13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""bzr upgrade logic."""
26
from bzrlib.bzrdir import (
30
from bzrlib.remote import RemoteBzrDir
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Experiment in converting existing bzr branches to weaves."""
19
# To make this properly useful
21
# 1. assign text version ids, and put those text versions into
22
# the inventory as they're converted.
24
# 2. keep track of the previous version of each file, rather than
25
# just using the last one imported
27
# 3. assign entry versions when files are added, renamed or moved.
29
# 4. when merged-in versions are observed, walk down through them
30
# to discover everything, then commit bottom-up
32
# 5. track ancestry as things are merged in, and commit that in each
35
# Perhaps it's best to first walk the whole graph and make a plan for
36
# what should be imported in what order? Need a kind of topological
37
# sort of all revisions. (Or do we, can we just before doing a revision
38
# see that all its parents have either been converted or abandoned?)
41
# Cannot import a revision until all its parents have been
42
# imported. in other words, we can only import revisions whose
43
# parents have all been imported. the first step must be to
44
# import a revision with no parents, of which there must be at
45
# least one. (So perhaps it's useful to store forward pointers
46
# from a list of parents to their children?)
48
# Another (equivalent?) approach is to build up the ordered
49
# ancestry list for the last revision, and walk through that. We
50
# are going to need that.
52
# We don't want to have to recurse all the way back down the list.
54
# Suppose we keep a queue of the revisions able to be processed at
55
# any point. This starts out with all the revisions having no
58
# This seems like a generally useful algorithm...
60
# The current algorithm is dumb (O(n**2)?) but will do the job, and
61
# takes less than a second on the bzr.dev branch.
63
# This currently does a kind of lazy conversion of file texts, where a
64
# new text is written in every version. That's unnecessary but for
65
# the moment saves us having to worry about when files need new
74
from bzrlib.branch import Branch, find_branch
75
from bzrlib.branch import BZR_BRANCH_FORMAT_5, BZR_BRANCH_FORMAT_6
76
import bzrlib.hashcache as hashcache
77
from bzrlib.weave import Weave
78
from bzrlib.weavefile import read_weave, write_weave
79
from bzrlib.ui import ui_factory
80
from bzrlib.atomicfile import AtomicFile
81
from bzrlib.xml4 import serializer_v4
82
from bzrlib.xml5 import serializer_v5
83
from bzrlib.trace import mutter, note, warning
84
from bzrlib.osutils import sha_strings, sha_string, pathjoin, abspath
33
87
class Convert(object):
35
def __init__(self, url=None, format=None, control_dir=None):
36
"""Convert a Bazaar control directory to a given format.
38
Either the url or control_dir parameter must be given.
40
:param url: the URL of the control directory or None if the
41
control_dir is explicitly given instead
42
:param format: the format to convert to or None for the default
43
:param control_dir: the control directory or None if it is
44
specified via the URL parameter instead
47
# XXX: Change to cleanup
48
warning_id = 'cross_format_fetch'
49
saved_warning = warning_id in ui.ui_factory.suppressed_warnings
50
if url is None and control_dir is None:
52
"either the url or control_dir parameter must be set.")
53
if control_dir is not None:
54
self.bzrdir = control_dir
56
self.bzrdir = BzrDir.open_unsupported(url)
57
if isinstance(self.bzrdir, RemoteBzrDir):
58
self.bzrdir._ensure_real()
59
self.bzrdir = self.bzrdir._real_bzrdir
60
if self.bzrdir.root_transport.is_readonly():
61
raise errors.UpgradeReadonly
62
self.transport = self.bzrdir.root_transport
63
ui.ui_factory.suppressed_warnings.add(warning_id)
68
ui.ui_factory.suppressed_warnings.remove(warning_id)
88
def __init__(self, base_dir):
90
self.converted_revs = set()
91
self.absent_revisions = set()
72
branch = self.bzrdir.open_branch()
73
if branch.user_url != self.bzrdir.user_url:
75
'This is a checkout. The branch (%s) needs to be upgraded'
76
' separately.' % (urlutils.unescape_for_display(
77
branch.user_url, 'utf-8')))
79
except (errors.NotBranchError, errors.IncompatibleRepositories):
80
# might not be a format we can open without upgrading; see e.g.
81
# https://bugs.launchpad.net/bzr/+bug/253891
83
if self.format is None:
85
rich_root = self.bzrdir.find_repository()._format.rich_root_data
86
except errors.NoRepositoryPresent:
87
rich_root = False # assume no rich roots
89
format_name = "default-rich-root"
91
format_name = "default"
92
format = format_registry.make_bzrdir(format_name)
95
if not self.bzrdir.needs_format_conversion(format):
96
raise errors.UpToDateFormat(self.bzrdir._format)
97
if not self.bzrdir.can_convert_format():
98
raise errors.BzrError("cannot upgrade from bzrdir format %s" %
100
self.bzrdir.check_conversion_target(format)
101
ui.ui_factory.note('starting upgrade of %s' %
102
urlutils.unescape_for_display(self.transport.base, 'utf-8'))
104
self.backup_oldpath, self.backup_newpath = self.bzrdir.backup_bzrdir()
105
while self.bzrdir.needs_format_conversion(format):
106
converter = self.bzrdir._format.get_converter(format)
107
self.bzrdir = converter.convert(self.bzrdir, None)
108
ui.ui_factory.note('finished')
111
"""Clean-up after a conversion.
113
This removes the backup.bzr directory.
115
transport = self.transport
116
backup_relpath = transport.relpath(self.backup_newpath)
117
child_pb = ui.ui_factory.nested_progress_bar()
118
child_pb.update('Deleting backup.bzr')
120
transport.delete_tree(backup_relpath)
125
def upgrade(url, format=None, clean_up=False, dry_run=False):
126
"""Upgrade locations to format.
128
This routine wraps the smart_upgrade() routine with a nicer UI.
129
In particular, it ensures all URLs can be opened before starting
130
and reports a summary at the end if more than one upgrade was attempted.
131
This routine is useful for command line tools. Other bzrlib clients
132
probably ought to use smart_upgrade() instead.
134
:param url: a URL of the locations to upgrade.
135
:param format: the format to convert to or None for the best default
136
:param clean-up: if True, the backup.bzr directory is removed if the
137
upgrade succeeded for a given repo/branch/tree
138
:param dry_run: show what would happen but don't actually do any upgrades
139
:return: the list of exceptions encountered
141
control_dirs = [BzrDir.open_unsupported(url)]
142
attempted, succeeded, exceptions = smart_upgrade(control_dirs,
143
format, clean_up=clean_up, dry_run=dry_run)
144
if len(attempted) > 1:
145
attempted_count = len(attempted)
146
succeeded_count = len(succeeded)
147
failed_count = attempted_count - succeeded_count
149
'\nSUMMARY: %d upgrades attempted, %d succeeded, %d failed'
150
% (attempted_count, succeeded_count, failed_count))
154
def smart_upgrade(control_dirs, format, clean_up=False,
156
"""Convert control directories to a new format intelligently.
158
If the control directory is a shared repository, dependent branches
159
are also converted provided the repository converted successfully.
160
If the conversion of a branch fails, remaining branches are still tried.
162
:param control_dirs: the BzrDirs to upgrade
163
:param format: the format to convert to or None for the best default
164
:param clean_up: if True, the backup.bzr directory is removed if the
165
upgrade succeeded for a given repo/branch/tree
166
:param dry_run: show what would happen but don't actually do any upgrades
167
:return: attempted-control-dirs, succeeded-control-dirs, exceptions
172
for control_dir in control_dirs:
173
attempted, succeeded, exceptions = _smart_upgrade_one(control_dir,
174
format, clean_up=clean_up, dry_run=dry_run)
175
all_attempted.extend(attempted)
176
all_succeeded.extend(succeeded)
177
all_exceptions.extend(exceptions)
178
return all_attempted, all_succeeded, all_exceptions
181
def _smart_upgrade_one(control_dir, format, clean_up=False,
183
"""Convert a control directory to a new format intelligently.
185
See smart_upgrade for parameter details.
187
# If the URL is a shared repository, find the dependent branches
190
repo = control_dir.open_repository()
191
except errors.NoRepositoryPresent:
192
# A branch or checkout using a shared repository higher up
195
# The URL is a repository. If it successfully upgrades,
196
# then upgrade the dependent branches as well.
198
dependents = repo.find_branches(using=True)
201
attempted = [control_dir]
202
succeeded, exceptions = _convert_items([control_dir], format, clean_up,
204
if succeeded and dependents:
205
ui.ui_factory.note('Found %d dependent branches - upgrading ...'
206
% (len(dependents),))
207
# Convert dependent branches
208
branch_cdirs = [b.bzrdir for b in dependents]
209
successes, problems = _convert_items(branch_cdirs, format, clean_up,
210
dry_run, label="branch")
211
attempted.extend(branch_cdirs)
212
succeeded.extend(successes)
213
exceptions.extend(problems)
216
return attempted, succeeded, exceptions
218
# FIXME: There are several problems below:
219
# - RemoteRepository doesn't support _unsupported (really ?)
220
# - raising AssertionError is rude and may not be necessary
222
# - the only caller uses only the label
223
def _get_object_and_label(control_dir):
224
"""Return the primary object and type label for a control directory.
226
:return: object, label where:
227
* object is a Branch, Repository or WorkingTree and
230
* repository - a repository
231
* tree - a lightweight checkout
235
br = control_dir.open_branch(unsupported=True,
236
ignore_fallbacks=True)
237
except NotImplementedError:
238
# RemoteRepository doesn't support the unsupported parameter
239
br = control_dir.open_branch(ignore_fallbacks=True)
240
except errors.NotBranchError:
245
repo = control_dir.open_repository()
246
except errors.NoRepositoryPresent:
249
return repo, "repository"
251
wt = control_dir.open_workingtree()
252
except (errors.NoWorkingTree, errors.NotLocalUrl):
256
raise AssertionError("unknown type of control directory %s", control_dir)
259
def _convert_items(items, format, clean_up, dry_run, label=None):
260
"""Convert a sequence of control directories to the given format.
262
:param items: the control directories to upgrade
263
:param format: the format to convert to or None for the best default
264
:param clean-up: if True, the backup.bzr directory is removed if the
265
upgrade succeeded for a given repo/branch/tree
266
:param dry_run: show what would happen but don't actually do any upgrades
267
:param label: the label for these items or None to calculate one
268
:return: items successfully upgraded, exceptions
272
child_pb = ui.ui_factory.nested_progress_bar()
273
child_pb.update('Upgrading bzrdirs', 0, len(items))
274
for i, control_dir in enumerate(items):
276
location = control_dir.root_transport.base
277
bzr_object, bzr_label = _get_object_and_label(control_dir)
278
type_label = label or bzr_label
279
child_pb.update("Upgrading %s" % (type_label), i+1, len(items))
280
ui.ui_factory.note('Upgrading %s %s ...' % (type_label,
281
urlutils.unescape_for_display(location, 'utf-8'),))
284
cv = Convert(control_dir=control_dir, format=format)
285
except Exception, ex:
286
trace.warning('conversion error: %s' % ex)
287
exceptions.append(ex)
290
# Do any required post processing
291
succeeded.append(control_dir)
294
ui.ui_factory.note('Removing backup ...')
297
except Exception, ex:
298
trace.warning('failed to clean-up %s: %s' % (location, ex))
299
exceptions.append(ex)
304
return succeeded, exceptions
98
if not self._open_branch():
100
note('starting upgrade of %s', os.path.abspath(self.base))
101
self._backup_control_dir()
102
self.pb = ui_factory.progress_bar()
103
if self.old_format == 4:
104
note('starting upgrade from format 4 to 5')
105
self._convert_to_weaves()
107
if self.old_format == 5:
108
note('starting upgrade from format 5 to 6')
109
self._convert_to_prefixed()
111
cache = hashcache.HashCache(abspath(self.base))
117
def _convert_to_prefixed(self):
118
from bzrlib.store import hash_prefix
119
for store_name in ["weaves", "revision-store"]:
120
note("adding prefixes to %s" % store_name)
121
store_dir = pathjoin(self.base, ".bzr", store_name)
122
for filename in os.listdir(store_dir):
123
if filename.endswith(".weave") or filename.endswith(".gz"):
124
file_id = os.path.splitext(filename)[0]
127
prefix_dir = pathjoin(store_dir, hash_prefix(file_id))
128
if not os.path.isdir(prefix_dir):
130
os.rename(pathjoin(store_dir, filename),
131
pathjoin(prefix_dir, filename))
132
self._set_new_format(BZR_BRANCH_FORMAT_6)
135
def _convert_to_weaves(self):
136
note('note: upgrade may be faster if all store files are ungzipped first')
137
if not os.path.isdir(self.base + '/.bzr/weaves'):
138
os.mkdir(self.base + '/.bzr/weaves')
139
self.inv_weave = Weave('inventory')
140
# holds in-memory weaves for all files
141
self.text_weaves = {}
142
os.remove(self.branch.controlfilename('branch-format'))
143
self._convert_working_inv()
144
rev_history = self.branch.revision_history()
145
# to_read is a stack holding the revisions we still need to process;
146
# appending to it adds new highest-priority revisions
147
self.known_revisions = set(rev_history)
148
self.to_read = rev_history[-1:]
150
rev_id = self.to_read.pop()
151
if (rev_id not in self.revisions
152
and rev_id not in self.absent_revisions):
153
self._load_one_rev(rev_id)
155
to_import = self._make_order()
156
for i, rev_id in enumerate(to_import):
157
self.pb.update('converting revision', i, len(to_import))
158
self._convert_one_rev(rev_id)
160
note('upgraded to weaves:')
161
note(' %6d revisions and inventories' % len(self.revisions))
162
note(' %6d revisions not present' % len(self.absent_revisions))
163
note(' %6d texts' % self.text_count)
164
self._write_all_weaves()
165
self._write_all_revs()
166
self._cleanup_spare_files()
167
self._set_new_format(BZR_BRANCH_FORMAT_5)
170
def _open_branch(self):
171
self.branch = Branch.open_downlevel(self.base)
172
self.old_format = self.branch._branch_format
173
if self.old_format == 6:
174
note('this branch is in the most current format')
176
if self.old_format not in (4, 5):
177
raise BzrError("cannot upgrade from branch format %r" %
178
self.branch._branch_format)
182
def _set_new_format(self, format):
183
self.branch.put_controlfile('branch-format', format)
186
def _cleanup_spare_files(self):
187
for n in 'merged-patches', 'pending-merged-patches':
188
p = self.branch.controlfilename(n)
189
if not os.path.exists(p):
191
## assert os.path.getsize(p) == 0
193
shutil.rmtree(self.base + '/.bzr/inventory-store')
194
shutil.rmtree(self.base + '/.bzr/text-store')
197
def _backup_control_dir(self):
198
orig = self.base + '/.bzr'
199
backup = orig + '.backup'
200
note('making backup of tree history')
201
shutil.copytree(orig, backup)
202
note('%s has been backed up to %s', orig, backup)
203
note('if conversion fails, you can move this directory back to .bzr')
204
note('if it succeeds, you can remove this directory if you wish')
207
def _convert_working_inv(self):
209
inv = serializer_v4.read_inventory(branch.controlfile('inventory', 'rb'))
210
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
211
branch.put_controlfile('inventory', new_inv_xml)
215
def _write_all_weaves(self):
216
write_a_weave(self.inv_weave, self.base + '/.bzr/inventory.weave')
219
for file_id, file_weave in self.text_weaves.items():
220
self.pb.update('writing weave', i, len(self.text_weaves))
221
write_a_weave(file_weave, self.base + '/.bzr/weaves/%s.weave' % file_id)
227
def _write_all_revs(self):
228
"""Write all revisions out in new form."""
229
shutil.rmtree(self.base + '/.bzr/revision-store')
230
os.mkdir(self.base + '/.bzr/revision-store')
232
for i, rev_id in enumerate(self.converted_revs):
233
self.pb.update('write revision', i, len(self.converted_revs))
234
f = file(self.base + '/.bzr/revision-store/%s' % rev_id, 'wb')
236
serializer_v5.write_revision(self.revisions[rev_id], f)
243
def _load_one_rev(self, rev_id):
244
"""Load a revision object into memory.
246
Any parents not either loaded or abandoned get queued to be
248
self.pb.update('loading revision',
250
len(self.known_revisions))
251
if not self.branch.revision_store.has_id(rev_id):
253
note('revision {%s} not present in branch; '
254
'will be converted as a ghost',
256
self.absent_revisions.add(rev_id)
258
rev_xml = self.branch.revision_store.get(rev_id).read()
259
rev = serializer_v4.read_revision_from_string(rev_xml)
260
for parent_id in rev.parent_ids:
261
self.known_revisions.add(parent_id)
262
self.to_read.append(parent_id)
263
self.revisions[rev_id] = rev
266
def _load_old_inventory(self, rev_id):
267
assert rev_id not in self.converted_revs
268
old_inv_xml = self.branch.inventory_store.get(rev_id).read()
269
inv = serializer_v4.read_inventory_from_string(old_inv_xml)
270
rev = self.revisions[rev_id]
271
if rev.inventory_sha1:
272
assert rev.inventory_sha1 == sha_string(old_inv_xml), \
273
'inventory sha mismatch for {%s}' % rev_id
277
def _load_updated_inventory(self, rev_id):
278
assert rev_id in self.converted_revs
279
inv_xml = self.inv_weave.get_text(rev_id)
280
inv = serializer_v5.read_inventory_from_string(inv_xml)
284
def _convert_one_rev(self, rev_id):
285
"""Convert revision and all referenced objects to new format."""
286
rev = self.revisions[rev_id]
287
inv = self._load_old_inventory(rev_id)
288
present_parents = [p for p in rev.parent_ids
289
if p not in self.absent_revisions]
290
self._convert_revision_contents(rev, inv, present_parents)
291
self._store_new_weave(rev, inv, present_parents)
292
self.converted_revs.add(rev_id)
295
def _store_new_weave(self, rev, inv, present_parents):
296
# the XML is now updated with text versions
300
if ie.kind == 'root_directory':
302
assert hasattr(ie, 'revision'), \
303
'no revision on {%s} in {%s}' % \
304
(file_id, rev.revision_id)
305
new_inv_xml = serializer_v5.write_inventory_to_string(inv)
306
new_inv_sha1 = sha_string(new_inv_xml)
307
self.inv_weave.add(rev.revision_id,
309
new_inv_xml.splitlines(True),
311
rev.inventory_sha1 = new_inv_sha1
313
def _convert_revision_contents(self, rev, inv, present_parents):
314
"""Convert all the files within a revision.
316
Also upgrade the inventory to refer to the text revision ids."""
317
rev_id = rev.revision_id
318
mutter('converting texts of revision {%s}',
320
parent_invs = map(self._load_updated_inventory, present_parents)
323
self._convert_file_version(rev, ie, parent_invs)
325
def _convert_file_version(self, rev, ie, parent_invs):
326
"""Convert one version of one file.
328
The file needs to be added into the weave if it is a merge
329
of >=2 parents or if it's changed from its parent.
331
if ie.kind == 'root_directory':
334
rev_id = rev.revision_id
335
w = self.text_weaves.get(file_id)
338
self.text_weaves[file_id] = w
340
previous_entries = ie.find_previous_heads(parent_invs, w)
341
for old_revision in previous_entries:
342
# if this fails, its a ghost ?
343
assert old_revision in self.converted_revs
344
self.snapshot_ie(previous_entries, ie, w, rev_id)
346
assert getattr(ie, 'revision', None) is not None
348
def snapshot_ie(self, previous_revisions, ie, w, rev_id):
349
# TODO: convert this logic, which is ~= snapshot to
350
# a call to:. This needs the path figured out. rather than a work_tree
351
# a v4 revision_tree can be given, or something that looks enough like
352
# one to give the file content to the entry if it needs it.
353
# and we need something that looks like a weave store for snapshot to
355
#ie.snapshot(rev, PATH, previous_revisions, REVISION_TREE, InMemoryWeaveStore(self.text_weaves))
356
if len(previous_revisions) == 1:
357
previous_ie = previous_revisions.values()[0]
358
if ie._unchanged(previous_ie):
359
ie.revision = previous_ie.revision
361
parent_indexes = map(w.lookup, previous_revisions)
363
file_lines = self.branch.text_store.get(ie.text_id).readlines()
364
assert sha_strings(file_lines) == ie.text_sha1
365
assert sum(map(len, file_lines)) == ie.text_size
366
w.add(rev_id, parent_indexes, file_lines, ie.text_sha1)
369
w.add(rev_id, parent_indexes, [], None)
371
##mutter('import text {%s} of {%s}',
372
## ie.text_id, file_id)
374
def _make_order(self):
375
"""Return a suitable order for importing revisions.
377
The order must be such that an revision is imported after all
378
its (present) parents.
380
todo = set(self.revisions.keys())
381
done = self.absent_revisions.copy()
384
# scan through looking for a revision whose parents
386
for rev_id in sorted(list(todo)):
387
rev = self.revisions[rev_id]
388
parent_ids = set(rev.parent_ids)
389
if parent_ids.issubset(done):
390
# can take this one now
397
def write_a_weave(weave, filename):
398
inv_wf = file(filename, 'wb')
400
write_weave(weave, inv_wf)
405
def upgrade(base_dir):
added
removed
bzrlib/upgrade.py
