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
30
30
CACHE_HEADER = "### bzr hashcache v5\n"
32
32
import os, stat, time
35
from bzrlib.osutils import sha_file
34
from bzrlib.osutils import sha_file, sha_string, pathjoin, safe_unicode
36
35
from bzrlib.trace import mutter, warning
37
36
from bzrlib.atomicfile import AtomicFile
37
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)
58
46
class HashCache(object):
116
104
self.needs_write = True
121
108
"""Scan all files and remove entries where the cache entry is obsolete.
123
110
Obsolete entries are those where the file has been modified or deleted
124
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.
126
115
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
129
118
for inum, path, cache_entry in prep:
130
abspath = os.sep.join([self.basedir, path])
131
fp = _fingerprint(abspath)
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
132
121
self.stat_count += 1
134
123
cache_fp = cache_entry[1]
139
128
self.needs_write = True
140
129
del self._cache[path]
143
def get_sha1(self, path):
131
def get_sha1(self, path, stat_value=None):
144
132
"""Return the sha1 of a file.
146
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)
147
138
self.stat_count += 1
148
file_fp = _fingerprint(abspath)
139
file_fp = self._fingerprint(abspath, stat_value)
151
142
# not a regular file or not existing
161
152
cache_sha1, cache_fp = None, None
163
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())
164
157
self.hit_count += 1
165
158
return cache_sha1
167
160
self.miss_count += 1
170
162
mode = file_fp[FP_MODE_COLUMN]
171
163
if stat.S_ISREG(mode):
172
digest = sha_file(file(abspath, 'rb', buffering=65000))
164
digest = self._really_sha1_file(abspath)
173
165
elif stat.S_ISLNK(mode):
174
link_target = os.readlink(abspath)
175
digest = sha.new(os.readlink(abspath)).hexdigest()
166
digest = sha_string(os.readlink(abspath))
177
168
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
179
now = int(time.time())
180
if file_fp[1] >= now or file_fp[2] >= now:
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:
181
175
# changed too recently; can't be cached. we can
182
176
# return the result and it could possibly be cached
184
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
186
189
self.removed_count += 1
187
190
self.needs_write = True
188
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])
190
196
self.update_count += 1
191
197
self.needs_write = True
192
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))
196
206
"""Write contents of cache to file."""
197
outf = AtomicFile(self.cache_file_name(), 'wb')
207
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
199
print >>outf, CACHE_HEADER,
209
outf.write(CACHE_HEADER)
201
211
for path, c in self._cache.iteritems():
202
assert '//' not in path, path
203
outf.write(path.encode('utf-8'))
205
print >>outf, c[0], # hex sha1
207
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))
211
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)
219
226
"""Reinstate cache from file.
264
270
self._cache[path] = (sha1, fp)
266
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)