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
from bzrlib import errors
35
from bzrlib import errors, osutils
36
from bzrlib import repository as _mod_repository
37
from bzrlib import revision
50
38
from bzrlib.branch import Branch
51
39
from bzrlib.bzrdir import BzrDir
52
from bzrlib.revision import NULL_REVISION
40
from bzrlib.errors import BzrCheckError
41
from bzrlib.repository import Repository
53
42
from bzrlib.symbol_versioning import deprecated_function, deprecated_in
54
from bzrlib.trace import note
43
from bzrlib.trace import log_error, note
56
45
from bzrlib.workingtree import WorkingTree
61
50
# The Check object interacts with InventoryEntry.check, etc.
63
def __init__(self, repository, check_repo=True):
52
def __init__(self, repository):
64
53
self.repository = repository
54
self.checked_text_cnt = 0
65
55
self.checked_rev_cnt = 0
57
self.repeated_text_cnt = 0
67
58
self.missing_parent_links = {}
68
59
self.missing_inventory_sha_cnt = 0
69
60
self.missing_revision_cnt = 0
61
# maps (file-id, version) -> sha1; used by InventoryFile._check
62
self.checked_texts = {}
70
63
self.checked_weaves = set()
71
64
self.unreferenced_versions = set()
72
65
self.inconsistent_parents = []
73
66
self.rich_roots = repository.supports_rich_root()
74
67
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
70
self.repository.lock_read()
91
71
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))
73
self.progress.update('retrieving inventory', 0, 2)
74
# do not put in init, as it should be done with progess,
75
# and inside the lock.
76
self.inventory_weave = self.repository.inventories
77
self.progress.update('checking revision graph', 1)
78
self.check_revision_graph()
81
while revno < len(self.planned_revisions):
82
rev_id = self.planned_revisions[revno]
83
self.progress.update('checking revision', revno,
84
len(self.planned_revisions))
86
self.check_one_rev(rev_id)
87
# check_weaves is done after the revision scan so that
88
# revision index is known to be valid.
146
91
self.progress.finished()
147
92
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.
94
def check_revision_graph(self):
173
95
if not self.repository.revision_graph_can_have_wrong_parents():
174
# The check against the index isn't needed.
96
# This check is not necessary.
175
97
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)
99
bad_revisions = self.repository._find_inconsistent_revision_parents()
100
self.revs_with_bad_parents_in_index = list(bad_revisions)
102
def plan_revisions(self):
103
repository = self.repository
104
self.planned_revisions = repository.all_revision_ids()
105
self.progress.clear()
106
inventoried = set(key[-1] for key in self.inventory_weave.keys())
107
awol = set(self.planned_revisions) - inventoried
109
raise BzrCheckError('Stored revisions missing from inventory'
110
'{%s}' % ','.join([f for f in awol]))
183
112
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
113
note('checked repository %s format %s',
191
self.repository.user_url,
192
self.repository._format)
114
self.repository.bzrdir.root_transport,
115
self.repository._format)
193
116
note('%6d revisions', self.checked_rev_cnt)
194
117
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,
118
note('%6d unique file texts', self.checked_text_cnt)
119
note('%6d repeated file texts', self.repeated_text_cnt)
120
note('%6d unreferenced text versions',
121
len(self.unreferenced_versions))
202
122
if self.missing_inventory_sha_cnt:
203
123
note('%6d revisions are missing inventory_sha1',
204
124
self.missing_inventory_sha_cnt)
237
161
' %s has wrong parents in index: '
238
162
'%r should be %r',
239
163
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.
165
def check_one_rev(self, rev_id):
166
"""Check one revision.
168
rev_id - the one to check
170
rev = self.repository.get_revision(rev_id)
249
172
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.
173
raise BzrCheckError('wrong internal revision id in revision {%s}'
260
176
for parent in rev.parent_ids:
261
177
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',
178
# rev has a parent we didn't know about.
179
missing_links = self.missing_parent_links.get(parent, [])
180
missing_links.append(rev_id)
181
self.missing_parent_links[parent] = missing_links
182
# list based so somewhat slow,
183
# TODO have a planned_revisions list and set.
184
if self.repository.has_revision(parent):
185
missing_ancestry = self.repository.get_ancestry(parent)
186
for missing in missing_ancestry:
187
if (missing is not None
188
and missing not in self.planned_revisions):
189
self.planned_revisions.append(missing)
191
self.ghosts.append(rev_id)
193
if rev.inventory_sha1:
194
# Loopback - this is currently circular logic as the
195
# knit get_inventory_sha1 call returns rev.inventory_sha1.
196
# Repository.py's get_inventory_sha1 should instead return
197
# inventories.get_record_stream([(revid,)]).next().sha1 or
199
inv_sha1 = self.repository.get_inventory_sha1(rev_id)
200
if inv_sha1 != rev.inventory_sha1:
201
raise BzrCheckError('Inventory sha1 hash doesn\'t match'
202
' value in revision {%s}' % rev_id)
203
self._check_revision_tree(rev_id)
267
204
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
206
def check_weaves(self):
287
207
"""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)
210
self.progress.update('checking inventory', 0, 2)
211
self.inventory_weave.check(progress_bar=self.progress)
212
self.progress.update('checking text storage', 1, 2)
213
self.repository.texts.check(progress_bar=self.progress)
214
weave_checker = self.repository._get_versioned_file_checker(
215
text_key_references=self.text_key_references)
307
216
result = weave_checker.check_file_version_parents(
308
self.repository.texts)
217
self.repository.texts, progress_bar=self.progress)
309
218
self.checked_weaves = weave_checker.file_ids
310
219
bad_parents, unused_versions = result
311
220
bad_parents = bad_parents.items()
352
280
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]
282
branch_result = branch.check()
361
285
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
288
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 = \
290
tree, branch, repo, relpath = \
408
291
BzrDir.open_containing_tree_branch_or_repository(path)
409
292
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,
293
tree = branch = repo = None
297
note("Checking working tree at '%s'."
298
% (tree.bzrdir.root_transport.base,))
301
log_error("No working tree found at specified location.")
303
if branch is not None:
306
# The branch is in a shared repository
307
repo = branch.repository
309
elif repo is not None:
310
branches = repo.find_branches(using=True)
316
note("Checking repository at '%s'."
317
% (repo.bzrdir.root_transport.base,))
318
result = repo.check()
453
319
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:
322
log_error("No branch found at specified location.")
324
for branch in branches:
325
note("Checking branch at '%s'."
326
% (branch.bzrdir.root_transport.base,))
327
check_branch(branch, verbose)
332
log_error("No branch found at specified location.")
334
log_error("No repository found at specified location.")
added
removed
bzrlib/check.py
