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)
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:
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
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
56
from bzrlib.workingtree import WorkingTree
59
"""Check a repository"""
61
# The Check object interacts with InventoryEntry.check, etc.
63
def __init__(self, repository, check_repo=True):
64
self.repository = repository
65
self.checked_rev_cnt = 0
67
self.missing_parent_links = {}
68
self.missing_inventory_sha_cnt = 0
69
self.missing_revision_cnt = 0
70
self.checked_weaves = set()
71
self.unreferenced_versions = set()
72
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
self.repository.lock_read()
91
self.progress = bzrlib.ui.ui_factory.nested_progress_bar()
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
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]
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)
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
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))
146
self.progress.finished()
147
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.
173
if not self.repository.revision_graph_can_have_wrong_parents():
174
# The check against the index isn't needed.
175
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)
183
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
note('checked repository %s format %s',
191
self.repository.user_url,
192
self.repository._format)
193
note('%6d revisions', self.checked_rev_cnt)
194
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,
202
if self.missing_inventory_sha_cnt:
203
note('%6d revisions are missing inventory_sha1',
204
self.missing_inventory_sha_cnt)
205
if self.missing_revision_cnt:
206
note('%6d revisions are mentioned but not present',
207
self.missing_revision_cnt)
209
note('%6d ghost revisions', len(self.ghosts))
211
for ghost in self.ghosts:
213
if len(self.missing_parent_links):
214
note('%6d revisions missing parents in ancestry',
215
len(self.missing_parent_links))
217
for link, linkers in self.missing_parent_links.items():
218
note(' %s should be in the ancestry for:', link)
219
for linker in linkers:
220
note(' * %s', linker)
221
if len(self.inconsistent_parents):
222
note('%6d inconsistent parents', len(self.inconsistent_parents))
224
for info in self.inconsistent_parents:
225
revision_id, file_id, found_parents, correct_parents = info
226
note(' * %s version %s has parents %r '
228
% (file_id, revision_id, found_parents,
230
if self.revs_with_bad_parents_in_index:
231
note('%6d revisions have incorrect parents in the revision index',
232
len(self.revs_with_bad_parents_in_index))
234
for item in self.revs_with_bad_parents_in_index:
235
revision_id, index_parents, actual_parents = item
237
' %s has wrong parents in index: '
239
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.
249
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.
260
for parent in rev.parent_ids:
261
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',
267
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
def check_weaves(self):
287
"""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)))
332
def check(branch, verbose):
333
"""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.
350
:raise BzrCheckError: if there's a consistency error.
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]
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
# stub this out for now because the main bzr branch has references
228
# to revisions that aren't present in the store -- mbp 20050804
229
# if (missing_inventory_sha_cnt
230
# or missing_revision_sha_cnt):
231
# print ' (use "bzr upgrade" to fix them)'
234
print '%d revisions have mismatched inventory ids:' % len(mismatch_inv_id)
235
for rev_id in mismatch_inv_id:
361
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:
added
removed
bzrlib/check.py
