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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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 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_c.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_c(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_c(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(PyString_AsString(path1),
292
PyString_Size(path1),
293
PyString_AsString(path2),
294
PyString_Size(path2))
297
cdef int _cmp_path_by_dirblock(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_c`` 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_c(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(cur_cstr, cur_size, path_cstr, path_size) < 0:
423
def _bisect_path_right_c(paths, path):
424
"""Return the index where to insert path into paths.
426
This uses a path-wise comparison so we get::
436
:param paths: A list of paths to search through
437
:param path: A single path to insert
438
:return: An offset where 'path' can be inserted.
439
:seealso: bisect.bisect_right
450
if not PyList_CheckExact(paths):
451
raise TypeError("you must pass a python list for 'paths' not: %s %r"
452
% (type(paths), paths))
453
if not PyString_CheckExact(path):
454
raise TypeError("you must pass a string for 'path' not: %s %r"
455
% (type(path), path))
460
path_cstr = PyString_AsString(path)
461
path_size = PyString_Size(path)
464
_mid = (_lo + _hi) / 2
465
cur = PyList_GetItem_object_void(paths, _mid)
466
cur_cstr = PyString_AS_STRING_void(cur)
467
cur_size = PyString_GET_SIZE_void(cur)
468
if _cmp_path_by_dirblock(path_cstr, path_size, cur_cstr, cur_size) < 0:
475
def bisect_dirblock_c(dirblocks, dirname, lo=0, hi=None, cache=None):
476
"""Return the index where to insert dirname into the dirblocks.
478
The return value idx is such that all directories blocks in dirblock[:idx]
479
have names < dirname, and all blocks in dirblock[idx:] have names >=
482
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
483
slice of a to be searched.
488
cdef char *dirname_cstr
489
cdef int dirname_size
494
if not PyList_CheckExact(dirblocks):
495
raise TypeError("you must pass a python list for 'dirblocks' not: %s %r"
496
% (type(dirblocks), dirblocks))
497
if not PyString_CheckExact(dirname):
498
raise TypeError("you must pass a string for dirname not: %s %r"
499
% (type(dirname), dirname))
506
dirname_cstr = PyString_AsString(dirname)
507
dirname_size = PyString_Size(dirname)
510
_mid = (_lo + _hi) / 2
511
# Grab the dirname for the current dirblock
512
# cur = dirblocks[_mid][0]
513
cur = PyTuple_GetItem_void_void(
514
PyList_GetItem_object_void(dirblocks, _mid), 0)
515
cur_cstr = PyString_AS_STRING_void(cur)
516
cur_size = PyString_GET_SIZE_void(cur)
517
if _cmp_by_dirs(cur_cstr, cur_size, dirname_cstr, dirname_size) < 0:
525
"""Maintain the current location, and return fields as you parse them."""
527
cdef object state # The DirState object
528
cdef object text # The overall string object
529
cdef char *text_cstr # Pointer to the beginning of text
530
cdef int text_size # Length of text
532
cdef char *end_cstr # End of text
533
cdef char *cur_cstr # Pointer to the current record
534
cdef char *next # Pointer to the end of this record
536
def __init__(self, text, state):
539
self.text_cstr = PyString_AsString(text)
540
self.text_size = PyString_Size(text)
541
self.end_cstr = self.text_cstr + self.text_size
542
self.cur_cstr = self.text_cstr
544
cdef char *get_next(self, int *size) except NULL:
545
"""Return a pointer to the start of the next field."""
547
cdef Py_ssize_t extra_len
549
if self.cur_cstr == NULL:
550
raise AssertionError('get_next() called when cur_str is NULL')
551
elif self.cur_cstr >= self.end_cstr:
552
raise AssertionError('get_next() called when there are no chars'
555
self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr - next)
556
if self.cur_cstr == NULL:
557
extra_len = self.end_cstr - next
558
raise errors.DirstateCorrupt(self.state,
559
'failed to find trailing NULL (\\0).'
560
' Trailing garbage: %r'
561
% safe_string_from_size(next, extra_len))
562
size[0] = self.cur_cstr - next
563
self.cur_cstr = self.cur_cstr + 1
566
cdef object get_next_str(self):
567
"""Get the next field as a Python string."""
570
next = self.get_next(&size)
571
return safe_string_from_size(next, size)
573
cdef int _init(self) except -1:
574
"""Get the pointer ready.
576
This assumes that the dirstate header has already been read, and we
577
already have the dirblock string loaded into memory.
578
This just initializes our memory pointers, etc for parsing of the
583
# The first field should be an empty string left over from the Header
584
first = self.get_next(&size)
585
if first[0] != c'\0' and size == 0:
586
raise AssertionError('First character should be null not: %s'
590
cdef object _get_entry(self, int num_trees, void **p_current_dirname,
592
"""Extract the next entry.
594
This parses the next entry based on the current location in
596
Each entry can be considered a "row" in the total table. And each row
597
has a fixed number of columns. It is generally broken up into "key"
598
columns, then "current" columns, and then "parent" columns.
600
:param num_trees: How many parent trees need to be parsed
601
:param p_current_dirname: A pointer to the current PyString
602
representing the directory name.
603
We pass this in as a void * so that pyrex doesn't have to
604
increment/decrement the PyObject reference counter for each
606
We use a pointer so that _get_entry can update it with the new
608
:param new_block: This is to let the caller know that it needs to
609
create a new directory block to store the next entry.
611
cdef object path_name_file_id_key
612
cdef char *entry_size_cstr
613
cdef unsigned long int entry_size
614
cdef char* executable_cstr
615
cdef int is_executable
616
cdef char* dirname_cstr
621
cdef object fingerprint
624
# Read the 'key' information (dirname, name, file_id)
625
dirname_cstr = self.get_next(&cur_size)
626
# Check to see if we have started a new directory block.
627
# If so, then we need to create a new dirname PyString, so that it can
628
# be used in all of the tuples. This saves time and memory, by re-using
629
# the same object repeatedly.
631
# Do the cheap 'length of string' check first. If the string is a
632
# different length, then we *have* to be a different directory.
633
if (cur_size != PyString_GET_SIZE_void(p_current_dirname[0])
634
or strncmp(dirname_cstr,
635
# Extract the char* from our current dirname string. We
636
# know it is a PyString, so we can use
637
# PyString_AS_STRING, we use the _void version because
638
# we are tricking Pyrex by using a void* rather than an
640
PyString_AS_STRING_void(p_current_dirname[0]),
642
dirname = safe_string_from_size(dirname_cstr, cur_size)
643
p_current_dirname[0] = <void*>dirname
648
# Build up the key that will be used.
649
# By using <object>(void *) Pyrex will automatically handle the
650
# Py_INCREF that we need.
651
path_name_file_id_key = (<object>p_current_dirname[0],
656
# Parse all of the per-tree information. current has the information in
657
# the same location as parent trees. The only difference is that 'info'
658
# is a 'packed_stat' for current, while it is a 'revision_id' for
660
# minikind, fingerprint, and info will be returned as regular python
662
# entry_size and is_executable will be parsed into a python Long and
663
# python Boolean, respectively.
664
# TODO: jam 20070718 Consider changin the entry_size conversion to
665
# prefer python Int when possible. They are generally faster to
666
# work with, and it will be rare that we have a file >2GB.
667
# Especially since this code is pretty much fixed at a max of
670
for i from 0 <= i < num_trees:
671
minikind = self.get_next_str()
672
fingerprint = self.get_next_str()
673
entry_size_cstr = self.get_next(&cur_size)
674
entry_size = strtoul(entry_size_cstr, NULL, 10)
675
executable_cstr = self.get_next(&cur_size)
676
is_executable = (executable_cstr[0] == c'y')
677
info = self.get_next_str()
678
PyList_Append(trees, (
680
fingerprint, # fingerprint
682
is_executable,# executable
683
info, # packed_stat or revision_id
686
# The returned tuple is (key, [trees])
687
ret = (path_name_file_id_key, trees)
688
# Ignore the trailing newline, but assert that it does exist, this
689
# ensures that we always finish parsing a line on an end-of-entry
691
trailing = self.get_next(&cur_size)
692
if cur_size != 1 or trailing[0] != c'\n':
693
raise errors.DirstateCorrupt(self.state,
694
'Bad parse, we expected to end on \\n, not: %d %s: %s'
695
% (cur_size, safe_string_from_size(trailing, cur_size),
699
def _parse_dirblocks(self):
700
"""Parse all dirblocks in the state file."""
702
cdef object current_block
704
cdef void * current_dirname
706
cdef int expected_entry_count
709
num_trees = self.state._num_present_parents() + 1
710
expected_entry_count = self.state._num_entries
712
# Ignore the first record
716
dirblocks = [('', current_block), ('', [])]
717
self.state._dirblocks = dirblocks
719
current_dirname = <void*>obj
723
# TODO: jam 2007-05-07 Consider pre-allocating some space for the
724
# members, and then growing and shrinking from there. If most
725
# directories have close to 10 entries in them, it would save a
726
# few mallocs if we default our list size to something
727
# reasonable. Or we could malloc it to something large (100 or
728
# so), and then truncate. That would give us a malloc + realloc,
729
# rather than lots of reallocs.
730
while self.cur_cstr < self.end_cstr:
731
entry = self._get_entry(num_trees, ¤t_dirname, &new_block)
733
# new block - different dirname
735
PyList_Append(dirblocks,
736
(<object>current_dirname, current_block))
737
PyList_Append(current_block, entry)
738
entry_count = entry_count + 1
739
if entry_count != expected_entry_count:
740
raise errors.DirstateCorrupt(self.state,
741
'We read the wrong number of entries.'
742
' We expected to read %s, but read %s'
743
% (expected_entry_count, entry_count))
744
self.state._split_root_dirblock_into_contents()
747
def _read_dirblocks_c(state):
748
"""Read in the dirblocks for the given DirState object.
750
This is tightly bound to the DirState internal representation. It should be
751
thought of as a member function, which is only separated out so that we can
752
re-write it in pyrex.
754
:param state: A DirState object.
756
:postcondition: The dirblocks will be loaded into the appropriate fields in
759
state._state_file.seek(state._end_of_header)
760
text = state._state_file.read()
761
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
763
reader = Reader(text, state)
765
reader._parse_dirblocks()
766
state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
769
cdef int minikind_from_mode(int mode):
770
# in order of frequency:
780
_encode = binascii.b2a_base64
783
from struct import pack
784
cdef _pack_stat(stat_value):
785
"""return a string representing the stat value's key fields.
787
:param stat_value: A stat oject with st_size, st_mtime, st_ctime, st_dev,
788
st_ino and st_mode fields.
790
cdef char result[6*4] # 6 long ints
792
aliased = <int *>result
793
aliased[0] = htonl(stat_value.st_size)
794
aliased[1] = htonl(int(stat_value.st_mtime))
795
aliased[2] = htonl(int(stat_value.st_ctime))
796
aliased[3] = htonl(stat_value.st_dev)
797
aliased[4] = htonl(stat_value.st_ino & 0xFFFFFFFF)
798
aliased[5] = htonl(stat_value.st_mode)
799
packed = PyString_FromStringAndSize(result, 6*4)
800
return _encode(packed)[:-1]
803
def update_entry(self, entry, abspath, stat_value):
804
"""Update the entry based on what is actually on disk.
806
This function only calculates the sha if it needs to - if the entry is
807
uncachable, or clearly different to the first parent's entry, no sha
808
is calculated, and None is returned.
810
:param entry: This is the dirblock entry for the file in question.
811
:param abspath: The path on disk for this file.
812
:param stat_value: (optional) if we already have done a stat on the
814
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
817
return _update_entry(self, entry, abspath, stat_value)
820
cdef _update_entry(self, entry, abspath, stat_value):
821
"""Update the entry based on what is actually on disk.
823
This function only calculates the sha if it needs to - if the entry is
824
uncachable, or clearly different to the first parent's entry, no sha
825
is calculated, and None is returned.
827
:param self: The dirstate object this is operating on.
828
:param entry: This is the dirblock entry for the file in question.
829
:param abspath: The path on disk for this file.
830
:param stat_value: The stat value done on the path.
831
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
834
# TODO - require pyrex 0.9.8, then use a pyd file to define access to the
835
# _st mode of the compiled stat objects.
836
cdef int minikind, saved_minikind
838
minikind = minikind_from_mode(stat_value.st_mode)
841
packed_stat = _pack_stat(stat_value)
842
details = PyList_GetItem_void_void(PyTuple_GetItem_void_void(<void *>entry, 1), 0)
843
saved_minikind = PyString_AsString_obj(<PyObject *>PyTuple_GetItem_void_void(details, 0))[0]
844
saved_link_or_sha1 = PyTuple_GetItem_void_object(details, 1)
845
saved_file_size = PyTuple_GetItem_void_object(details, 2)
846
saved_executable = PyTuple_GetItem_void_object(details, 3)
847
saved_packed_stat = PyTuple_GetItem_void_object(details, 4)
848
# Deal with pyrex decrefing the objects
849
Py_INCREF(saved_link_or_sha1)
850
Py_INCREF(saved_file_size)
851
Py_INCREF(saved_executable)
852
Py_INCREF(saved_packed_stat)
853
#(saved_minikind, saved_link_or_sha1, saved_file_size,
854
# saved_executable, saved_packed_stat) = entry[1][0]
856
if (minikind == saved_minikind
857
and packed_stat == saved_packed_stat):
858
# The stat hasn't changed since we saved, so we can re-use the
863
# size should also be in packed_stat
864
if saved_file_size == stat_value.st_size:
865
return saved_link_or_sha1
867
# If we have gotten this far, that means that we need to actually
868
# process this entry.
871
executable = self._is_executable(stat_value.st_mode,
873
if self._cutoff_time is None:
874
self._sha_cutoff_time()
875
if (stat_value.st_mtime < self._cutoff_time
876
and stat_value.st_ctime < self._cutoff_time
877
and len(entry[1]) > 1
878
and entry[1][1][0] != 'a'):
879
# Could check for size changes for further optimised
880
# avoidance of sha1's. However the most prominent case of
881
# over-shaing is during initial add, which this catches.
882
link_or_sha1 = self._sha1_file(abspath)
883
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
884
executable, packed_stat)
886
entry[1][0] = ('f', '', stat_value.st_size,
887
executable, DirState.NULLSTAT)
888
elif minikind == c'd':
890
entry[1][0] = ('d', '', 0, False, packed_stat)
891
if saved_minikind != c'd':
892
# This changed from something into a directory. Make sure we
893
# have a directory block for it. This doesn't happen very
894
# often, so this doesn't have to be super fast.
895
block_index, entry_index, dir_present, file_present = \
896
self._get_block_entry_index(entry[0][0], entry[0][1], 0)
897
self._ensure_block(block_index, entry_index,
898
pathjoin(entry[0][0], entry[0][1]))
899
elif minikind == c'l':
900
link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
901
if self._cutoff_time is None:
902
self._sha_cutoff_time()
903
if (stat_value.st_mtime < self._cutoff_time
904
and stat_value.st_ctime < self._cutoff_time):
905
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
908
entry[1][0] = ('l', '', stat_value.st_size,
909
False, DirState.NULLSTAT)
910
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
914
cdef char _minikind_from_string(object string):
915
"""Convert a python string to a char."""
916
return PyString_AsString(string)[0]
919
cdef object _kind_absent
920
cdef object _kind_file
921
cdef object _kind_directory
922
cdef object _kind_symlink
923
cdef object _kind_relocated
924
cdef object _kind_tree_reference
925
_kind_absent = "absent"
927
_kind_directory = "directory"
928
_kind_symlink = "symlink"
929
_kind_relocated = "relocated"
930
_kind_tree_reference = "tree-reference"
933
cdef object _minikind_to_kind(char minikind):
934
"""Create a string kind for minikind."""
935
cdef char _minikind[1]
938
elif minikind == c'd':
939
return _kind_directory
940
elif minikind == c'a':
942
elif minikind == c'r':
943
return _kind_relocated
944
elif minikind == c'l':
946
elif minikind == c't':
947
return _kind_tree_reference
948
_minikind[0] = minikind
949
raise KeyError(PyString_FromStringAndSize(_minikind, 1))
952
cdef int _versioned_minikind(char minikind):
953
"""Return non-zero if minikind is in fltd"""
954
return (minikind == c'f' or
960
cdef class ProcessEntryC:
962
cdef object old_dirname_to_file_id # dict
963
cdef object new_dirname_to_file_id # dict
964
cdef readonly object uninteresting
965
cdef object last_source_parent
966
cdef object last_target_parent
967
cdef object include_unchanged
968
cdef object use_filesystem_for_exec
969
cdef object utf8_decode
970
cdef readonly object searched_specific_files
971
cdef object search_specific_files
973
# Current iteration variables:
974
cdef object current_root
975
cdef object current_root_unicode
976
cdef object root_entries
977
cdef int root_entries_pos, root_entries_len
978
cdef object root_abspath
979
cdef int source_index, target_index
980
cdef int want_unversioned
982
cdef object dir_iterator
984
cdef object current_block
985
cdef int current_block_pos
986
cdef object current_block_list
987
cdef object current_dir_info
988
cdef object current_dir_list
990
cdef object root_dir_info
991
cdef object bisect_left
996
def __init__(self, include_unchanged, use_filesystem_for_exec,
997
search_specific_files, state, source_index, target_index,
998
want_unversioned, tree):
999
self.old_dirname_to_file_id = {}
1000
self.new_dirname_to_file_id = {}
1001
# Just a sentry, so that _process_entry can say that this
1002
# record is handled, but isn't interesting to process (unchanged)
1003
self.uninteresting = object()
1004
# Using a list so that we can access the values and change them in
1005
# nested scope. Each one is [path, file_id, entry]
1006
self.last_source_parent = [None, None]
1007
self.last_target_parent = [None, None]
1008
self.include_unchanged = include_unchanged
1009
self.use_filesystem_for_exec = use_filesystem_for_exec
1010
self.utf8_decode = cache_utf8._utf8_decode
1011
# for all search_indexs in each path at or under each element of
1012
# search_specific_files, if the detail is relocated: add the id, and add the
1013
# relocated path as one to search if its not searched already. If the
1014
# detail is not relocated, add the id.
1015
self.searched_specific_files = set()
1016
self.search_specific_files = search_specific_files
1018
self.current_root = None
1019
self.current_root_unicode = None
1020
self.root_entries = None
1021
self.root_entries_pos = 0
1022
self.root_entries_len = 0
1023
self.root_abspath = None
1024
if source_index is None:
1025
self.source_index = -1
1027
self.source_index = source_index
1028
self.target_index = target_index
1029
self.want_unversioned = want_unversioned
1031
self.dir_iterator = None
1032
self.block_index = -1
1033
self.current_block = None
1034
self.current_block_list = None
1035
self.current_block_pos = -1
1036
self.current_dir_info = None
1037
self.current_dir_list = None
1039
self.root_dir_info = None
1040
self.bisect_left = bisect.bisect_left
1041
self.pathjoin = osutils.pathjoin
1042
self.fstat = os.fstat
1043
self.sha_file = osutils.sha_file
1045
cdef _process_entry(self, entry, path_info):
1046
"""Compare an entry and real disk to generate delta information.
1048
:param path_info: top_relpath, basename, kind, lstat, abspath for
1049
the path of entry. If None, then the path is considered absent.
1050
(Perhaps we should pass in a concrete entry for this ?)
1051
Basename is returned as a utf8 string because we expect this
1052
tuple will be ignored, and don't want to take the time to
1054
:return: None if the these don't match
1055
A tuple of information about the change, or
1056
the object 'uninteresting' if these match, but are
1057
basically identical.
1059
cdef char target_minikind
1060
cdef char source_minikind
1062
cdef int content_change
1063
cdef object details_list
1065
details_list = entry[1]
1066
if -1 == self.source_index:
1067
source_details = DirState.NULL_PARENT_DETAILS
1069
source_details = details_list[self.source_index]
1070
target_details = details_list[self.target_index]
1071
target_minikind = _minikind_from_string(target_details[0])
1072
if path_info is not None and _versioned_minikind(target_minikind):
1073
if self.target_index != 0:
1074
raise AssertionError("Unsupported target index %d" %
1076
link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1077
# The entry may have been modified by update_entry
1078
target_details = details_list[self.target_index]
1079
target_minikind = _minikind_from_string(target_details[0])
1082
# the rest of this function is 0.3 seconds on 50K paths, or
1083
# 0.000006 seconds per call.
1084
source_minikind = _minikind_from_string(source_details[0])
1085
if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1086
and _versioned_minikind(target_minikind)):
1087
# claimed content in both: diff
1088
# r | fdlt | | add source to search, add id path move and perform
1089
# | | | diff check on source-target
1090
# r | fdlt | a | dangling file that was present in the basis.
1092
if source_minikind != c'r':
1093
old_dirname = entry[0][0]
1094
old_basename = entry[0][1]
1095
old_path = path = None
1097
# add the source to the search path to find any children it
1098
# has. TODO ? : only add if it is a container ?
1099
if not osutils.is_inside_any(self.searched_specific_files,
1101
self.search_specific_files.add(source_details[1])
1102
# generate the old path; this is needed for stating later
1104
old_path = source_details[1]
1105
old_dirname, old_basename = os.path.split(old_path)
1106
path = self.pathjoin(entry[0][0], entry[0][1])
1107
old_entry = self.state._get_entry(self.source_index,
1109
# update the source details variable to be the real
1111
if old_entry == (None, None):
1112
raise errors.CorruptDirstate(self.state._filename,
1113
"entry '%s/%s' is considered renamed from %r"
1114
" but source does not exist\n"
1115
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1116
source_details = old_entry[1][self.source_index]
1117
source_minikind = _minikind_from_string(source_details[0])
1118
if path_info is None:
1119
# the file is missing on disk, show as removed.
1124
# source and target are both versioned and disk file is present.
1125
target_kind = path_info[2]
1126
if target_kind == 'directory':
1128
old_path = path = self.pathjoin(old_dirname, old_basename)
1129
file_id = entry[0][2]
1130
self.new_dirname_to_file_id[path] = file_id
1131
if source_minikind != c'd':
1134
# directories have no fingerprint
1137
elif target_kind == 'file':
1138
if source_minikind != c'f':
1141
# If the size is the same, check the sha:
1142
if target_details[2] == source_details[2]:
1143
if link_or_sha1 is None:
1145
file_obj = file(path_info[4], 'rb')
1147
# XXX: TODO: Use lower level file IO rather
1148
# than python objects for sha-misses.
1149
statvalue = self.fstat(file_obj.fileno())
1150
link_or_sha1 = self.sha_file(file_obj)
1153
self.state._observed_sha1(entry, link_or_sha1,
1155
content_change = (link_or_sha1 != source_details[1])
1157
# Size changed, so must be different
1159
# Target details is updated at update_entry time
1160
if self.use_filesystem_for_exec:
1161
# We don't need S_ISREG here, because we are sure
1162
# we are dealing with a file.
1163
target_exec = bool(S_IXUSR & path_info[3].st_mode)
1165
target_exec = target_details[3]
1166
elif target_kind == 'symlink':
1167
if source_minikind != c'l':
1170
content_change = (link_or_sha1 != source_details[1])
1172
elif target_kind == 'tree-reference':
1173
if source_minikind != c't':
1179
raise Exception, "unknown kind %s" % path_info[2]
1180
if source_minikind == c'd':
1182
old_path = path = self.pathjoin(old_dirname, old_basename)
1184
file_id = entry[0][2]
1185
self.old_dirname_to_file_id[old_path] = file_id
1186
# parent id is the entry for the path in the target tree
1187
if old_dirname == self.last_source_parent[0]:
1188
source_parent_id = self.last_source_parent[1]
1191
source_parent_id = self.old_dirname_to_file_id[old_dirname]
1193
source_parent_entry = self.state._get_entry(self.source_index,
1194
path_utf8=old_dirname)
1195
source_parent_id = source_parent_entry[0][2]
1196
if source_parent_id == entry[0][2]:
1197
# This is the root, so the parent is None
1198
source_parent_id = None
1200
self.last_source_parent[0] = old_dirname
1201
self.last_source_parent[1] = source_parent_id
1202
new_dirname = entry[0][0]
1203
if new_dirname == self.last_target_parent[0]:
1204
target_parent_id = self.last_target_parent[1]
1207
target_parent_id = self.new_dirname_to_file_id[new_dirname]
1209
# TODO: We don't always need to do the lookup, because the
1210
# parent entry will be the same as the source entry.
1211
target_parent_entry = self.state._get_entry(self.target_index,
1212
path_utf8=new_dirname)
1213
if target_parent_entry == (None, None):
1214
raise AssertionError(
1215
"Could not find target parent in wt: %s\nparent of: %s"
1216
% (new_dirname, entry))
1217
target_parent_id = target_parent_entry[0][2]
1218
if target_parent_id == entry[0][2]:
1219
# This is the root, so the parent is None
1220
target_parent_id = None
1222
self.last_target_parent[0] = new_dirname
1223
self.last_target_parent[1] = target_parent_id
1225
source_exec = source_details[3]
1226
if (self.include_unchanged
1228
or source_parent_id != target_parent_id
1229
or old_basename != entry[0][1]
1230
or source_exec != target_exec
1232
if old_path is None:
1233
path = self.pathjoin(old_dirname, old_basename)
1235
old_path_u = self.utf8_decode(old_path)[0]
1238
old_path_u = self.utf8_decode(old_path)[0]
1239
if old_path == path:
1242
path_u = self.utf8_decode(path)[0]
1243
source_kind = _minikind_to_kind(source_minikind)
1244
return (entry[0][2],
1245
(old_path_u, path_u),
1248
(source_parent_id, target_parent_id),
1249
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1250
(source_kind, target_kind),
1251
(source_exec, target_exec))
1253
return self.uninteresting
1254
elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1255
# looks like a new file
1256
path = self.pathjoin(entry[0][0], entry[0][1])
1257
# parent id is the entry for the path in the target tree
1258
# TODO: these are the same for an entire directory: cache em.
1259
parent_id = self.state._get_entry(self.target_index,
1260
path_utf8=entry[0][0])[0][2]
1261
if parent_id == entry[0][2]:
1263
if path_info is not None:
1265
if self.use_filesystem_for_exec:
1266
# We need S_ISREG here, because we aren't sure if this
1269
S_ISREG(path_info[3].st_mode)
1270
and S_IXUSR & path_info[3].st_mode)
1272
target_exec = target_details[3]
1273
return (entry[0][2],
1274
(None, self.utf8_decode(path)[0]),
1278
(None, self.utf8_decode(entry[0][1])[0]),
1279
(None, path_info[2]),
1280
(None, target_exec))
1282
# Its a missing file, report it as such.
1283
return (entry[0][2],
1284
(None, self.utf8_decode(path)[0]),
1288
(None, self.utf8_decode(entry[0][1])[0]),
1291
elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1292
# unversioned, possibly, or possibly not deleted: we dont care.
1293
# if its still on disk, *and* theres no other entry at this
1294
# path [we dont know this in this routine at the moment -
1295
# perhaps we should change this - then it would be an unknown.
1296
old_path = self.pathjoin(entry[0][0], entry[0][1])
1297
# parent id is the entry for the path in the target tree
1298
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1299
if parent_id == entry[0][2]:
1301
return (entry[0][2],
1302
(self.utf8_decode(old_path)[0], None),
1306
(self.utf8_decode(entry[0][1])[0], None),
1307
(_minikind_to_kind(source_minikind), None),
1308
(source_details[3], None))
1309
elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1310
# a rename; could be a true rename, or a rename inherited from
1311
# a renamed parent. TODO: handle this efficiently. Its not
1312
# common case to rename dirs though, so a correct but slow
1313
# implementation will do.
1314
if not osutils.is_inside_any(self.searched_specific_files, target_details[1]):
1315
self.search_specific_files.add(target_details[1])
1316
elif ((source_minikind == c'r' or source_minikind == c'a') and
1317
(target_minikind == c'r' or target_minikind == c'a')):
1318
# neither of the selected trees contain this path,
1319
# so skip over it. This is not currently directly tested, but
1320
# is indirectly via test_too_much.TestCommands.test_conflicts.
1323
raise AssertionError("don't know how to compare "
1324
"source_minikind=%r, target_minikind=%r"
1325
% (source_minikind, target_minikind))
1326
## import pdb;pdb.set_trace()
1332
def iter_changes(self):
1335
cdef void _update_current_block(self):
1336
if (self.block_index < len(self.state._dirblocks) and
1337
osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1338
self.current_block = self.state._dirblocks[self.block_index]
1339
self.current_block_list = self.current_block[1]
1340
self.current_block_pos = 0
1342
self.current_block = None
1343
self.current_block_list = None
1346
# Simple thunk to allow tail recursion without pyrex confusion
1347
return self._iter_next()
1349
cdef _iter_next(self):
1350
"""Iterate over the changes."""
1351
# This function single steps through an iterator. As such while loops
1352
# are often exited by 'return' - the code is structured so that the
1353
# next call into the function will return to the same while loop. Note
1354
# that all flow control needed to re-reach that step is reexecuted,
1355
# which can be a performance problem. It has not yet been tuned to
1356
# minimise this; a state machine is probably the simplest restructuring
1357
# to both minimise this overhead and make the code considerably more
1361
# compare source_index and target_index at or under each element of search_specific_files.
1362
# follow the following comparison table. Note that we only want to do diff operations when
1363
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1367
# Source | Target | disk | action
1368
# r | fdlt | | add source to search, add id path move and perform
1369
# | | | diff check on source-target
1370
# r | fdlt | a | dangling file that was present in the basis.
1372
# r | a | | add source to search
1374
# r | r | | this path is present in a non-examined tree, skip.
1375
# r | r | a | this path is present in a non-examined tree, skip.
1376
# a | fdlt | | add new id
1377
# a | fdlt | a | dangling locally added file, skip
1378
# a | a | | not present in either tree, skip
1379
# a | a | a | not present in any tree, skip
1380
# a | r | | not present in either tree at this path, skip as it
1381
# | | | may not be selected by the users list of paths.
1382
# a | r | a | not present in either tree at this path, skip as it
1383
# | | | may not be selected by the users list of paths.
1384
# fdlt | fdlt | | content in both: diff them
1385
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
1386
# fdlt | a | | unversioned: output deleted id for now
1387
# fdlt | a | a | unversioned and deleted: output deleted id
1388
# fdlt | r | | relocated in this tree, so add target to search.
1389
# | | | Dont diff, we will see an r,fd; pair when we reach
1390
# | | | this id at the other path.
1391
# fdlt | r | a | relocated in this tree, so add target to search.
1392
# | | | Dont diff, we will see an r,fd; pair when we reach
1393
# | | | this id at the other path.
1395
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1396
# keeping a cache of directories that we have seen.
1397
cdef object current_dirname, current_blockname
1398
cdef char * current_dirname_c, * current_blockname_c
1399
cdef int advance_entry, advance_path
1400
cdef int path_handled
1401
uninteresting = self.uninteresting
1402
searched_specific_files = self.searched_specific_files
1403
# Are we walking a root?
1404
while self.root_entries_pos < self.root_entries_len:
1405
entry = self.root_entries[self.root_entries_pos]
1406
self.root_entries_pos = self.root_entries_pos + 1
1407
result = self._process_entry(entry, self.root_dir_info)
1408
if result is not None and result is not self.uninteresting:
1410
# Have we finished the prior root, or never started one ?
1411
if self.current_root is None:
1412
# TODO: the pending list should be lexically sorted? the
1413
# interface doesn't require it.
1415
self.current_root = self.search_specific_files.pop()
1417
raise StopIteration()
1418
self.current_root_unicode = self.current_root.decode('utf8')
1419
self.searched_specific_files.add(self.current_root)
1420
# process the entries for this containing directory: the rest will be
1421
# found by their parents recursively.
1422
self.root_entries = self.state._entries_for_path(self.current_root)
1423
self.root_entries_len = len(self.root_entries)
1424
self.root_abspath = self.tree.abspath(self.current_root_unicode)
1426
root_stat = os.lstat(self.root_abspath)
1428
if e.errno == errno.ENOENT:
1429
# the path does not exist: let _process_entry know that.
1430
self.root_dir_info = None
1432
# some other random error: hand it up.
1435
self.root_dir_info = ('', self.current_root,
1436
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1438
if self.root_dir_info[2] == 'directory':
1439
if self.tree._directory_is_tree_reference(
1440
self.current_root_unicode):
1441
self.root_dir_info = self.root_dir_info[:2] + \
1442
('tree-reference',) + self.root_dir_info[3:]
1443
if not self.root_entries and not self.root_dir_info:
1444
# this specified path is not present at all, skip it.
1445
# (tail recursion, can do a loop once the full structure is
1447
return self._iter_next()
1449
self.root_entries_pos = 0
1450
# XXX Clarity: This loop is duplicated a out the self.current_root
1451
# is None guard above: if we return from it, it completes there
1452
# (and the following if block cannot trigger because
1453
# path_handled must be true, so the if block is not # duplicated.
1454
while self.root_entries_pos < self.root_entries_len:
1455
entry = self.root_entries[self.root_entries_pos]
1456
self.root_entries_pos = self.root_entries_pos + 1
1457
result = self._process_entry(entry, self.root_dir_info)
1458
if result is not None:
1460
if result is not self.uninteresting:
1462
# handle unversioned specified paths:
1463
if self.want_unversioned and not path_handled and self.root_dir_info:
1464
new_executable = bool(
1465
stat.S_ISREG(self.root_dir_info[3].st_mode)
1466
and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1468
(None, self.current_root_unicode),
1472
(None, splitpath(self.current_root_unicode)[-1]),
1473
(None, self.root_dir_info[2]),
1474
(None, new_executable)
1476
# If we reach here, the outer flow continues, which enters into the
1477
# per-root setup logic.
1478
if self.current_dir_info is None and self.current_block is None:
1479
# setup iteration of this root:
1480
self.current_dir_list = None
1481
if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1482
self.current_dir_info = None
1484
self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1485
prefix=self.current_root)
1488
self.current_dir_info = self.dir_iterator.next()
1489
self.current_dir_list = self.current_dir_info[1]
1491
# there may be directories in the inventory even though
1492
# this path is not a file on disk: so mark it as end of
1494
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1495
self.current_dir_info = None
1496
elif sys.platform == 'win32':
1497
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1498
# python 2.5 has e.errno == EINVAL,
1499
# and e.winerror == ERROR_DIRECTORY
1501
e_winerror = e.winerror
1502
except AttributeError:
1504
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1505
if (e.errno in win_errors or e_winerror in win_errors):
1506
self.current_dir_info = None
1508
# Will this really raise the right exception ?
1513
if self.current_dir_info[0][0] == '':
1514
# remove .bzr from iteration
1515
bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1516
if self.current_dir_list[bzr_index][0] != '.bzr':
1517
raise AssertionError()
1518
del self.current_dir_list[bzr_index]
1519
initial_key = (self.current_root, '', '')
1520
self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1521
if self.block_index == 0:
1522
# we have processed the total root already, but because the
1523
# initial key matched it we should skip it here.
1524
self.block_index = self.block_index + 1
1525
self._update_current_block()
1526
# walk until both the directory listing and the versioned metadata
1528
while (self.current_dir_info is not None
1529
or self.current_block is not None):
1530
# Uncommon case - a missing directory or an unversioned directory:
1531
if (self.current_dir_info and self.current_block
1532
and self.current_dir_info[0][0] != self.current_block[0]):
1533
# Work around pyrex broken heuristic - current_dirname has
1534
# the same scope as current_dirname_c
1535
current_dirname = self.current_dir_info[0][0]
1536
current_dirname_c = PyString_AS_STRING_void(
1537
<void *>current_dirname)
1538
current_blockname = self.current_block[0]
1539
current_blockname_c = PyString_AS_STRING_void(
1540
<void *>current_blockname)
1541
# In the python generator we evaluate this if block once per
1542
# dir+block; because we reenter in the pyrex version its being
1543
# evaluated once per path: we could cache the result before
1544
# doing the while loop and probably save time.
1545
if _cmp_by_dirs(current_dirname_c,
1546
PyString_Size(current_dirname),
1547
current_blockname_c,
1548
PyString_Size(current_blockname)) < 0:
1549
# filesystem data refers to paths not covered by the
1550
# dirblock. this has two possibilities:
1551
# A) it is versioned but empty, so there is no block for it
1552
# B) it is not versioned.
1554
# if (A) then we need to recurse into it to check for
1555
# new unknown files or directories.
1556
# if (B) then we should ignore it, because we don't
1557
# recurse into unknown directories.
1558
# We are doing a loop
1559
while self.path_index < len(self.current_dir_list):
1560
current_path_info = self.current_dir_list[self.path_index]
1561
# dont descend into this unversioned path if it is
1563
if current_path_info[2] in ('directory',
1565
del self.current_dir_list[self.path_index]
1566
self.path_index = self.path_index - 1
1567
self.path_index = self.path_index + 1
1568
if self.want_unversioned:
1569
if current_path_info[2] == 'directory':
1570
if self.tree._directory_is_tree_reference(
1571
self.utf8_decode(current_path_info[0])[0]):
1572
current_path_info = current_path_info[:2] + \
1573
('tree-reference',) + current_path_info[3:]
1574
new_executable = bool(
1575
stat.S_ISREG(current_path_info[3].st_mode)
1576
and stat.S_IEXEC & current_path_info[3].st_mode)
1578
(None, self.utf8_decode(current_path_info[0])[0]),
1582
(None, self.utf8_decode(current_path_info[1])[0]),
1583
(None, current_path_info[2]),
1584
(None, new_executable))
1585
# This dir info has been handled, go to the next
1587
self.current_dir_list = None
1589
self.current_dir_info = self.dir_iterator.next()
1590
self.current_dir_list = self.current_dir_info[1]
1591
except StopIteration:
1592
self.current_dir_info = None
1594
# We have a dirblock entry for this location, but there
1595
# is no filesystem path for this. This is most likely
1596
# because a directory was removed from the disk.
1597
# We don't have to report the missing directory,
1598
# because that should have already been handled, but we
1599
# need to handle all of the files that are contained
1601
while self.current_block_pos < len(self.current_block_list):
1602
current_entry = self.current_block_list[self.current_block_pos]
1603
self.current_block_pos = self.current_block_pos + 1
1604
# entry referring to file not present on disk.
1605
# advance the entry only, after processing.
1606
result = self._process_entry(current_entry, None)
1607
if result is not None:
1608
if result is not self.uninteresting:
1610
self.block_index = self.block_index + 1
1611
self._update_current_block()
1612
continue # next loop-on-block/dir
1613
result = self._loop_one_block()
1614
if result is not None:
1616
if len(self.search_specific_files):
1617
# More supplied paths to process
1618
self.current_root = None
1619
return self._iter_next()
1620
raise StopIteration()
1622
cdef object _maybe_tree_ref(self, current_path_info):
1623
if self.tree._directory_is_tree_reference(
1624
self.utf8_decode(current_path_info[0])[0]):
1625
return current_path_info[:2] + \
1626
('tree-reference',) + current_path_info[3:]
1628
return current_path_info
1630
cdef object _loop_one_block(self):
1631
# current_dir_info and current_block refer to the same directory -
1632
# this is the common case code.
1633
# Assign local variables for current path and entry:
1634
cdef object current_entry
1635
cdef object current_path_info
1636
cdef int path_handled
1639
# cdef char * temp_str
1640
# cdef Py_ssize_t temp_str_length
1641
# PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1642
# if not strncmp(temp_str, "directory", temp_str_length):
1643
if (self.current_block is not None and
1644
self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1645
current_entry = PyList_GET_ITEM(self.current_block_list,
1646
self.current_block_pos)
1648
Py_INCREF(current_entry)
1650
current_entry = None
1651
if (self.current_dir_info is not None and
1652
self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1653
current_path_info = PyList_GET_ITEM(self.current_dir_list,
1656
Py_INCREF(current_path_info)
1657
disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1659
Py_INCREF(disk_kind)
1660
if disk_kind == "directory":
1661
current_path_info = self._maybe_tree_ref(current_path_info)
1663
current_path_info = None
1664
while (current_entry is not None or current_path_info is not None):
1669
if current_entry is None:
1670
# unversioned - the check for path_handled when the path
1671
# is advanced will yield this path if needed.
1673
elif current_path_info is None:
1674
# no path is fine: the per entry code will handle it.
1675
result = self._process_entry(current_entry, current_path_info)
1676
if result is not None:
1677
if result is self.uninteresting:
1680
minikind = _minikind_from_string(
1681
current_entry[1][self.target_index][0])
1682
cmp_result = cmp(current_path_info[1], current_entry[0][1])
1683
if (cmp_result or minikind == c'a' or minikind == c'r'):
1684
# The current path on disk doesn't match the dirblock
1685
# record. Either the dirblock record is marked as
1686
# absent/renamed, or the file on disk is not present at all
1687
# in the dirblock. Either way, report about the dirblock
1688
# entry, and let other code handle the filesystem one.
1690
# Compare the basename for these files to determine
1693
# extra file on disk: pass for now, but only
1694
# increment the path, not the entry
1697
# entry referring to file not present on disk.
1698
# advance the entry only, after processing.
1699
result = self._process_entry(current_entry, None)
1700
if result is not None:
1701
if result is self.uninteresting:
1705
# paths are the same,and the dirstate entry is not
1706
# absent or renamed.
1707
result = self._process_entry(current_entry, current_path_info)
1708
if result is not None:
1710
if result is self.uninteresting:
1712
# >- loop control starts here:
1714
if advance_entry and current_entry is not None:
1715
self.current_block_pos = self.current_block_pos + 1
1716
if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1717
current_entry = self.current_block_list[self.current_block_pos]
1719
current_entry = None
1721
if advance_path and current_path_info is not None:
1722
if not path_handled:
1723
# unversioned in all regards
1724
if self.want_unversioned:
1725
new_executable = bool(
1726
stat.S_ISREG(current_path_info[3].st_mode)
1727
and stat.S_IEXEC & current_path_info[3].st_mode)
1729
relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1730
except UnicodeDecodeError:
1731
raise errors.BadFilenameEncoding(
1732
current_path_info[0], osutils._fs_enc)
1733
if result is not None:
1734
raise AssertionError(
1735
"result is not None: %r" % result)
1737
(None, relpath_unicode),
1741
(None, self.utf8_decode(current_path_info[1])[0]),
1742
(None, current_path_info[2]),
1743
(None, new_executable))
1744
# dont descend into this unversioned path if it is
1746
if current_path_info[2] in ('directory'):
1747
del self.current_dir_list[self.path_index]
1748
self.path_index = self.path_index - 1
1749
# dont descend the disk iterator into any tree
1751
if current_path_info[2] == 'tree-reference':
1752
del self.current_dir_list[self.path_index]
1753
self.path_index = self.path_index - 1
1754
self.path_index = self.path_index + 1
1755
if self.path_index < len(self.current_dir_list):
1756
current_path_info = self.current_dir_list[self.path_index]
1757
if current_path_info[2] == 'directory':
1758
current_path_info = self._maybe_tree_ref(
1761
current_path_info = None
1762
if result is not None:
1763
# Found a result on this pass, yield it
1765
if self.current_block is not None:
1766
self.block_index = self.block_index + 1
1767
self._update_current_block()
1768
if self.current_dir_info is not None:
1770
self.current_dir_list = None
1772
self.current_dir_info = self.dir_iterator.next()
1773
self.current_dir_list = self.current_dir_info[1]
1774
except StopIteration:
1775
self.current_dir_info = None
added
removed
bzrlib/_dirstate_helpers_c.pyx
