_id = None
 
2096
                    else:
 
2097
                        last_target_parent[0] = new_dirname
 
2098
                        last_target_parent[1] = target_parent_id
 
2099
 
 
2100
                source_exec = source_details[3]
 
2101
                if (include_unchanged
 
2102
                    or content_change
 
2103
                    or source_parent_id != target_parent_id
 
2104
                    or old_basename != entry[0][1]
 
2105
                    or source_exec != target_exec
 
2106
                    ):
 
2107
                    if old_path is None:
 
2108
                        old_path = path = pathjoin(old_dirname, old_basename)
 
2109
                        old_path_u = utf8_decode(old_path)[0]
 
2110
                        path_u = old_path_u
 
2111
                    else:
 
2112
                        old_path_u = utf8_decode(old_path)[0]
 
2113
                        if old_path == path:
 
2114
                            path_u = old_path_u
 
2115
                        else:
 
2116
                            path_u = utf8_decode(path)[0]
 
2117
                    source_kind = _minikind_to_kind[source_minikind]
 
2118
                    return (entry[0][2],
 
2119
                           (old_path_u, path_u),
 
2120
                           content_change,
 
2121
                           (True, True),
 
2122
                           (source_parent_id, target_parent_id),
 
2123
                           (utf8_decode(old_basename)[0], utf8_decode(entry[0][1])[0]),
 
2124
                           (source_kind, target_kind),
 
2125
                           (source_exec, target_exec))
 
2126
                else:
 
2127
                    return uninteresting
 
2128
            elif source_minikind in 'a' and target_minikind in 'fdlt':
 
2129
                # looks like a new file
 
2130
                if path_info is not None:
 
2131
                    path = pathjoin(entry[0][0], entry[0][1])
 
2132
                    # parent id is the entry for the path in the target tree
 
2133
                    # TODO: these are the same for an entire directory: cache em.
 
2134
                    parent_id = state._get_entry(target_index,
 
2135
                                                 path_utf8=entry[0][0])[0][2]
 
2136
                    if parent_id == entry[0][2]:
 
2137
                        parent_id = None
 
2138
                    if use_filesystem_for_exec:
 
2139
                        # We need S_ISREG here, because we aren't sure if this
 
2140
                        # is a file or not.
 
2141
                        target_exec = bool(
 
2142
                            stat.S_ISREG(path_info[3].st_mode)
 
2143
                            and stat.S_IEXEC & path_info[3].st_mode)
 
2144
                    else:
 
2145
                        target_exec = target_details[3]
 
2146
                    return (entry[0][2],
 
2147
                           (None, utf8_decode(path)[0]),
 
2148
                           True,
 
2149
                           (False, True),
 
2150
                           (None, parent_id),
 
2151
                           (None, utf8_decode(entry[0][1])[0]),
 
2152
                           (None, path_info[2]),
 
2153
                           (None, target_exec))
 
2154
                else:
 
2155
                    # but its not on disk: we deliberately treat this as just
 
2156
                    # never-present. (Why ?! - RBC 20070224)
 
2157
                    pass
 
2158
            elif source_minikind in 'fdlt' and target_minikind in 'a':
 
2159
                # unversioned, possibly, or possibly not deleted: we dont care.
 
2160
                # if its still on disk, *and* theres no other entry at this
 
2161
                # path [we dont know this in this routine at the moment -
 
2162
                # perhaps we should change this - then it would be an unknown.
 
2163
                old_path = pathjoin(entry[0][0], entry[0][1])
 
2164
                # parent id is the entry for the path in the target tree
 
2165
                parent_id = state._get_entry(source_index, path_utf8=entry[0][0])[0][2]
 
2166
                if parent_id == entry[0][2]:
 
2167
                    parent_id = None
 
2168
                return (entry[0][2],
 
2169
                       (utf8_decode(old_path)[0], None),
 
2170
                       True,
 
2171
                       (True, False),
 
2172
                       (parent_id, None),
 
2173
                       (utf8_decode(entry[0][1])[0], None),
 
2174
                       (_minikind_to_kind[source_minikind], None),
 
2175
                       (source_details[3], None))
 
2176
            elif source_minikind in 'fdlt' and target_minikind in 'r':
 
2177
                # a rename; could be a true rename, or a rename inherited from
 
2178
                # a renamed parent. TODO: handle this efficiently. Its not
 
2179
                # common case to rename dirs though, so a correct but slow
 
2180
                # implementation will do.
 
2181
                if not osutils.is_inside_any(searched_specific_files, target_details[1]):
 
2182
                    search_specific_files.add(target_details[1])
 
2183
            elif source_minikind in 'ra' and target_minikind in 'ra':
 
2184
                # neither of the selected trees contain this file,
 
2185
                # so skip over it. This is not currently directly tested, but
 
2186
                # is indirectly via test_too_much.TestCommands.test_conflicts.
 
2187
                pass
 
2188
            else:
 
2189
                raise AssertionError("don't know how to compare "
 
2190
                    "source_minikind=%r, target_minikind=%r"
 
2191
                    % (source_minikind, target_minikind))
 
2192
                ## import pdb;pdb.set_trace()
 
2193
            return None
 
2194
 
 
2195
        while search_specific_files:
 
2196
            # TODO: the pending list should be lexically sorted?  the
 
2197
            # interface doesn't require it.
 
2198
            current_root = search_specific_files.pop()
 
2199
            current_root_unicode = current_root.decode('utf8')
 
2200
            searched_specific_files.add(current_root)
 
2201
            # process the entries for this containing directory: the rest will be
 
2202
            # found by their parents recursively.
 
2203
            root_entries = _entries_for_path(current_root)
 
2204
            root_abspath = self.target.abspath(current_root_unicode)
 
2205
            try:
 
2206
                root_stat = os.lstat(root_abspath)
 
2207
            except OSError, e:
 
2208
                if e.errno == errno.ENOENT:
 
2209
                    # the path does not exist: let _process_entry know that.
 
2210
                    root_dir_info = None
 
2211
                else:
 
2212
                    # some other random error: hand it up.
 
2213
                    raise
 
2214
            else:
 
2215
                root_dir_info = ('', current_root,
 
2216
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
 
2217
                    root_abspath)
 
2218
                if root_dir_info[2] == 'directory':
 
2219
                    if self.target._directory_is_tree_reference(
 
2220
                        current_root.decode('utf8')):
 
2221
                        root_dir_info = root_dir_info[:2] + \
 
2222
                            ('tree-reference',) + root_dir_info[3:]
 
2223
 
 
2224
            if not root_entries and not root_dir_info:
 
2225
                # this specified path is not present at all, skip it.
 
2226
                continue
 
2227
            path_handled = False
 
2228
            for entry in root_entries:
 
2229
                result = _process_entry(entry, root_dir_info)
 
2230
                if result is not None:
 
2231
                    path_handled = True
 
2232
                    if result is not uninteresting:
 
2233
                        yield result
 
2234
            if want_unversioned and not path_handled and root_dir_info:
 
2235
                new_executable = bool(
 
2236
                    stat.S_ISREG(root_dir_info[3].st_mode)
 
2237
                    and stat.S_IEXEC & root_dir_info[3].st_mode)
 
2238
                yield (None,
 
2239
                       (None, current_root_unicode),
 
2240
                       True,
 
2241
                       (False, False),
 
2242
                       (None, None),
 
2243
                       (None, splitpath(current_root_unicode)[-1]),
 
2244
                       (None, root_dir_info[2]),
 
2245
                       (None, new_executable)
 
2246
                      )
 
2247
            initial_key = (current_root, '', '')
 
2248
            block_index, _ = state._find_block_index_from_key(initial_key)
 
2249
            if block_index == 0:
 
2250
                # we have processed the total root already, but because the
 
2251
                # initial key matched it we should skip it here.
 
2252
                block_index +=1
 
2253
            if root_dir_info and root_dir_info[2] == 'tree-reference':
 
2254
                current_dir_info = None
 
2255
            else:
 
2256
                dir_iterator = osutils._walkdirs_utf8(root_abspath, prefix=current_root)
 
2257
                try:
 
2258
                    current_dir_info = dir_iterator.next()
 
2259
                except OSError, e:
 
2260
                    # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
 
2261
                    # python 2.5 has e.errno == EINVAL,
 
2262
                    #            and e.winerror == ERROR_DIRECTORY
 
2263
                    e_winerror = getattr(e, 'winerror', None)
 
2264
                    win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
 
2265
                    # there may be directories in the inventory even though
 
2266
                    # this path is not a file on disk: so mark it as end of
 
2267
                    # iterator
 
2268
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
 
2269
                        current_dir_info = None
 
2270
                    elif (sys.platform == 'win32'
 
2271
                          and (e.errno in win_errors
 
2272
                               or e_winerror in win_errors)):
 
2273
                        current_dir_info = None
 
2274
                    else:
 
2275
                        raise
 
2276
                else:
 
2277
                    if current_dir_info[0][0] == '':
 
2278
                        # remove .bzr from iteration
 
2279
                        bzr_index = bisect_left(current_dir_info[1], ('.bzr',))
 
2280
                        if current_dir_info[1][bzr_index][0] != '.bzr':
 
2281
                            raise AssertionError()
 
2282
                        del current_dir_info[1][bzr_index]
 
2283
            # walk until both the directory listing and the versioned metadata
 
2284
            # are exhausted. 
 
2285
            if (block_index < len(state._dirblocks) and
 
2286
                osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2287
                current_block = state._dirblocks[block_index]
 
2288
            else:
 
2289
                current_block = None
 
2290
            while (current_dir_info is not None or
 
2291
                   current_block is not None):
 
2292
                if (current_dir_info and current_block
 
2293
                    and current_dir_info[0][0] != current_block[0]):
 
2294
                    if cmp_by_dirs(current_dir_info[0][0], current_block[0]) < 0:
 
2295
                        # filesystem data refers to paths not covered by the dirblock.
 
2296
                        # this has two possibilities:
 
2297
                        # A) it is versioned but empty, so there is no block for it
 
2298
                        # B) it is not versioned.
 
2299
 
 
2300
                        # if (A) then we need to recurse into it to check for
 
2301
                        # new unknown files or directories.
 
2302
                        # if (B) then we should ignore it, because we don't
 
2303
                        # recurse into unknown directories.
 
2304
                        path_index = 0
 
2305
                        while path_index < len(current_dir_info[1]):
 
2306
                                current_path_info = current_dir_info[1][path_index]
 
2307
                                if want_unversioned:
 
2308
                                    if current_path_info[2] == 'directory':
 
2309
                                        if self.target._directory_is_tree_reference(
 
2310
                                            current_path_info[0].decode('utf8')):
 
2311
                                            current_path_info = current_path_info[:2] + \
 
2312
                                                ('tree-reference',) + current_path_info[3:]
 
2313
                                    new_executable = bool(
 
2314
                                        stat.S_ISREG(current_path_info[3].st_mode)
 
2315
                                        and stat.S_IEXEC & current_path_info[3].st_mode)
 
2316
                                    yield (None,
 
2317
                                        (None, utf8_decode(current_path_info[0])[0]),
 
2318
                                        True,
 
2319
                                        (False, False),
 
2320
                                        (None, None),
 
2321
                                        (None, utf8_decode(current_path_info[1])[0]),
 
2322
                                        (None, current_path_info[2]),
 
2323
                                        (None, new_executable))
 
2324
                                # dont descend into this unversioned path if it is
 
2325
                                # a dir
 
2326
                                if current_path_info[2] in ('directory',
 
2327
                                                            'tree-reference'):
 
2328
                                    del current_dir_info[1][path_index]
 
2329
                                    path_index -= 1
 
2330
                                path_index += 1
 
2331
 
 
2332
                        # This dir info has been handled, go to the next
 
2333
                        try:
 
2334
                            current_dir_info = dir_iterator.next()
 
2335
                        except StopIteration:
 
2336
                            current_dir_info = None
 
2337
                    else:
 
2338
                        # We have a dirblock entry for this location, but there
 
2339
                        # is no filesystem path for this. This is most likely
 
2340
                        # because a directory was removed from the disk.
 
2341
                        # We don't have to report the missing directory,
 
2342
                        # because that should have already been handled, but we
 
2343
                        # need to handle all of the files that are contained
 
2344
                        # within.
 
2345
                        for current_entry in current_block[1]:
 
2346
                            # entry referring to file not present on disk.
 
2347
                            # advance the entry only, after processing.
 
2348
                            result = _process_entry(current_entry, None)
 
2349
                            if result is not None:
 
2350
                                if result is not uninteresting:
 
2351
                                    yield result
 
2352
                        block_index +=1
 
2353
                        if (block_index < len(state._dirblocks) and
 
2354
                            osutils.is_inside(current_root,
 
2355
                                              state._dirblocks[block_index][0])):
 
2356
                            current_block = state._dirblocks[block_index]
 
2357
                        else:
 
2358
                            current_block = None
 
2359
                    continue
 
2360
                entry_index = 0
 
2361
                if current_block and entry_index < len(current_block[1]):
 
2362
                    current_entry = current_block[1][entry_index]
 
2363
                else:
 
2364
                    current_entry = None
 
2365
                advance_entry = True
 
2366
                path_index = 0
 
2367
                if current_dir_info and path_index < len(current_dir_info[1]):
 
2368
                    current_path_info = current_dir_info[1][path_index]
 
2369
                    if current_path_info[2] == 'directory':
 
2370
                        if self.target._directory_is_tree_reference(
 
2371
                            current_path_info[0].decode('utf8')):
 
2372
                            current_path_info = current_path_info[:2] + \
 
2373
                                ('tree-reference',) + current_path_info[3:]
 
2374
                else:
 
2375
                    current_path_info = None
 
2376
                advance_path = True
 
2377
                path_handled = False
 
2378
                while (current_entry is not None or
 
2379
                    current_path_info is not None):
 
2380
                    if current_entry is None:
 
2381
                        # the check for path_handled when the path is adnvaced
 
2382
                        # will yield this path if needed.
 
2383
                        pass
 
2384
                    elif current_path_info is None:
 
2385
                        # no path is fine: the per entry code will handle it.
 
2386
                        result = _process_entry(current_entry, current_path_info)
 
2387
                        if result is not None:
 
2388
                            if result is not uninteresting:
 
2389
                                yield result
 
2390
                    elif (current_entry[0][1] != current_path_info[1]
 
2391
                          or current_entry[1][target_index][0] in 'ar'):
 
2392
                        # The current path on disk doesn't match the dirblock
 
2393
                        # record. Either the dirblock is marked as absent, or
 
2394
                        # the file on disk is not present at all in the
 
2395
                        # dirblock. Either way, report about the dirblock
 
2396
                        # entry, and let other code handle the filesystem one.
 
2397
 
 
2398
                        # Compare the basename for these files to determine
 
2399
                        # which comes first
 
2400
                        if current_path_info[1] < current_entry[0][1]:
 
2401
                            # extra file on disk: pass for now, but only
 
2402
                            # increment the path, not the entry
 
2403
                            advance_entry = False
 
2404
                        else:
 
2405
                            # entry referring to file not present on disk.
 
2406
                            # advance the entry only, after processing.
 
2407
                            result = _process_entry(current_entry, None)
 
2408
                            if result is not None:
 
2409
                                if result is not uninteresting:
 
2410
                                    yield result
 
2411
                            advance_path = False
 
2412
                    else:
 
2413
                        result = _process_entry(current_entry, current_path_info)
 
2414
                        if result is not None:
 
2415
                            path_handled = True
 
2416
                            if result is not uninteresting:
 
2417
                                yield result
 
2418
                    if advance_entry and current_entry is not None:
 
2419
                        entry_index += 1
 
2420
                        if entry_index < len(current_block[1]):
 
2421
                            current_entry = current_block[1][entry_index]
 
2422
                        else:
 
2423
                            current_entry = None
 
2424
                    else:
 
2425
                        advance_entry = True # reset the advance flaga
 
2426
                    if advance_path and current_path_info is not None:
 
2427
                        if not path_handled:
 
2428
                            # unversioned in all regards
 
2429
                            if want_unversioned:
 
2430
                                new_executable = bool(
 
2431
                                    stat.S_ISREG(current_path_info[3].st_mode)
 
2432
                                    and stat.S_IEXEC & current_path_info[3].st_mode)
 
2433
                                yield (None,
 
2434
                                    (None, utf8_decode(current_path_info[0])[0]),
 
2435
                                    True,
 
2436
                                    (False, False),
 
2437
                                    (None, None),
 
2438
                                    (None, utf8_decode(current_path_info[1])[0]),
 
2439
                                    (None, current_path_info[2]),
 
2440
                                    (None, new_executable))
 
2441
                            # dont descend into this unversioned path if it is
 
2442
                            # a dir
 
2443
                            if current_path_info[2] in ('directory'):
 
2444
                                del current_dir_info[1][path_index]
 
2445
                                path_index -= 1
 
2446
                        # dont descend the disk iterator into any tree 
 
2447
                        # paths.
 
2448
                        if current_path_info[2] == 'tree-reference':
 
2449
                            del current_dir_info[1][path_index]
 
2450
                            path_index -= 1
 
2451
                        path_index += 1
 
2452
                        if path_index < len(current_dir_info[1]):
 
2453
                            current_path_info = current_dir_info[1][path_index]
 
2454
                            if current_path_info[2] == 'directory':
 
2455
                                if self.target._directory_is_tree_reference(
 
2456
                                    current_path_info[0].decode('utf8')):
 
2457
                                    current_path_info = current_path_info[:2] + \
 
2458
                                        ('tree-reference',) + current_path_info[3:]
 
2459
                        else:
 
2460
                            current_path_info = None
 
2461
                        path_handled = False
 
2462
                    else:
 
2463
                        advance_path = True # reset the advance flagg.
 
2464
                if current_block is not None:
 
2465
                    block_index += 1
 
2466
                    if (block_index < len(state._dirblocks) and
 
2467
                        osutils.is_inside(current_root, state._dirblocks[block_index][0])):
 
2468
                        current_block = state._dirblocks[block_index]
 
2469
                    else:
 
2470
                        current_block = None
 
2471
                if current_dir_info is not None:
 
2472
                    try:
 
2473
                        current_dir_info = dir_iterator.next()
 
2474
                    except StopIteration:
 
2475
                        current_dir_info = None
 
2476
 
 
2477
 
 
2478
    @staticmethod
 
2479
    def is_compatible(source, target):
 
2480
        # the target must be a dirstate working tree
 
2481
        if not isinstance(target, WorkingTree4):
 
2482
            return False
 
2483
        # the source must be a revtreee or dirstate rev tree.
 
2484
        if not isinstance(source,
 
2485
            (revisiontree.RevisionTree, DirStateRevisionTree)):
 
2486
            return False
 
2487
        # the source revid must be in the target dirstate
 
2488
        if not (source._revision_id == NULL_REVISION or
 
2489
            source._revision_id in target.get_parent_ids()):
 
2490
            # TODO: what about ghosts? it may well need to 
 
2491
            # check for them explicitly.
 
2492
            return False
 
2493
        return True
 
2494
 
 
2495
InterTree.register_optimiser(InterDirStateTree)
 
2496
 
 
2497
 
 
2498
class Converter3to4(object):
 
2499
    """Perform an in-place upgrade of format 3 to format 4 trees."""
 
2500
 
 
2501
    def __init__(self):
 
2502
        self.target_format = WorkingTreeFormat4()
 
2503
 
 
2504
    def convert(self, tree):
 
2505
        # lock the control files not the tree, so that we dont get tree
 
2506
        # on-unlock behaviours, and so that noone else diddles with the 
 
2507
        # tree during upgrade.
 
2508
        tree._control_files.lock_write()
 
2509
        try:
 
2510
            tree.read_working_inventory()
 
2511
            self.create_dirstate_data(tree)
 
2512
            self.update_format(tree)
 
2513
            self.remove_xml_files(tree)
 
2514
        finally:
 
2515
            tree._control_files.unlock()
 
2516
 
 
2517
    def create_dirstate_data(self, tree):
 
2518
        """Create the dirstate based data for tree."""
 
2519
        local_path = tree.bzrdir.get_workingtree_transport(None
 
2520
            ).local_abspath('dirstate')
 
2521
        state = dirstate.DirState.from_tree(tree, local_path)
 
2522
        state.save()
 
2523
        state.unlock()
 
2524
 
 
2525
    def remove_xml_files(self, tree):
 
2526
        """Remove the oldformat 3 data."""
 
2527
        transport = tree.bzrdir.get_workingtree_transport(None)
 
2528
        for path in ['basis-inventory-cache', 'inventory', 'last-revision',
 
2529
            'pending-merges', 'stat-cache']:
 
2530
            try:
 
2531
                transport.delete(path)
 
2532
            except errors.NoSuchFile:
 
2533
                # some files are optional - just deal.
 
2534
                pass
 
2535
 
 
2536
    def update_format(self, tree):
 
2537
        """Change the format marker."""
 
2538
        tree._transport.put_bytes('format',
 
2539
            self.target_format.get_format_string(),
 
2540
            mode=tree._control_files._file_mode)