1
# Copyright (C) 2005, 2006 by Canonical Ltd
1
# (C) 2005 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
32
32
import os, stat, time
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
35
from bzrlib.osutils import sha_file
36
36
from bzrlib.trace import mutter, warning
37
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
42
def _fingerprint(abspath):
44
fs = os.lstat(abspath)
46
# might be missing, etc
49
if stat.S_ISDIR(fs.st_mode):
52
# we discard any high precision because it's not reliable; perhaps we
53
# could do better on some systems?
54
return (fs.st_size, long(fs.st_mtime),
55
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)
47
58
class HashCache(object):
104
116
self.needs_write = True
108
121
"""Scan all files and remove entries where the cache entry is obsolete.
110
123
Obsolete entries are those where the file has been modified or deleted
111
124
since the entry was inserted.
113
# FIXME optimisation opportunity, on linux [and check other oses]:
114
# rather than iteritems order, stat in inode order.
115
126
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
118
129
for inum, path, cache_entry in prep:
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
130
abspath = os.sep.join([self.basedir, path])
131
fp = _fingerprint(abspath)
121
132
self.stat_count += 1
123
134
cache_fp = cache_entry[1]
128
139
self.needs_write = True
129
140
del self._cache[path]
131
143
def get_sha1(self, path):
132
144
"""Return the sha1 of a file.
134
abspath = pathjoin(self.root, path)
146
abspath = os.sep.join([self.basedir, path])
135
147
self.stat_count += 1
136
file_fp = self._fingerprint(abspath)
148
file_fp = _fingerprint(abspath)
139
151
# not a regular file or not existing
149
161
cache_sha1, cache_fp = None, None
151
163
if cache_fp == file_fp:
152
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
153
## mutter("now = %s", time.time())
154
164
self.hit_count += 1
155
165
return cache_sha1
157
167
self.miss_count += 1
159
170
mode = file_fp[FP_MODE_COLUMN]
160
171
if stat.S_ISREG(mode):
161
digest = self._really_sha1_file(abspath)
172
digest = sha_file(file(abspath, 'rb', buffering=65000))
162
173
elif stat.S_ISLNK(mode):
174
link_target = os.readlink(abspath)
163
175
digest = sha.new(os.readlink(abspath)).hexdigest()
165
177
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
167
# window of 3 seconds to allow for 2s resolution on windows,
168
# unsynchronized file servers, etc.
169
cutoff = self._cutoff_time()
170
if file_fp[FP_MTIME_COLUMN] >= cutoff \
171
or file_fp[FP_CTIME_COLUMN] >= cutoff:
179
now = int(time.time())
180
if file_fp[1] >= now or file_fp[2] >= now:
172
181
# changed too recently; can't be cached. we can
173
182
# return the result and it could possibly be cached
176
# the point is that we only want to cache when we are sure that any
177
# subsequent modifications of the file can be detected. If a
178
# modification neither changes the inode, the device, the size, nor
179
# the mode, then we can only distinguish it by time; therefore we
180
# need to let sufficient time elapse before we may cache this entry
181
# again. If we didn't do this, then, for example, a very quick 1
182
# byte replacement in the file might go undetected.
183
## mutter('%r modified too recently; not caching', path)
184
self.danger_count += 1
184
self.danger_count += 1
186
186
self.removed_count += 1
187
187
self.needs_write = True
188
188
del self._cache[path]
190
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
191
## path, time.time(), file_fp[FP_MTIME_COLUMN],
192
## file_fp[FP_CTIME_COLUMN])
193
190
self.update_count += 1
194
191
self.needs_write = True
195
192
self._cache[path] = (digest, file_fp)
198
def _really_sha1_file(self, abspath):
199
"""Calculate the SHA1 of a file by reading the full text"""
200
return sha_file(file(abspath, 'rb', buffering=65000))
203
196
"""Write contents of cache to file."""
204
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
197
outf = AtomicFile(self.cache_file_name(), 'wb')
206
199
print >>outf, CACHE_HEADER,
214
207
print >>outf, "%d" % fld,
217
211
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)
223
213
if not outf.closed:
227
219
"""Reinstate cache from file.
237
229
inf = file(fn, 'rb', buffering=65000)
238
230
except IOError, e:
239
mutter("failed to open %s: %s", fn, e)
231
mutter("failed to open %s: %s" % (fn, e))
240
232
# better write it now so it is valid
241
233
self.needs_write = True
244
237
hdr = inf.readline()
245
238
if hdr != CACHE_HEADER:
246
mutter('cache header marker not found at top of %s;'
247
' discarding cache', fn)
239
mutter('cache header marker not found at top of %s; discarding cache'
248
241
self.needs_write = True
271
264
self._cache[path] = (sha1, fp)
273
266
self.needs_write = False
275
def _cutoff_time(self):
276
"""Return cutoff time.
278
Files modified more recently than this time are at risk of being
279
undetectably modified and so can't be cached.
281
return int(time.time()) - 3
283
def _fingerprint(self, abspath):
285
fs = os.lstat(abspath)
287
# might be missing, etc
289
if stat.S_ISDIR(fs.st_mode):
291
# we discard any high precision because it's not reliable; perhaps we
292
# could do better on some systems?
293
return (fs.st_size, long(fs.st_mtime),
294
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)