1
# (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006 by 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.
28
30
CACHE_HEADER = "### bzr hashcache v5\n"
30
32
import os, stat, time
32
from bzrlib.osutils import sha_file
35
from bzrlib.osutils import sha_file, pathjoin, safe_unicode
33
36
from bzrlib.trace import mutter, warning
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)
37
from bzrlib.atomicfile import AtomicFile
38
from bzrlib.errors import BzrError
53
47
class HashCache(object):
111
104
self.needs_write = True
116
108
"""Scan all files and remove entries where the cache entry is obsolete.
118
110
Obsolete entries are those where the file has been modified or deleted
119
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.
121
115
prep = [(ce[1][3], path, ce) for (path, ce) in self._cache.iteritems()]
124
118
for inum, path, cache_entry in prep:
125
abspath = os.sep.join([self.basedir, path])
126
fp = _fingerprint(abspath)
119
abspath = pathjoin(self.root, path)
120
fp = self._fingerprint(abspath)
127
121
self.stat_count += 1
129
123
cache_fp = cache_entry[1]
157
149
cache_sha1, cache_fp = None, None
159
151
if cache_fp == file_fp:
152
## mutter("hashcache hit for %s %r -> %s", path, file_fp, cache_sha1)
153
## mutter("now = %s", time.time())
160
154
self.hit_count += 1
161
155
return cache_sha1
163
157
self.miss_count += 1
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:
159
mode = file_fp[FP_MODE_COLUMN]
160
if stat.S_ISREG(mode):
161
digest = self._really_sha1_file(abspath)
162
elif stat.S_ISLNK(mode):
163
digest = sha.new(os.readlink(abspath)).hexdigest()
165
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:
168
172
# changed too recently; can't be cached. we can
169
173
# return the result and it could possibly be cached
171
self.danger_count += 1
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
173
186
self.removed_count += 1
174
187
self.needs_write = True
175
188
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])
177
193
self.update_count += 1
178
194
self.needs_write = True
179
195
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))
187
203
"""Write contents of cache to file."""
188
from atomicfile import AtomicFile
190
outf = AtomicFile(self.cache_file_name(), 'wb')
204
outf = AtomicFile(self.cache_file_name(), 'wb', new_mode=self._mode)
192
print >>outf, CACHE_HEADER,
206
outf.write(CACHE_HEADER)
194
208
for path, c in self._cache.iteritems():
195
209
assert '//' not in path, path
196
outf.write(path.encode('utf-8'))
198
print >>outf, c[0], # hex sha1
200
print >>outf, "%d" % fld,
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))
204
215
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)
212
224
"""Reinstate cache from file.
254
268
self._cache[path] = (sha1, fp)
256
270
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):
282
fs = os.lstat(abspath)
284
# might be missing, etc
286
if stat.S_ISDIR(fs.st_mode):
288
# we discard any high precision because it's not reliable; perhaps we
289
# could do better on some systems?
290
return (fs.st_size, long(fs.st_mtime),
291
long(fs.st_ctime), fs.st_ino, fs.st_dev, fs.st_mode)