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
from bzrlib.trace import note, warning
21
def _update_store_entry(obj, obj_id, branch, store_name, store):
22
"""This is just a meta-function, which handles both revision entries
23
and inventory entries.
25
from bzrlib.trace import mutter
26
import tempfile, os, errno
27
from osutils import rename
28
obj_tmp = tempfile.TemporaryFile()
29
obj.write_xml(obj_tmp)
32
tmpfd, tmp_path = tempfile.mkstemp(prefix=obj_id, suffix='.gz',
33
dir=branch.controlfilename(store_name))
36
orig_obj_path = branch.controlfilename([store_name, obj_id+'.gz'])
37
# Remove the old entry out of the way
38
rename(orig_obj_path, tmp_path)
40
# TODO: We may need to handle the case where the old
41
# entry was not compressed (and thus did not end with .gz)
43
store.add(obj_tmp, obj_id) # Add the new one
44
os.remove(tmp_path) # Remove the old name
45
mutter(' Updated %s entry {%s}' % (store_name, obj_id))
47
# On any exception, restore the old entry
48
rename(tmp_path, orig_obj_path)
51
if os.path.exists(tmp_path):
52
# Unfortunately, the next command might throw
53
# an exception, which will mask a previous exception.
57
def _update_revision_entry(rev, branch):
58
"""After updating the values in a revision, make sure to
59
write out the data, but try to do it in an atomic manner.
61
:param rev: The Revision object to store
62
:param branch: The Branch object where this Revision is to be stored.
64
_update_store_entry(rev, rev.revision_id, branch,
65
'revision-store', branch.revision_store)
67
def _update_inventory_entry(inv, inv_id, branch):
68
"""When an inventory has been modified (such as by adding a unique tree root)
69
this atomically re-generates the file.
71
:param inv: The Inventory
72
:param inv_id: The inventory id for this inventory
73
:param branch: The Branch where this entry will be stored.
75
_update_store_entry(inv, inv_id, branch,
76
'inventory-store', branch.inventory_store)
80
"""Run consistency checks on a branch.
82
TODO: Also check non-mainline revisions mentioned as parents.
84
TODO: Check for extra files in the control directory.
86
from bzrlib.trace import mutter
87
from bzrlib.errors import BzrCheckError, NoSuchRevision
88
from bzrlib.osutils import fingerprint_file
89
from bzrlib.inventory import ROOT_ID
90
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.
50
from __future__ import absolute_import
56
from bzrlib.branch import Branch
57
from bzrlib.controldir import ControlDir
58
from bzrlib.revision import NULL_REVISION
59
from bzrlib.trace import note
60
from bzrlib.workingtree import WorkingTree
61
from bzrlib.i18n import gettext
64
"""Check a repository"""
66
def __init__(self, repository, check_repo=True):
67
self.repository = repository
69
def report_results(self, verbose):
70
raise NotImplementedError(self.report_results)
73
class VersionedFileCheck(Check):
74
"""Check a versioned file repository"""
76
# The Check object interacts with InventoryEntry.check, etc.
78
def __init__(self, repository, check_repo=True):
79
self.repository = repository
80
self.checked_rev_cnt = 0
82
self.missing_parent_links = {}
83
self.missing_inventory_sha_cnt = 0
84
self.missing_revision_cnt = 0
85
self.checked_weaves = set()
86
self.unreferenced_versions = set()
87
self.inconsistent_parents = []
88
self.rich_roots = repository.supports_rich_root()
89
self.text_key_references = {}
90
self.check_repo = check_repo
91
self.other_results = []
92
# Plain text lines to include in the report
93
self._report_items = []
94
# Keys we are looking for; may be large and need spilling to disk.
95
# key->(type(revision/inventory/text/signature/map), sha1, first-referer)
96
self.pending_keys = {}
97
# Ancestors map for all of revisions being checked; while large helper
98
# functions we call would create it anyway, so better to have once and
102
def check(self, callback_refs=None, check_repo=True):
103
if callback_refs is None:
105
self.repository.lock_read()
106
self.progress = ui.ui_factory.nested_progress_bar()
108
self.progress.update(gettext('check'), 0, 4)
110
self.progress.update(gettext('checking revisions'), 0)
111
self.check_revisions()
112
self.progress.update(gettext('checking commit contents'), 1)
113
self.repository._check_inventories(self)
114
self.progress.update(gettext('checking file graphs'), 2)
115
# check_weaves is done after the revision scan so that
116
# revision index is known to be valid.
118
self.progress.update(gettext('checking branches and trees'), 3)
120
repo = self.repository
121
# calculate all refs, and callback the objects requesting them.
123
wanting_items = set()
124
# Current crude version calculates everything and calls
125
# everything at once. Doing a queue and popping as things are
126
# satisfied would be cheaper on memory [but few people have
127
# huge numbers of working trees today. TODO: fix before
131
for ref, wantlist in callback_refs.iteritems():
132
wanting_items.update(wantlist)
135
refs[ref] = repo.revision_tree(value)
136
elif kind == 'lefthand-distance':
138
elif kind == 'revision-existence':
139
existences.add(value)
141
raise AssertionError(
142
'unknown ref kind for ref %s' % ref)
143
node_distances = repo.get_graph().find_lefthand_distances(distances)
144
for key, distance in node_distances.iteritems():
145
refs[('lefthand-distance', key)] = distance
146
if key in existences and distance > 0:
147
refs[('revision-existence', key)] = True
148
existences.remove(key)
149
parent_map = repo.get_graph().get_parent_map(existences)
150
for key in parent_map:
151
refs[('revision-existence', key)] = True
152
existences.remove(key)
153
for key in existences:
154
refs[('revision-existence', key)] = False
155
for item in wanting_items:
156
if isinstance(item, WorkingTree):
158
if isinstance(item, Branch):
159
self.other_results.append(item.check(refs))
161
self.progress.finished()
162
self.repository.unlock()
164
def _check_revisions(self, revisions_iterator):
165
"""Check revision objects by decorating a generator.
167
:param revisions_iterator: An iterator of(revid, Revision-or-None).
168
:return: A generator of the contents of revisions_iterator.
170
self.planned_revisions = set()
171
for revid, revision in revisions_iterator:
172
yield revid, revision
173
self._check_one_rev(revid, revision)
174
# Flatten the revisions we found to guarantee consistent later
176
self.planned_revisions = list(self.planned_revisions)
177
# TODO: extract digital signatures as items to callback on too.
179
def check_revisions(self):
180
"""Scan revisions, checking data directly available as we go."""
181
revision_iterator = self.repository._iter_revisions(None)
182
revision_iterator = self._check_revisions(revision_iterator)
183
# We read the all revisions here:
184
# - doing this allows later code to depend on the revision index.
185
# - we can fill out existence flags at this point
186
# - we can read the revision inventory sha at this point
187
# - we can check properties and serialisers etc.
188
if not self.repository._format.revision_graph_can_have_wrong_parents:
189
# The check against the index isn't needed.
190
self.revs_with_bad_parents_in_index = None
191
for thing in revision_iterator:
194
bad_revisions = self.repository._find_inconsistent_revision_parents(
196
self.revs_with_bad_parents_in_index = list(bad_revisions)
198
def report_results(self, verbose):
200
self._report_repo_results(verbose)
201
for result in self.other_results:
202
result.report_results(verbose)
204
def _report_repo_results(self, verbose):
205
note(gettext('checked repository {0} format {1}').format(
206
self.repository.user_url,
207
self.repository._format))
208
note(gettext('%6d revisions'), self.checked_rev_cnt)
209
note(gettext('%6d file-ids'), len(self.checked_weaves))
211
note(gettext('%6d unreferenced text versions'),
212
len(self.unreferenced_versions))
213
if verbose and len(self.unreferenced_versions):
214
for file_id, revision_id in self.unreferenced_versions:
215
note(gettext('unreferenced version: {{{0}}} in {1}').format(revision_id,
217
if self.missing_inventory_sha_cnt:
218
note(gettext('%6d revisions are missing inventory_sha1'),
219
self.missing_inventory_sha_cnt)
220
if self.missing_revision_cnt:
221
note(gettext('%6d revisions are mentioned but not present'),
222
self.missing_revision_cnt)
224
note(gettext('%6d ghost revisions'), len(self.ghosts))
226
for ghost in self.ghosts:
228
if len(self.missing_parent_links):
229
note(gettext('%6d revisions missing parents in ancestry'),
230
len(self.missing_parent_links))
232
for link, linkers in self.missing_parent_links.items():
233
note(gettext(' %s should be in the ancestry for:'), link)
234
for linker in linkers:
235
note(' * %s', linker)
236
if len(self.inconsistent_parents):
237
note(gettext('%6d inconsistent parents'), len(self.inconsistent_parents))
239
for info in self.inconsistent_parents:
240
revision_id, file_id, found_parents, correct_parents = info
241
note(gettext(' * {0} version {1} has parents {2!r} '
242
'but should have {3!r}').format(
243
file_id, revision_id, found_parents,
245
if self.revs_with_bad_parents_in_index:
247
'%6d revisions have incorrect parents in the revision index'),
248
len(self.revs_with_bad_parents_in_index))
250
for item in self.revs_with_bad_parents_in_index:
251
revision_id, index_parents, actual_parents = item
253
' {0} has wrong parents in index: '
254
'{1!r} should be {2!r}').format(
255
revision_id, index_parents, actual_parents))
256
for item in self._report_items:
259
def _check_one_rev(self, rev_id, rev):
260
"""Cross-check one revision.
262
:param rev_id: A revision id to check.
263
:param rev: A revision or None to indicate a missing revision.
265
if rev.revision_id != rev_id:
266
self._report_items.append(gettext(
267
'Mismatched internal revid {{{0}}} and index revid {{{1}}}').format(
268
rev.revision_id, rev_id))
269
rev_id = rev.revision_id
270
# Check this revision tree etc, and count as seen when we encounter a
272
self.planned_revisions.add(rev_id)
274
self.ghosts.discard(rev_id)
275
# Count all parents as ghosts if we haven't seen them yet.
276
for parent in rev.parent_ids:
277
if not parent in self.planned_revisions:
278
self.ghosts.add(parent)
280
self.ancestors[rev_id] = tuple(rev.parent_ids) or (NULL_REVISION,)
281
self.add_pending_item(rev_id, ('inventories', rev_id), 'inventory',
283
self.checked_rev_cnt += 1
285
def add_pending_item(self, referer, key, kind, sha1):
286
"""Add a reference to a sha1 to be cross checked against a key.
288
:param referer: The referer that expects key to have sha1.
289
:param key: A storage key e.g. ('texts', 'foo@bar-20040504-1234')
290
:param kind: revision/inventory/text/map/signature
291
:param sha1: A hex sha1 or None if no sha1 is known.
293
existing = self.pending_keys.get(key)
295
if sha1 != existing[1]:
296
self._report_items.append(gettext('Multiple expected sha1s for {0}. {{{1}}}'
297
' expects {{{2}}}, {{{3}}} expects {{{4}}}').format(
298
key, referer, sha1, existing[1], existing[0]))
300
self.pending_keys[key] = (kind, sha1, referer)
302
def check_weaves(self):
303
"""Check all the weaves we can get our hands on.
306
storebar = ui.ui_factory.nested_progress_bar()
308
self._check_weaves(storebar)
312
def _check_weaves(self, storebar):
313
storebar.update('text-index', 0, 2)
314
if self.repository._format.fast_deltas:
315
# We haven't considered every fileid instance so far.
316
weave_checker = self.repository._get_versioned_file_checker(
317
ancestors=self.ancestors)
319
weave_checker = self.repository._get_versioned_file_checker(
320
text_key_references=self.text_key_references,
321
ancestors=self.ancestors)
322
storebar.update('file-graph', 1)
323
result = weave_checker.check_file_version_parents(
324
self.repository.texts)
325
self.checked_weaves = weave_checker.file_ids
326
bad_parents, unused_versions = result
327
bad_parents = bad_parents.items()
328
for text_key, (stored_parents, correct_parents) in bad_parents:
329
# XXX not ready for id join/split operations.
330
weave_id = text_key[0]
331
revision_id = text_key[-1]
332
weave_parents = tuple([parent[-1] for parent in stored_parents])
333
correct_parents = tuple([parent[-1] for parent in correct_parents])
334
self.inconsistent_parents.append(
335
(revision_id, weave_id, weave_parents, correct_parents))
336
self.unreferenced_versions.update(unused_versions)
338
def _add_entry_to_text_key_references(self, inv, entry):
339
if not self.rich_roots and entry.name == '':
341
key = (entry.file_id, entry.revision)
342
self.text_key_references.setdefault(key, False)
343
if entry.revision == inv.revision_id:
344
self.text_key_references[key] = True
347
def scan_branch(branch, needed_refs, to_unlock):
348
"""Scan a branch for refs.
350
:param branch: The branch to schedule for checking.
351
:param needed_refs: Refs we are accumulating.
352
:param to_unlock: The unlock list accumulating.
354
note(gettext("Checking branch at '%s'.") % (branch.base,))
92
355
branch.lock_read()
97
missing_inventory_sha_cnt = 0
98
missing_revision_sha_cnt = 0
99
missing_revision_cnt = 0
101
history = branch.revision_history()
103
revcount = len(history)
106
# for all texts checked, text_id -> sha1
109
progress = bzrlib.ui.ui_factory.progress_bar()
111
for rev_id in history:
113
progress.update('checking revision', revno, revcount)
114
# mutter(' revision {%s}' % rev_id)
115
rev = branch.get_revision(rev_id)
116
if rev.revision_id != rev_id:
117
raise BzrCheckError('wrong internal revision id in revision {%s}'
120
# check the previous history entry is a parent of this entry
122
if last_rev_id is None:
123
raise BzrCheckError("revision {%s} has %d parents, but is the "
124
"start of the branch"
125
% (rev_id, len(rev.parents)))
126
for prr in rev.parents:
127
if prr.revision_id == last_rev_id:
130
raise BzrCheckError("previous revision {%s} not listed among "
132
% (last_rev_id, rev_id))
134
for prr in rev.parents:
135
if prr.revision_sha1 is None:
136
missing_revision_sha_cnt += 1
138
prid = prr.revision_id
141
actual_sha = branch.get_revision_sha1(prid)
142
except NoSuchRevision:
143
missing_revision_cnt += 1
144
mutter("parent {%s} of {%s} not present in store",
148
if prr.revision_sha1 != actual_sha:
149
raise BzrCheckError("mismatched revision sha1 for "
150
"parent {%s} of {%s}: %s vs %s"
152
prr.revision_sha1, actual_sha))
154
raise BzrCheckError("revision {%s} has no parents listed but preceded "
156
% (rev_id, last_rev_id))
158
if rev.inventory_id != rev_id:
159
mismatch_inv_id.append(rev_id)
161
## TODO: Check all the required fields are present on the revision.
163
if rev.inventory_sha1:
164
inv_sha1 = branch.get_inventory_sha1(rev.inventory_id)
165
if inv_sha1 != rev.inventory_sha1:
166
raise BzrCheckError('Inventory sha1 hash doesn\'t match'
167
' value in revision {%s}' % rev_id)
169
missing_inventory_sha_cnt += 1
170
mutter("no inventory_sha1 on revision {%s}" % rev_id)
172
inv = branch.get_inventory(rev.inventory_id)
176
## p('revision %d/%d file ids' % (revno, revcount))
178
if file_id in seen_ids:
179
raise BzrCheckError('duplicated file_id {%s} '
180
'in inventory for revision {%s}'
182
seen_ids[file_id] = True
192
if ie.parent_id != None:
193
if ie.parent_id not in seen_ids:
194
raise BzrCheckError('missing parent {%s} in inventory for revision {%s}'
195
% (ie.parent_id, rev_id))
197
if ie.kind == 'file':
198
if ie.text_id in checked_texts:
199
fp = checked_texts[ie.text_id]
201
if not ie.text_id in branch.text_store:
202
raise BzrCheckError('text {%s} not in text_store' % ie.text_id)
204
tf = branch.text_store[ie.text_id]
205
fp = fingerprint_file(tf)
206
checked_texts[ie.text_id] = fp
208
if ie.text_size != fp['size']:
209
raise BzrCheckError('text {%s} wrong size' % ie.text_id)
210
if ie.text_sha1 != fp['sha1']:
211
raise BzrCheckError('text {%s} wrong sha1' % ie.text_id)
212
elif ie.kind == 'directory':
213
if ie.text_sha1 != None or ie.text_size != None or ie.text_id != None:
214
raise BzrCheckError('directory {%s} has text in revision {%s}'
218
for path, ie in inv.iter_entries():
219
if path in seen_names:
220
raise BzrCheckError('duplicated path %s '
221
'in inventory for revision {%s}'
223
seen_names[path] = True
356
to_unlock.append(branch)
357
branch_refs = branch._get_check_refs()
358
for ref in branch_refs:
359
reflist = needed_refs.setdefault(ref, [])
360
reflist.append(branch)
363
def scan_tree(base_tree, tree, needed_refs, to_unlock):
364
"""Scan a tree for refs.
366
:param base_tree: The original tree check opened, used to detect duplicate
368
:param tree: The tree to schedule for checking.
369
:param needed_refs: Refs we are accumulating.
370
:param to_unlock: The unlock list accumulating.
372
if base_tree is not None and tree.basedir == base_tree.basedir:
374
note(gettext("Checking working tree at '%s'.") % (tree.basedir,))
376
to_unlock.append(tree)
377
tree_refs = tree._get_check_refs()
378
for ref in tree_refs:
379
reflist = needed_refs.setdefault(ref, [])
383
def check_dwim(path, verbose, do_branch=False, do_repo=False, do_tree=False):
384
"""Check multiple objects.
386
If errors occur they are accumulated and reported as far as possible, and
387
an exception raised at the end of the process.
390
base_tree, branch, repo, relpath = \
391
ControlDir.open_containing_tree_branch_or_repository(path)
392
except errors.NotBranchError:
393
base_tree = branch = repo = None
398
if base_tree is not None:
399
# If the tree is a lightweight checkout we won't see it in
400
# repo.find_branches - add now.
402
scan_tree(None, base_tree, needed_refs, to_unlock)
403
branch = base_tree.branch
404
if branch is not None:
407
# The branch is in a shared repository
408
repo = branch.repository
411
to_unlock.append(repo)
412
branches = repo.find_branches(using=True)
414
if do_branch or do_tree:
415
for branch in branches:
418
tree = branch.bzrdir.open_workingtree()
420
except (errors.NotLocalUrl, errors.NoWorkingTree):
423
scan_tree(base_tree, tree, needed_refs, to_unlock)
425
scan_branch(branch, needed_refs, to_unlock)
426
if do_branch and not branches:
427
note(gettext("No branch found at specified location."))
428
if do_tree and base_tree is None and not saw_tree:
429
note(gettext("No working tree found at specified location."))
430
if do_repo or do_branch or do_tree:
432
note(gettext("Checking repository at '%s'.")
434
result = repo.check(None, callback_refs=needed_refs,
436
result.report_results(verbose)
439
note(gettext("No working tree found at specified location."))
441
note(gettext("No branch found at specified location."))
443
note(gettext("No repository found at specified location."))
231
note('checked %d revisions, %d file texts' % (revcount, len(checked_texts)))
233
if missing_inventory_sha_cnt:
234
note('%d revisions are missing inventory_sha1' % missing_inventory_sha_cnt)
236
if missing_revision_sha_cnt:
237
note('%d parent links are missing revision_sha1' % missing_revision_sha_cnt)
239
if missing_revision_cnt:
240
note('%d revisions are mentioned but not present' % missing_revision_cnt)
242
# stub this out for now because the main bzr branch has references
243
# to revisions that aren't present in the store -- mbp 20050804
244
# if (missing_inventory_sha_cnt
245
# or missing_revision_sha_cnt):
246
# print ' (use "bzr upgrade" to fix them)'
249
warning('%d revisions have mismatched inventory ids:' % len(mismatch_inv_id))
250
for rev_id in mismatch_inv_id:
251
warning(' %s', rev_id)
445
for thing in to_unlock:
added
removed
bzrlib/check.py
