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
34
from bzrlib.osutils import sha_file
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
35
36
from bzrlib.trace import mutter, warning
36
37
from bzrlib.atomicfile import AtomicFile
41
def _fingerprint(abspath):
43
fs = os.lstat(abspath)
45
# might be missing, etc
48
if stat.S_ISDIR(fs.st_mode):
51
# we discard any high precision because it's not reliable; perhaps we
52
# could do better on some systems?
53
return (fs.st_size, long(fs.st_mtime),
54
long(fs.st_ctime), fs.st_ino, fs.st_dev)
38
from bzrlib.errors import BzrError
57
47
class HashCache(object):
115
105
self.needs_write = True
120
109
"""Scan all files and remove entries where the cache entry is obsolete.
122
111
Obsolete entries are those where the file has been modified or deleted
123
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.
125
116
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
128
119
for inum, path, cache_entry in prep:
129
abspath = os.sep.join([self.basedir, path])
130
fp = _fingerprint(abspath)
120
abspath = pathjoin(self.root, path)
121
fp = self._fingerprint(abspath)
131
122
self.stat_count += 1
133
124
cache_fp = cache_entry[1]
138
129
self.needs_write = True
139
130
del self._cache[path]
143
def get_sha1(self, path):
132
def get_sha1(self, path, stat_value=None):
144
133
"""Return the sha1 of a file.
146
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)
147
139
self.stat_count += 1
148
file_fp = _fingerprint(abspath)
140
file_fp = self._fingerprint(abspath, stat_value)
151
143
# not a regular file or not existing
161
153
cache_sha1, cache_fp = None, None
163
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())
164
158
self.hit_count += 1
165
159
return cache_sha1
167
161
self.miss_count += 1
168
digest = sha_file(file(abspath, 'rb', buffering=65000))
170
now = int(time.time())
171
if file_fp[1] >= now or file_fp[2] >= now:
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:
172
176
# changed too recently; can't be cached. we can
173
177
# return the result and it could possibly be cached
175
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
177
190
self.removed_count += 1
178
191
self.needs_write = True
179
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])
181
197
self.update_count += 1
182
198
self.needs_write = True
183
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))
191
207
"""Write contents of cache to file."""
192
outf = AtomicFile(self.cache_file_name(), 'wb')
208
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
194
print >>outf, CACHE_HEADER,
210
outf.write(CACHE_HEADER)
196
212
for path, c in self._cache.iteritems():
197
assert '//' not in path, path
198
outf.write(path.encode('utf-8'))
200
print >>outf, c[0], # hex sha1
202
print >>outf, "%d" % fld,
213
line_info = [path.encode('utf-8'), '// ', c[0], ' ']
214
line_info.append(' '.join([str(fld) for fld in c[1]]))
215
line_info.append('\n')
216
outf.write(''.join(line_info))
206
218
self.needs_write = False
219
## mutter("write hash cache: %s hits=%d misses=%d stat=%d recent=%d updates=%d",
220
## self.cache_file_name(), self.hit_count, self.miss_count,
222
## self.danger_count, self.update_count)
214
227
"""Reinstate cache from file.
259
271
self._cache[path] = (sha1, fp)
261
273
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, stat_value=None):
284
if stat_value is None:
286
stat_value = os.lstat(abspath)
288
# might be missing, etc
290
if stat.S_ISDIR(stat_value.st_mode):
292
# we discard any high precision because it's not reliable; perhaps we
293
# could do better on some systems?
294
return (stat_value.st_size, long(stat_value.st_mtime),
295
long(stat_value.st_ctime), stat_value.st_ino,
296
stat_value.st_dev, stat_value.st_mode)