1
# (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
# TODO: Up-front, stat all files in order and remove those which are deleted or
18
# out-of-date. Don't actually re-read them until they're needed. That ought
19
# to bring all the inodes into core so that future stats to them are fast, and
17
# TODO: Up-front, stat all files in order and remove those which are deleted or
18
# out-of-date. Don't actually re-read them until they're needed. That ought
19
# to bring all the inodes into core so that future stats to them are fast, and
20
20
# it preserves the nice property that any caller will always get up-to-date
21
21
# data except in unavoidable cases.
23
23
# TODO: Perhaps return more details on the file to avoid statting it
24
24
# again: nonexistent, file type, size, etc
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
28
30
CACHE_HEADER = "### bzr hashcache v5\n"
30
32
import os, stat, time
32
from bzrlib.osutils import sha_file
34
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
33
35
from bzrlib.trace import mutter, warning
37
def _fingerprint(abspath):
39
fs = os.lstat(abspath)
41
# might be missing, etc
44
if stat.S_ISDIR(fs.st_mode):
47
# we discard any high precision because it's not reliable; perhaps we
48
# could do better on some systems?
49
return (fs.st_size, long(fs.st_mtime),
50
long(fs.st_ctime), fs.st_ino, fs.st_dev)
36
from bzrlib.atomicfile import AtomicFile
37
from bzrlib.errors import BzrError
53
46
class HashCache(object):
111
104
self.needs_write = True
116
108
"""Scan all files and remove entries where the cache entry is obsolete.
118
110
Obsolete entries are those where the file has been modified or deleted
119
since the entry was inserted.
111
since the entry was inserted.
113
# FIXME optimisation opportunity, on linux [and check other oses]:
114
# rather than iteritems order, stat in inode order.
121
115
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
124
118
for inum, path, cache_entry in prep:
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
127
121
self.stat_count += 1
129
123
cache_fp = cache_entry[1]
131
125
if (not fp) or (cache_fp != fp):
132
126
# not here or not a regular file anymore
133
127
self.removed_count += 1
134
128
self.needs_write = True
135
129
del self._cache[path]
139
def get_sha1(self, path):
131
def get_sha1(self, path, stat_value=None):
140
132
"""Return the sha1 of a file.
142
abspath = os.sep.join([self.basedir, path])
134
if path.__class__ is str:
135
abspath = pathjoin(self.root_utf8, path)
137
abspath = pathjoin(self.root, path)
143
138
self.stat_count += 1
144
file_fp = _fingerprint(abspath)
139
file_fp = self._fingerprint(abspath, stat_value)
147
142
# not a regular file or not existing
148
143
if path in self._cache:
149
144
self.removed_count += 1
150
145
self.needs_write = True
151
146
del self._cache[path]
154
149
if path in self._cache:
155
150
cache_sha1, cache_fp = self._cache[path]
157
152
cache_sha1, cache_fp = None, None
159
154
if cache_fp == file_fp:
155
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
156
## mutter("now = %s", time.time())
160
157
self.hit_count += 1
161
158
return cache_sha1
163
160
self.miss_count += 1
164
digest = sha_file(file(abspath, 'rb', buffering=65000))
166
now = int(time.time())
167
if file_fp[1] >= now or file_fp[2] >= now:
162
mode = file_fp[FP_MODE_COLUMN]
163
if stat.S_ISREG(mode):
164
digest = self._really_sha1_file(abspath)
165
elif stat.S_ISLNK(mode):
166
digest = sha_string(os.readlink(abspath))
168
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
170
# window of 3 seconds to allow for 2s resolution on windows,
171
# unsynchronized file servers, etc.
172
cutoff = self._cutoff_time()
173
if file_fp[FP_MTIME_COLUMN] >= cutoff \
174
or file_fp[FP_CTIME_COLUMN] >= cutoff:
168
175
# changed too recently; can't be cached. we can
169
176
# return the result and it could possibly be cached
171
self.danger_count += 1
179
# the point is that we only want to cache when we are sure that any
180
# subsequent modifications of the file can be detected. If a
181
# modification neither changes the inode, the device, the size, nor
182
# the mode, then we can only distinguish it by time; therefore we
183
# need to let sufficient time elapse before we may cache this entry
184
# again. If we didn't do this, then, for example, a very quick 1
185
# byte replacement in the file might go undetected.
186
## mutter('%r modified too recently; not caching', path)
187
self.danger_count += 1
173
189
self.removed_count += 1
174
190
self.needs_write = True
175
191
del self._cache[path]
193
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
194
## path, time.time(), file_fp[FP_MTIME_COLUMN],
195
## file_fp[FP_CTIME_COLUMN])
177
196
self.update_count += 1
178
197
self.needs_write = True
179
198
self._cache[path] = (digest, file_fp)
201
def _really_sha1_file(self, abspath):
202
"""Calculate the SHA1 of a file by reading the full text"""
203
return sha_file(file(abspath, 'rb', buffering=65000))
187
206
"""Write contents of cache to file."""
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
192
print >>outf, CACHE_HEADER,
209
outf.write(CACHE_HEADER)
194
211
for path, c in self._cache.iteritems():
195
assert '//' not in path, path
196
outf.write(path.encode('utf-8'))
198
print >>outf, c[0], # hex sha1
200
print >>outf, "%d" % fld,
212
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
213
line_info.append(' '.join([str(fld) for fld in c[1]]))
214
line_info.append('\n')
215
outf.write(''.join(line_info))
204
217
self.needs_write = False
218
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
219
## self.cache_file_name(), self.hit_count, self.miss_count,
221
## self.danger_count, self.update_count)
212
226
"""Reinstate cache from file.
214
228
Overwrites existing cache.
216
If the cache file has the wrong version marker, this just clears
230
If the cache file has the wrong version marker, this just clears
254
270
self._cache[path] = (sha1, fp)
256
272
self.needs_write = False
274
def _cutoff_time(self):
275
"""Return cutoff time.
277
Files modified more recently than this time are at risk of being
278
undetectably modified and so can't be cached.
280
return int(time.time()) - 3
282
def _fingerprint(self, abspath, stat_value=None):
283
if stat_value is None:
285
stat_value = os.lstat(abspath)
287
# might be missing, etc
289
if stat.S_ISDIR(stat_value.st_mode):
291
# we discard any high precision because it's not reliable; perhaps we
292
# could do better on some systems?
293
return (stat_value.st_size, long(stat_value.st_mtime),
294
long(stat_value.st_ctime), stat_value.st_ino,
295
stat_value.st_dev, stat_value.st_mode)