1
# (C) 2005 Canonical Ltd
1
# Copyright (C) 2005-2010 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
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 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# out-of-date.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# out-of-date.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
28
CACHE_HEADER = "### bzr hashcache v5\n"
32
from bzrlib.osutils import sha_file
33
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)
26
# TODO: Perhaps use a Python pickle instead of a text file; might be faster.
29
CACHE_HEADER = b"### bzr hashcache v5\n"
38
filters as _mod_filters,
53
49
class HashCache(object):
111
116
self.needs_write = True
116
120
"""Scan all files and remove entries where the cache entry is obsolete.
118
122
Obsolete entries are those where the file has been modified or deleted
119
since the entry was inserted.
123
since the entry was inserted.
121
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
124
for inum, path, cache_entry in prep:
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
125
# Stat in inode order as optimisation for at least linux.
126
def inode_order(path_and_cache):
127
return path_and_cache[1][1][3]
128
for path, cache_val in sorted(self._cache.items(), key=inode_order):
129
abspath = osutils.pathjoin(self.root, path)
130
fp = self._fingerprint(abspath)
127
131
self.stat_count += 1
129
cache_fp = cache_entry[1]
131
if (not fp) or (cache_fp != fp):
133
if not fp or cache_val[1] != fp:
132
134
# not here or not a regular file anymore
133
135
self.removed_count += 1
134
136
self.needs_write = True
135
137
del self._cache[path]
139
def get_sha1(self, path):
139
def get_sha1(self, path, stat_value=None):
140
140
"""Return the sha1 of a file.
142
abspath = os.sep.join([self.basedir, path])
142
abspath = osutils.pathjoin(self.root, path)
143
143
self.stat_count += 1
144
file_fp = _fingerprint(abspath)
144
file_fp = self._fingerprint(abspath, stat_value)
147
147
# not a regular file or not existing
148
148
if path in self._cache:
149
149
self.removed_count += 1
150
150
self.needs_write = True
151
151
del self._cache[path]
154
154
if path in self._cache:
155
155
cache_sha1, cache_fp = self._cache[path]
159
159
if cache_fp == file_fp:
160
160
self.hit_count += 1
161
161
return cache_sha1
163
163
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:
165
mode = file_fp[FP_MODE_COLUMN]
166
if stat.S_ISREG(mode):
167
if self._filter_provider is None:
170
filters = self._filter_provider(path=path)
171
digest = self._really_sha1_file(abspath, filters)
172
elif stat.S_ISLNK(mode):
173
target = osutils.readlink(abspath)
174
digest = osutils.sha_string(target.encode('UTF-8'))
176
raise errors.BzrError("file %r: unknown file stat mode: %o"
179
# window of 3 seconds to allow for 2s resolution on windows,
180
# unsynchronized file servers, etc.
181
cutoff = self._cutoff_time()
182
if file_fp[FP_MTIME_COLUMN] >= cutoff \
183
or file_fp[FP_CTIME_COLUMN] >= cutoff:
168
184
# changed too recently; can't be cached. we can
169
185
# return the result and it could possibly be cached
171
self.danger_count += 1
188
# the point is that we only want to cache when we are sure that any
189
# subsequent modifications of the file can be detected. If a
190
# modification neither changes the inode, the device, the size, nor
191
# the mode, then we can only distinguish it by time; therefore we
192
# need to let sufficient time elapse before we may cache this entry
193
# again. If we didn't do this, then, for example, a very quick 1
194
# byte replacement in the file might go undetected.
195
## mutter('%r modified too recently; not caching', path)
196
self.danger_count += 1
173
198
self.removed_count += 1
174
199
self.needs_write = True
175
200
del self._cache[path]
202
# mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
203
## path, time.time(), file_fp[FP_MTIME_COLUMN],
204
# file_fp[FP_CTIME_COLUMN])
177
205
self.update_count += 1
178
206
self.needs_write = True
179
207
self._cache[path] = (digest, file_fp)
210
def _really_sha1_file(self, abspath, filters):
211
"""Calculate the SHA1 of a file by reading the full text"""
212
return _mod_filters.internal_size_sha_file_byname(abspath, filters)[1]
187
215
"""Write contents of cache to file."""
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
192
print >>outf, CACHE_HEADER,
194
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,
216
with atomicfile.AtomicFile(self.cache_file_name(), 'wb',
217
new_mode=self._mode) as outf:
218
outf.write(CACHE_HEADER)
220
for path, c in self._cache.items():
221
line_info = [path.encode('utf-8'), b'// ', c[0], b' ']
222
line_info.append(b'%d %d %d %d %d %d' % c[1])
223
line_info.append(b'\n')
224
outf.write(b''.join(line_info))
204
225
self.needs_write = False
226
# mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
227
# self.cache_file_name(), self.hit_count, self.miss_count,
229
# self.danger_count, self.update_count)
212
232
"""Reinstate cache from file.
214
234
Overwrites existing cache.
216
If the cache file has the wrong version marker, this just clears
236
If the cache file has the wrong version marker, this just clears
220
240
fn = self.cache_file_name()
222
inf = file(fn, 'rb', buffering=65000)
224
mutter("failed to open %s: %s" % (fn, e))
242
inf = open(fn, 'rb', buffering=65000)
244
trace.mutter("failed to open %s: %s", fn, str(e))
225
245
# better write it now so it is valid
226
246
self.needs_write = True
231
if hdr != CACHE_HEADER:
232
mutter('cache header marker not found at top of %s; discarding cache'
234
self.needs_write = True
239
path = l[:pos].decode('utf-8')
240
if path in self._cache:
241
warning('duplicated path %r in cache' % path)
245
fields = l[pos:].split(' ')
247
warning("bad line in hashcache: %r" % l)
252
warning("bad sha1 in hashcache: %r" % sha1)
255
fp = tuple(map(long, fields[1:]))
257
self._cache[path] = (sha1, fp)
251
if hdr != CACHE_HEADER:
252
trace.mutter('cache header marker not found at top of %s;'
253
' discarding cache', fn)
254
self.needs_write = True
258
pos = l.index(b'// ')
259
path = l[:pos].decode('utf-8')
260
if path in self._cache:
261
trace.warning('duplicated path %r in cache' % path)
265
fields = l[pos:].split(b' ')
267
trace.warning("bad line in hashcache: %r" % l)
272
trace.warning("bad sha1 in hashcache: %r" % sha1)
275
fp = tuple(map(int, fields[1:]))
277
self._cache[path] = (sha1, fp)
259
279
self.needs_write = False
281
def _cutoff_time(self):
282
"""Return cutoff time.
284
Files modified more recently than this time are at risk of being
285
undetectably modified and so can't be cached.
287
return int(time.time()) - 3
289
def _fingerprint(self, abspath, stat_value=None):
290
if stat_value is None:
292
stat_value = os.lstat(abspath)
294
# might be missing, etc
296
if stat.S_ISDIR(stat_value.st_mode):
298
# we discard any high precision because it's not reliable; perhaps we
299
# could do better on some systems?
300
return (stat_value.st_size, int(stat_value.st_mtime),
301
int(stat_value.st_ctime), stat_value.st_ino,
302
stat_value.st_dev, stat_value.st_mode)