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
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.
30
28
CACHE_HEADER = "### bzr hashcache v5\n"
32
30
import os, stat, time
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
32
from bzrlib.osutils import sha_file
36
33
from bzrlib.trace import mutter, warning
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
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)
47
53
class HashCache(object):
105
111
self.needs_write = True
109
116
"""Scan all files and remove entries where the cache entry is obsolete.
111
118
Obsolete entries are those where the file has been modified or deleted
112
119
since the entry was inserted.
114
# FIXME optimisation opportunity, on linux [and check other oses]:
115
# rather than iteritems order, stat in inode order.
116
121
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
119
124
for inum, path, cache_entry in prep:
120
abspath = pathjoin(self.root, path)
121
fp = self._fingerprint(abspath)
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
122
127
self.stat_count += 1
124
129
cache_fp = cache_entry[1]
129
134
self.needs_write = True
130
135
del self._cache[path]
132
def get_sha1(self, path, stat_value=None):
139
def get_sha1(self, path):
133
140
"""Return the sha1 of a file.
135
if path.__class__ is str:
136
abspath = pathjoin(self.root_utf8, path)
138
abspath = pathjoin(self.root, path)
142
abspath = os.sep.join([self.basedir, path])
139
143
self.stat_count += 1
140
file_fp = self._fingerprint(abspath, stat_value)
144
file_fp = _fingerprint(abspath)
143
147
# not a regular file or not existing
153
157
cache_sha1, cache_fp = None, None
155
159
if cache_fp == file_fp:
156
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
157
## mutter("now = %s", time.time())
158
160
self.hit_count += 1
159
161
return cache_sha1
161
163
self.miss_count += 1
163
mode = file_fp[FP_MODE_COLUMN]
164
if stat.S_ISREG(mode):
165
digest = self._really_sha1_file(abspath)
166
elif stat.S_ISLNK(mode):
167
digest = sha.new(os.readlink(abspath)).hexdigest()
169
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
171
# window of 3 seconds to allow for 2s resolution on windows,
172
# unsynchronized file servers, etc.
173
cutoff = self._cutoff_time()
174
if file_fp[FP_MTIME_COLUMN] >= cutoff \
175
or file_fp[FP_CTIME_COLUMN] >= cutoff:
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:
176
168
# changed too recently; can't be cached. we can
177
169
# return the result and it could possibly be cached
180
# the point is that we only want to cache when we are sure that any
181
# subsequent modifications of the file can be detected. If a
182
# modification neither changes the inode, the device, the size, nor
183
# the mode, then we can only distinguish it by time; therefore we
184
# need to let sufficient time elapse before we may cache this entry
185
# again. If we didn't do this, then, for example, a very quick 1
186
# byte replacement in the file might go undetected.
187
## mutter('%r modified too recently; not caching', path)
188
self.danger_count += 1
171
self.danger_count += 1
190
173
self.removed_count += 1
191
174
self.needs_write = True
192
175
del self._cache[path]
194
## mutter('%r added to cache: now=%f, mtime=%d, ctime=%d',
195
## path, time.time(), file_fp[FP_MTIME_COLUMN],
196
## file_fp[FP_CTIME_COLUMN])
197
177
self.update_count += 1
198
178
self.needs_write = True
199
179
self._cache[path] = (digest, file_fp)
202
def _really_sha1_file(self, abspath):
203
"""Calculate the SHA1 of a file by reading the full text"""
204
return sha_file(file(abspath, 'rb', buffering=65000))
207
187
"""Write contents of cache to file."""
208
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
210
outf.write(CACHE_HEADER)
192
print >>outf, CACHE_HEADER,
212
194
for path, c in self._cache.iteritems():
213
195
assert '//' not in path, path
214
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
215
line_info.append(' '.join([str(fld) for fld in c[1]]))
216
line_info.append('\n')
217
outf.write(''.join(line_info))
196
outf.write(path.encode('utf-8'))
198
print >>outf, c[0], # hex sha1
200
print >>outf, "%d" % fld,
219
204
self.needs_write = False
220
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
221
## self.cache_file_name(), self.hit_count, self.miss_count,
223
## self.danger_count, self.update_count)
228
212
"""Reinstate cache from file.
272
254
self._cache[path] = (sha1, fp)
274
256
self.needs_write = False
276
def _cutoff_time(self):
277
"""Return cutoff time.
279
Files modified more recently than this time are at risk of being
280
undetectably modified and so can't be cached.
282
return int(time.time()) - 3
284
def _fingerprint(self, abspath, stat_value=None):
285
if stat_value is None:
287
stat_value = os.lstat(abspath)
289
# might be missing, etc
291
if stat.S_ISDIR(stat_value.st_mode):
293
# we discard any high precision because it's not reliable; perhaps we
294
# could do better on some systems?
295
return (stat_value.st_size, long(stat_value.st_mtime),
296
long(stat_value.st_ctime), stat_value.st_ino,
297
stat_value.st_dev, stat_value.st_mode)