1
# Copyright (C) 2004, 2005 by Martin Pool
2
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006 Canonical Ltd
4
3
# This program is free software; you can redistribute it and/or modify
5
4
# it under the terms of the GNU General Public License as published by
6
5
# the Free Software Foundation; either version 2 of the License, or
7
6
# (at your option) any later version.
9
8
# This program is distributed in the hope that it will be useful,
10
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
11
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12
11
# GNU General Public License for more details.
14
13
# You should have received a copy of the GNU General Public License
15
14
# along with this program; if not, write to the Free Software
16
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
def _update_store_entry(obj, obj_id, branch, store_name, store):
20
"""This is just a meta-function, which handles both revision entries
21
and inventory entries.
23
from bzrlib.trace import mutter
24
import tempfile, os, errno
25
from osutils import rename
26
obj_tmp = tempfile.TemporaryFile()
27
obj.write_xml(obj_tmp)
30
tmpfd, tmp_path = tempfile.mkstemp(prefix=obj_id, suffix='.gz',
31
dir=branch.controlfilename(store_name))
34
orig_obj_path = branch.controlfilename([store_name, obj_id+'.gz'])
35
# Remove the old entry out of the way
36
rename(orig_obj_path, tmp_path)
38
# TODO: We may need to handle the case where the old
39
# entry was not compressed (and thus did not end with .gz)
41
store.add(obj_tmp, obj_id) # Add the new one
42
os.remove(tmp_path) # Remove the old name
43
mutter(' Updated %s entry {%s}' % (store_name, obj_id))
45
# On any exception, restore the old entry
46
rename(tmp_path, orig_obj_path)
49
if os.path.exists(tmp_path):
50
# Unfortunately, the next command might throw
51
# an exception, which will mask a previous exception.
55
def _update_revision_entry(rev, branch):
56
"""After updating the values in a revision, make sure to
57
write out the data, but try to do it in an atomic manner.
59
:param rev: The Revision object to store
60
:param branch: The Branch object where this Revision is to be stored.
62
_update_store_entry(rev, rev.revision_id, branch,
63
'revision-store', branch.revision_store)
65
def _update_inventory_entry(inv, inv_id, branch):
66
"""When an inventory has been modified (such as by adding a unique tree root)
67
this atomically re-generates the file.
69
:param inv: The Inventory
70
:param inv_id: The inventory id for this inventory
71
:param branch: The Branch where this entry will be stored.
73
_update_store_entry(inv, inv_id, branch,
74
'inventory-store', branch.inventory_store)
77
"""Run consistency checks on a branch.
79
TODO: Also check non-mainline revisions mentioned as parents.
81
TODO: Check for extra files in the control directory.
83
from bzrlib.trace import mutter
84
from bzrlib.errors import BzrCheckError
85
from bzrlib.osutils import fingerprint_file
86
from bzrlib.progress import ProgressBar
87
from bzrlib.inventory import ROOT_ID
88
from bzrlib.branch import gen_root_id
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
# TODO: Check ancestries are correct for every revision: includes
18
# every committed so far, and in a reasonable order.
20
# TODO: Also check non-mainline revisions mentioned as parents.
22
# TODO: Check for extra files in the control directory.
24
# TODO: Check revision, inventory and entry objects have all
27
# TODO: Get every revision in the revision-store even if they're not
28
# referenced by history and make sure they're all valid.
30
# TODO: Perhaps have a way to record errors other than by raising exceptions;
31
# would perhaps be enough to accumulate exception objects in a list without
32
# raising them. If there's more than one exception it'd be good to see them
35
"""Checking of bzr objects.
37
check_refs is a concept used for optimising check. Objects that depend on other
38
objects (e.g. tree on repository) can list the objects they would be requesting
39
so that when the dependent object is checked, matches can be pulled out and
40
evaluated in-line rather than re-reading the same data many times.
41
check_refs are tuples (kind, value). Currently defined kinds are:
43
* 'trees', where value is a revid and the looked up objects are revision trees.
44
* 'lefthand-distance', where value is a revid and the looked up objects are the
45
distance along the lefthand path to NULL for that revid.
46
* 'revision-existence', where value is a revid, and the result is True or False
47
indicating that the revision was found/not found.
54
from bzrlib.branch import Branch
55
from bzrlib.bzrdir import BzrDir
56
from bzrlib.revision import NULL_REVISION
57
from bzrlib.trace import note
58
from bzrlib.workingtree import WorkingTree
62
"""Check a repository"""
64
def __init__(self, repository, check_repo=True):
65
self.repository = repository
67
def report_results(self, verbose):
68
raise NotImplementedError(self.report_results)
71
class VersionedFileCheck(Check):
72
"""Check a versioned file repository"""
74
# The Check object interacts with InventoryEntry.check, etc.
76
def __init__(self, repository, check_repo=True):
77
self.repository = repository
78
self.checked_rev_cnt = 0
80
self.missing_parent_links = {}
81
self.missing_inventory_sha_cnt = 0
82
self.missing_revision_cnt = 0
83
self.checked_weaves = set()
84
self.unreferenced_versions = set()
85
self.inconsistent_parents = []
86
self.rich_roots = repository.supports_rich_root()
87
self.text_key_references = {}
88
self.check_repo = check_repo
89
self.other_results = []
90
# Plain text lines to include in the report
91
self._report_items = []
92
# Keys we are looking for; may be large and need spilling to disk.
93
# key->(type(revision/inventory/text/signature/map), sha1, first-referer)
94
self.pending_keys = {}
95
# Ancestors map for all of revisions being checked; while large helper
96
# functions we call would create it anyway, so better to have once and
100
def check(self, callback_refs=None, check_repo=True):
101
if callback_refs is None:
103
self.repository.lock_read()
104
self.progress = ui.ui_factory.nested_progress_bar()
106
self.progress.update('check', 0, 4)
108
self.progress.update('checking revisions', 0)
109
self.check_revisions()
110
self.progress.update('checking commit contents', 1)
111
self.repository._check_inventories(self)
112
self.progress.update('checking file graphs', 2)
113
# check_weaves is done after the revision scan so that
114
# revision index is known to be valid.
116
self.progress.update('checking branches and trees', 3)
118
repo = self.repository
119
# calculate all refs, and callback the objects requesting them.
121
wanting_items = set()
122
# Current crude version calculates everything and calls
123
# everything at once. Doing a queue and popping as things are
124
# satisfied would be cheaper on memory [but few people have
125
# huge numbers of working trees today. TODO: fix before
129
for ref, wantlist in callback_refs.iteritems():
130
wanting_items.update(wantlist)
133
refs[ref] = repo.revision_tree(value)
134
elif kind == 'lefthand-distance':
136
elif kind == 'revision-existence':
137
existences.add(value)
139
raise AssertionError(
140
'unknown ref kind for ref %s' % ref)
141
node_distances = repo.get_graph().find_lefthand_distances(distances)
142
for key, distance in node_distances.iteritems():
143
refs[('lefthand-distance', key)] = distance
144
if key in existences and distance > 0:
145
refs[('revision-existence', key)] = True
146
existences.remove(key)
147
parent_map = repo.get_graph().get_parent_map(existences)
148
for key in parent_map:
149
refs[('revision-existence', key)] = True
150
existences.remove(key)
151
for key in existences:
152
refs[('revision-existence', key)] = False
153
for item in wanting_items:
154
if isinstance(item, WorkingTree):
156
if isinstance(item, Branch):
157
self.other_results.append(item.check(refs))
159
self.progress.finished()
160
self.repository.unlock()
162
def _check_revisions(self, revisions_iterator):
163
"""Check revision objects by decorating a generator.
165
:param revisions_iterator: An iterator of(revid, Revision-or-None).
166
:return: A generator of the contents of revisions_iterator.
168
self.planned_revisions = set()
169
for revid, revision in revisions_iterator:
170
yield revid, revision
171
self._check_one_rev(revid, revision)
172
# Flatten the revisions we found to guarantee consistent later
174
self.planned_revisions = list(self.planned_revisions)
175
# TODO: extract digital signatures as items to callback on too.
177
def check_revisions(self):
178
"""Scan revisions, checking data directly available as we go."""
179
revision_iterator = self.repository._iter_revisions(None)
180
revision_iterator = self._check_revisions(revision_iterator)
181
# We read the all revisions here:
182
# - doing this allows later code to depend on the revision index.
183
# - we can fill out existence flags at this point
184
# - we can read the revision inventory sha at this point
185
# - we can check properties and serialisers etc.
186
if not self.repository._format.revision_graph_can_have_wrong_parents:
187
# The check against the index isn't needed.
188
self.revs_with_bad_parents_in_index = None
189
for thing in revision_iterator:
192
bad_revisions = self.repository._find_inconsistent_revision_parents(
194
self.revs_with_bad_parents_in_index = list(bad_revisions)
196
def report_results(self, verbose):
198
self._report_repo_results(verbose)
199
for result in self.other_results:
200
result.report_results(verbose)
202
def _report_repo_results(self, verbose):
203
note('checked repository %s format %s',
204
self.repository.user_url,
205
self.repository._format)
206
note('%6d revisions', self.checked_rev_cnt)
207
note('%6d file-ids', len(self.checked_weaves))
209
note('%6d unreferenced text versions',
210
len(self.unreferenced_versions))
211
if verbose and len(self.unreferenced_versions):
212
for file_id, revision_id in self.unreferenced_versions:
213
note('unreferenced version: {%s} in %s', revision_id,
215
if self.missing_inventory_sha_cnt:
216
note('%6d revisions are missing inventory_sha1',
217
self.missing_inventory_sha_cnt)
218
if self.missing_revision_cnt:
219
note('%6d revisions are mentioned but not present',
220
self.missing_revision_cnt)
222
note('%6d ghost revisions', len(self.ghosts))
224
for ghost in self.ghosts:
226
if len(self.missing_parent_links):
227
note('%6d revisions missing parents in ancestry',
228
len(self.missing_parent_links))
230
for link, linkers in self.missing_parent_links.items():
231
note(' %s should be in the ancestry for:', link)
232
for linker in linkers:
233
note(' * %s', linker)
234
if len(self.inconsistent_parents):
235
note('%6d inconsistent parents', len(self.inconsistent_parents))
237
for info in self.inconsistent_parents:
238
revision_id, file_id, found_parents, correct_parents = info
239
note(' * %s version %s has parents %r '
241
% (file_id, revision_id, found_parents,
243
if self.revs_with_bad_parents_in_index:
244
note('%6d revisions have incorrect parents in the revision index',
245
len(self.revs_with_bad_parents_in_index))
247
for item in self.revs_with_bad_parents_in_index:
248
revision_id, index_parents, actual_parents = item
250
' %s has wrong parents in index: '
252
revision_id, index_parents, actual_parents)
253
for item in self._report_items:
256
def _check_one_rev(self, rev_id, rev):
257
"""Cross-check one revision.
259
:param rev_id: A revision id to check.
260
:param rev: A revision or None to indicate a missing revision.
262
if rev.revision_id != rev_id:
263
self._report_items.append(
264
'Mismatched internal revid {%s} and index revid {%s}' % (
265
rev.revision_id, rev_id))
266
rev_id = rev.revision_id
267
# Check this revision tree etc, and count as seen when we encounter a
269
self.planned_revisions.add(rev_id)
271
self.ghosts.discard(rev_id)
272
# Count all parents as ghosts if we haven't seen them yet.
273
for parent in rev.parent_ids:
274
if not parent in self.planned_revisions:
275
self.ghosts.add(parent)
277
self.ancestors[rev_id] = tuple(rev.parent_ids) or (NULL_REVISION,)
278
self.add_pending_item(rev_id, ('inventories', rev_id), 'inventory',
280
self.checked_rev_cnt += 1
282
def add_pending_item(self, referer, key, kind, sha1):
283
"""Add a reference to a sha1 to be cross checked against a key.
285
:param referer: The referer that expects key to have sha1.
286
:param key: A storage key e.g. ('texts', 'foo@bar-20040504-1234')
287
:param kind: revision/inventory/text/map/signature
288
:param sha1: A hex sha1 or None if no sha1 is known.
290
existing = self.pending_keys.get(key)
292
if sha1 != existing[1]:
293
self._report_items.append('Multiple expected sha1s for %s. {%s}'
294
' expects {%s}, {%s} expects {%s}', (
295
key, referer, sha1, existing[1], existing[0]))
297
self.pending_keys[key] = (kind, sha1, referer)
299
def check_weaves(self):
300
"""Check all the weaves we can get our hands on.
303
storebar = ui.ui_factory.nested_progress_bar()
305
self._check_weaves(storebar)
309
def _check_weaves(self, storebar):
310
storebar.update('text-index', 0, 2)
311
if self.repository._format.fast_deltas:
312
# We haven't considered every fileid instance so far.
313
weave_checker = self.repository._get_versioned_file_checker(
314
ancestors=self.ancestors)
316
weave_checker = self.repository._get_versioned_file_checker(
317
text_key_references=self.text_key_references,
318
ancestors=self.ancestors)
319
storebar.update('file-graph', 1)
320
result = weave_checker.check_file_version_parents(
321
self.repository.texts)
322
self.checked_weaves = weave_checker.file_ids
323
bad_parents, unused_versions = result
324
bad_parents = bad_parents.items()
325
for text_key, (stored_parents, correct_parents) in bad_parents:
326
# XXX not ready for id join/split operations.
327
weave_id = text_key[0]
328
revision_id = text_key[-1]
329
weave_parents = tuple([parent[-1] for parent in stored_parents])
330
correct_parents = tuple([parent[-1] for parent in correct_parents])
331
self.inconsistent_parents.append(
332
(revision_id, weave_id, weave_parents, correct_parents))
333
self.unreferenced_versions.update(unused_versions)
335
def _add_entry_to_text_key_references(self, inv, entry):
336
if not self.rich_roots and entry.name == '':
338
key = (entry.file_id, entry.revision)
339
self.text_key_references.setdefault(key, False)
340
if entry.revision == inv.revision_id:
341
self.text_key_references[key] = True
344
def scan_branch(branch, needed_refs, to_unlock):
345
"""Scan a branch for refs.
347
:param branch: The branch to schedule for checking.
348
:param needed_refs: Refs we are accumulating.
349
:param to_unlock: The unlock list accumulating.
351
note("Checking branch at '%s'." % (branch.base,))
90
352
branch.lock_read()
93
pb = ProgressBar(show_spinner=True)
96
missing_inventory_sha_cnt = 0
97
missing_revision_sha_cnt = 0
99
history = branch.revision_history()
101
revcount = len(history)
104
# for all texts checked, text_id -> sha1
107
for rev_id in history:
109
pb.update('checking revision', revno, revcount)
110
mutter(' revision {%s}' % rev_id)
111
rev = branch.get_revision(rev_id)
112
if rev.revision_id != rev_id:
113
raise BzrCheckError('wrong internal revision id in revision {%s}'
116
# check the previous history entry is a parent of this entry
118
if last_rev_id is None:
119
raise BzrCheckError("revision {%s} has %d parents, but is the "
120
"start of the branch"
121
% (rev_id, len(rev.parents)))
122
for prr in rev.parents:
123
if prr.revision_id == last_rev_id:
126
raise BzrCheckError("previous revision {%s} not listed among "
128
% (last_rev_id, rev_id))
130
for prr in rev.parents:
131
if prr.revision_sha1 is None:
132
missing_revision_sha_cnt += 1
134
prid = prr.revision_id
135
actual_sha = branch.get_revision_sha1(prid)
136
if prr.revision_sha1 != actual_sha:
137
raise BzrCheckError("mismatched revision sha1 for "
138
"parent {%s} of {%s}: %s vs %s"
140
prr.revision_sha1, actual_sha))
142
raise BzrCheckError("revision {%s} has no parents listed but preceded "
144
% (rev_id, last_rev_id))
146
if rev.inventory_id != rev_id:
147
mismatch_inv_id.append(rev_id)
149
## TODO: Check all the required fields are present on the revision.
151
if rev.inventory_sha1:
152
inv_sha1 = branch.get_inventory_sha1(rev.inventory_id)
153
if inv_sha1 != rev.inventory_sha1:
154
raise BzrCheckError('Inventory sha1 hash doesn\'t match'
155
' value in revision {%s}' % rev_id)
157
missing_inventory_sha_cnt += 1
158
mutter("no inventory_sha1 on revision {%s}" % rev_id)
160
inv = branch.get_inventory(rev.inventory_id)
164
## p('revision %d/%d file ids' % (revno, revcount))
166
if file_id in seen_ids:
167
raise BzrCheckError('duplicated file_id {%s} '
168
'in inventory for revision {%s}'
170
seen_ids[file_id] = True
180
if ie.parent_id != None:
181
if ie.parent_id not in seen_ids:
182
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
183
% (ie.parent_id, rev_id))
185
if ie.kind == 'file':
186
if ie.text_id in checked_texts:
187
fp = checked_texts[ie.text_id]
189
if not ie.text_id in branch.text_store:
190
raise BzrCheckError('text {%s} not in text_store' % ie.text_id)
192
tf = branch.text_store[ie.text_id]
193
fp = fingerprint_file(tf)
194
checked_texts[ie.text_id] = fp
196
if ie.text_size != fp['size']:
197
raise BzrCheckError('text {%s} wrong size' % ie.text_id)
198
if ie.text_sha1 != fp['sha1']:
199
raise BzrCheckError('text {%s} wrong sha1' % ie.text_id)
200
elif ie.kind == 'directory':
201
if ie.text_sha1 != None or ie.text_size != None or ie.text_id != None:
202
raise BzrCheckError('directory {%s} has text in revision {%s}'
206
for path, ie in inv.iter_entries():
207
if path in seen_names:
208
raise BzrCheckError('duplicated path %s '
209
'in inventory for revision {%s}'
211
seen_names[path] = True
353
to_unlock.append(branch)
354
branch_refs = branch._get_check_refs()
355
for ref in branch_refs:
356
reflist = needed_refs.setdefault(ref, [])
357
reflist.append(branch)
360
def scan_tree(base_tree, tree, needed_refs, to_unlock):
361
"""Scan a tree for refs.
363
:param base_tree: The original tree check opened, used to detect duplicate
365
:param tree: The tree to schedule for checking.
366
:param needed_refs: Refs we are accumulating.
367
:param to_unlock: The unlock list accumulating.
369
if base_tree is not None and tree.basedir == base_tree.basedir:
371
note("Checking working tree at '%s'." % (tree.basedir,))
373
to_unlock.append(tree)
374
tree_refs = tree._get_check_refs()
375
for ref in tree_refs:
376
reflist = needed_refs.setdefault(ref, [])
380
def check_dwim(path, verbose, do_branch=False, do_repo=False, do_tree=False):
381
"""Check multiple objects.
383
If errors occur they are accumulated and reported as far as possible, and
384
an exception raised at the end of the process.
387
base_tree, branch, repo, relpath = \
388
BzrDir.open_containing_tree_branch_or_repository(path)
389
except errors.NotBranchError:
390
base_tree = branch = repo = None
395
if base_tree is not None:
396
# If the tree is a lightweight checkout we won't see it in
397
# repo.find_branches - add now.
399
scan_tree(None, base_tree, needed_refs, to_unlock)
400
branch = base_tree.branch
401
if branch is not None:
404
# The branch is in a shared repository
405
repo = branch.repository
408
to_unlock.append(repo)
409
branches = repo.find_branches(using=True)
411
if do_branch or do_tree:
412
for branch in branches:
415
tree = branch.bzrdir.open_workingtree()
417
except (errors.NotLocalUrl, errors.NoWorkingTree):
420
scan_tree(base_tree, tree, needed_refs, to_unlock)
422
scan_branch(branch, needed_refs, to_unlock)
423
if do_branch and not branches:
424
note("No branch found at specified location.")
425
if do_tree and base_tree is None and not saw_tree:
426
note("No working tree found at specified location.")
427
if do_repo or do_branch or do_tree:
429
note("Checking repository at '%s'."
431
result = repo.check(None, callback_refs=needed_refs,
433
result.report_results(verbose)
436
note("No working tree found at specified location.")
438
note("No branch found at specified location.")
440
note("No repository found at specified location.")
219
print 'checked %d revisions, %d file texts' % (revcount, len(checked_texts))
221
if missing_inventory_sha_cnt:
222
print '%d revisions are missing inventory_sha1' % missing_inventory_sha_cnt
224
if missing_revision_sha_cnt:
225
print '%d parent links are missing revision_sha1' % missing_revision_sha_cnt
227
if (missing_inventory_sha_cnt
228
or missing_revision_sha_cnt):
229
print ' (use "bzr upgrade" to fix them)'
232
print '%d revisions have mismatched inventory ids:' % len(mismatch_inv_id)
233
for rev_id in mismatch_inv_id:
442
for thing in to_unlock:
added
removed
bzrlib/check.py
