1
# Copyright (C) 2007, 2008 Canonical Ltd
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.
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.
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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Helper functions for DirState.
19
This is the python implementation for DirState functions.
29
from bzrlib import cache_utf8, errors, osutils
30
from bzrlib.dirstate import DirState
31
from bzrlib.osutils import parent_directories, pathjoin, splitpath
34
# This is the Windows equivalent of ENOTDIR
35
# It is defined in pywin32.winerror, but we don't want a strong dependency for
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
43
#python2.4 support, and other platform-dependent includes
44
cdef extern from "python-compat.h":
45
unsigned long htonl(unsigned long)
47
# Give Pyrex some function definitions for it to understand.
48
# All of these are just hints to Pyrex, so that it can try to convert python
49
# objects into similar C objects. (such as PyInt => int).
50
# In anything defined 'cdef extern from XXX' the real C header will be
51
# imported, and the real definition will be used from there. So these are just
52
# hints, and do not need to match exactly to the C definitions.
55
ctypedef unsigned long size_t
57
cdef extern from "_dirstate_helpers_pyx.h":
62
cdef extern from "stdlib.h":
63
unsigned long int strtoul(char *nptr, char **endptr, int base)
66
cdef extern from 'sys/stat.h':
69
# On win32, this actually comes from "python-compat.h"
73
# These functions allow us access to a bit of the 'bare metal' of python
74
# objects, rather than going through the object abstraction. (For example,
75
# PyList_Append, rather than getting the 'append' attribute of the object, and
76
# creating a tuple, and then using PyCallObject).
77
# Functions that return (or take) a void* are meant to grab a C PyObject*. This
78
# differs from the Pyrex 'object'. If you declare a variable as 'object' Pyrex
79
# will automatically Py_INCREF and Py_DECREF when appropriate. But for some
80
# inner loops, we don't need to do that at all, as the reference only lasts for
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
cdef extern from "Python.h":
86
ctypedef int Py_ssize_t
87
ctypedef struct PyObject:
89
int PyList_Append(object lst, object item) except -1
90
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
int PyList_CheckExact(object)
94
Py_ssize_t PyList_GET_SIZE (object p)
96
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)
101
char *PyString_AsString(object p)
102
char *PyString_AsString_obj "PyString_AsString" (PyObject *string)
103
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
object PyString_FromString(char *)
106
object PyString_FromStringAndSize(char *, Py_ssize_t)
107
int PyString_Size(object p)
108
int PyString_GET_SIZE_void "PyString_GET_SIZE" (void *p)
109
int PyString_CheckExact(object p)
110
void Py_INCREF(object o)
111
void Py_DECREF(object o)
114
cdef extern from "string.h":
115
int strncmp(char *s1, char *s2, int len)
116
void *memchr(void *s, int c, size_t len)
117
int memcmp(void *b1, void *b2, size_t len)
118
# ??? memrchr is a GNU extension :(
119
# void *memrchr(void *s, int c, size_t len)
122
cdef void* _my_memrchr(void *s, int c, size_t n):
123
# memrchr seems to be a GNU extension, so we have to implement it ourselves
136
def _py_memrchr(s, c):
137
"""Just to expose _my_memrchr for testing.
139
:param s: The Python string to search
140
:param c: The character to search for
141
:return: The offset to the last instance of 'c' in s
148
_s = PyString_AsString(s)
149
length = PyString_Size(s)
151
_c = PyString_AsString(c)
152
assert PyString_Size(c) == 1,\
153
'Must be a single character string, not %s' % (c,)
154
found = _my_memrchr(_s, _c[0], length)
157
return <char*>found - <char*>_s
159
cdef object safe_string_from_size(char *s, Py_ssize_t size):
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'
165
return PyString_FromStringAndSize(s, size)
168
cdef int _is_aligned(void *ptr):
169
"""Is this pointer aligned to an integer size offset?
171
:return: 1 if this pointer is aligned, 0 otherwise.
173
return ((<intptr_t>ptr) & ((sizeof(int))-1)) == 0
176
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2):
177
cdef unsigned char *cur1
178
cdef unsigned char *cur2
179
cdef unsigned char *end1
180
cdef unsigned char *end2
186
if path1 == path2 and size1 == size2:
189
end1 = <unsigned char*>path1+size1
190
end2 = <unsigned char*>path2+size2
192
# Use 32-bit comparisons for the matching portion of the string.
193
# Almost all CPU's are faster at loading and comparing 32-bit integers,
194
# than they are at 8-bit integers.
195
# 99% of the time, these will be aligned, but in case they aren't just skip
197
if _is_aligned(path1) and _is_aligned(path2):
198
cur_int1 = <int*>path1
199
cur_int2 = <int*>path2
200
end_int1 = <int*>(path1 + size1 - (size1 % sizeof(int)))
201
end_int2 = <int*>(path2 + size2 - (size2 % sizeof(int)))
203
while cur_int1 < end_int1 and cur_int2 < end_int2:
204
if cur_int1[0] != cur_int2[0]:
206
cur_int1 = cur_int1 + 1
207
cur_int2 = cur_int2 + 1
209
cur1 = <unsigned char*>cur_int1
210
cur2 = <unsigned char*>cur_int2
212
cur1 = <unsigned char*>path1
213
cur2 = <unsigned char*>path2
215
while cur1 < end1 and cur2 < end2:
216
if cur1[0] == cur2[0]:
217
# This character matches, just go to the next one
221
# The current characters do not match
223
return -1 # Reached the end of path1 segment first
224
elif cur2[0] == c'/':
225
return 1 # Reached the end of path2 segment first
226
elif cur1[0] < cur2[0]:
231
# We reached the end of at least one of the strings
233
return 1 # Not at the end of cur1, must be at the end of cur2
235
return -1 # At the end of cur1, but not at cur2
236
# We reached the end of both strings
240
def cmp_by_dirs(path1, path2):
241
"""Compare two paths directory by directory.
243
This is equivalent to doing::
245
cmp(path1.split('/'), path2.split('/'))
247
The idea is that you should compare path components separately. This
248
differs from plain ``cmp(path1, path2)`` for paths like ``'a-b'`` and
249
``a/b``. "a-b" comes after "a" but would come before "a/b" lexically.
251
:param path1: first path
252
:param path2: second path
253
:return: negative number if ``path1`` comes first,
254
0 if paths are equal,
255
and positive number if ``path2`` sorts first
257
if not PyString_CheckExact(path1):
258
raise TypeError("'path1' must be a plain string, not %s: %r"
259
% (type(path1), path1))
260
if not PyString_CheckExact(path2):
261
raise TypeError("'path2' must be a plain string, not %s: %r"
262
% (type(path2), path2))
263
return _cmp_by_dirs(PyString_AsString(path1),
264
PyString_Size(path1),
265
PyString_AsString(path2),
266
PyString_Size(path2))
269
def _cmp_path_by_dirblock(path1, path2):
270
"""Compare two paths based on what directory they are in.
272
This generates a sort order, such that all children of a directory are
273
sorted together, and grandchildren are in the same order as the
274
children appear. But all grandchildren come after all children.
276
In other words, all entries in a directory are sorted together, and
277
directorys are sorted in cmp_by_dirs order.
279
:param path1: first path
280
:param path2: the second path
281
:return: negative number if ``path1`` comes first,
283
and a positive number if ``path2`` sorts first
285
if not PyString_CheckExact(path1):
286
raise TypeError("'path1' must be a plain string, not %s: %r"
287
% (type(path1), path1))
288
if not PyString_CheckExact(path2):
289
raise TypeError("'path2' must be a plain string, not %s: %r"
290
% (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))
297
cdef int _cmp_path_by_dirblock_intern(char *path1, int path1_len,
298
char *path2, int path2_len):
299
"""Compare two paths by what directory they are in.
301
see ``_cmp_path_by_dirblock`` for details.
304
cdef int dirname1_len
306
cdef int dirname2_len
308
cdef int basename1_len
310
cdef int basename2_len
314
if path1_len == 0 and path2_len == 0:
317
if path1 == path2 and path1_len == path2_len:
326
basename1 = <char*>_my_memrchr(path1, c'/', path1_len)
328
if basename1 == NULL:
330
basename1_len = path1_len
335
dirname1_len = basename1 - path1
336
basename1 = basename1 + 1
337
basename1_len = path1_len - dirname1_len - 1
339
basename2 = <char*>_my_memrchr(path2, c'/', path2_len)
341
if basename2 == NULL:
343
basename2_len = path2_len
348
dirname2_len = basename2 - path2
349
basename2 = basename2 + 1
350
basename2_len = path2_len - dirname2_len - 1
352
cmp_val = _cmp_by_dirs(dirname1, dirname1_len,
353
dirname2, dirname2_len)
357
cur_len = basename1_len
358
if basename2_len < basename1_len:
359
cur_len = basename2_len
361
cmp_val = memcmp(basename1, basename2, cur_len)
364
if basename1_len == basename2_len:
366
if basename1_len < basename2_len:
371
def _bisect_path_left(paths, path):
372
"""Return the index where to insert path into paths.
374
This uses a path-wise comparison so we get::
384
:param paths: A list of paths to search through
385
:param path: A single path to insert
386
:return: An offset where 'path' can be inserted.
387
:seealso: bisect.bisect_left
398
if not PyList_CheckExact(paths):
399
raise TypeError("you must pass a python list for 'paths' not: %s %r"
400
% (type(paths), paths))
401
if not PyString_CheckExact(path):
402
raise TypeError("you must pass a string for 'path' not: %s %r"
403
% (type(path), path))
408
path_cstr = PyString_AsString(path)
409
path_size = PyString_Size(path)
412
_mid = (_lo + _hi) / 2
413
cur = PyList_GetItem_object_void(paths, _mid)
414
cur_cstr = PyString_AS_STRING_void(cur)
415
cur_size = PyString_GET_SIZE_void(cur)
416
if _cmp_path_by_dirblock_intern(cur_cstr, cur_size,
417
path_cstr, path_size) < 0:
424
def _bisect_path_right(paths, path):
425
"""Return the index where to insert path into paths.
427
This uses a path-wise comparison so we get::
437
:param paths: A list of paths to search through
438
:param path: A single path to insert
439
:return: An offset where 'path' can be inserted.
440
:seealso: bisect.bisect_right
451
if not PyList_CheckExact(paths):
452
raise TypeError("you must pass a python list for 'paths' not: %s %r"
453
% (type(paths), paths))
454
if not PyString_CheckExact(path):
455
raise TypeError("you must pass a string for 'path' not: %s %r"
456
% (type(path), path))
461
path_cstr = PyString_AsString(path)
462
path_size = PyString_Size(path)
465
_mid = (_lo + _hi) / 2
466
cur = PyList_GetItem_object_void(paths, _mid)
467
cur_cstr = PyString_AS_STRING_void(cur)
468
cur_size = PyString_GET_SIZE_void(cur)
469
if _cmp_path_by_dirblock_intern(path_cstr, path_size,
470
cur_cstr, cur_size) < 0:
477
def bisect_dirblock(dirblocks, dirname, lo=0, hi=None, cache=None):
478
"""Return the index where to insert dirname into the dirblocks.
480
The return value idx is such that all directories blocks in dirblock[:idx]
481
have names < dirname, and all blocks in dirblock[idx:] have names >=
484
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
485
slice of a to be searched.
490
cdef char *dirname_cstr
491
cdef int dirname_size
496
if not PyList_CheckExact(dirblocks):
497
raise TypeError("you must pass a python list for 'dirblocks' not: %s %r"
498
% (type(dirblocks), dirblocks))
499
if not PyString_CheckExact(dirname):
500
raise TypeError("you must pass a string for dirname not: %s %r"
501
% (type(dirname), dirname))
508
dirname_cstr = PyString_AsString(dirname)
509
dirname_size = PyString_Size(dirname)
512
_mid = (_lo + _hi) / 2
513
# Grab the dirname for the current dirblock
514
# cur = dirblocks[_mid][0]
515
cur = PyTuple_GetItem_void_void(
516
PyList_GetItem_object_void(dirblocks, _mid), 0)
517
cur_cstr = PyString_AS_STRING_void(cur)
518
cur_size = PyString_GET_SIZE_void(cur)
519
if _cmp_by_dirs(cur_cstr, cur_size, dirname_cstr, dirname_size) < 0:
527
"""Maintain the current location, and return fields as you parse them."""
529
cdef object state # The DirState object
530
cdef object text # The overall string object
531
cdef char *text_cstr # Pointer to the beginning of text
532
cdef int text_size # Length of text
534
cdef char *end_cstr # End of text
535
cdef char *cur_cstr # Pointer to the current record
536
cdef char *next # Pointer to the end of this record
538
def __init__(self, text, state):
541
self.text_cstr = PyString_AsString(text)
542
self.text_size = PyString_Size(text)
543
self.end_cstr = self.text_cstr + self.text_size
544
self.cur_cstr = self.text_cstr
546
cdef char *get_next(self, int *size) except NULL:
547
"""Return a pointer to the start of the next field."""
549
cdef Py_ssize_t extra_len
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'
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))
564
size[0] = self.cur_cstr - next
565
self.cur_cstr = self.cur_cstr + 1
568
cdef object get_next_str(self):
569
"""Get the next field as a Python string."""
572
next = self.get_next(&size)
573
return safe_string_from_size(next, size)
575
cdef int _init(self) except -1:
576
"""Get the pointer ready.
578
This assumes that the dirstate header has already been read, and we
579
already have the dirblock string loaded into memory.
580
This just initializes our memory pointers, etc for parsing of the
585
# The first field should be an empty string left over from the Header
586
first = self.get_next(&size)
587
if first[0] != c'\0' and size == 0:
588
raise AssertionError('First character should be null not: %s'
592
cdef object _get_entry(self, int num_trees, void **p_current_dirname,
594
"""Extract the next entry.
596
This parses the next entry based on the current location in
598
Each entry can be considered a "row" in the total table. And each row
599
has a fixed number of columns. It is generally broken up into "key"
600
columns, then "current" columns, and then "parent" columns.
602
:param num_trees: How many parent trees need to be parsed
603
:param p_current_dirname: A pointer to the current PyString
604
representing the directory name.
605
We pass this in as a void * so that pyrex doesn't have to
606
increment/decrement the PyObject reference counter for each
608
We use a pointer so that _get_entry can update it with the new
610
:param new_block: This is to let the caller know that it needs to
611
create a new directory block to store the next entry.
613
cdef object path_name_file_id_key
614
cdef char *entry_size_cstr
615
cdef unsigned long int entry_size
616
cdef char* executable_cstr
617
cdef int is_executable
618
cdef char* dirname_cstr
623
cdef object fingerprint
626
# Read the 'key' information (dirname, name, file_id)
627
dirname_cstr = self.get_next(&cur_size)
628
# Check to see if we have started a new directory block.
629
# If so, then we need to create a new dirname PyString, so that it can
630
# be used in all of the tuples. This saves time and memory, by re-using
631
# the same object repeatedly.
633
# Do the cheap 'length of string' check first. If the string is a
634
# different length, then we *have* to be a different directory.
635
if (cur_size != PyString_GET_SIZE_void(p_current_dirname[0])
636
or strncmp(dirname_cstr,
637
# Extract the char* from our current dirname string. We
638
# know it is a PyString, so we can use
639
# PyString_AS_STRING, we use the _void version because
640
# we are tricking Pyrex by using a void* rather than an
642
PyString_AS_STRING_void(p_current_dirname[0]),
644
dirname = safe_string_from_size(dirname_cstr, cur_size)
645
p_current_dirname[0] = <void*>dirname
650
# Build up the key that will be used.
651
# By using <object>(void *) Pyrex will automatically handle the
652
# Py_INCREF that we need.
653
path_name_file_id_key = (<object>p_current_dirname[0],
658
# Parse all of the per-tree information. current has the information in
659
# the same location as parent trees. The only difference is that 'info'
660
# is a 'packed_stat' for current, while it is a 'revision_id' for
662
# minikind, fingerprint, and info will be returned as regular python
664
# entry_size and is_executable will be parsed into a python Long and
665
# python Boolean, respectively.
666
# TODO: jam 20070718 Consider changin the entry_size conversion to
667
# prefer python Int when possible. They are generally faster to
668
# work with, and it will be rare that we have a file >2GB.
669
# Especially since this code is pretty much fixed at a max of
672
for i from 0 <= i < num_trees:
673
minikind = self.get_next_str()
674
fingerprint = self.get_next_str()
675
entry_size_cstr = self.get_next(&cur_size)
676
entry_size = strtoul(entry_size_cstr, NULL, 10)
677
executable_cstr = self.get_next(&cur_size)
678
is_executable = (executable_cstr[0] == c'y')
679
info = self.get_next_str()
680
PyList_Append(trees, (
682
fingerprint, # fingerprint
684
is_executable,# executable
685
info, # packed_stat or revision_id
688
# The returned tuple is (key, [trees])
689
ret = (path_name_file_id_key, trees)
690
# Ignore the trailing newline, but assert that it does exist, this
691
# ensures that we always finish parsing a line on an end-of-entry
693
trailing = self.get_next(&cur_size)
694
if cur_size != 1 or trailing[0] != c'\n':
695
raise errors.DirstateCorrupt(self.state,
696
'Bad parse, we expected to end on \\n, not: %d %s: %s'
697
% (cur_size, safe_string_from_size(trailing, cur_size),
701
def _parse_dirblocks(self):
702
"""Parse all dirblocks in the state file."""
704
cdef object current_block
706
cdef void * current_dirname
708
cdef int expected_entry_count
711
num_trees = self.state._num_present_parents() + 1
712
expected_entry_count = self.state._num_entries
714
# Ignore the first record
718
dirblocks = [('', current_block), ('', [])]
719
self.state._dirblocks = dirblocks
721
current_dirname = <void*>obj
725
# TODO: jam 2007-05-07 Consider pre-allocating some space for the
726
# members, and then growing and shrinking from there. If most
727
# directories have close to 10 entries in them, it would save a
728
# few mallocs if we default our list size to something
729
# reasonable. Or we could malloc it to something large (100 or
730
# so), and then truncate. That would give us a malloc + realloc,
731
# rather than lots of reallocs.
732
while self.cur_cstr < self.end_cstr:
733
entry = self._get_entry(num_trees, ¤t_dirname, &new_block)
735
# new block - different dirname
737
PyList_Append(dirblocks,
738
(<object>current_dirname, current_block))
739
PyList_Append(current_block, entry)
740
entry_count = entry_count + 1
741
if entry_count != expected_entry_count:
742
raise errors.DirstateCorrupt(self.state,
743
'We read the wrong number of entries.'
744
' We expected to read %s, but read %s'
745
% (expected_entry_count, entry_count))
746
self.state._split_root_dirblock_into_contents()
749
def _read_dirblocks(state):
750
"""Read in the dirblocks for the given DirState object.
752
This is tightly bound to the DirState internal representation. It should be
753
thought of as a member function, which is only separated out so that we can
754
re-write it in pyrex.
756
:param state: A DirState object.
758
:postcondition: The dirblocks will be loaded into the appropriate fields in
761
state._state_file.seek(state._end_of_header)
762
text = state._state_file.read()
763
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
765
reader = Reader(text, state)
767
reader._parse_dirblocks()
768
state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
771
cdef int minikind_from_mode(int mode):
772
# in order of frequency:
782
_encode = binascii.b2a_base64
785
from struct import pack
786
cdef _pack_stat(stat_value):
787
"""return a string representing the stat value's key fields.
789
:param stat_value: A stat oject with st_size, st_mtime, st_ctime, st_dev,
790
st_ino and st_mode fields.
792
cdef char result[6*4] # 6 long ints
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]
805
def update_entry(self, entry, abspath, stat_value):
806
"""Update the entry based on what is actually on disk.
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.
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
816
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
819
return _update_entry(self, entry, abspath, stat_value)
822
cdef _update_entry(self, entry, abspath, stat_value):
823
"""Update the entry based on what is actually on disk.
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.
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
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
840
minikind = minikind_from_mode(stat_value.st_mode)
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':
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]
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
867
# size should also be in packed_stat
868
if saved_file_size == stat_value.st_size:
869
return saved_link_or_sha1
871
# If we have gotten this far, that means that we need to actually
872
# process this entry.
875
executable = self._is_executable(stat_value.st_mode,
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)
890
entry[1][0] = ('f', '', stat_value.st_size,
891
executable, DirState.NULLSTAT)
892
elif minikind == c'd':
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,
912
entry[1][0] = ('l', '', stat_value.st_size,
913
False, DirState.NULLSTAT)
914
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
918
cdef char _minikind_from_string(object string):
919
"""Convert a python string to a char."""
920
return PyString_AsString(string)[0]
923
cdef object _kind_absent
924
cdef object _kind_file
925
cdef object _kind_directory
926
cdef object _kind_symlink
927
cdef object _kind_relocated
928
cdef object _kind_tree_reference
929
_kind_absent = "absent"
931
_kind_directory = "directory"
932
_kind_symlink = "symlink"
933
_kind_relocated = "relocated"
934
_kind_tree_reference = "tree-reference"
937
cdef object _minikind_to_kind(char minikind):
938
"""Create a string kind for minikind."""
939
cdef char _minikind[1]
942
elif minikind == c'd':
943
return _kind_directory
944
elif minikind == c'a':
946
elif minikind == c'r':
947
return _kind_relocated
948
elif minikind == c'l':
950
elif minikind == c't':
951
return _kind_tree_reference
952
_minikind[0] = minikind
953
raise KeyError(PyString_FromStringAndSize(_minikind, 1))
956
cdef int _versioned_minikind(char minikind):
957
"""Return non-zero if minikind is in fltd"""
958
return (minikind == c'f' or
964
cdef class ProcessEntryC:
966
cdef int doing_consistency_expansion
967
cdef object old_dirname_to_file_id # dict
968
cdef object new_dirname_to_file_id # dict
969
cdef object last_source_parent
970
cdef object last_target_parent
971
cdef int include_unchanged
973
cdef object use_filesystem_for_exec
974
cdef object utf8_decode
975
cdef readonly object searched_specific_files
976
cdef readonly object searched_exact_paths
977
cdef object search_specific_files
978
# The parents up to the root of the paths we are searching.
979
# After all normal paths are returned, these specific items are returned.
980
cdef object search_specific_file_parents
982
# Current iteration variables:
983
cdef object current_root
984
cdef object current_root_unicode
985
cdef object root_entries
986
cdef int root_entries_pos, root_entries_len
987
cdef object root_abspath
988
cdef int source_index, target_index
989
cdef int want_unversioned
991
cdef object dir_iterator
993
cdef object current_block
994
cdef int current_block_pos
995
cdef object current_block_list
996
cdef object current_dir_info
997
cdef object current_dir_list
998
cdef object _pending_consistent_entries # list
1000
cdef object root_dir_info
1001
cdef object bisect_left
1002
cdef object pathjoin
1004
# A set of the ids we've output when doing partial output.
1005
cdef object seen_ids
1006
cdef object sha_file
1008
def __init__(self, include_unchanged, use_filesystem_for_exec,
1009
search_specific_files, state, source_index, target_index,
1010
want_unversioned, tree):
1011
self.doing_consistency_expansion = 0
1012
self.old_dirname_to_file_id = {}
1013
self.new_dirname_to_file_id = {}
1014
# Are we doing a partial iter_changes?
1015
self.partial = set(['']).__ne__(search_specific_files)
1016
# Using a list so that we can access the values and change them in
1017
# nested scope. Each one is [path, file_id, entry]
1018
self.last_source_parent = [None, None]
1019
self.last_target_parent = [None, None]
1020
if include_unchanged is None:
1021
self.include_unchanged = False
1023
self.include_unchanged = int(include_unchanged)
1024
self.use_filesystem_for_exec = use_filesystem_for_exec
1025
self.utf8_decode = cache_utf8._utf8_decode
1026
# for all search_indexs in each path at or under each element of
1027
# search_specific_files, if the detail is relocated: add the id, and
1028
# add the relocated path as one to search if its not searched already.
1029
# If the detail is not relocated, add the id.
1030
self.searched_specific_files = set()
1031
# When we search exact paths without expanding downwards, we record
1033
self.searched_exact_paths = set()
1034
self.search_specific_files = search_specific_files
1035
# The parents up to the root of the paths we are searching.
1036
# After all normal paths are returned, these specific items are returned.
1037
self.search_specific_file_parents = set()
1038
# The ids we've sent out in the delta.
1039
self.seen_ids = set()
1041
self.current_root = None
1042
self.current_root_unicode = None
1043
self.root_entries = None
1044
self.root_entries_pos = 0
1045
self.root_entries_len = 0
1046
self.root_abspath = None
1047
if source_index is None:
1048
self.source_index = -1
1050
self.source_index = source_index
1051
self.target_index = target_index
1052
self.want_unversioned = want_unversioned
1054
self.dir_iterator = None
1055
self.block_index = -1
1056
self.current_block = None
1057
self.current_block_list = None
1058
self.current_block_pos = -1
1059
self.current_dir_info = None
1060
self.current_dir_list = None
1061
self._pending_consistent_entries = []
1063
self.root_dir_info = None
1064
self.bisect_left = bisect.bisect_left
1065
self.pathjoin = osutils.pathjoin
1066
self.fstat = os.fstat
1067
self.sha_file = osutils.sha_file
1068
if target_index != 0:
1069
# A lot of code in here depends on target_index == 0
1070
raise errors.BzrError('unsupported target index')
1072
cdef _process_entry(self, entry, path_info):
1073
"""Compare an entry and real disk to generate delta information.
1075
:param path_info: top_relpath, basename, kind, lstat, abspath for
1076
the path of entry. If None, then the path is considered absent in
1077
the target (Perhaps we should pass in a concrete entry for this ?)
1078
Basename is returned as a utf8 string because we expect this
1079
tuple will be ignored, and don't want to take the time to
1081
:return: (iter_changes_result, changed). If the entry has not been
1082
handled then changed is None. Otherwise it is False if no content
1083
or metadata changes have occured, and True if any content or
1084
metadata change has occurred. If self.include_unchanged is True then
1085
if changed is not None, iter_changes_result will always be a result
1086
tuple. Otherwise, iter_changes_result is None unless changed is
1089
cdef char target_minikind
1090
cdef char source_minikind
1092
cdef int content_change
1093
cdef object details_list
1095
details_list = entry[1]
1096
if -1 == self.source_index:
1097
source_details = DirState.NULL_PARENT_DETAILS
1099
source_details = details_list[self.source_index]
1100
target_details = details_list[self.target_index]
1101
target_minikind = _minikind_from_string(target_details[0])
1102
if path_info is not None and _versioned_minikind(target_minikind):
1103
if self.target_index != 0:
1104
raise AssertionError("Unsupported target index %d" %
1106
link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1107
# The entry may have been modified by update_entry
1108
target_details = details_list[self.target_index]
1109
target_minikind = _minikind_from_string(target_details[0])
1112
# the rest of this function is 0.3 seconds on 50K paths, or
1113
# 0.000006 seconds per call.
1114
source_minikind = _minikind_from_string(source_details[0])
1115
if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1116
and _versioned_minikind(target_minikind)):
1117
# claimed content in both: diff
1118
# r | fdlt | | add source to search, add id path move and perform
1119
# | | | diff check on source-target
1120
# r | fdlt | a | dangling file that was present in the basis.
1122
if source_minikind != c'r':
1123
old_dirname = entry[0][0]
1124
old_basename = entry[0][1]
1125
old_path = path = None
1127
# add the source to the search path to find any children it
1128
# has. TODO ? : only add if it is a container ?
1129
if (not self.doing_consistency_expansion and
1130
not osutils.is_inside_any(self.searched_specific_files,
1131
source_details[1])):
1132
self.search_specific_files.add(source_details[1])
1133
# expanding from a user requested path, parent expansion
1134
# for delta consistency happens later.
1135
# generate the old path; this is needed for stating later
1137
old_path = source_details[1]
1138
old_dirname, old_basename = os.path.split(old_path)
1139
path = self.pathjoin(entry[0][0], entry[0][1])
1140
old_entry = self.state._get_entry(self.source_index,
1142
# update the source details variable to be the real
1144
if old_entry == (None, None):
1145
raise errors.CorruptDirstate(self.state._filename,
1146
"entry '%s/%s' is considered renamed from %r"
1147
" but source does not exist\n"
1148
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1149
source_details = old_entry[1][self.source_index]
1150
source_minikind = _minikind_from_string(source_details[0])
1151
if path_info is None:
1152
# the file is missing on disk, show as removed.
1157
# source and target are both versioned and disk file is present.
1158
target_kind = path_info[2]
1159
if target_kind == 'directory':
1161
old_path = path = self.pathjoin(old_dirname, old_basename)
1162
file_id = entry[0][2]
1163
self.new_dirname_to_file_id[path] = file_id
1164
if source_minikind != c'd':
1167
# directories have no fingerprint
1170
elif target_kind == 'file':
1171
if source_minikind != c'f':
1174
# Check the sha. We can't just rely on the size as
1175
# content filtering may mean differ sizes actually
1176
# map to the same content
1177
if link_or_sha1 is None:
1179
statvalue, link_or_sha1 = \
1180
self.state._sha1_provider.stat_and_sha1(
1182
self.state._observed_sha1(entry, link_or_sha1,
1184
content_change = (link_or_sha1 != source_details[1])
1185
# Target details is updated at update_entry time
1186
if self.use_filesystem_for_exec:
1187
# We don't need S_ISREG here, because we are sure
1188
# we are dealing with a file.
1189
target_exec = bool(S_IXUSR & path_info[3].st_mode)
1191
target_exec = target_details[3]
1192
elif target_kind == 'symlink':
1193
if source_minikind != c'l':
1196
content_change = (link_or_sha1 != source_details[1])
1198
elif target_kind == 'tree-reference':
1199
if source_minikind != c't':
1206
path = self.pathjoin(old_dirname, old_basename)
1207
raise errors.BadFileKindError(path, path_info[2])
1208
if source_minikind == c'd':
1210
old_path = path = self.pathjoin(old_dirname, old_basename)
1212
file_id = entry[0][2]
1213
self.old_dirname_to_file_id[old_path] = file_id
1214
# parent id is the entry for the path in the target tree
1215
if old_basename and old_dirname == self.last_source_parent[0]:
1216
# use a cached hit for non-root source entries.
1217
source_parent_id = self.last_source_parent[1]
1220
source_parent_id = self.old_dirname_to_file_id[old_dirname]
1222
source_parent_entry = self.state._get_entry(self.source_index,
1223
path_utf8=old_dirname)
1224
source_parent_id = source_parent_entry[0][2]
1225
if source_parent_id == entry[0][2]:
1226
# This is the root, so the parent is None
1227
source_parent_id = None
1229
self.last_source_parent[0] = old_dirname
1230
self.last_source_parent[1] = source_parent_id
1231
new_dirname = entry[0][0]
1232
if entry[0][1] and new_dirname == self.last_target_parent[0]:
1233
# use a cached hit for non-root target entries.
1234
target_parent_id = self.last_target_parent[1]
1237
target_parent_id = self.new_dirname_to_file_id[new_dirname]
1239
# TODO: We don't always need to do the lookup, because the
1240
# parent entry will be the same as the source entry.
1241
target_parent_entry = self.state._get_entry(self.target_index,
1242
path_utf8=new_dirname)
1243
if target_parent_entry == (None, None):
1244
raise AssertionError(
1245
"Could not find target parent in wt: %s\nparent of: %s"
1246
% (new_dirname, entry))
1247
target_parent_id = target_parent_entry[0][2]
1248
if target_parent_id == entry[0][2]:
1249
# This is the root, so the parent is None
1250
target_parent_id = None
1252
self.last_target_parent[0] = new_dirname
1253
self.last_target_parent[1] = target_parent_id
1255
source_exec = source_details[3]
1256
changed = (content_change
1257
or source_parent_id != target_parent_id
1258
or old_basename != entry[0][1]
1259
or source_exec != target_exec
1261
if not changed and not self.include_unchanged:
1264
if old_path is None:
1265
path = self.pathjoin(old_dirname, old_basename)
1267
old_path_u = self.utf8_decode(old_path)[0]
1270
old_path_u = self.utf8_decode(old_path)[0]
1271
if old_path == path:
1274
path_u = self.utf8_decode(path)[0]
1275
source_kind = _minikind_to_kind(source_minikind)
1276
return (entry[0][2],
1277
(old_path_u, path_u),
1280
(source_parent_id, target_parent_id),
1281
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1282
(source_kind, target_kind),
1283
(source_exec, target_exec)), changed
1284
elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1285
# looks like a new file
1286
path = self.pathjoin(entry[0][0], entry[0][1])
1287
# parent id is the entry for the path in the target tree
1288
# TODO: these are the same for an entire directory: cache em.
1289
parent_entry = self.state._get_entry(self.target_index,
1290
path_utf8=entry[0][0])
1291
if parent_entry is None:
1292
raise errors.DirstateCorrupt(self.state,
1293
"We could not find the parent entry in index %d"
1294
" for the entry: %s"
1295
% (self.target_index, entry[0]))
1296
parent_id = parent_entry[0][2]
1297
if parent_id == entry[0][2]:
1299
if path_info is not None:
1301
if self.use_filesystem_for_exec:
1302
# We need S_ISREG here, because we aren't sure if this
1305
S_ISREG(path_info[3].st_mode)
1306
and S_IXUSR & path_info[3].st_mode)
1308
target_exec = target_details[3]
1309
return (entry[0][2],
1310
(None, self.utf8_decode(path)[0]),
1314
(None, self.utf8_decode(entry[0][1])[0]),
1315
(None, path_info[2]),
1316
(None, target_exec)), True
1318
# Its a missing file, report it as such.
1319
return (entry[0][2],
1320
(None, self.utf8_decode(path)[0]),
1324
(None, self.utf8_decode(entry[0][1])[0]),
1326
(None, False)), True
1327
elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1328
# unversioned, possibly, or possibly not deleted: we dont care.
1329
# if its still on disk, *and* theres no other entry at this
1330
# path [we dont know this in this routine at the moment -
1331
# perhaps we should change this - then it would be an unknown.
1332
old_path = self.pathjoin(entry[0][0], entry[0][1])
1333
# parent id is the entry for the path in the target tree
1334
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1335
if parent_id == entry[0][2]:
1337
return (entry[0][2],
1338
(self.utf8_decode(old_path)[0], None),
1342
(self.utf8_decode(entry[0][1])[0], None),
1343
(_minikind_to_kind(source_minikind), None),
1344
(source_details[3], None)), True
1345
elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1346
# a rename; could be a true rename, or a rename inherited from
1347
# a renamed parent. TODO: handle this efficiently. Its not
1348
# common case to rename dirs though, so a correct but slow
1349
# implementation will do.
1350
if (not self.doing_consistency_expansion and
1351
not osutils.is_inside_any(self.searched_specific_files,
1352
target_details[1])):
1353
self.search_specific_files.add(target_details[1])
1354
# We don't expand the specific files parents list here as
1355
# the path is absent in target and won't create a delta with
1357
elif ((source_minikind == c'r' or source_minikind == c'a') and
1358
(target_minikind == c'r' or target_minikind == c'a')):
1359
# neither of the selected trees contain this path,
1360
# so skip over it. This is not currently directly tested, but
1361
# is indirectly via test_too_much.TestCommands.test_conflicts.
1364
raise AssertionError("don't know how to compare "
1365
"source_minikind=%r, target_minikind=%r"
1366
% (source_minikind, target_minikind))
1367
## import pdb;pdb.set_trace()
1373
def iter_changes(self):
1376
cdef void _gather_result_for_consistency(self, result):
1377
"""Check a result we will yield to make sure we are consistent later.
1379
This gathers result's parents into a set to output later.
1381
:param result: A result tuple.
1383
if not self.partial or not result[0]:
1385
self.seen_ids.add(result[0])
1386
new_path = result[1][1]
1388
# Not the root and not a delete: queue up the parents of the path.
1389
self.search_specific_file_parents.update(
1390
osutils.parent_directories(new_path.encode('utf8')))
1391
# Add the root directory which parent_directories does not
1393
self.search_specific_file_parents.add('')
1395
cdef void _update_current_block(self):
1396
if (self.block_index < len(self.state._dirblocks) and
1397
osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1398
self.current_block = self.state._dirblocks[self.block_index]
1399
self.current_block_list = self.current_block[1]
1400
self.current_block_pos = 0
1402
self.current_block = None
1403
self.current_block_list = None
1406
# Simple thunk to allow tail recursion without pyrex confusion
1407
return self._iter_next()
1409
cdef _iter_next(self):
1410
"""Iterate over the changes."""
1411
# This function single steps through an iterator. As such while loops
1412
# are often exited by 'return' - the code is structured so that the
1413
# next call into the function will return to the same while loop. Note
1414
# that all flow control needed to re-reach that step is reexecuted,
1415
# which can be a performance problem. It has not yet been tuned to
1416
# minimise this; a state machine is probably the simplest restructuring
1417
# to both minimise this overhead and make the code considerably more
1421
# compare source_index and target_index at or under each element of search_specific_files.
1422
# follow the following comparison table. Note that we only want to do diff operations when
1423
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1427
# Source | Target | disk | action
1428
# r | fdlt | | add source to search, add id path move and perform
1429
# | | | diff check on source-target
1430
# r | fdlt | a | dangling file that was present in the basis.
1432
# r | a | | add source to search
1434
# r | r | | this path is present in a non-examined tree, skip.
1435
# r | r | a | this path is present in a non-examined tree, skip.
1436
# a | fdlt | | add new id
1437
# a | fdlt | a | dangling locally added file, skip
1438
# a | a | | not present in either tree, skip
1439
# a | a | a | not present in any tree, skip
1440
# a | r | | not present in either tree at this path, skip as it
1441
# | | | may not be selected by the users list of paths.
1442
# a | r | a | not present in either tree at this path, skip as it
1443
# | | | may not be selected by the users list of paths.
1444
# fdlt | fdlt | | content in both: diff them
1445
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
1446
# fdlt | a | | unversioned: output deleted id for now
1447
# fdlt | a | a | unversioned and deleted: output deleted id
1448
# fdlt | r | | relocated in this tree, so add target to search.
1449
# | | | Dont diff, we will see an r,fd; pair when we reach
1450
# | | | this id at the other path.
1451
# fdlt | r | a | 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.
1455
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1456
# keeping a cache of directories that we have seen.
1457
cdef object current_dirname, current_blockname
1458
cdef char * current_dirname_c, * current_blockname_c
1459
cdef int advance_entry, advance_path
1460
cdef int path_handled
1461
searched_specific_files = self.searched_specific_files
1462
# Are we walking a root?
1463
while self.root_entries_pos < self.root_entries_len:
1464
entry = self.root_entries[self.root_entries_pos]
1465
self.root_entries_pos = self.root_entries_pos + 1
1466
result, changed = self._process_entry(entry, self.root_dir_info)
1467
if changed is not None:
1469
self._gather_result_for_consistency(result)
1470
if changed or self.include_unchanged:
1472
# Have we finished the prior root, or never started one ?
1473
if self.current_root is None:
1474
# TODO: the pending list should be lexically sorted? the
1475
# interface doesn't require it.
1477
self.current_root = self.search_specific_files.pop()
1479
raise StopIteration()
1480
self.searched_specific_files.add(self.current_root)
1481
# process the entries for this containing directory: the rest will be
1482
# found by their parents recursively.
1483
self.root_entries = self.state._entries_for_path(self.current_root)
1484
self.root_entries_len = len(self.root_entries)
1485
self.current_root_unicode = self.current_root.decode('utf8')
1486
self.root_abspath = self.tree.abspath(self.current_root_unicode)
1488
root_stat = os.lstat(self.root_abspath)
1490
if e.errno == errno.ENOENT:
1491
# the path does not exist: let _process_entry know that.
1492
self.root_dir_info = None
1494
# some other random error: hand it up.
1497
self.root_dir_info = ('', self.current_root,
1498
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1500
if self.root_dir_info[2] == 'directory':
1501
if self.tree._directory_is_tree_reference(
1502
self.current_root_unicode):
1503
self.root_dir_info = self.root_dir_info[:2] + \
1504
('tree-reference',) + self.root_dir_info[3:]
1505
if not self.root_entries and not self.root_dir_info:
1506
# this specified path is not present at all, skip it.
1507
# (tail recursion, can do a loop once the full structure is
1509
return self._iter_next()
1511
self.root_entries_pos = 0
1512
# XXX Clarity: This loop is duplicated a out the self.current_root
1513
# is None guard above: if we return from it, it completes there
1514
# (and the following if block cannot trigger because
1515
# path_handled must be true, so the if block is not # duplicated.
1516
while self.root_entries_pos < self.root_entries_len:
1517
entry = self.root_entries[self.root_entries_pos]
1518
self.root_entries_pos = self.root_entries_pos + 1
1519
result, changed = self._process_entry(entry, self.root_dir_info)
1520
if changed is not None:
1523
self._gather_result_for_consistency(result)
1524
if changed or self.include_unchanged:
1526
# handle unversioned specified paths:
1527
if self.want_unversioned and not path_handled and self.root_dir_info:
1528
new_executable = bool(
1529
stat.S_ISREG(self.root_dir_info[3].st_mode)
1530
and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1532
(None, self.current_root_unicode),
1536
(None, splitpath(self.current_root_unicode)[-1]),
1537
(None, self.root_dir_info[2]),
1538
(None, new_executable)
1540
# If we reach here, the outer flow continues, which enters into the
1541
# per-root setup logic.
1542
if (self.current_dir_info is None and self.current_block is None and not
1543
self.doing_consistency_expansion):
1544
# setup iteration of this root:
1545
self.current_dir_list = None
1546
if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1547
self.current_dir_info = None
1549
self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1550
prefix=self.current_root)
1553
self.current_dir_info = self.dir_iterator.next()
1554
self.current_dir_list = self.current_dir_info[1]
1556
# there may be directories in the inventory even though
1557
# this path is not a file on disk: so mark it as end of
1559
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1560
self.current_dir_info = None
1561
elif sys.platform == 'win32':
1562
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1563
# python 2.5 has e.errno == EINVAL,
1564
# and e.winerror == ERROR_DIRECTORY
1566
e_winerror = e.winerror
1567
except AttributeError:
1569
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1570
if (e.errno in win_errors or e_winerror in win_errors):
1571
self.current_dir_info = None
1573
# Will this really raise the right exception ?
1578
if self.current_dir_info[0][0] == '':
1579
# remove .bzr from iteration
1580
bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1581
if self.current_dir_list[bzr_index][0] != '.bzr':
1582
raise AssertionError()
1583
del self.current_dir_list[bzr_index]
1584
initial_key = (self.current_root, '', '')
1585
self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1586
if self.block_index == 0:
1587
# we have processed the total root already, but because the
1588
# initial key matched it we should skip it here.
1589
self.block_index = self.block_index + 1
1590
self._update_current_block()
1591
# walk until both the directory listing and the versioned metadata
1593
while (self.current_dir_info is not None
1594
or self.current_block is not None):
1595
# Uncommon case - a missing directory or an unversioned directory:
1596
if (self.current_dir_info and self.current_block
1597
and self.current_dir_info[0][0] != self.current_block[0]):
1598
# Work around pyrex broken heuristic - current_dirname has
1599
# the same scope as current_dirname_c
1600
current_dirname = self.current_dir_info[0][0]
1601
current_dirname_c = PyString_AS_STRING_void(
1602
<void *>current_dirname)
1603
current_blockname = self.current_block[0]
1604
current_blockname_c = PyString_AS_STRING_void(
1605
<void *>current_blockname)
1606
# In the python generator we evaluate this if block once per
1607
# dir+block; because we reenter in the pyrex version its being
1608
# evaluated once per path: we could cache the result before
1609
# doing the while loop and probably save time.
1610
if _cmp_by_dirs(current_dirname_c,
1611
PyString_Size(current_dirname),
1612
current_blockname_c,
1613
PyString_Size(current_blockname)) < 0:
1614
# filesystem data refers to paths not covered by the
1615
# dirblock. this has two possibilities:
1616
# A) it is versioned but empty, so there is no block for it
1617
# B) it is not versioned.
1619
# if (A) then we need to recurse into it to check for
1620
# new unknown files or directories.
1621
# if (B) then we should ignore it, because we don't
1622
# recurse into unknown directories.
1623
# We are doing a loop
1624
while self.path_index < len(self.current_dir_list):
1625
current_path_info = self.current_dir_list[self.path_index]
1626
# dont descend into this unversioned path if it is
1628
if current_path_info[2] in ('directory',
1630
del self.current_dir_list[self.path_index]
1631
self.path_index = self.path_index - 1
1632
self.path_index = self.path_index + 1
1633
if self.want_unversioned:
1634
if current_path_info[2] == 'directory':
1635
if self.tree._directory_is_tree_reference(
1636
self.utf8_decode(current_path_info[0])[0]):
1637
current_path_info = current_path_info[:2] + \
1638
('tree-reference',) + current_path_info[3:]
1639
new_executable = bool(
1640
stat.S_ISREG(current_path_info[3].st_mode)
1641
and stat.S_IEXEC & current_path_info[3].st_mode)
1643
(None, self.utf8_decode(current_path_info[0])[0]),
1647
(None, self.utf8_decode(current_path_info[1])[0]),
1648
(None, current_path_info[2]),
1649
(None, new_executable))
1650
# This dir info has been handled, go to the next
1652
self.current_dir_list = None
1654
self.current_dir_info = self.dir_iterator.next()
1655
self.current_dir_list = self.current_dir_info[1]
1656
except StopIteration:
1657
self.current_dir_info = None
1659
# We have a dirblock entry for this location, but there
1660
# is no filesystem path for this. This is most likely
1661
# because a directory was removed from the disk.
1662
# We don't have to report the missing directory,
1663
# because that should have already been handled, but we
1664
# need to handle all of the files that are contained
1666
while self.current_block_pos < len(self.current_block_list):
1667
current_entry = self.current_block_list[self.current_block_pos]
1668
self.current_block_pos = self.current_block_pos + 1
1669
# entry referring to file not present on disk.
1670
# advance the entry only, after processing.
1671
result, changed = self._process_entry(current_entry, None)
1672
if changed is not None:
1674
self._gather_result_for_consistency(result)
1675
if changed or self.include_unchanged:
1677
self.block_index = self.block_index + 1
1678
self._update_current_block()
1679
continue # next loop-on-block/dir
1680
result = self._loop_one_block()
1681
if result is not None:
1683
if len(self.search_specific_files):
1684
# More supplied paths to process
1685
self.current_root = None
1686
return self._iter_next()
1687
# Start expanding more conservatively, adding paths the user may not
1688
# have intended but required for consistent deltas.
1689
self.doing_consistency_expansion = 1
1690
if not self._pending_consistent_entries:
1691
self._pending_consistent_entries = self._next_consistent_entries()
1692
while self._pending_consistent_entries:
1693
result, changed = self._pending_consistent_entries.pop()
1694
if changed is not None:
1696
raise StopIteration()
1698
cdef object _maybe_tree_ref(self, current_path_info):
1699
if self.tree._directory_is_tree_reference(
1700
self.utf8_decode(current_path_info[0])[0]):
1701
return current_path_info[:2] + \
1702
('tree-reference',) + current_path_info[3:]
1704
return current_path_info
1706
cdef object _loop_one_block(self):
1707
# current_dir_info and current_block refer to the same directory -
1708
# this is the common case code.
1709
# Assign local variables for current path and entry:
1710
cdef object current_entry
1711
cdef object current_path_info
1712
cdef int path_handled
1715
# cdef char * temp_str
1716
# cdef Py_ssize_t temp_str_length
1717
# PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1718
# if not strncmp(temp_str, "directory", temp_str_length):
1719
if (self.current_block is not None and
1720
self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1721
current_entry = PyList_GET_ITEM(self.current_block_list,
1722
self.current_block_pos)
1724
Py_INCREF(current_entry)
1726
current_entry = None
1727
if (self.current_dir_info is not None and
1728
self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1729
current_path_info = PyList_GET_ITEM(self.current_dir_list,
1732
Py_INCREF(current_path_info)
1733
disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1735
Py_INCREF(disk_kind)
1736
if disk_kind == "directory":
1737
current_path_info = self._maybe_tree_ref(current_path_info)
1739
current_path_info = None
1740
while (current_entry is not None or current_path_info is not None):
1745
if current_entry is None:
1746
# unversioned - the check for path_handled when the path
1747
# is advanced will yield this path if needed.
1749
elif current_path_info is None:
1750
# no path is fine: the per entry code will handle it.
1751
result, changed = self._process_entry(current_entry,
1754
minikind = _minikind_from_string(
1755
current_entry[1][self.target_index][0])
1756
cmp_result = cmp(current_path_info[1], current_entry[0][1])
1757
if (cmp_result or minikind == c'a' or minikind == c'r'):
1758
# The current path on disk doesn't match the dirblock
1759
# record. Either the dirblock record is marked as
1760
# absent/renamed, or the file on disk is not present at all
1761
# in the dirblock. Either way, report about the dirblock
1762
# entry, and let other code handle the filesystem one.
1764
# Compare the basename for these files to determine
1767
# extra file on disk: pass for now, but only
1768
# increment the path, not the entry
1771
# entry referring to file not present on disk.
1772
# advance the entry only, after processing.
1773
result, changed = self._process_entry(current_entry,
1777
# paths are the same,and the dirstate entry is not
1778
# absent or renamed.
1779
result, changed = self._process_entry(current_entry,
1781
if changed is not None:
1783
if not changed and not self.include_unchanged:
1785
# >- loop control starts here:
1787
if advance_entry and current_entry is not None:
1788
self.current_block_pos = self.current_block_pos + 1
1789
if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1790
current_entry = self.current_block_list[self.current_block_pos]
1792
current_entry = None
1794
if advance_path and current_path_info is not None:
1795
if not path_handled:
1796
# unversioned in all regards
1797
if self.want_unversioned:
1798
new_executable = bool(
1799
stat.S_ISREG(current_path_info[3].st_mode)
1800
and stat.S_IEXEC & current_path_info[3].st_mode)
1802
relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1803
except UnicodeDecodeError:
1804
raise errors.BadFilenameEncoding(
1805
current_path_info[0], osutils._fs_enc)
1806
if changed is not None:
1807
raise AssertionError(
1808
"result is not None: %r" % result)
1810
(None, relpath_unicode),
1814
(None, self.utf8_decode(current_path_info[1])[0]),
1815
(None, current_path_info[2]),
1816
(None, new_executable))
1818
# dont descend into this unversioned path if it is
1820
if current_path_info[2] in ('directory'):
1821
del self.current_dir_list[self.path_index]
1822
self.path_index = self.path_index - 1
1823
# dont descend the disk iterator into any tree
1825
if current_path_info[2] == 'tree-reference':
1826
del self.current_dir_list[self.path_index]
1827
self.path_index = self.path_index - 1
1828
self.path_index = self.path_index + 1
1829
if self.path_index < len(self.current_dir_list):
1830
current_path_info = self.current_dir_list[self.path_index]
1831
if current_path_info[2] == 'directory':
1832
current_path_info = self._maybe_tree_ref(
1835
current_path_info = None
1836
if changed is not None:
1837
# Found a result on this pass, yield it
1839
self._gather_result_for_consistency(result)
1840
if changed or self.include_unchanged:
1842
if self.current_block is not None:
1843
self.block_index = self.block_index + 1
1844
self._update_current_block()
1845
if self.current_dir_info is not None:
1847
self.current_dir_list = None
1849
self.current_dir_info = self.dir_iterator.next()
1850
self.current_dir_list = self.current_dir_info[1]
1851
except StopIteration:
1852
self.current_dir_info = None
1854
cdef object _next_consistent_entries(self):
1855
"""Grabs the next specific file parent case to consider.
1857
:return: A list of the results, each of which is as for _process_entry.
1860
while self.search_specific_file_parents:
1861
# Process the parent directories for the paths we were iterating.
1862
# Even in extremely large trees this should be modest, so currently
1863
# no attempt is made to optimise.
1864
path_utf8 = self.search_specific_file_parents.pop()
1865
if path_utf8 in self.searched_exact_paths:
1866
# We've examined this path.
1868
if osutils.is_inside_any(self.searched_specific_files, path_utf8):
1869
# We've examined this path.
1871
path_entries = self.state._entries_for_path(path_utf8)
1872
# We need either one or two entries. If the path in
1873
# self.target_index has moved (so the entry in source_index is in
1874
# 'ar') then we need to also look for the entry for this path in
1875
# self.source_index, to output the appropriate delete-or-rename.
1876
selected_entries = []
1878
for candidate_entry in path_entries:
1879
# Find entries present in target at this path:
1880
if candidate_entry[1][self.target_index][0] not in 'ar':
1882
selected_entries.append(candidate_entry)
1883
# Find entries present in source at this path:
1884
elif (self.source_index is not None and
1885
candidate_entry[1][self.source_index][0] not in 'ar'):
1887
if candidate_entry[1][self.target_index][0] == 'a':
1888
# Deleted, emit it here.
1889
selected_entries.append(candidate_entry)
1891
# renamed, emit it when we process the directory it
1893
self.search_specific_file_parents.add(
1894
candidate_entry[1][self.target_index][1])
1896
raise AssertionError(
1897
"Missing entry for specific path parent %r, %r" % (
1898
path_utf8, path_entries))
1899
path_info = self._path_info(path_utf8, path_utf8.decode('utf8'))
1900
for entry in selected_entries:
1901
if entry[0][2] in self.seen_ids:
1903
result, changed = self._process_entry(entry, path_info)
1905
raise AssertionError(
1906
"Got entry<->path mismatch for specific path "
1907
"%r entry %r path_info %r " % (
1908
path_utf8, entry, path_info))
1909
# Only include changes - we're outside the users requested
1912
self._gather_result_for_consistency(result)
1913
if (result[6][0] == 'directory' and
1914
result[6][1] != 'directory'):
1915
# This stopped being a directory, the old children have
1917
if entry[1][self.source_index][0] == 'r':
1918
# renamed, take the source path
1919
entry_path_utf8 = entry[1][self.source_index][1]
1921
entry_path_utf8 = path_utf8
1922
initial_key = (entry_path_utf8, '', '')
1923
block_index, _ = self.state._find_block_index_from_key(
1925
if block_index == 0:
1926
# The children of the root are in block index 1.
1927
block_index = block_index + 1
1928
current_block = None
1929
if block_index < len(self.state._dirblocks):
1930
current_block = self.state._dirblocks[block_index]
1931
if not osutils.is_inside(
1932
entry_path_utf8, current_block[0]):
1933
# No entries for this directory at all.
1934
current_block = None
1935
if current_block is not None:
1936
for entry in current_block[1]:
1937
if entry[1][self.source_index][0] in 'ar':
1938
# Not in the source tree, so doesn't have to be
1941
# Path of the entry itself.
1942
self.search_specific_file_parents.add(
1943
self.pathjoin(*entry[0][:2]))
1944
if changed or self.include_unchanged:
1945
results.append((result, changed))
1946
self.searched_exact_paths.add(path_utf8)
1949
cdef object _path_info(self, utf8_path, unicode_path):
1950
"""Generate path_info for unicode_path.
1952
:return: None if unicode_path does not exist, or a path_info tuple.
1954
abspath = self.tree.abspath(unicode_path)
1956
stat = os.lstat(abspath)
1958
if e.errno == errno.ENOENT:
1959
# the path does not exist.
1963
utf8_basename = utf8_path.rsplit('/', 1)[-1]
1964
dir_info = (utf8_path, utf8_basename,
1965
osutils.file_kind_from_stat_mode(stat.st_mode), stat,
1967
if dir_info[2] == 'directory':
1968
if self.tree._directory_is_tree_reference(
1970
self.root_dir_info = self.root_dir_info[:2] + \
1971
('tree-reference',) + self.root_dir_info[3:]
added
removed
bzrlib/_dirstate_helpers_pyx.pyx
