~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/_dirstate_helpers_c.pyx

  • Committer: Daniel Watkins
  • Date: 2007-08-02 18:04:35 UTC
  • mto: This revision was merged to the branch mainline in revision 2705.
  • Revision ID: d.m.watkins@warwick.ac.uk-20070802180435-cs6xalg5tkze209b
tests.blackbox.test_aliases now uses internals where appropriate.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2007, 2008 Canonical Ltd
 
1
# Copyright (C) 2007 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
12
12
#
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
 
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16
16
 
17
17
"""Helper functions for DirState.
18
18
 
19
19
This is the python implementation for DirState functions.
20
20
"""
21
21
 
22
 
import binascii
23
 
import bisect
24
 
import errno
25
 
import os
26
 
import stat
27
 
import sys
28
 
 
29
 
from bzrlib import cache_utf8, errors, osutils
30
22
from bzrlib.dirstate import DirState
31
 
from bzrlib.osutils import parent_directories, pathjoin, splitpath
32
 
 
33
 
 
34
 
# This is the Windows equivalent of ENOTDIR
35
 
# It is defined in pywin32.winerror, but we don't want a strong dependency for
36
 
# just an error code.
37
 
# XXX: Perhaps we could get it from a windows header ?
38
 
cdef int ERROR_PATH_NOT_FOUND
39
 
ERROR_PATH_NOT_FOUND = 3
40
 
cdef int ERROR_DIRECTORY
41
 
ERROR_DIRECTORY = 267
42
 
 
43
 
#python2.4 support, and other platform-dependent includes
44
 
cdef extern from "python-compat.h":
45
 
    unsigned long htonl(unsigned long)
 
23
 
46
24
 
47
25
# Give Pyrex some function definitions for it to understand.
48
26
# All of these are just hints to Pyrex, so that it can try to convert python
54
32
cdef extern from *:
55
33
    ctypedef unsigned long size_t
56
34
 
57
 
cdef extern from "_dirstate_helpers_pyx.h":
 
35
cdef extern from "stdint.h":
58
36
    ctypedef int intptr_t
59
37
 
60
38
 
61
 
 
62
39
cdef extern from "stdlib.h":
63
40
    unsigned long int strtoul(char *nptr, char **endptr, int base)
64
41
 
65
42
 
66
 
cdef extern from 'sys/stat.h':
67
 
    int S_ISDIR(int mode)
68
 
    int S_ISREG(int mode)
69
 
    # On win32, this actually comes from "python-compat.h"
70
 
    int S_ISLNK(int mode)
71
 
    int S_IXUSR
72
 
 
73
43
# These functions allow us access to a bit of the 'bare metal' of python
74
44
# objects, rather than going through the object abstraction. (For example,
75
45
# PyList_Append, rather than getting the 'append' attribute of the object, and
79
49
# will automatically Py_INCREF and Py_DECREF when appropriate. But for some
80
50
# inner loops, we don't need to do that at all, as the reference only lasts for
81
51
# a very short time.
82
 
# Note that the C API GetItem calls borrow references, so pyrex does the wrong
83
 
# thing if you declare e.g. object PyList_GetItem(object lst, int index) - you
84
 
# need to manually Py_INCREF yourself.
85
52
cdef extern from "Python.h":
86
 
    ctypedef int Py_ssize_t
87
 
    ctypedef struct PyObject:
88
 
        pass
89
53
    int PyList_Append(object lst, object item) except -1
90
54
    void *PyList_GetItem_object_void "PyList_GET_ITEM" (object lst, int index)
91
 
    void *PyList_GetItem_void_void "PyList_GET_ITEM" (void * lst, int index)
92
 
    object PyList_GET_ITEM(object lst, Py_ssize_t index)
93
55
    int PyList_CheckExact(object)
94
 
    Py_ssize_t PyList_GET_SIZE (object p)
95
56
 
96
57
    void *PyTuple_GetItem_void_void "PyTuple_GET_ITEM" (void* tpl, int index)
97
 
    object PyTuple_GetItem_void_object "PyTuple_GET_ITEM" (void* tpl, int index)
98
 
    object PyTuple_GET_ITEM(object tpl, Py_ssize_t index)
99
 
 
100
58
 
101
59
    char *PyString_AsString(object p)
102
 
    char *PyString_AsString_obj "PyString_AsString" (PyObject *string)
103
60
    char *PyString_AS_STRING_void "PyString_AS_STRING" (void *p)
104
 
    int PyString_AsStringAndSize(object str, char **buffer, Py_ssize_t *length) except -1
105
61
    object PyString_FromString(char *)
106
 
    object PyString_FromStringAndSize(char *, Py_ssize_t)
 
62
    object PyString_FromStringAndSize(char *, int)
107
63
    int PyString_Size(object p)
108
64
    int PyString_GET_SIZE_void "PyString_GET_SIZE" (void *p)
109
65
    int PyString_CheckExact(object p)
110
 
    void Py_INCREF(object o)
111
 
    void Py_DECREF(object o)
112
66
 
113
67
 
114
68
cdef extern from "string.h":
119
73
    # void *memrchr(void *s, int c, size_t len)
120
74
 
121
75
 
122
 
cdef void* _my_memrchr(void *s, int c, size_t n): # cannot_raise
 
76
cdef void* _my_memrchr(void *s, int c, size_t n):
123
77
    # memrchr seems to be a GNU extension, so we have to implement it ourselves
124
78
    cdef char *pos
125
79
    cdef char *start
154
108
    found = _my_memrchr(_s, _c[0], length)
155
109
    if found == NULL:
156
110
        return None
157
 
    return <char*>found - <char*>_s
158
 
 
159
 
cdef object safe_string_from_size(char *s, Py_ssize_t size):
160
 
    if size < 0:
161
 
        # XXX: On 64-bit machines the <int> cast causes a C compiler warning.
162
 
        raise AssertionError(
163
 
            'tried to create a string with an invalid size: %d @0x%x'
164
 
            % (size, <int>s))
165
 
    return PyString_FromStringAndSize(s, size)
166
 
 
167
 
 
168
 
cdef int _is_aligned(void *ptr): # cannot_raise
 
111
    return found - _s
 
112
 
 
113
 
 
114
cdef int _is_aligned(void *ptr):
169
115
    """Is this pointer aligned to an integer size offset?
170
116
 
171
117
    :return: 1 if this pointer is aligned, 0 otherwise.
173
119
    return ((<intptr_t>ptr) & ((sizeof(int))-1)) == 0
174
120
 
175
121
 
176
 
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2): # cannot_raise
 
122
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2):
177
123
    cdef unsigned char *cur1
178
124
    cdef unsigned char *cur2
179
125
    cdef unsigned char *end1
237
183
    return 0
238
184
 
239
185
 
240
 
def cmp_by_dirs(path1, path2):
 
186
def cmp_by_dirs_c(path1, path2):
241
187
    """Compare two paths directory by directory.
242
188
 
243
189
    This is equivalent to doing::
250
196
 
251
197
    :param path1: first path
252
198
    :param path2: second path
253
 
    :return: negative number if ``path1`` comes first,
 
199
    :return: positive number if ``path1`` comes first,
254
200
        0 if paths are equal,
255
 
        and positive number if ``path2`` sorts first
 
201
        and negative number if ``path2`` sorts first
256
202
    """
257
203
    if not PyString_CheckExact(path1):
258
204
        raise TypeError("'path1' must be a plain string, not %s: %r"
266
212
                        PyString_Size(path2))
267
213
 
268
214
 
269
 
def _cmp_path_by_dirblock(path1, path2):
 
215
def _cmp_path_by_dirblock_c(path1, path2):
270
216
    """Compare two paths based on what directory they are in.
271
217
 
272
218
    This generates a sort order, such that all children of a directory are
278
224
 
279
225
    :param path1: first path
280
226
    :param path2: the second path
281
 
    :return: negative number if ``path1`` comes first,
 
227
    :return: positive number if ``path1`` comes first,
282
228
        0 if paths are equal
283
 
        and a positive number if ``path2`` sorts first
 
229
        and a negative number if ``path2`` sorts first
284
230
    """
285
231
    if not PyString_CheckExact(path1):
286
232
        raise TypeError("'path1' must be a plain string, not %s: %r"
288
234
    if not PyString_CheckExact(path2):
289
235
        raise TypeError("'path2' must be a plain string, not %s: %r"
290
236
                        % (type(path2), path2))
291
 
    return _cmp_path_by_dirblock_intern(PyString_AsString(path1),
292
 
                                        PyString_Size(path1),
293
 
                                        PyString_AsString(path2),
294
 
                                        PyString_Size(path2))
295
 
 
296
 
 
297
 
cdef int _cmp_path_by_dirblock_intern(char *path1, int path1_len,
298
 
                                      char *path2, int path2_len): # cannot_raise
 
237
    return _cmp_path_by_dirblock(PyString_AsString(path1),
 
238
                                 PyString_Size(path1),
 
239
                                 PyString_AsString(path2),
 
240
                                 PyString_Size(path2))
 
241
 
 
242
 
 
243
cdef int _cmp_path_by_dirblock(char *path1, int path1_len,
 
244
                               char *path2, int path2_len):
299
245
    """Compare two paths by what directory they are in.
300
246
 
301
 
    see ``_cmp_path_by_dirblock`` for details.
 
247
    see ``_cmp_path_by_dirblock_c`` for details.
302
248
    """
303
249
    cdef char *dirname1
304
250
    cdef int dirname1_len
368
314
    return 1
369
315
 
370
316
 
371
 
def _bisect_path_left(paths, path):
 
317
def _bisect_path_left_c(paths, path):
372
318
    """Return the index where to insert path into paths.
373
319
 
374
320
    This uses a path-wise comparison so we get::
413
359
        cur = PyList_GetItem_object_void(paths, _mid)
414
360
        cur_cstr = PyString_AS_STRING_void(cur)
415
361
        cur_size = PyString_GET_SIZE_void(cur)
416
 
        if _cmp_path_by_dirblock_intern(cur_cstr, cur_size,
417
 
                                        path_cstr, path_size) < 0:
 
362
        if _cmp_path_by_dirblock(cur_cstr, cur_size, path_cstr, path_size) < 0:
418
363
            _lo = _mid + 1
419
364
        else:
420
365
            _hi = _mid
421
366
    return _lo
422
367
 
423
368
 
424
 
def _bisect_path_right(paths, path):
 
369
def _bisect_path_right_c(paths, path):
425
370
    """Return the index where to insert path into paths.
426
371
 
427
372
    This uses a path-wise comparison so we get::
466
411
        cur = PyList_GetItem_object_void(paths, _mid)
467
412
        cur_cstr = PyString_AS_STRING_void(cur)
468
413
        cur_size = PyString_GET_SIZE_void(cur)
469
 
        if _cmp_path_by_dirblock_intern(path_cstr, path_size,
470
 
                                        cur_cstr, cur_size) < 0:
 
414
        if _cmp_path_by_dirblock(path_cstr, path_size, cur_cstr, cur_size) < 0:
471
415
            _hi = _mid
472
416
        else:
473
417
            _lo = _mid + 1
474
418
    return _lo
475
419
 
476
420
 
477
 
def bisect_dirblock(dirblocks, dirname, lo=0, hi=None, cache=None):
 
421
def bisect_dirblock_c(dirblocks, dirname, lo=0, hi=None, cache=None):
478
422
    """Return the index where to insert dirname into the dirblocks.
479
423
 
480
424
    The return value idx is such that all directories blocks in dirblock[:idx]
526
470
cdef class Reader:
527
471
    """Maintain the current location, and return fields as you parse them."""
528
472
 
529
 
    cdef object state # The DirState object
530
473
    cdef object text # The overall string object
531
474
    cdef char *text_cstr # Pointer to the beginning of text
532
475
    cdef int text_size # Length of text
535
478
    cdef char *cur_cstr # Pointer to the current record
536
479
    cdef char *next # Pointer to the end of this record
537
480
 
538
 
    def __init__(self, text, state):
539
 
        self.state = state
 
481
    def __new__(self, text):
540
482
        self.text = text
541
483
        self.text_cstr = PyString_AsString(text)
542
484
        self.text_size = PyString_Size(text)
543
485
        self.end_cstr = self.text_cstr + self.text_size
544
486
        self.cur_cstr = self.text_cstr
545
487
 
546
 
    cdef char *get_next(self, int *size) except NULL:
 
488
    cdef char *get_next(self, int *size):
547
489
        """Return a pointer to the start of the next field."""
548
490
        cdef char *next
549
 
        cdef Py_ssize_t extra_len
550
 
 
551
 
        if self.cur_cstr == NULL:
552
 
            raise AssertionError('get_next() called when cur_str is NULL')
553
 
        elif self.cur_cstr >= self.end_cstr:
554
 
            raise AssertionError('get_next() called when there are no chars'
555
 
                                 ' left')
556
491
        next = self.cur_cstr
557
 
        self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr - next)
558
 
        if self.cur_cstr == NULL:
559
 
            extra_len = self.end_cstr - next
560
 
            raise errors.DirstateCorrupt(self.state,
561
 
                'failed to find trailing NULL (\\0).'
562
 
                ' Trailing garbage: %r'
563
 
                % safe_string_from_size(next, extra_len))
 
492
        self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr-next)
564
493
        size[0] = self.cur_cstr - next
565
494
        self.cur_cstr = self.cur_cstr + 1
566
495
        return next
570
499
        cdef int size
571
500
        cdef char *next
572
501
        next = self.get_next(&size)
573
 
        return safe_string_from_size(next, size)
 
502
        return PyString_FromStringAndSize(next, size)
574
503
 
575
504
    cdef int _init(self) except -1:
576
505
        """Get the pointer ready.
641
570
                       # <object>
642
571
                       PyString_AS_STRING_void(p_current_dirname[0]),
643
572
                       cur_size+1) != 0):
644
 
            dirname = safe_string_from_size(dirname_cstr, cur_size)
 
573
            dirname = PyString_FromStringAndSize(dirname_cstr, cur_size)
645
574
            p_current_dirname[0] = <void*>dirname
646
575
            new_block[0] = 1
647
576
        else:
692
621
        # marker.
693
622
        trailing = self.get_next(&cur_size)
694
623
        if cur_size != 1 or trailing[0] != c'\n':
695
 
            raise errors.DirstateCorrupt(self.state,
 
624
            raise AssertionError(
696
625
                'Bad parse, we expected to end on \\n, not: %d %s: %s'
697
 
                % (cur_size, safe_string_from_size(trailing, cur_size),
 
626
                % (cur_size, PyString_FromStringAndSize(trailing, cur_size),
698
627
                   ret))
699
628
        return ret
700
629
 
701
 
    def _parse_dirblocks(self):
 
630
    def _parse_dirblocks(self, state):
702
631
        """Parse all dirblocks in the state file."""
703
632
        cdef int num_trees
704
633
        cdef object current_block
708
637
        cdef int expected_entry_count
709
638
        cdef int entry_count
710
639
 
711
 
        num_trees = self.state._num_present_parents() + 1
712
 
        expected_entry_count = self.state._num_entries
 
640
        num_trees = state._num_present_parents() + 1
 
641
        expected_entry_count = state._num_entries
713
642
 
714
643
        # Ignore the first record
715
644
        self._init()
716
645
 
717
646
        current_block = []
718
 
        dirblocks = [('', current_block), ('', [])]
719
 
        self.state._dirblocks = dirblocks
 
647
        state._dirblocks = [('', current_block), ('', [])]
720
648
        obj = ''
721
649
        current_dirname = <void*>obj
722
650
        new_block = 0
734
662
            if new_block:
735
663
                # new block - different dirname
736
664
                current_block = []
737
 
                PyList_Append(dirblocks,
 
665
                PyList_Append(state._dirblocks,
738
666
                              (<object>current_dirname, current_block))
739
667
            PyList_Append(current_block, entry)
740
668
            entry_count = entry_count + 1
741
669
        if entry_count != expected_entry_count:
742
 
            raise errors.DirstateCorrupt(self.state,
743
 
                    'We read the wrong number of entries.'
 
670
            raise AssertionError('We read the wrong number of entries.'
744
671
                    ' We expected to read %s, but read %s'
745
672
                    % (expected_entry_count, entry_count))
746
 
        self.state._split_root_dirblock_into_contents()
747
 
 
748
 
 
749
 
def _read_dirblocks(state):
 
673
        state._split_root_dirblock_into_contents()
 
674
 
 
675
 
 
676
def _read_dirblocks_c(state):
750
677
    """Read in the dirblocks for the given DirState object.
751
678
 
752
679
    This is tightly bound to the DirState internal representation. It should be
762
689
    text = state._state_file.read()
763
690
    # TODO: check the crc checksums. crc_measured = zlib.crc32(text)
764
691
 
765
 
    reader = Reader(text, state)
 
692
    reader = Reader(text)
766
693
 
767
 
    reader._parse_dirblocks()
 
694
    reader._parse_dirblocks(state)
768
695
    state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
769
 
 
770
 
 
771
 
cdef int minikind_from_mode(int mode): # cannot_raise
772
 
    # in order of frequency:
773
 
    if S_ISREG(mode):
774
 
        return c"f"
775
 
    if S_ISDIR(mode):
776
 
        return c"d"
777
 
    if S_ISLNK(mode):
778
 
        return c"l"
779
 
    return 0
780
 
 
781
 
 
782
 
_encode = binascii.b2a_base64
783
 
 
784
 
 
785
 
from struct import pack
786
 
cdef _pack_stat(stat_value):
787
 
    """return a string representing the stat value's key fields.
788
 
 
789
 
    :param stat_value: A stat oject with st_size, st_mtime, st_ctime, st_dev,
790
 
        st_ino and st_mode fields.
791
 
    """
792
 
    cdef char result[6*4] # 6 long ints
793
 
    cdef int *aliased
794
 
    aliased = <int *>result
795
 
    aliased[0] = htonl(stat_value.st_size)
796
 
    aliased[1] = htonl(int(stat_value.st_mtime))
797
 
    aliased[2] = htonl(int(stat_value.st_ctime))
798
 
    aliased[3] = htonl(stat_value.st_dev)
799
 
    aliased[4] = htonl(stat_value.st_ino & 0xFFFFFFFF)
800
 
    aliased[5] = htonl(stat_value.st_mode)
801
 
    packed = PyString_FromStringAndSize(result, 6*4)
802
 
    return _encode(packed)[:-1]
803
 
 
804
 
 
805
 
def update_entry(self, entry, abspath, stat_value):
806
 
    """Update the entry based on what is actually on disk.
807
 
 
808
 
    This function only calculates the sha if it needs to - if the entry is
809
 
    uncachable, or clearly different to the first parent's entry, no sha
810
 
    is calculated, and None is returned.
811
 
 
812
 
    :param entry: This is the dirblock entry for the file in question.
813
 
    :param abspath: The path on disk for this file.
814
 
    :param stat_value: (optional) if we already have done a stat on the
815
 
        file, re-use it.
816
 
    :return: None, or The sha1 hexdigest of the file (40 bytes) or link
817
 
        target of a symlink.
818
 
    """
819
 
    return _update_entry(self, entry, abspath, stat_value)
820
 
 
821
 
 
822
 
cdef _update_entry(self, entry, abspath, stat_value):
823
 
    """Update the entry based on what is actually on disk.
824
 
 
825
 
    This function only calculates the sha if it needs to - if the entry is
826
 
    uncachable, or clearly different to the first parent's entry, no sha
827
 
    is calculated, and None is returned.
828
 
 
829
 
    :param self: The dirstate object this is operating on.
830
 
    :param entry: This is the dirblock entry for the file in question.
831
 
    :param abspath: The path on disk for this file.
832
 
    :param stat_value: The stat value done on the path.
833
 
    :return: None, or The sha1 hexdigest of the file (40 bytes) or link
834
 
        target of a symlink.
835
 
    """
836
 
    # TODO - require pyrex 0.9.8, then use a pyd file to define access to the
837
 
    # _st mode of the compiled stat objects.
838
 
    cdef int minikind, saved_minikind
839
 
    cdef void * details
840
 
    minikind = minikind_from_mode(stat_value.st_mode)
841
 
    if 0 == minikind:
842
 
        return None
843
 
    packed_stat = _pack_stat(stat_value)
844
 
    details = PyList_GetItem_void_void(PyTuple_GetItem_void_void(<void *>entry, 1), 0)
845
 
    saved_minikind = PyString_AsString_obj(<PyObject *>PyTuple_GetItem_void_void(details, 0))[0]
846
 
    if minikind == c'd' and saved_minikind == c't':
847
 
        minikind = c't'
848
 
    saved_link_or_sha1 = PyTuple_GetItem_void_object(details, 1)
849
 
    saved_file_size = PyTuple_GetItem_void_object(details, 2)
850
 
    saved_executable = PyTuple_GetItem_void_object(details, 3)
851
 
    saved_packed_stat = PyTuple_GetItem_void_object(details, 4)
852
 
    # Deal with pyrex decrefing the objects
853
 
    Py_INCREF(saved_link_or_sha1)
854
 
    Py_INCREF(saved_file_size)
855
 
    Py_INCREF(saved_executable)
856
 
    Py_INCREF(saved_packed_stat)
857
 
    #(saved_minikind, saved_link_or_sha1, saved_file_size,
858
 
    # saved_executable, saved_packed_stat) = entry[1][0]
859
 
 
860
 
    if (minikind == saved_minikind
861
 
        and packed_stat == saved_packed_stat):
862
 
        # The stat hasn't changed since we saved, so we can re-use the
863
 
        # saved sha hash.
864
 
        if minikind == c'd':
865
 
            return None
866
 
 
867
 
        # size should also be in packed_stat
868
 
        if saved_file_size == stat_value.st_size:
869
 
            return saved_link_or_sha1
870
 
 
871
 
    # If we have gotten this far, that means that we need to actually
872
 
    # process this entry.
873
 
    link_or_sha1 = None
874
 
    if minikind == c'f':
875
 
        executable = self._is_executable(stat_value.st_mode,
876
 
                                         saved_executable)
877
 
        if self._cutoff_time is None:
878
 
            self._sha_cutoff_time()
879
 
        if (stat_value.st_mtime < self._cutoff_time
880
 
            and stat_value.st_ctime < self._cutoff_time
881
 
            and len(entry[1]) > 1
882
 
            and entry[1][1][0] != 'a'):
883
 
                # Could check for size changes for further optimised
884
 
                # avoidance of sha1's. However the most prominent case of
885
 
                # over-shaing is during initial add, which this catches.
886
 
            link_or_sha1 = self._sha1_file(abspath)
887
 
            entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
888
 
                           executable, packed_stat)
889
 
        else:
890
 
            entry[1][0] = ('f', '', stat_value.st_size,
891
 
                           executable, DirState.NULLSTAT)
892
 
    elif minikind == c'd':
893
 
        link_or_sha1 = None
894
 
        entry[1][0] = ('d', '', 0, False, packed_stat)
895
 
        if saved_minikind != c'd':
896
 
            # This changed from something into a directory. Make sure we
897
 
            # have a directory block for it. This doesn't happen very
898
 
            # often, so this doesn't have to be super fast.
899
 
            block_index, entry_index, dir_present, file_present = \
900
 
                self._get_block_entry_index(entry[0][0], entry[0][1], 0)
901
 
            self._ensure_block(block_index, entry_index,
902
 
                               pathjoin(entry[0][0], entry[0][1]))
903
 
    elif minikind == c'l':
904
 
        link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
905
 
        if self._cutoff_time is None:
906
 
            self._sha_cutoff_time()
907
 
        if (stat_value.st_mtime < self._cutoff_time
908
 
            and stat_value.st_ctime < self._cutoff_time):
909
 
            entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
910
 
                           False, packed_stat)
911
 
        else:
912
 
            entry[1][0] = ('l', '', stat_value.st_size,
913
 
                           False, DirState.NULLSTAT)
914
 
    self._dirblock_state = DirState.IN_MEMORY_MODIFIED
915
 
    return link_or_sha1
916
 
 
917
 
 
918
 
# TODO: Do we want to worry about exceptions here?
919
 
cdef char _minikind_from_string(object string) except? -1:
920
 
    """Convert a python string to a char."""
921
 
    return PyString_AsString(string)[0]
922
 
 
923
 
 
924
 
cdef object _kind_absent
925
 
cdef object _kind_file
926
 
cdef object _kind_directory
927
 
cdef object _kind_symlink
928
 
cdef object _kind_relocated
929
 
cdef object _kind_tree_reference
930
 
_kind_absent = "absent"
931
 
_kind_file = "file"
932
 
_kind_directory = "directory"
933
 
_kind_symlink = "symlink"
934
 
_kind_relocated = "relocated"
935
 
_kind_tree_reference = "tree-reference"
936
 
 
937
 
 
938
 
cdef object _minikind_to_kind(char minikind):
939
 
    """Create a string kind for minikind."""
940
 
    cdef char _minikind[1]
941
 
    if minikind == c'f':
942
 
        return _kind_file
943
 
    elif minikind == c'd':
944
 
        return _kind_directory
945
 
    elif minikind == c'a':
946
 
        return _kind_absent
947
 
    elif minikind == c'r':
948
 
        return _kind_relocated
949
 
    elif minikind == c'l':
950
 
        return _kind_symlink
951
 
    elif minikind == c't':
952
 
        return _kind_tree_reference
953
 
    _minikind[0] = minikind
954
 
    raise KeyError(PyString_FromStringAndSize(_minikind, 1))
955
 
 
956
 
 
957
 
cdef int _versioned_minikind(char minikind): # cannot_raise
958
 
    """Return non-zero if minikind is in fltd"""
959
 
    return (minikind == c'f' or
960
 
            minikind == c'd' or
961
 
            minikind == c'l' or
962
 
            minikind == c't')
963
 
 
964
 
 
965
 
cdef class ProcessEntryC:
966
 
 
967
 
    cdef int doing_consistency_expansion
968
 
    cdef object old_dirname_to_file_id # dict
969
 
    cdef object new_dirname_to_file_id # dict
970
 
    cdef object last_source_parent
971
 
    cdef object last_target_parent
972
 
    cdef int include_unchanged
973
 
    cdef int partial
974
 
    cdef object use_filesystem_for_exec
975
 
    cdef object utf8_decode
976
 
    cdef readonly object searched_specific_files
977
 
    cdef readonly object searched_exact_paths
978
 
    cdef object search_specific_files
979
 
    # The parents up to the root of the paths we are searching.
980
 
    # After all normal paths are returned, these specific items are returned.
981
 
    cdef object search_specific_file_parents
982
 
    cdef object state
983
 
    # Current iteration variables:
984
 
    cdef object current_root
985
 
    cdef object current_root_unicode
986
 
    cdef object root_entries
987
 
    cdef int root_entries_pos, root_entries_len
988
 
    cdef object root_abspath
989
 
    cdef int source_index, target_index
990
 
    cdef int want_unversioned
991
 
    cdef object tree
992
 
    cdef object dir_iterator
993
 
    cdef int block_index
994
 
    cdef object current_block
995
 
    cdef int current_block_pos
996
 
    cdef object current_block_list
997
 
    cdef object current_dir_info
998
 
    cdef object current_dir_list
999
 
    cdef object _pending_consistent_entries # list
1000
 
    cdef int path_index
1001
 
    cdef object root_dir_info
1002
 
    cdef object bisect_left
1003
 
    cdef object pathjoin
1004
 
    cdef object fstat
1005
 
    # A set of the ids we've output when doing partial output.
1006
 
    cdef object seen_ids
1007
 
    cdef object sha_file
1008
 
 
1009
 
    def __init__(self, include_unchanged, use_filesystem_for_exec,
1010
 
        search_specific_files, state, source_index, target_index,
1011
 
        want_unversioned, tree):
1012
 
        self.doing_consistency_expansion = 0
1013
 
        self.old_dirname_to_file_id = {}
1014
 
        self.new_dirname_to_file_id = {}
1015
 
        # Are we doing a partial iter_changes?
1016
 
        self.partial = set(['']).__ne__(search_specific_files)
1017
 
        # Using a list so that we can access the values and change them in
1018
 
        # nested scope. Each one is [path, file_id, entry]
1019
 
        self.last_source_parent = [None, None]
1020
 
        self.last_target_parent = [None, None]
1021
 
        if include_unchanged is None:
1022
 
            self.include_unchanged = False
1023
 
        else:
1024
 
            self.include_unchanged = int(include_unchanged)
1025
 
        self.use_filesystem_for_exec = use_filesystem_for_exec
1026
 
        self.utf8_decode = cache_utf8._utf8_decode
1027
 
        # for all search_indexs in each path at or under each element of
1028
 
        # search_specific_files, if the detail is relocated: add the id, and
1029
 
        # add the relocated path as one to search if its not searched already.
1030
 
        # If the detail is not relocated, add the id.
1031
 
        self.searched_specific_files = set()
1032
 
        # When we search exact paths without expanding downwards, we record
1033
 
        # that here.
1034
 
        self.searched_exact_paths = set()
1035
 
        self.search_specific_files = search_specific_files
1036
 
        # The parents up to the root of the paths we are searching.
1037
 
        # After all normal paths are returned, these specific items are returned.
1038
 
        self.search_specific_file_parents = set()
1039
 
        # The ids we've sent out in the delta.
1040
 
        self.seen_ids = set()
1041
 
        self.state = state
1042
 
        self.current_root = None
1043
 
        self.current_root_unicode = None
1044
 
        self.root_entries = None
1045
 
        self.root_entries_pos = 0
1046
 
        self.root_entries_len = 0
1047
 
        self.root_abspath = None
1048
 
        if source_index is None:
1049
 
            self.source_index = -1
1050
 
        else:
1051
 
            self.source_index = source_index
1052
 
        self.target_index = target_index
1053
 
        self.want_unversioned = want_unversioned
1054
 
        self.tree = tree
1055
 
        self.dir_iterator = None
1056
 
        self.block_index = -1
1057
 
        self.current_block = None
1058
 
        self.current_block_list = None
1059
 
        self.current_block_pos = -1
1060
 
        self.current_dir_info = None
1061
 
        self.current_dir_list = None
1062
 
        self._pending_consistent_entries = []
1063
 
        self.path_index = 0
1064
 
        self.root_dir_info = None
1065
 
        self.bisect_left = bisect.bisect_left
1066
 
        self.pathjoin = osutils.pathjoin
1067
 
        self.fstat = os.fstat
1068
 
        self.sha_file = osutils.sha_file
1069
 
        if target_index != 0:
1070
 
            # A lot of code in here depends on target_index == 0
1071
 
            raise errors.BzrError('unsupported target index')
1072
 
 
1073
 
    cdef _process_entry(self, entry, path_info):
1074
 
        """Compare an entry and real disk to generate delta information.
1075
 
 
1076
 
        :param path_info: top_relpath, basename, kind, lstat, abspath for
1077
 
            the path of entry. If None, then the path is considered absent in 
1078
 
            the target (Perhaps we should pass in a concrete entry for this ?)
1079
 
            Basename is returned as a utf8 string because we expect this
1080
 
            tuple will be ignored, and don't want to take the time to
1081
 
            decode.
1082
 
        :return: (iter_changes_result, changed). If the entry has not been
1083
 
            handled then changed is None. Otherwise it is False if no content
1084
 
            or metadata changes have occured, and True if any content or
1085
 
            metadata change has occurred. If self.include_unchanged is True then
1086
 
            if changed is not None, iter_changes_result will always be a result
1087
 
            tuple. Otherwise, iter_changes_result is None unless changed is
1088
 
            True.
1089
 
        """
1090
 
        cdef char target_minikind
1091
 
        cdef char source_minikind
1092
 
        cdef object file_id
1093
 
        cdef int content_change
1094
 
        cdef object details_list
1095
 
        file_id = None
1096
 
        details_list = entry[1]
1097
 
        if -1 == self.source_index:
1098
 
            source_details = DirState.NULL_PARENT_DETAILS
1099
 
        else:
1100
 
            source_details = details_list[self.source_index]
1101
 
        target_details = details_list[self.target_index]
1102
 
        target_minikind = _minikind_from_string(target_details[0])
1103
 
        if path_info is not None and _versioned_minikind(target_minikind):
1104
 
            if self.target_index != 0:
1105
 
                raise AssertionError("Unsupported target index %d" %
1106
 
                                     self.target_index)
1107
 
            link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1108
 
            # The entry may have been modified by update_entry
1109
 
            target_details = details_list[self.target_index]
1110
 
            target_minikind = _minikind_from_string(target_details[0])
1111
 
        else:
1112
 
            link_or_sha1 = None
1113
 
        # the rest of this function is 0.3 seconds on 50K paths, or
1114
 
        # 0.000006 seconds per call.
1115
 
        source_minikind = _minikind_from_string(source_details[0])
1116
 
        if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1117
 
            and _versioned_minikind(target_minikind)):
1118
 
            # claimed content in both: diff
1119
 
            #   r    | fdlt   |      | add source to search, add id path move and perform
1120
 
            #        |        |      | diff check on source-target
1121
 
            #   r    | fdlt   |  a   | dangling file that was present in the basis.
1122
 
            #        |        |      | ???
1123
 
            if source_minikind != c'r':
1124
 
                old_dirname = entry[0][0]
1125
 
                old_basename = entry[0][1]
1126
 
                old_path = path = None
1127
 
            else:
1128
 
                # add the source to the search path to find any children it
1129
 
                # has.  TODO ? : only add if it is a container ?
1130
 
                if (not self.doing_consistency_expansion and 
1131
 
                    not osutils.is_inside_any(self.searched_specific_files,
1132
 
                                             source_details[1])):
1133
 
                    self.search_specific_files.add(source_details[1])
1134
 
                    # expanding from a user requested path, parent expansion
1135
 
                    # for delta consistency happens later.
1136
 
                # generate the old path; this is needed for stating later
1137
 
                # as well.
1138
 
                old_path = source_details[1]
1139
 
                old_dirname, old_basename = os.path.split(old_path)
1140
 
                path = self.pathjoin(entry[0][0], entry[0][1])
1141
 
                old_entry = self.state._get_entry(self.source_index,
1142
 
                                             path_utf8=old_path)
1143
 
                # update the source details variable to be the real
1144
 
                # location.
1145
 
                if old_entry == (None, None):
1146
 
                    raise errors.CorruptDirstate(self.state._filename,
1147
 
                        "entry '%s/%s' is considered renamed from %r"
1148
 
                        " but source does not exist\n"
1149
 
                        "entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1150
 
                source_details = old_entry[1][self.source_index]
1151
 
                source_minikind = _minikind_from_string(source_details[0])
1152
 
            if path_info is None:
1153
 
                # the file is missing on disk, show as removed.
1154
 
                content_change = 1
1155
 
                target_kind = None
1156
 
                target_exec = False
1157
 
            else:
1158
 
                # source and target are both versioned and disk file is present.
1159
 
                target_kind = path_info[2]
1160
 
                if target_kind == 'directory':
1161
 
                    if path is None:
1162
 
                        old_path = path = self.pathjoin(old_dirname, old_basename)
1163
 
                    file_id = entry[0][2]
1164
 
                    self.new_dirname_to_file_id[path] = file_id
1165
 
                    if source_minikind != c'd':
1166
 
                        content_change = 1
1167
 
                    else:
1168
 
                        # directories have no fingerprint
1169
 
                        content_change = 0
1170
 
                    target_exec = False
1171
 
                elif target_kind == 'file':
1172
 
                    if source_minikind != c'f':
1173
 
                        content_change = 1
1174
 
                    else:
1175
 
                        # Check the sha. We can't just rely on the size as
1176
 
                        # content filtering may mean differ sizes actually
1177
 
                        # map to the same content
1178
 
                        if link_or_sha1 is None:
1179
 
                            # Stat cache miss:
1180
 
                            statvalue, link_or_sha1 = \
1181
 
                                self.state._sha1_provider.stat_and_sha1(
1182
 
                                path_info[4])
1183
 
                            self.state._observed_sha1(entry, link_or_sha1,
1184
 
                                statvalue)
1185
 
                        content_change = (link_or_sha1 != source_details[1])
1186
 
                    # Target details is updated at update_entry time
1187
 
                    if self.use_filesystem_for_exec:
1188
 
                        # We don't need S_ISREG here, because we are sure
1189
 
                        # we are dealing with a file.
1190
 
                        target_exec = bool(S_IXUSR & path_info[3].st_mode)
1191
 
                    else:
1192
 
                        target_exec = target_details[3]
1193
 
                elif target_kind == 'symlink':
1194
 
                    if source_minikind != c'l':
1195
 
                        content_change = 1
1196
 
                    else:
1197
 
                        content_change = (link_or_sha1 != source_details[1])
1198
 
                    target_exec = False
1199
 
                elif target_kind == 'tree-reference':
1200
 
                    if source_minikind != c't':
1201
 
                        content_change = 1
1202
 
                    else:
1203
 
                        content_change = 0
1204
 
                    target_exec = False
1205
 
                else:
1206
 
                    if path is None:
1207
 
                        path = self.pathjoin(old_dirname, old_basename)
1208
 
                    raise errors.BadFileKindError(path, path_info[2])
1209
 
            if source_minikind == c'd':
1210
 
                if path is None:
1211
 
                    old_path = path = self.pathjoin(old_dirname, old_basename)
1212
 
                if file_id is None:
1213
 
                    file_id = entry[0][2]
1214
 
                self.old_dirname_to_file_id[old_path] = file_id
1215
 
            # parent id is the entry for the path in the target tree
1216
 
            if old_basename and old_dirname == self.last_source_parent[0]:
1217
 
                # use a cached hit for non-root source entries.
1218
 
                source_parent_id = self.last_source_parent[1]
1219
 
            else:
1220
 
                try:
1221
 
                    source_parent_id = self.old_dirname_to_file_id[old_dirname]
1222
 
                except KeyError:
1223
 
                    source_parent_entry = self.state._get_entry(self.source_index,
1224
 
                                                           path_utf8=old_dirname)
1225
 
                    source_parent_id = source_parent_entry[0][2]
1226
 
                if source_parent_id == entry[0][2]:
1227
 
                    # This is the root, so the parent is None
1228
 
                    source_parent_id = None
1229
 
                else:
1230
 
                    self.last_source_parent[0] = old_dirname
1231
 
                    self.last_source_parent[1] = source_parent_id
1232
 
            new_dirname = entry[0][0]
1233
 
            if entry[0][1] and new_dirname == self.last_target_parent[0]:
1234
 
                # use a cached hit for non-root target entries.
1235
 
                target_parent_id = self.last_target_parent[1]
1236
 
            else:
1237
 
                try:
1238
 
                    target_parent_id = self.new_dirname_to_file_id[new_dirname]
1239
 
                except KeyError:
1240
 
                    # TODO: We don't always need to do the lookup, because the
1241
 
                    #       parent entry will be the same as the source entry.
1242
 
                    target_parent_entry = self.state._get_entry(self.target_index,
1243
 
                                                           path_utf8=new_dirname)
1244
 
                    if target_parent_entry == (None, None):
1245
 
                        raise AssertionError(
1246
 
                            "Could not find target parent in wt: %s\nparent of: %s"
1247
 
                            % (new_dirname, entry))
1248
 
                    target_parent_id = target_parent_entry[0][2]
1249
 
                if target_parent_id == entry[0][2]:
1250
 
                    # This is the root, so the parent is None
1251
 
                    target_parent_id = None
1252
 
                else:
1253
 
                    self.last_target_parent[0] = new_dirname
1254
 
                    self.last_target_parent[1] = target_parent_id
1255
 
 
1256
 
            source_exec = source_details[3]
1257
 
            changed = (content_change
1258
 
                or source_parent_id != target_parent_id
1259
 
                or old_basename != entry[0][1]
1260
 
                or source_exec != target_exec
1261
 
                )
1262
 
            if not changed and not self.include_unchanged:
1263
 
                return None, False
1264
 
            else:
1265
 
                if old_path is None:
1266
 
                    path = self.pathjoin(old_dirname, old_basename)
1267
 
                    old_path = path
1268
 
                    old_path_u = self.utf8_decode(old_path)[0]
1269
 
                    path_u = old_path_u
1270
 
                else:
1271
 
                    old_path_u = self.utf8_decode(old_path)[0]
1272
 
                    if old_path == path:
1273
 
                        path_u = old_path_u
1274
 
                    else:
1275
 
                        path_u = self.utf8_decode(path)[0]
1276
 
                source_kind = _minikind_to_kind(source_minikind)
1277
 
                return (entry[0][2],
1278
 
                       (old_path_u, path_u),
1279
 
                       content_change,
1280
 
                       (True, True),
1281
 
                       (source_parent_id, target_parent_id),
1282
 
                       (self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1283
 
                       (source_kind, target_kind),
1284
 
                       (source_exec, target_exec)), changed
1285
 
        elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1286
 
            # looks like a new file
1287
 
            path = self.pathjoin(entry[0][0], entry[0][1])
1288
 
            # parent id is the entry for the path in the target tree
1289
 
            # TODO: these are the same for an entire directory: cache em.
1290
 
            parent_entry = self.state._get_entry(self.target_index,
1291
 
                                                 path_utf8=entry[0][0])
1292
 
            if parent_entry is None:
1293
 
                raise errors.DirstateCorrupt(self.state,
1294
 
                    "We could not find the parent entry in index %d"
1295
 
                    " for the entry: %s"
1296
 
                    % (self.target_index, entry[0]))
1297
 
            parent_id = parent_entry[0][2]
1298
 
            if parent_id == entry[0][2]:
1299
 
                parent_id = None
1300
 
            if path_info is not None:
1301
 
                # Present on disk:
1302
 
                if self.use_filesystem_for_exec:
1303
 
                    # We need S_ISREG here, because we aren't sure if this
1304
 
                    # is a file or not.
1305
 
                    target_exec = bool(
1306
 
                        S_ISREG(path_info[3].st_mode)
1307
 
                        and S_IXUSR & path_info[3].st_mode)
1308
 
                else:
1309
 
                    target_exec = target_details[3]
1310
 
                return (entry[0][2],
1311
 
                       (None, self.utf8_decode(path)[0]),
1312
 
                       True,
1313
 
                       (False, True),
1314
 
                       (None, parent_id),
1315
 
                       (None, self.utf8_decode(entry[0][1])[0]),
1316
 
                       (None, path_info[2]),
1317
 
                       (None, target_exec)), True
1318
 
            else:
1319
 
                # Its a missing file, report it as such.
1320
 
                return (entry[0][2],
1321
 
                       (None, self.utf8_decode(path)[0]),
1322
 
                       False,
1323
 
                       (False, True),
1324
 
                       (None, parent_id),
1325
 
                       (None, self.utf8_decode(entry[0][1])[0]),
1326
 
                       (None, None),
1327
 
                       (None, False)), True
1328
 
        elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1329
 
            # unversioned, possibly, or possibly not deleted: we dont care.
1330
 
            # if its still on disk, *and* theres no other entry at this
1331
 
            # path [we dont know this in this routine at the moment -
1332
 
            # perhaps we should change this - then it would be an unknown.
1333
 
            old_path = self.pathjoin(entry[0][0], entry[0][1])
1334
 
            # parent id is the entry for the path in the target tree
1335
 
            parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1336
 
            if parent_id == entry[0][2]:
1337
 
                parent_id = None
1338
 
            return (entry[0][2],
1339
 
                   (self.utf8_decode(old_path)[0], None),
1340
 
                   True,
1341
 
                   (True, False),
1342
 
                   (parent_id, None),
1343
 
                   (self.utf8_decode(entry[0][1])[0], None),
1344
 
                   (_minikind_to_kind(source_minikind), None),
1345
 
                   (source_details[3], None)), True
1346
 
        elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1347
 
            # a rename; could be a true rename, or a rename inherited from
1348
 
            # a renamed parent. TODO: handle this efficiently. Its not
1349
 
            # common case to rename dirs though, so a correct but slow
1350
 
            # implementation will do.
1351
 
            if (not self.doing_consistency_expansion and 
1352
 
                not osutils.is_inside_any(self.searched_specific_files,
1353
 
                    target_details[1])):
1354
 
                self.search_specific_files.add(target_details[1])
1355
 
                # We don't expand the specific files parents list here as
1356
 
                # the path is absent in target and won't create a delta with
1357
 
                # missing parent.
1358
 
        elif ((source_minikind == c'r' or source_minikind == c'a') and
1359
 
              (target_minikind == c'r' or target_minikind == c'a')):
1360
 
            # neither of the selected trees contain this path,
1361
 
            # so skip over it. This is not currently directly tested, but
1362
 
            # is indirectly via test_too_much.TestCommands.test_conflicts.
1363
 
            pass
1364
 
        else:
1365
 
            raise AssertionError("don't know how to compare "
1366
 
                "source_minikind=%r, target_minikind=%r"
1367
 
                % (source_minikind, target_minikind))
1368
 
            ## import pdb;pdb.set_trace()
1369
 
        return None, None
1370
 
 
1371
 
    def __iter__(self):
1372
 
        return self
1373
 
 
1374
 
    def iter_changes(self):
1375
 
        return self
1376
 
 
1377
 
    cdef int _gather_result_for_consistency(self, result) except -1:
1378
 
        """Check a result we will yield to make sure we are consistent later.
1379
 
        
1380
 
        This gathers result's parents into a set to output later.
1381
 
 
1382
 
        :param result: A result tuple.
1383
 
        """
1384
 
        if not self.partial or not result[0]:
1385
 
            return 0
1386
 
        self.seen_ids.add(result[0])
1387
 
        new_path = result[1][1]
1388
 
        if new_path:
1389
 
            # Not the root and not a delete: queue up the parents of the path.
1390
 
            self.search_specific_file_parents.update(
1391
 
                osutils.parent_directories(new_path.encode('utf8')))
1392
 
            # Add the root directory which parent_directories does not
1393
 
            # provide.
1394
 
            self.search_specific_file_parents.add('')
1395
 
        return 0
1396
 
 
1397
 
    cdef int _update_current_block(self) except -1:
1398
 
        if (self.block_index < len(self.state._dirblocks) and
1399
 
            osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1400
 
            self.current_block = self.state._dirblocks[self.block_index]
1401
 
            self.current_block_list = self.current_block[1]
1402
 
            self.current_block_pos = 0
1403
 
        else:
1404
 
            self.current_block = None
1405
 
            self.current_block_list = None
1406
 
        return 0
1407
 
 
1408
 
    def __next__(self):
1409
 
        # Simple thunk to allow tail recursion without pyrex confusion
1410
 
        return self._iter_next()
1411
 
 
1412
 
    cdef _iter_next(self):
1413
 
        """Iterate over the changes."""
1414
 
        # This function single steps through an iterator. As such while loops
1415
 
        # are often exited by 'return' - the code is structured so that the
1416
 
        # next call into the function will return to the same while loop. Note
1417
 
        # that all flow control needed to re-reach that step is reexecuted,
1418
 
        # which can be a performance problem. It has not yet been tuned to
1419
 
        # minimise this; a state machine is probably the simplest restructuring
1420
 
        # to both minimise this overhead and make the code considerably more
1421
 
        # understandable.
1422
 
 
1423
 
        # sketch: 
1424
 
        # compare source_index and target_index at or under each element of search_specific_files.
1425
 
        # follow the following comparison table. Note that we only want to do diff operations when
1426
 
        # the target is fdl because thats when the walkdirs logic will have exposed the pathinfo 
1427
 
        # for the target.
1428
 
        # cases:
1429
 
        # 
1430
 
        # Source | Target | disk | action
1431
 
        #   r    | fdlt   |      | add source to search, add id path move and perform
1432
 
        #        |        |      | diff check on source-target
1433
 
        #   r    | fdlt   |  a   | dangling file that was present in the basis. 
1434
 
        #        |        |      | ???
1435
 
        #   r    |  a     |      | add source to search
1436
 
        #   r    |  a     |  a   | 
1437
 
        #   r    |  r     |      | this path is present in a non-examined tree, skip.
1438
 
        #   r    |  r     |  a   | this path is present in a non-examined tree, skip.
1439
 
        #   a    | fdlt   |      | add new id
1440
 
        #   a    | fdlt   |  a   | dangling locally added file, skip
1441
 
        #   a    |  a     |      | not present in either tree, skip
1442
 
        #   a    |  a     |  a   | not present in any tree, skip
1443
 
        #   a    |  r     |      | not present in either tree at this path, skip as it
1444
 
        #        |        |      | may not be selected by the users list of paths.
1445
 
        #   a    |  r     |  a   | not present in either tree at this path, skip as it
1446
 
        #        |        |      | may not be selected by the users list of paths.
1447
 
        #  fdlt  | fdlt   |      | content in both: diff them
1448
 
        #  fdlt  | fdlt   |  a   | deleted locally, but not unversioned - show as deleted ?
1449
 
        #  fdlt  |  a     |      | unversioned: output deleted id for now
1450
 
        #  fdlt  |  a     |  a   | unversioned and deleted: output deleted id
1451
 
        #  fdlt  |  r     |      | relocated in this tree, so add target to search.
1452
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1453
 
        #        |        |      | this id at the other path.
1454
 
        #  fdlt  |  r     |  a   | relocated in this tree, so add target to search.
1455
 
        #        |        |      | Dont diff, we will see an r,fd; pair when we reach
1456
 
        #        |        |      | this id at the other path.
1457
 
 
1458
 
        # TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1459
 
        #       keeping a cache of directories that we have seen.
1460
 
        cdef object current_dirname, current_blockname
1461
 
        cdef char * current_dirname_c, * current_blockname_c
1462
 
        cdef int advance_entry, advance_path
1463
 
        cdef int path_handled
1464
 
        searched_specific_files = self.searched_specific_files
1465
 
        # Are we walking a root?
1466
 
        while self.root_entries_pos < self.root_entries_len:
1467
 
            entry = self.root_entries[self.root_entries_pos]
1468
 
            self.root_entries_pos = self.root_entries_pos + 1
1469
 
            result, changed = self._process_entry(entry, self.root_dir_info)
1470
 
            if changed is not None:
1471
 
                if changed:
1472
 
                    self._gather_result_for_consistency(result)
1473
 
                if changed or self.include_unchanged:
1474
 
                    return result
1475
 
        # Have we finished the prior root, or never started one ?
1476
 
        if self.current_root is None:
1477
 
            # TODO: the pending list should be lexically sorted?  the
1478
 
            # interface doesn't require it.
1479
 
            try:
1480
 
                self.current_root = self.search_specific_files.pop()
1481
 
            except KeyError:
1482
 
                raise StopIteration()
1483
 
            self.searched_specific_files.add(self.current_root)
1484
 
            # process the entries for this containing directory: the rest will be
1485
 
            # found by their parents recursively.
1486
 
            self.root_entries = self.state._entries_for_path(self.current_root)
1487
 
            self.root_entries_len = len(self.root_entries)
1488
 
            self.current_root_unicode = self.current_root.decode('utf8')
1489
 
            self.root_abspath = self.tree.abspath(self.current_root_unicode)
1490
 
            try:
1491
 
                root_stat = os.lstat(self.root_abspath)
1492
 
            except OSError, e:
1493
 
                if e.errno == errno.ENOENT:
1494
 
                    # the path does not exist: let _process_entry know that.
1495
 
                    self.root_dir_info = None
1496
 
                else:
1497
 
                    # some other random error: hand it up.
1498
 
                    raise
1499
 
            else:
1500
 
                self.root_dir_info = ('', self.current_root,
1501
 
                    osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1502
 
                    self.root_abspath)
1503
 
                if self.root_dir_info[2] == 'directory':
1504
 
                    if self.tree._directory_is_tree_reference(
1505
 
                        self.current_root_unicode):
1506
 
                        self.root_dir_info = self.root_dir_info[:2] + \
1507
 
                            ('tree-reference',) + self.root_dir_info[3:]
1508
 
            if not self.root_entries and not self.root_dir_info:
1509
 
                # this specified path is not present at all, skip it.
1510
 
                # (tail recursion, can do a loop once the full structure is
1511
 
                # known).
1512
 
                return self._iter_next()
1513
 
            path_handled = 0
1514
 
            self.root_entries_pos = 0
1515
 
            # XXX Clarity: This loop is duplicated a out the self.current_root
1516
 
            # is None guard above: if we return from it, it completes there
1517
 
            # (and the following if block cannot trigger because
1518
 
            # path_handled must be true, so the if block is not # duplicated.
1519
 
            while self.root_entries_pos < self.root_entries_len:
1520
 
                entry = self.root_entries[self.root_entries_pos]
1521
 
                self.root_entries_pos = self.root_entries_pos + 1
1522
 
                result, changed = self._process_entry(entry, self.root_dir_info)
1523
 
                if changed is not None:
1524
 
                    path_handled = -1
1525
 
                    if changed:
1526
 
                        self._gather_result_for_consistency(result)
1527
 
                    if changed or self.include_unchanged:
1528
 
                        return result
1529
 
            # handle unversioned specified paths:
1530
 
            if self.want_unversioned and not path_handled and self.root_dir_info:
1531
 
                new_executable = bool(
1532
 
                    stat.S_ISREG(self.root_dir_info[3].st_mode)
1533
 
                    and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1534
 
                return (None,
1535
 
                       (None, self.current_root_unicode),
1536
 
                       True,
1537
 
                       (False, False),
1538
 
                       (None, None),
1539
 
                       (None, splitpath(self.current_root_unicode)[-1]),
1540
 
                       (None, self.root_dir_info[2]),
1541
 
                       (None, new_executable)
1542
 
                      )
1543
 
            # If we reach here, the outer flow continues, which enters into the
1544
 
            # per-root setup logic.
1545
 
        if (self.current_dir_info is None and self.current_block is None and not
1546
 
            self.doing_consistency_expansion):
1547
 
            # setup iteration of this root:
1548
 
            self.current_dir_list = None
1549
 
            if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1550
 
                self.current_dir_info = None
1551
 
            else:
1552
 
                self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1553
 
                    prefix=self.current_root)
1554
 
                self.path_index = 0
1555
 
                try:
1556
 
                    self.current_dir_info = self.dir_iterator.next()
1557
 
                    self.current_dir_list = self.current_dir_info[1]
1558
 
                except OSError, e:
1559
 
                    # there may be directories in the inventory even though
1560
 
                    # this path is not a file on disk: so mark it as end of
1561
 
                    # iterator
1562
 
                    if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1563
 
                        self.current_dir_info = None
1564
 
                    elif sys.platform == 'win32':
1565
 
                        # on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1566
 
                        # python 2.5 has e.errno == EINVAL,
1567
 
                        #            and e.winerror == ERROR_DIRECTORY
1568
 
                        try:
1569
 
                            e_winerror = e.winerror
1570
 
                        except AttributeError:
1571
 
                            e_winerror = None
1572
 
                        win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1573
 
                        if (e.errno in win_errors or e_winerror in win_errors):
1574
 
                            self.current_dir_info = None
1575
 
                        else:
1576
 
                            # Will this really raise the right exception ?
1577
 
                            raise
1578
 
                    else:
1579
 
                        raise
1580
 
                else:
1581
 
                    if self.current_dir_info[0][0] == '':
1582
 
                        # remove .bzr from iteration
1583
 
                        bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1584
 
                        if self.current_dir_list[bzr_index][0] != '.bzr':
1585
 
                            raise AssertionError()
1586
 
                        del self.current_dir_list[bzr_index]
1587
 
            initial_key = (self.current_root, '', '')
1588
 
            self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1589
 
            if self.block_index == 0:
1590
 
                # we have processed the total root already, but because the
1591
 
                # initial key matched it we should skip it here.
1592
 
                self.block_index = self.block_index + 1
1593
 
            self._update_current_block()
1594
 
        # walk until both the directory listing and the versioned metadata
1595
 
        # are exhausted. 
1596
 
        while (self.current_dir_info is not None
1597
 
            or self.current_block is not None):
1598
 
            # Uncommon case - a missing directory or an unversioned directory:
1599
 
            if (self.current_dir_info and self.current_block
1600
 
                and self.current_dir_info[0][0] != self.current_block[0]):
1601
 
                # Work around pyrex broken heuristic - current_dirname has
1602
 
                # the same scope as current_dirname_c
1603
 
                current_dirname = self.current_dir_info[0][0]
1604
 
                current_dirname_c = PyString_AS_STRING_void(
1605
 
                    <void *>current_dirname)
1606
 
                current_blockname = self.current_block[0]
1607
 
                current_blockname_c = PyString_AS_STRING_void(
1608
 
                    <void *>current_blockname)
1609
 
                # In the python generator we evaluate this if block once per
1610
 
                # dir+block; because we reenter in the pyrex version its being
1611
 
                # evaluated once per path: we could cache the result before
1612
 
                # doing the while loop and probably save time.
1613
 
                if _cmp_by_dirs(current_dirname_c,
1614
 
                    PyString_Size(current_dirname),
1615
 
                    current_blockname_c,
1616
 
                    PyString_Size(current_blockname)) < 0:
1617
 
                    # filesystem data refers to paths not covered by the
1618
 
                    # dirblock.  this has two possibilities:
1619
 
                    # A) it is versioned but empty, so there is no block for it
1620
 
                    # B) it is not versioned.
1621
 
 
1622
 
                    # if (A) then we need to recurse into it to check for
1623
 
                    # new unknown files or directories.
1624
 
                    # if (B) then we should ignore it, because we don't
1625
 
                    # recurse into unknown directories.
1626
 
                    # We are doing a loop
1627
 
                    while self.path_index < len(self.current_dir_list):
1628
 
                        current_path_info = self.current_dir_list[self.path_index]
1629
 
                        # dont descend into this unversioned path if it is
1630
 
                        # a dir
1631
 
                        if current_path_info[2] in ('directory',
1632
 
                                                    'tree-reference'):
1633
 
                            del self.current_dir_list[self.path_index]
1634
 
                            self.path_index = self.path_index - 1
1635
 
                        self.path_index = self.path_index + 1
1636
 
                        if self.want_unversioned:
1637
 
                            if current_path_info[2] == 'directory':
1638
 
                                if self.tree._directory_is_tree_reference(
1639
 
                                    self.utf8_decode(current_path_info[0])[0]):
1640
 
                                    current_path_info = current_path_info[:2] + \
1641
 
                                        ('tree-reference',) + current_path_info[3:]
1642
 
                            new_executable = bool(
1643
 
                                stat.S_ISREG(current_path_info[3].st_mode)
1644
 
                                and stat.S_IEXEC & current_path_info[3].st_mode)
1645
 
                            return (None,
1646
 
                                (None, self.utf8_decode(current_path_info[0])[0]),
1647
 
                                True,
1648
 
                                (False, False),
1649
 
                                (None, None),
1650
 
                                (None, self.utf8_decode(current_path_info[1])[0]),
1651
 
                                (None, current_path_info[2]),
1652
 
                                (None, new_executable))
1653
 
                    # This dir info has been handled, go to the next
1654
 
                    self.path_index = 0
1655
 
                    self.current_dir_list = None
1656
 
                    try:
1657
 
                        self.current_dir_info = self.dir_iterator.next()
1658
 
                        self.current_dir_list = self.current_dir_info[1]
1659
 
                    except StopIteration:
1660
 
                        self.current_dir_info = None
1661
 
                else: #(dircmp > 0)
1662
 
                    # We have a dirblock entry for this location, but there
1663
 
                    # is no filesystem path for this. This is most likely
1664
 
                    # because a directory was removed from the disk.
1665
 
                    # We don't have to report the missing directory,
1666
 
                    # because that should have already been handled, but we
1667
 
                    # need to handle all of the files that are contained
1668
 
                    # within.
1669
 
                    while self.current_block_pos < len(self.current_block_list):
1670
 
                        current_entry = self.current_block_list[self.current_block_pos]
1671
 
                        self.current_block_pos = self.current_block_pos + 1
1672
 
                        # entry referring to file not present on disk.
1673
 
                        # advance the entry only, after processing.
1674
 
                        result, changed = self._process_entry(current_entry, None)
1675
 
                        if changed is not None:
1676
 
                            if changed:
1677
 
                                self._gather_result_for_consistency(result)
1678
 
                            if changed or self.include_unchanged:
1679
 
                                return result
1680
 
                    self.block_index = self.block_index + 1
1681
 
                    self._update_current_block()
1682
 
                continue # next loop-on-block/dir
1683
 
            result = self._loop_one_block()
1684
 
            if result is not None:
1685
 
                return result
1686
 
        if len(self.search_specific_files):
1687
 
            # More supplied paths to process
1688
 
            self.current_root = None
1689
 
            return self._iter_next()
1690
 
        # Start expanding more conservatively, adding paths the user may not
1691
 
        # have intended but required for consistent deltas.
1692
 
        self.doing_consistency_expansion = 1
1693
 
        if not self._pending_consistent_entries:
1694
 
            self._pending_consistent_entries = self._next_consistent_entries()
1695
 
        while self._pending_consistent_entries:
1696
 
            result, changed = self._pending_consistent_entries.pop()
1697
 
            if changed is not None:
1698
 
                return result
1699
 
        raise StopIteration()
1700
 
 
1701
 
    cdef object _maybe_tree_ref(self, current_path_info):
1702
 
        if self.tree._directory_is_tree_reference(
1703
 
            self.utf8_decode(current_path_info[0])[0]):
1704
 
            return current_path_info[:2] + \
1705
 
                ('tree-reference',) + current_path_info[3:]
1706
 
        else:
1707
 
            return current_path_info
1708
 
 
1709
 
    cdef object _loop_one_block(self):
1710
 
            # current_dir_info and current_block refer to the same directory -
1711
 
            # this is the common case code.
1712
 
            # Assign local variables for current path and entry:
1713
 
            cdef object current_entry
1714
 
            cdef object current_path_info
1715
 
            cdef int path_handled
1716
 
            cdef char minikind
1717
 
            cdef int cmp_result
1718
 
            # cdef char * temp_str
1719
 
            # cdef Py_ssize_t temp_str_length
1720
 
            # PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1721
 
            # if not strncmp(temp_str, "directory", temp_str_length):
1722
 
            if (self.current_block is not None and
1723
 
                self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1724
 
                current_entry = PyList_GET_ITEM(self.current_block_list,
1725
 
                    self.current_block_pos)
1726
 
                # accomodate pyrex
1727
 
                Py_INCREF(current_entry)
1728
 
            else:
1729
 
                current_entry = None
1730
 
            if (self.current_dir_info is not None and
1731
 
                self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1732
 
                current_path_info = PyList_GET_ITEM(self.current_dir_list,
1733
 
                    self.path_index)
1734
 
                # accomodate pyrex
1735
 
                Py_INCREF(current_path_info)
1736
 
                disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1737
 
                # accomodate pyrex
1738
 
                Py_INCREF(disk_kind)
1739
 
                if disk_kind == "directory":
1740
 
                    current_path_info = self._maybe_tree_ref(current_path_info)
1741
 
            else:
1742
 
                current_path_info = None
1743
 
            while (current_entry is not None or current_path_info is not None):
1744
 
                advance_entry = -1
1745
 
                advance_path = -1
1746
 
                result = None
1747
 
                path_handled = 0
1748
 
                if current_entry is None:
1749
 
                    # unversioned -  the check for path_handled when the path
1750
 
                    # is advanced will yield this path if needed.
1751
 
                    pass
1752
 
                elif current_path_info is None:
1753
 
                    # no path is fine: the per entry code will handle it.
1754
 
                    result, changed = self._process_entry(current_entry,
1755
 
                        current_path_info)
1756
 
                else:
1757
 
                    minikind = _minikind_from_string(
1758
 
                        current_entry[1][self.target_index][0])
1759
 
                    cmp_result = cmp(current_path_info[1], current_entry[0][1])
1760
 
                    if (cmp_result or minikind == c'a' or minikind == c'r'):
1761
 
                        # The current path on disk doesn't match the dirblock
1762
 
                        # record. Either the dirblock record is marked as
1763
 
                        # absent/renamed, or the file on disk is not present at all
1764
 
                        # in the dirblock. Either way, report about the dirblock
1765
 
                        # entry, and let other code handle the filesystem one.
1766
 
 
1767
 
                        # Compare the basename for these files to determine
1768
 
                        # which comes first
1769
 
                        if cmp_result < 0:
1770
 
                            # extra file on disk: pass for now, but only
1771
 
                            # increment the path, not the entry
1772
 
                            advance_entry = 0
1773
 
                        else:
1774
 
                            # entry referring to file not present on disk.
1775
 
                            # advance the entry only, after processing.
1776
 
                            result, changed = self._process_entry(current_entry,
1777
 
                                None)
1778
 
                            advance_path = 0
1779
 
                    else:
1780
 
                        # paths are the same,and the dirstate entry is not
1781
 
                        # absent or renamed.
1782
 
                        result, changed = self._process_entry(current_entry,
1783
 
                            current_path_info)
1784
 
                        if changed is not None:
1785
 
                            path_handled = -1
1786
 
                            if not changed and not self.include_unchanged:
1787
 
                                changed = None
1788
 
                # >- loop control starts here:
1789
 
                # >- entry
1790
 
                if advance_entry and current_entry is not None:
1791
 
                    self.current_block_pos = self.current_block_pos + 1
1792
 
                    if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1793
 
                        current_entry = self.current_block_list[self.current_block_pos]
1794
 
                    else:
1795
 
                        current_entry = None
1796
 
                # >- path
1797
 
                if advance_path and current_path_info is not None:
1798
 
                    if not path_handled:
1799
 
                        # unversioned in all regards
1800
 
                        if self.want_unversioned:
1801
 
                            new_executable = bool(
1802
 
                                stat.S_ISREG(current_path_info[3].st_mode)
1803
 
                                and stat.S_IEXEC & current_path_info[3].st_mode)
1804
 
                            try:
1805
 
                                relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1806
 
                            except UnicodeDecodeError:
1807
 
                                raise errors.BadFilenameEncoding(
1808
 
                                    current_path_info[0], osutils._fs_enc)
1809
 
                            if changed is not None:
1810
 
                                raise AssertionError(
1811
 
                                    "result is not None: %r" % result)
1812
 
                            result = (None,
1813
 
                                (None, relpath_unicode),
1814
 
                                True,
1815
 
                                (False, False),
1816
 
                                (None, None),
1817
 
                                (None, self.utf8_decode(current_path_info[1])[0]),
1818
 
                                (None, current_path_info[2]),
1819
 
                                (None, new_executable))
1820
 
                            changed = True
1821
 
                        # dont descend into this unversioned path if it is
1822
 
                        # a dir
1823
 
                        if current_path_info[2] in ('directory'):
1824
 
                            del self.current_dir_list[self.path_index]
1825
 
                            self.path_index = self.path_index - 1
1826
 
                    # dont descend the disk iterator into any tree 
1827
 
                    # paths.
1828
 
                    if current_path_info[2] == 'tree-reference':
1829
 
                        del self.current_dir_list[self.path_index]
1830
 
                        self.path_index = self.path_index - 1
1831
 
                    self.path_index = self.path_index + 1
1832
 
                    if self.path_index < len(self.current_dir_list):
1833
 
                        current_path_info = self.current_dir_list[self.path_index]
1834
 
                        if current_path_info[2] == 'directory':
1835
 
                            current_path_info = self._maybe_tree_ref(
1836
 
                                current_path_info)
1837
 
                    else:
1838
 
                        current_path_info = None
1839
 
                if changed is not None:
1840
 
                    # Found a result on this pass, yield it
1841
 
                    if changed:
1842
 
                        self._gather_result_for_consistency(result)
1843
 
                    if changed or self.include_unchanged:
1844
 
                        return result
1845
 
            if self.current_block is not None:
1846
 
                self.block_index = self.block_index + 1
1847
 
                self._update_current_block()
1848
 
            if self.current_dir_info is not None:
1849
 
                self.path_index = 0
1850
 
                self.current_dir_list = None
1851
 
                try:
1852
 
                    self.current_dir_info = self.dir_iterator.next()
1853
 
                    self.current_dir_list = self.current_dir_info[1]
1854
 
                except StopIteration:
1855
 
                    self.current_dir_info = None
1856
 
 
1857
 
    cdef object _next_consistent_entries(self):
1858
 
        """Grabs the next specific file parent case to consider.
1859
 
        
1860
 
        :return: A list of the results, each of which is as for _process_entry.
1861
 
        """
1862
 
        results = []
1863
 
        while self.search_specific_file_parents:
1864
 
            # Process the parent directories for the paths we were iterating.
1865
 
            # Even in extremely large trees this should be modest, so currently
1866
 
            # no attempt is made to optimise.
1867
 
            path_utf8 = self.search_specific_file_parents.pop()
1868
 
            if path_utf8 in self.searched_exact_paths:
1869
 
                # We've examined this path.
1870
 
                continue
1871
 
            if osutils.is_inside_any(self.searched_specific_files, path_utf8):
1872
 
                # We've examined this path.
1873
 
                continue
1874
 
            path_entries = self.state._entries_for_path(path_utf8)
1875
 
            # We need either one or two entries. If the path in
1876
 
            # self.target_index has moved (so the entry in source_index is in
1877
 
            # 'ar') then we need to also look for the entry for this path in
1878
 
            # self.source_index, to output the appropriate delete-or-rename.
1879
 
            selected_entries = []
1880
 
            found_item = False
1881
 
            for candidate_entry in path_entries:
1882
 
                # Find entries present in target at this path:
1883
 
                if candidate_entry[1][self.target_index][0] not in 'ar':
1884
 
                    found_item = True
1885
 
                    selected_entries.append(candidate_entry)
1886
 
                # Find entries present in source at this path:
1887
 
                elif (self.source_index is not None and
1888
 
                    candidate_entry[1][self.source_index][0] not in 'ar'):
1889
 
                    found_item = True
1890
 
                    if candidate_entry[1][self.target_index][0] == 'a':
1891
 
                        # Deleted, emit it here.
1892
 
                        selected_entries.append(candidate_entry)
1893
 
                    else:
1894
 
                        # renamed, emit it when we process the directory it
1895
 
                        # ended up at.
1896
 
                        self.search_specific_file_parents.add(
1897
 
                            candidate_entry[1][self.target_index][1])
1898
 
            if not found_item:
1899
 
                raise AssertionError(
1900
 
                    "Missing entry for specific path parent %r, %r" % (
1901
 
                    path_utf8, path_entries))
1902
 
            path_info = self._path_info(path_utf8, path_utf8.decode('utf8'))
1903
 
            for entry in selected_entries:
1904
 
                if entry[0][2] in self.seen_ids:
1905
 
                    continue
1906
 
                result, changed = self._process_entry(entry, path_info)
1907
 
                if changed is None:
1908
 
                    raise AssertionError(
1909
 
                        "Got entry<->path mismatch for specific path "
1910
 
                        "%r entry %r path_info %r " % (
1911
 
                        path_utf8, entry, path_info))
1912
 
                # Only include changes - we're outside the users requested
1913
 
                # expansion.
1914
 
                if changed:
1915
 
                    self._gather_result_for_consistency(result)
1916
 
                    if (result[6][0] == 'directory' and
1917
 
                        result[6][1] != 'directory'):
1918
 
                        # This stopped being a directory, the old children have
1919
 
                        # to be included.
1920
 
                        if entry[1][self.source_index][0] == 'r':
1921
 
                            # renamed, take the source path
1922
 
                            entry_path_utf8 = entry[1][self.source_index][1]
1923
 
                        else:
1924
 
                            entry_path_utf8 = path_utf8
1925
 
                        initial_key = (entry_path_utf8, '', '')
1926
 
                        block_index, _ = self.state._find_block_index_from_key(
1927
 
                            initial_key)
1928
 
                        if block_index == 0:
1929
 
                            # The children of the root are in block index 1.
1930
 
                            block_index = block_index + 1
1931
 
                        current_block = None
1932
 
                        if block_index < len(self.state._dirblocks):
1933
 
                            current_block = self.state._dirblocks[block_index]
1934
 
                            if not osutils.is_inside(
1935
 
                                entry_path_utf8, current_block[0]):
1936
 
                                # No entries for this directory at all.
1937
 
                                current_block = None
1938
 
                        if current_block is not None:
1939
 
                            for entry in current_block[1]:
1940
 
                                if entry[1][self.source_index][0] in 'ar':
1941
 
                                    # Not in the source tree, so doesn't have to be
1942
 
                                    # included.
1943
 
                                    continue
1944
 
                                # Path of the entry itself.
1945
 
                                self.search_specific_file_parents.add(
1946
 
                                    self.pathjoin(*entry[0][:2]))
1947
 
                if changed or self.include_unchanged:
1948
 
                    results.append((result, changed))
1949
 
            self.searched_exact_paths.add(path_utf8)
1950
 
        return results
1951
 
 
1952
 
    cdef object _path_info(self, utf8_path, unicode_path):
1953
 
        """Generate path_info for unicode_path.
1954
 
 
1955
 
        :return: None if unicode_path does not exist, or a path_info tuple.
1956
 
        """
1957
 
        abspath = self.tree.abspath(unicode_path)
1958
 
        try:
1959
 
            stat = os.lstat(abspath)
1960
 
        except OSError, e:
1961
 
            if e.errno == errno.ENOENT:
1962
 
                # the path does not exist.
1963
 
                return None
1964
 
            else:
1965
 
                raise
1966
 
        utf8_basename = utf8_path.rsplit('/', 1)[-1]
1967
 
        dir_info = (utf8_path, utf8_basename,
1968
 
            osutils.file_kind_from_stat_mode(stat.st_mode), stat,
1969
 
            abspath)
1970
 
        if dir_info[2] == 'directory':
1971
 
            if self.tree._directory_is_tree_reference(
1972
 
                unicode_path):
1973
 
                self.root_dir_info = self.root_dir_info[:2] + \
1974
 
                    ('tree-reference',) + self.root_dir_info[3:]
1975
 
        return dir_info