32
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:
42
* 'trees', where value is a revid and the looked up objects are revision trees.
43
* 'lefthand-distance', where value is a revid and the looked up objects are the
44
distance along the lefthand path to NULL for that revid.
45
* 'revision-existence', where value is a revid, and the result is True or False
46
indicating that the revision was found/not found.
49
35
from bzrlib import errors
50
from bzrlib.branch import Branch
51
from bzrlib.bzrdir import BzrDir
52
from bzrlib.revision import NULL_REVISION
53
from bzrlib.symbol_versioning import deprecated_function, deprecated_in
54
from bzrlib.trace import note
36
from bzrlib import repository as _mod_repository
37
from bzrlib import revision
38
from bzrlib.errors import BzrCheckError
56
from bzrlib.workingtree import WorkingTree
40
from bzrlib.trace import log_error, note
58
42
class Check(object):
59
43
"""Check a repository"""
61
45
# The Check object interacts with InventoryEntry.check, etc.
63
def __init__(self, repository, check_repo=True):
47
def __init__(self, repository):
64
48
self.repository = repository
49
self.checked_text_cnt = 0
65
50
self.checked_rev_cnt = 0
52
self.repeated_text_cnt = 0
67
53
self.missing_parent_links = {}
68
54
self.missing_inventory_sha_cnt = 0
69
55
self.missing_revision_cnt = 0
70
self.checked_weaves = set()
71
self.unreferenced_versions = set()
56
# maps (file-id, version) -> sha1; used by InventoryFile._check
57
self.checked_texts = {}
58
self.checked_weaves = {}
59
self.revision_versions = _mod_repository._RevisionTextVersionCache(
61
self.unreferenced_ancestors = set()
72
62
self.inconsistent_parents = []
73
self.rich_roots = repository.supports_rich_root()
74
self.text_key_references = {}
75
self.check_repo = check_repo
76
self.other_results = []
77
# Plain text lines to include in the report
78
self._report_items = []
79
# Keys we are looking for; may be large and need spilling to disk.
80
# key->(type(revision/inventory/text/signature/map), sha1, first-referer)
81
self.pending_keys = {}
82
# Ancestors map for all of revisions being checked; while large helper
83
# functions we call would create it anyway, so better to have once and
87
def check(self, callback_refs=None, check_repo=True):
88
if callback_refs is None:
90
65
self.repository.lock_read()
91
66
self.progress = bzrlib.ui.ui_factory.nested_progress_bar()
93
self.progress.update('check', 0, 4)
95
self.progress.update('checking revisions', 0)
96
self.check_revisions()
97
self.progress.update('checking commit contents', 1)
98
self.repository._check_inventories(self)
99
self.progress.update('checking file graphs', 2)
100
# check_weaves is done after the revision scan so that
101
# revision index is known to be valid.
103
self.progress.update('checking branches and trees', 3)
105
repo = self.repository
106
# calculate all refs, and callback the objects requesting them.
108
wanting_items = set()
109
# Current crude version calculates everything and calls
110
# everything at once. Doing a queue and popping as things are
111
# satisfied would be cheaper on memory [but few people have
112
# huge numbers of working trees today. TODO: fix before
116
for ref, wantlist in callback_refs.iteritems():
117
wanting_items.update(wantlist)
120
refs[ref] = repo.revision_tree(value)
121
elif kind == 'lefthand-distance':
123
elif kind == 'revision-existence':
124
existences.add(value)
126
raise AssertionError(
127
'unknown ref kind for ref %s' % ref)
128
node_distances = repo.get_graph().find_lefthand_distances(distances)
129
for key, distance in node_distances.iteritems():
130
refs[('lefthand-distance', key)] = distance
131
if key in existences and distance > 0:
132
refs[('revision-existence', key)] = True
133
existences.remove(key)
134
parent_map = repo.get_graph().get_parent_map(existences)
135
for key in parent_map:
136
refs[('revision-existence', key)] = True
137
existences.remove(key)
138
for key in existences:
139
refs[('revision-existence', key)] = False
140
for item in wanting_items:
141
if isinstance(item, WorkingTree):
143
if isinstance(item, Branch):
144
self.other_results.append(item.check(refs))
68
self.progress.update('retrieving inventory', 0, 2)
69
# do not put in init, as it should be done with progess,
70
# and inside the lock.
71
self.inventory_weave = self.repository.get_inventory_weave()
72
self.progress.update('checking revision graph', 1)
73
self.check_revision_graph()
76
while revno < len(self.planned_revisions):
77
rev_id = self.planned_revisions[revno]
78
self.progress.update('checking revision', revno,
79
len(self.planned_revisions))
81
self.check_one_rev(rev_id)
82
# check_weaves is done after the revision scan so that
83
# revision_versions is pre-populated
146
86
self.progress.finished()
147
87
self.repository.unlock()
149
def _check_revisions(self, revisions_iterator):
150
"""Check revision objects by decorating a generator.
152
:param revisions_iterator: An iterator of(revid, Revision-or-None).
153
:return: A generator of the contents of revisions_iterator.
155
self.planned_revisions = set()
156
for revid, revision in revisions_iterator:
157
yield revid, revision
158
self._check_one_rev(revid, revision)
159
# Flatten the revisions we found to guarantee consistent later
161
self.planned_revisions = list(self.planned_revisions)
162
# TODO: extract digital signatures as items to callback on too.
164
def check_revisions(self):
165
"""Scan revisions, checking data directly available as we go."""
166
revision_iterator = self.repository._iter_revisions(None)
167
revision_iterator = self._check_revisions(revision_iterator)
168
# We read the all revisions here:
169
# - doing this allows later code to depend on the revision index.
170
# - we can fill out existence flags at this point
171
# - we can read the revision inventory sha at this point
172
# - we can check properties and serialisers etc.
89
def check_revision_graph(self):
173
90
if not self.repository.revision_graph_can_have_wrong_parents():
174
# The check against the index isn't needed.
91
# This check is not necessary.
175
92
self.revs_with_bad_parents_in_index = None
176
for thing in revision_iterator:
179
bad_revisions = self.repository._find_inconsistent_revision_parents(
181
self.revs_with_bad_parents_in_index = list(bad_revisions)
94
bad_revisions = self.repository._find_inconsistent_revision_parents()
95
self.revs_with_bad_parents_in_index = list(bad_revisions)
97
def plan_revisions(self):
98
repository = self.repository
99
self.planned_revisions = repository.all_revision_ids()
100
self.progress.clear()
101
inventoried = set(self.inventory_weave.versions())
102
awol = set(self.planned_revisions) - inventoried
104
raise BzrCheckError('Stored revisions missing from inventory'
105
'{%s}' % ','.join([f for f in awol]))
183
107
def report_results(self, verbose):
185
self._report_repo_results(verbose)
186
for result in self.other_results:
187
result.report_results(verbose)
189
def _report_repo_results(self, verbose):
190
108
note('checked repository %s format %s',
191
self.repository.user_url,
192
self.repository._format)
109
self.repository.bzrdir.root_transport,
110
self.repository._format)
193
111
note('%6d revisions', self.checked_rev_cnt)
194
112
note('%6d file-ids', len(self.checked_weaves))
196
note('%6d unreferenced text versions',
197
len(self.unreferenced_versions))
198
if verbose and len(self.unreferenced_versions):
199
for file_id, revision_id in self.unreferenced_versions:
200
note('unreferenced version: {%s} in %s', revision_id,
113
note('%6d unique file texts', self.checked_text_cnt)
114
note('%6d repeated file texts', self.repeated_text_cnt)
115
note('%6d unreferenced text ancestors',
116
len(self.unreferenced_ancestors))
202
117
if self.missing_inventory_sha_cnt:
203
118
note('%6d revisions are missing inventory_sha1',
204
119
self.missing_inventory_sha_cnt)
237
156
' %s has wrong parents in index: '
238
157
'%r should be %r',
239
158
revision_id, index_parents, actual_parents)
240
for item in self._report_items:
243
def _check_one_rev(self, rev_id, rev):
244
"""Cross-check one revision.
246
:param rev_id: A revision id to check.
247
:param rev: A revision or None to indicate a missing revision.
160
def check_one_rev(self, rev_id):
161
"""Check one revision.
163
rev_id - the one to check
165
rev = self.repository.get_revision(rev_id)
249
167
if rev.revision_id != rev_id:
250
self._report_items.append(
251
'Mismatched internal revid {%s} and index revid {%s}' % (
252
rev.revision_id, rev_id))
253
rev_id = rev.revision_id
254
# Check this revision tree etc, and count as seen when we encounter a
256
self.planned_revisions.add(rev_id)
258
self.ghosts.discard(rev_id)
259
# Count all parents as ghosts if we haven't seen them yet.
168
raise BzrCheckError('wrong internal revision id in revision {%s}'
260
171
for parent in rev.parent_ids:
261
172
if not parent in self.planned_revisions:
262
self.ghosts.add(parent)
264
self.ancestors[rev_id] = tuple(rev.parent_ids) or (NULL_REVISION,)
265
self.add_pending_item(rev_id, ('inventories', rev_id), 'inventory',
173
missing_links = self.missing_parent_links.get(parent, [])
174
missing_links.append(rev_id)
175
self.missing_parent_links[parent] = missing_links
176
# list based so somewhat slow,
177
# TODO have a planned_revisions list and set.
178
if self.repository.has_revision(parent):
179
missing_ancestry = self.repository.get_ancestry(parent)
180
for missing in missing_ancestry:
181
if (missing is not None
182
and missing not in self.planned_revisions):
183
self.planned_revisions.append(missing)
185
self.ghosts.append(rev_id)
187
if rev.inventory_sha1:
188
inv_sha1 = self.repository.get_inventory_sha1(rev_id)
189
if inv_sha1 != rev.inventory_sha1:
190
raise BzrCheckError('Inventory sha1 hash doesn\'t match'
191
' value in revision {%s}' % rev_id)
192
self._check_revision_tree(rev_id)
267
193
self.checked_rev_cnt += 1
269
def add_pending_item(self, referer, key, kind, sha1):
270
"""Add a reference to a sha1 to be cross checked against a key.
272
:param referer: The referer that expects key to have sha1.
273
:param key: A storage key e.g. ('texts', 'foo@bar-20040504-1234')
274
:param kind: revision/inventory/text/map/signature
275
:param sha1: A hex sha1 or None if no sha1 is known.
277
existing = self.pending_keys.get(key)
279
if sha1 != existing[1]:
280
self._report_items.append('Multiple expected sha1s for %s. {%s}'
281
' expects {%s}, {%s} expects {%s}', (
282
key, referer, sha1, existing[1], existing[0]))
284
self.pending_keys[key] = (kind, sha1, referer)
286
195
def check_weaves(self):
287
196
"""Check all the weaves we can get our hands on.
290
storebar = bzrlib.ui.ui_factory.nested_progress_bar()
292
self._check_weaves(storebar)
296
def _check_weaves(self, storebar):
297
storebar.update('text-index', 0, 2)
298
if self.repository._format.fast_deltas:
299
# We haven't considered every fileid instance so far.
300
weave_checker = self.repository._get_versioned_file_checker(
301
ancestors=self.ancestors)
303
weave_checker = self.repository._get_versioned_file_checker(
304
text_key_references=self.text_key_references,
305
ancestors=self.ancestors)
306
storebar.update('file-graph', 1)
307
result = weave_checker.check_file_version_parents(
308
self.repository.texts)
309
self.checked_weaves = weave_checker.file_ids
310
bad_parents, unused_versions = result
311
bad_parents = bad_parents.items()
312
for text_key, (stored_parents, correct_parents) in bad_parents:
313
# XXX not ready for id join/split operations.
314
weave_id = text_key[0]
315
revision_id = text_key[-1]
316
weave_parents = tuple([parent[-1] for parent in stored_parents])
317
correct_parents = tuple([parent[-1] for parent in correct_parents])
318
self.inconsistent_parents.append(
319
(revision_id, weave_id, weave_parents, correct_parents))
320
self.unreferenced_versions.update(unused_versions)
322
def _add_entry_to_text_key_references(self, inv, entry):
323
if not self.rich_roots and entry.name == '':
325
key = (entry.file_id, entry.revision)
326
self.text_key_references.setdefault(key, False)
327
if entry.revision == inv.revision_id:
328
self.text_key_references[key] = True
331
@deprecated_function(deprecated_in((1,6,0)))
200
if self.repository.weave_store.listable():
201
weave_ids = list(self.repository.weave_store)
202
n_weaves = len(weave_ids) + 1
203
self.progress.update('checking versionedfile', 0, n_weaves)
204
self.inventory_weave.check(progress_bar=self.progress)
205
files_in_revisions = {}
206
revisions_of_files = {}
207
for i, weave_id in enumerate(weave_ids):
208
self.progress.update('checking versionedfile', i, n_weaves)
209
w = self.repository.weave_store.get_weave(weave_id,
210
self.repository.get_transaction())
211
# No progress here, because it looks ugly.
214
weave_checker = self.repository.get_versioned_file_checker(
215
self.planned_revisions, self.revision_versions)
216
result = weave_checker.check_file_version_parents(w, weave_id)
218
for revision_id, (weave_parents,correct_parents) in result.items():
219
self.inconsistent_parents.append(
220
(revision_id, weave_id, weave_parents, correct_parents))
221
unreferenced_parents = set(weave_parents)-set(correct_parents)
222
for unreferenced_parent in unreferenced_parents:
223
self.unreferenced_ancestors.add(
224
(weave_id, unreferenced_parent))
225
self.checked_weaves[weave_id] = True
227
def _check_revision_tree(self, rev_id):
228
tree = self.repository.revision_tree(rev_id)
229
self.revision_versions.add_revision_text_versions(tree)
233
if file_id in seen_ids:
234
raise BzrCheckError('duplicated file_id {%s} '
235
'in inventory for revision {%s}'
237
seen_ids[file_id] = True
240
ie.check(self, rev_id, inv, tree)
242
for path, ie in inv.iter_entries():
243
if path in seen_names:
244
raise BzrCheckError('duplicated path %s '
245
'in inventory for revision {%s}'
247
seen_names[path] = True
332
250
def check(branch, verbose):
333
251
"""Run consistency checks on a branch.
335
Results are reported through logging.
337
Deprecated in 1.6. Please use check_dwim instead.
339
:raise BzrCheckError: if there's a consistency error.
341
check_branch(branch, verbose)
344
@deprecated_function(deprecated_in((1,16,0)))
345
def check_branch(branch, verbose):
346
"""Run consistency checks on a branch.
348
Results are reported through logging.
253
Results are reported through logging.
350
255
:raise BzrCheckError: if there's a consistency error.
352
257
branch.lock_read()
355
for ref in branch._get_check_refs():
356
needed_refs.setdefault(ref, []).append(branch)
357
result = branch.repository.check([branch.last_revision()], needed_refs)
358
branch_result = result.other_results[0]
259
branch_result = branch.check()
260
repo_result = branch.repository.check([branch.last_revision()])
361
263
branch_result.report_results(verbose)
364
def scan_branch(branch, needed_refs, to_unlock):
365
"""Scan a branch for refs.
367
:param branch: The branch to schedule for checking.
368
:param needed_refs: Refs we are accumulating.
369
:param to_unlock: The unlock list accumulating.
371
note("Checking branch at '%s'." % (branch.base,))
373
to_unlock.append(branch)
374
branch_refs = branch._get_check_refs()
375
for ref in branch_refs:
376
reflist = needed_refs.setdefault(ref, [])
377
reflist.append(branch)
380
def scan_tree(base_tree, tree, needed_refs, to_unlock):
381
"""Scan a tree for refs.
383
:param base_tree: The original tree check opened, used to detect duplicate
385
:param tree: The tree to schedule for checking.
386
:param needed_refs: Refs we are accumulating.
387
:param to_unlock: The unlock list accumulating.
389
if base_tree is not None and tree.basedir == base_tree.basedir:
391
note("Checking working tree at '%s'." % (tree.basedir,))
393
to_unlock.append(tree)
394
tree_refs = tree._get_check_refs()
395
for ref in tree_refs:
396
reflist = needed_refs.setdefault(ref, [])
400
def check_dwim(path, verbose, do_branch=False, do_repo=False, do_tree=False):
401
"""Check multiple objects.
403
If errors occur they are accumulated and reported as far as possible, and
404
an exception raised at the end of the process.
407
base_tree, branch, repo, relpath = \
408
BzrDir.open_containing_tree_branch_or_repository(path)
409
except errors.NotBranchError:
410
base_tree = branch = repo = None
415
if base_tree is not None:
416
# If the tree is a lightweight checkout we won't see it in
417
# repo.find_branches - add now.
419
scan_tree(None, base_tree, needed_refs, to_unlock)
420
branch = base_tree.branch
421
if branch is not None:
424
# The branch is in a shared repository
425
repo = branch.repository
428
to_unlock.append(repo)
429
branches = repo.find_branches(using=True)
431
if do_branch or do_tree:
432
for branch in branches:
435
tree = branch.bzrdir.open_workingtree()
437
except (errors.NotLocalUrl, errors.NoWorkingTree):
440
scan_tree(base_tree, tree, needed_refs, to_unlock)
442
scan_branch(branch, needed_refs, to_unlock)
443
if do_branch and not branches:
444
note("No branch found at specified location.")
445
if do_tree and base_tree is None and not saw_tree:
446
note("No working tree found at specified location.")
447
if do_repo or do_branch or do_tree:
449
note("Checking repository at '%s'."
451
result = repo.check(None, callback_refs=needed_refs,
453
result.report_results(verbose)
456
note("No working tree found at specified location.")
458
note("No branch found at specified location.")
460
note("No repository found at specified location.")
462
for thing in to_unlock:
264
repo_result.report_results(verbose)