1
# Copyright (C) 2005, 2006 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
114
self.needs_write = True
108
119
"""Scan all files and remove entries where the cache entry is obsolete.
110
121
Obsolete entries are those where the file has been modified or deleted
111
122
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
124
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
118
127
for inum, path, cache_entry in prep:
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
128
abspath = os.sep.join([self.basedir, path])
129
fp = _fingerprint(abspath)
121
130
self.stat_count += 1
123
132
cache_fp = cache_entry[1]
149
159
cache_sha1, cache_fp = None, None
151
161
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
162
self.hit_count += 1
155
163
return cache_sha1
157
165
self.miss_count += 1
159
168
mode = file_fp[FP_MODE_COLUMN]
160
169
if stat.S_ISREG(mode):
161
digest = self._really_sha1_file(abspath)
170
digest = sha_file(file(abspath, 'rb', buffering=65000))
162
171
elif stat.S_ISLNK(mode):
172
link_target = os.readlink(abspath)
163
173
digest = sha.new(os.readlink(abspath)).hexdigest()
165
175
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:
177
now = int(time.time())
178
if file_fp[1] >= now or file_fp[2] >= now:
172
179
# changed too recently; can't be cached. we can
173
180
# 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
182
self.danger_count += 1
186
184
self.removed_count += 1
187
185
self.needs_write = True
188
186
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
188
self.update_count += 1
194
189
self.needs_write = True
195
190
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
194
"""Write contents of cache to file."""
204
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
195
outf = AtomicFile(self.cache_file_name(), 'wb')
206
outf.write(CACHE_HEADER)
197
print >>outf, CACHE_HEADER,
208
199
for path, c in self._cache.iteritems():
209
200
assert '//' not in path, path
210
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
211
line_info.append(' '.join([str(fld) for fld in c[1]]))
212
line_info.append('\n')
213
outf.write(''.join(line_info))
201
outf.write(path.encode('utf-8'))
203
print >>outf, c[0], # hex sha1
205
print >>outf, "%d" % fld,
215
209
self.needs_write = False
216
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
217
## self.cache_file_name(), self.hit_count, self.miss_count,
219
## self.danger_count, self.update_count)
224
215
"""Reinstate cache from file.
268
260
self._cache[path] = (sha1, fp)
270
262
self.needs_write = False
272
def _cutoff_time(self):
273
"""Return cutoff time.
275
Files modified more recently than this time are at risk of being
276
undetectably modified and so can't be cached.
278
return int(time.time()) - 3
280
def _fingerprint(self, abspath, stat_value=None):
281
if stat_value is None:
283
stat_value = os.lstat(abspath)
285
# might be missing, etc
287
if stat.S_ISDIR(stat_value.st_mode):
289
# we discard any high precision because it's not reliable; perhaps we
290
# could do better on some systems?
291
return (stat_value.st_size, long(stat_value.st_mtime),
292
long(stat_value.st_ctime), stat_value.st_ino,
293
stat_value.st_dev, stat_value.st_mode)
added
removed
bzrlib/hashcache.py
