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
32
32
import os, stat, time
35
from bzrlib.osutils import sha_file
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
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)
58
47
class HashCache(object):
114
105
self.needs_write = True
119
109
"""Scan all files and remove entries where the cache entry is obsolete.
121
111
Obsolete entries are those where the file has been modified or deleted
122
112
since the entry was inserted.
114
# FIXME optimisation opportunity, on linux [and check other oses]:
115
# rather than iteritems order, stat in inode order.
124
116
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
127
119
for inum, path, cache_entry in prep:
128
abspath = os.sep.join([self.basedir, path])
129
fp = _fingerprint(abspath)
120
abspath = pathjoin(self.root, path)
121
fp = self._fingerprint(abspath)
130
122
self.stat_count += 1
132
124
cache_fp = cache_entry[1]
137
129
self.needs_write = True
138
130
del self._cache[path]
141
def get_sha1(self, path):
132
def get_sha1(self, path, stat_value=None):
142
133
"""Return the sha1 of a file.
144
abspath = os.sep.join([self.basedir, path])
135
if path.__class__ is str:
136
abspath = pathjoin(self.root_utf8, path)
138
abspath = pathjoin(self.root, path)
145
139
self.stat_count += 1
146
file_fp = _fingerprint(abspath)
140
file_fp = self._fingerprint(abspath, stat_value)
149
143
# not a regular file or not existing
159
153
cache_sha1, cache_fp = None, None
161
155
if cache_fp == file_fp:
156
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
157
## mutter("now = %s", time.time())
162
158
self.hit_count += 1
163
159
return cache_sha1
165
161
self.miss_count += 1
168
163
mode = file_fp[FP_MODE_COLUMN]
169
164
if stat.S_ISREG(mode):
170
digest = sha_file(file(abspath, 'rb', buffering=65000))
165
digest = self._really_sha1_file(abspath)
171
166
elif stat.S_ISLNK(mode):
172
link_target = os.readlink(abspath)
173
167
digest = sha.new(os.readlink(abspath)).hexdigest()
175
169
raise BzrError("file %r: unknown file stat mode: %o"%(abspath,mode))
177
now = int(time.time())
178
if file_fp[1] >= now or file_fp[2] >= now:
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:
179
176
# changed too recently; can't be cached. we can
180
177
# return the result and it could possibly be cached
182
self.danger_count += 1
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
184
190
self.removed_count += 1
185
191
self.needs_write = True
186
192
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])
188
197
self.update_count += 1
189
198
self.needs_write = True
190
199
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))
194
207
"""Write contents of cache to file."""
195
outf = AtomicFile(self.cache_file_name(), 'wb')
208
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
197
print >>outf, CACHE_HEADER,
210
outf.write(CACHE_HEADER)
199
212
for path, c in self._cache.iteritems():
200
213
assert '//' not in path, path
201
outf.write(path.encode('utf-8'))
203
print >>outf, c[0], # hex sha1
205
print >>outf, "%d" % fld,
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))
209
219
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)
215
228
"""Reinstate cache from file.
260
272
self._cache[path] = (sha1, fp)
262
274
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)