1
# Copyright (C) 2005 Canonical Ltd
1
# Copyright (C) 2005-2011 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
# FIXME: "bzr commit doc/format" commits doc/format.txt!
21
def commit(branch, message,
28
allow_pointless=True):
29
"""Commit working copy as a new revision.
31
The basic approach is to add all the file texts into the
32
store, then the inventory, then make a new revision pointing
33
to that inventory and store that.
35
This is not quite safe if the working copy changes during the
36
commit; for the moment that is simply not allowed. A better
37
approach is to make a temporary copy of the files before
38
computing their hashes, and then add those hashes in turn to
39
the inventory. This should mean at least that there are no
40
broken hash pointers. There is no way we can get a snapshot
41
of the whole directory at an instant. This would also have to
42
be robust against files disappearing, moving, etc. So the
43
whole thing is a bit hard.
45
This raises PointlessCommit if there are no changes, no new merges,
46
and allow_pointless is false.
48
timestamp -- if not None, seconds-since-epoch for a
49
postdated/predated commit.
52
If true, commit only those files.
55
If set, use this as the new revision id.
56
Useful for test or import commands that need to tightly
57
control what revisions are assigned. If you duplicate
58
a revision id that exists elsewhere it is your own fault.
59
If null (default), a time/random revision id is generated.
64
from bzrlib.osutils import local_time_offset, username
65
from bzrlib.branch import gen_file_id
66
from bzrlib.errors import BzrError, PointlessCommit
67
from bzrlib.revision import Revision, RevisionReference
68
from bzrlib.trace import mutter, note
69
from bzrlib.xml import pack_xml
74
# First walk over the working inventory; and both update that
75
# and also build a new revision inventory. The revision
76
# inventory needs to hold the text-id, sha1 and size of the
77
# actual file versions committed in the revision. (These are
78
# not present in the working inventory.) We also need to
79
# detect missing/deleted files, and remove them from the
82
work_tree = branch.working_tree()
83
work_inv = work_tree.inventory
84
basis = branch.basis_tree()
85
basis_inv = basis.inventory
88
note('looking for changes...')
90
pending_merges = branch.pending_merges()
92
missing_ids, new_inv, any_changes = \
93
_gather_commit(branch,
100
if not (any_changes or allow_pointless or pending_merges):
101
raise PointlessCommit()
103
for file_id in missing_ids:
104
# Any files that have been deleted are now removed from the
105
# working inventory. Files that were not selected for commit
106
# are left as they were in the working inventory and ommitted
107
# from the revision inventory.
109
# have to do this later so we don't mess up the iterator.
110
# since parents may be removed before their children we
113
# FIXME: There's probably a better way to do this; perhaps
114
# the workingtree should know how to filter itbranch.
115
if work_inv.has_id(file_id):
116
del work_inv[file_id]
120
rev_id = _gen_revision_id(time.time())
123
inv_tmp = tempfile.TemporaryFile()
124
pack_xml(new_inv, inv_tmp)
126
branch.inventory_store.add(inv_tmp, inv_id)
127
mutter('new inventory_id is {%s}' % inv_id)
129
# We could also just sha hash the inv_tmp file
130
# however, in the case that branch.inventory_store.add()
131
# ever actually does anything special
132
inv_sha1 = branch.get_inventory_sha1(inv_id)
134
branch._write_inventory(work_inv)
136
if timestamp == None:
137
timestamp = time.time()
139
if committer == None:
140
committer = username()
143
timezone = local_time_offset()
145
mutter("building commit log message")
146
rev = Revision(timestamp=timestamp,
151
inventory_sha1=inv_sha1,
155
precursor_id = branch.last_patch()
157
precursor_sha1 = branch.get_revision_sha1(precursor_id)
158
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
159
for merge_rev in pending_merges:
160
rev.parents.append(RevisionReference(merge_rev))
162
rev_tmp = tempfile.TemporaryFile()
163
pack_xml(rev, rev_tmp)
165
branch.revision_store.add(rev_tmp, rev_id)
166
mutter("new revision_id is {%s}" % rev_id)
168
## XXX: Everything up to here can simply be orphaned if we abort
169
## the commit; it will leave junk files behind but that doesn't
172
## TODO: Read back the just-generated changeset, and make sure it
173
## applies and recreates the right state.
175
## TODO: Also calculate and store the inventory SHA1
176
mutter("committing patch r%d" % (branch.revno() + 1))
178
branch.append_revision(rev_id)
180
branch.set_pending_merges([])
183
note("commited r%d" % branch.revno())
189
def _gen_revision_id(when):
190
"""Return new revision-id."""
191
from binascii import hexlify
192
from osutils import rand_bytes, compact_date, user_email
194
s = '%s-%s-' % (user_email(), compact_date(when))
195
s += hexlify(rand_bytes(8))
199
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
201
"""Build inventory preparatory to commit.
203
Returns missing_ids, new_inv, any_changes.
205
This adds any changed files into the text store, and sets their
206
test-id, sha and size in the returned inventory appropriately.
209
Modified to hold a list of files that have been deleted from
210
the working directory; these should be removed from the
213
from bzrlib.inventory import Inventory
214
from bzrlib.osutils import isdir, isfile, sha_string, quotefn, \
215
local_time_offset, username, kind_marker, is_inside_any
217
from bzrlib.branch import gen_file_id
218
from bzrlib.errors import BzrError
219
from bzrlib.revision import Revision
220
from bzrlib.trace import mutter, note
223
inv = Inventory(work_inv.root.file_id)
226
for path, entry in work_inv.iter_entries():
227
## TODO: Check that the file kind has not changed from the previous
228
## revision of this file (if any).
230
p = branch.abspath(path)
231
file_id = entry.file_id
232
mutter('commit prep file %s, id %r ' % (p, file_id))
234
if specific_files and not is_inside_any(specific_files, path):
235
mutter(' skipping file excluded from commit')
236
if basis_inv.has_id(file_id):
237
# carry over with previous state
238
inv.add(basis_inv[file_id].copy())
240
# omit this from committed inventory
244
if not work_tree.has_id(file_id):
246
print('deleted %s%s' % (path, kind_marker(entry.kind)))
248
mutter(" file is missing, removing from inventory")
249
missing_ids.append(file_id)
252
# this is present in the new inventory; may be new, modified or
254
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
260
old_kind = old_ie.kind
261
if old_kind != entry.kind:
262
raise BzrError("entry %r changed kind from %r to %r"
263
% (file_id, old_kind, entry.kind))
265
if entry.kind == 'directory':
267
raise BzrError("%s is entered as directory but not a directory"
269
elif entry.kind == 'file':
271
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
273
new_sha1 = work_tree.get_file_sha1(file_id)
276
and old_ie.text_sha1 == new_sha1):
277
## assert content == basis.get_file(file_id).read()
278
entry.text_id = old_ie.text_id
279
entry.text_sha1 = new_sha1
280
entry.text_size = old_ie.text_size
281
mutter(' unchanged from previous text_id {%s}' %
284
content = file(p, 'rb').read()
286
# calculate the sha again, just in case the file contents
287
# changed since we updated the cache
288
entry.text_sha1 = sha_string(content)
289
entry.text_size = len(content)
291
entry.text_id = gen_file_id(entry.name)
292
branch.text_store.add(content, entry.text_id)
293
mutter(' stored with text_id {%s}' % entry.text_id)
296
marked = path + kind_marker(entry.kind)
298
print 'added', marked
300
elif old_ie == entry:
302
elif (old_ie.name == entry.name
303
and old_ie.parent_id == entry.parent_id):
304
print 'modified', marked
307
print 'renamed', marked
310
return missing_ids, inv, any_changes
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
# The newly committed revision is going to have a shape corresponding
19
# to that of the working tree. Files that are not in the
20
# working tree and that were in the predecessor are reported as
21
# removed --- this can include files that were either removed from the
22
# inventory or deleted in the working tree. If they were only
23
# deleted from disk, they are removed from the working inventory.
25
# We then consider the remaining entries, which will be in the new
26
# version. Directory entries are simply copied across. File entries
27
# must be checked to see if a new version of the file should be
28
# recorded. For each parent revision tree, we check to see what
29
# version of the file was present. If the file was present in at
30
# least one tree, and if it was the same version in all the trees,
31
# then we can just refer to that version. Otherwise, a new version
32
# representing the merger of the file versions must be added.
34
# TODO: Update hashcache before and after - or does the WorkingTree
37
# TODO: Rather than mashing together the ancestry and storing it back,
38
# perhaps the weave should have single method which does it all in one
39
# go, avoiding a lot of redundant work.
41
# TODO: Perhaps give a warning if one of the revisions marked as
42
# merged is already in the ancestry, and then don't record it as a
45
# TODO: If the file is newly merged but unchanged from the version it
46
# merges from, then it should still be reported as newly added
47
# relative to the basis revision.
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
# the rest of the code; add a deprecation of the old name.
59
from bzrlib.branch import Branch
60
from bzrlib.cleanup import OperationWithCleanups
62
from bzrlib.errors import (BzrError, PointlessCommit,
66
from bzrlib.osutils import (get_user_encoding,
68
minimum_path_selection,
71
from bzrlib.trace import mutter, note, is_quiet
72
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
73
from bzrlib import symbol_versioning
74
from bzrlib.urlutils import unescape_for_display
75
from bzrlib.i18n import gettext
77
class NullCommitReporter(object):
78
"""I report on progress of a commit."""
80
def started(self, revno, revid, location=None):
82
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
83
"to started.", DeprecationWarning,
87
def snapshot_change(self, change, path):
90
def completed(self, revno, rev_id):
93
def deleted(self, path):
96
def missing(self, path):
99
def renamed(self, change, old_path, new_path):
102
def is_verbose(self):
106
class ReportCommitToLog(NullCommitReporter):
108
def _note(self, format, *args):
111
Subclasses may choose to override this method.
115
def snapshot_change(self, change, path):
116
if path == '' and change in (gettext('added'), gettext('modified')):
118
self._note("%s %s", change, path)
120
def started(self, revno, rev_id, location=None):
121
if location is not None:
122
location = ' to: ' + unescape_for_display(location, 'utf-8')
124
# When started was added, location was only made optional by
125
# accident. Matt Nordhoff 20071129
126
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
127
"to started.", DeprecationWarning,
130
self._note(gettext('Committing%s'), location)
132
def completed(self, revno, rev_id):
133
self._note(gettext('Committed revision %d.'), revno)
134
# self._note goes to the console too; so while we want to log the
135
# rev_id, we can't trivially only log it. (See bug 526425). Long
136
# term we should rearrange the reporting structure, but for now
137
# we just mutter seperately. We mutter the revid and revno together
138
# so that concurrent bzr invocations won't lead to confusion.
139
mutter('Committed revid %s as revno %d.', rev_id, revno)
141
def deleted(self, path):
142
self._note(gettext('deleted %s'), path)
144
def missing(self, path):
145
self._note(gettext('missing %s'), path)
147
def renamed(self, change, old_path, new_path):
148
self._note('%s %s => %s', change, old_path, new_path)
150
def is_verbose(self):
154
class Commit(object):
155
"""Task of committing a new revision.
157
This is a MethodObject: it accumulates state as the commit is
158
prepared, and then it is discarded. It doesn't represent
159
historical revisions, just the act of recording a new one.
162
Modified to hold a list of files that have been deleted from
163
the working directory; these should be removed from the
169
"""Create a Commit object.
171
:param reporter: the default reporter to use or None to decide later
173
self.reporter = reporter
177
def update_revprops(revprops, branch, authors=None, author=None,
178
local=False, possible_master_transports=None):
181
if possible_master_transports is None:
182
possible_master_transports = []
183
if not 'branch-nick' in revprops:
184
revprops['branch-nick'] = branch._get_nick(
186
possible_master_transports)
187
if authors is not None:
188
if author is not None:
189
raise AssertionError('Specifying both author and authors '
190
'is not allowed. Specify just authors instead')
191
if 'author' in revprops or 'authors' in revprops:
192
# XXX: maybe we should just accept one of them?
193
raise AssertionError('author property given twice')
195
for individual in authors:
196
if '\n' in individual:
197
raise AssertionError('\\n is not a valid character '
198
'in an author identity')
199
revprops['authors'] = '\n'.join(authors)
200
if author is not None:
201
symbol_versioning.warn('The parameter author was deprecated'
202
' in version 1.13. Use authors instead',
204
if 'author' in revprops or 'authors' in revprops:
205
# XXX: maybe we should just accept one of them?
206
raise AssertionError('author property given twice')
208
raise AssertionError('\\n is not a valid character '
209
'in an author identity')
210
revprops['authors'] = author
220
allow_pointless=True,
228
message_callback=None,
231
possible_master_transports=None,
233
"""Commit working copy as a new revision.
235
:param message: the commit message (it or message_callback is required)
236
:param message_callback: A callback: message = message_callback(cmt_obj)
238
:param timestamp: if not None, seconds-since-epoch for a
239
postdated/predated commit.
241
:param specific_files: If not None, commit only those files. An empty
242
list means 'commit no files'.
244
:param rev_id: If set, use this as the new revision id.
245
Useful for test or import commands that need to tightly
246
control what revisions are assigned. If you duplicate
247
a revision id that exists elsewhere it is your own fault.
248
If null (default), a time/random revision id is generated.
250
:param allow_pointless: If true (default), commit even if nothing
251
has changed and no merges are recorded.
253
:param strict: If true, don't allow a commit if the working tree
254
contains unknown files.
256
:param revprops: Properties for new revision
257
:param local: Perform a local only commit.
258
:param reporter: the reporter to use or None for the default
259
:param verbose: if True and the reporter is not None, report everything
260
:param recursive: If set to 'down', commit in any subtrees that have
261
pending changes of any sort during this commit.
262
:param exclude: None or a list of relative paths to exclude from the
263
commit. Pending changes to excluded files will be ignored by the
265
:param lossy: When committing to a foreign VCS, ignore any
266
data that can not be natively represented.
268
operation = OperationWithCleanups(self._commit)
269
self.revprops = revprops or {}
270
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
271
self.config = config or self.config
272
return operation.run(
277
specific_files=specific_files,
279
allow_pointless=allow_pointless,
282
working_tree=working_tree,
285
message_callback=message_callback,
288
possible_master_transports=possible_master_transports,
291
def _commit(self, operation, message, timestamp, timezone, committer,
292
specific_files, rev_id, allow_pointless, strict, verbose,
293
working_tree, local, reporter, message_callback, recursive,
294
exclude, possible_master_transports, lossy):
295
mutter('preparing to commit')
297
if working_tree is None:
298
raise BzrError("working_tree must be passed into commit().")
300
self.work_tree = working_tree
301
self.branch = self.work_tree.branch
302
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
303
if not self.branch.repository.supports_rich_root():
304
raise errors.RootNotRich()
305
if message_callback is None:
306
if message is not None:
307
if isinstance(message, str):
308
message = message.decode(get_user_encoding())
309
message_callback = lambda x: message
311
raise BzrError("The message or message_callback keyword"
312
" parameter is required for commit().")
314
self.bound_branch = None
315
self.any_entries_deleted = False
316
if exclude is not None:
317
self.exclude = sorted(
318
minimum_path_selection(exclude))
322
self.master_branch = None
323
self.recursive = recursive
325
# self.specific_files is None to indicate no filter, or any iterable to
326
# indicate a filter - [] means no files at all, as per iter_changes.
327
if specific_files is not None:
328
self.specific_files = sorted(
329
minimum_path_selection(specific_files))
331
self.specific_files = None
333
self.allow_pointless = allow_pointless
334
self.message_callback = message_callback
335
self.timestamp = timestamp
336
self.timezone = timezone
337
self.committer = committer
339
self.verbose = verbose
341
self.work_tree.lock_write()
342
operation.add_cleanup(self.work_tree.unlock)
343
self.parents = self.work_tree.get_parent_ids()
344
# We can use record_iter_changes IFF iter_changes is compatible with
345
# the command line parameters, and the repository has fast delta
346
# generation. See bug 347649.
347
self.use_record_iter_changes = (
349
not self.branch.repository._format.supports_tree_reference and
350
(self.branch.repository._format.fast_deltas or
351
len(self.parents) < 2))
352
self.pb = ui.ui_factory.nested_progress_bar()
353
operation.add_cleanup(self.pb.finished)
354
self.basis_revid = self.work_tree.last_revision()
355
self.basis_tree = self.work_tree.basis_tree()
356
self.basis_tree.lock_read()
357
operation.add_cleanup(self.basis_tree.unlock)
358
# Cannot commit with conflicts present.
359
if len(self.work_tree.conflicts()) > 0:
360
raise ConflictsInTree
362
# Setup the bound branch variables as needed.
363
self._check_bound_branch(operation, possible_master_transports)
365
# Check that the working tree is up to date
366
old_revno, old_revid, new_revno = self._check_out_of_date_tree()
368
# Complete configuration setup
369
if reporter is not None:
370
self.reporter = reporter
371
elif self.reporter is None:
372
self.reporter = self._select_reporter()
373
if self.config is None:
374
self.config = self.branch.get_config()
376
self._set_specific_file_ids()
378
# Setup the progress bar. As the number of files that need to be
379
# committed in unknown, progress is reported as stages.
380
# We keep track of entries separately though and include that
381
# information in the progress bar during the relevant stages.
382
self.pb_stage_name = ""
383
self.pb_stage_count = 0
384
self.pb_stage_total = 5
385
if self.bound_branch:
386
# 2 extra stages: "Uploading data to master branch" and "Merging
387
# tags to master branch"
388
self.pb_stage_total += 2
389
self.pb.show_pct = False
390
self.pb.show_spinner = False
391
self.pb.show_eta = False
392
self.pb.show_count = True
393
self.pb.show_bar = True
395
self._gather_parents()
396
# After a merge, a selected file commit is not supported.
397
# See 'bzr help merge' for an explanation as to why.
398
if len(self.parents) > 1 and self.specific_files is not None:
399
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
400
# Excludes are a form of selected file commit.
401
if len(self.parents) > 1 and self.exclude:
402
raise errors.CannotCommitSelectedFileMerge(self.exclude)
404
# Collect the changes
405
self._set_progress_stage("Collecting changes", counter=True)
407
self.builder = self.branch.get_commit_builder(self.parents,
408
self.config, timestamp, timezone, committer, self.revprops,
410
if not self.builder.supports_record_entry_contents and self.exclude:
412
raise errors.ExcludesUnsupported(self.branch.repository)
414
if self.builder.updates_branch and self.bound_branch:
416
raise AssertionError(
417
"bound branches not supported for commit builders "
418
"that update the branch")
421
self.builder.will_record_deletes()
422
# find the location being committed to
423
if self.bound_branch:
424
master_location = self.master_branch.base
426
master_location = self.branch.base
428
# report the start of the commit
429
self.reporter.started(new_revno, self.rev_id, master_location)
431
self._update_builder_with_changes()
432
self._check_pointless()
434
# TODO: Now the new inventory is known, check for conflicts.
435
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
436
# weave lines, because nothing should be recorded until it is known
437
# that commit will succeed.
438
self._set_progress_stage("Saving data locally")
439
self.builder.finish_inventory()
441
# Prompt the user for a commit message if none provided
442
message = message_callback(self)
443
self.message = message
445
# Add revision data to the local branch
446
self.rev_id = self.builder.commit(self.message)
449
mutter("aborting commit write group because of exception:")
450
trace.log_exception_quietly()
454
self._update_branches(old_revno, old_revid, new_revno)
456
# Make the working tree be up to date with the branch. This
457
# includes automatic changes scheduled to be made to the tree, such
458
# as updating its basis and unversioning paths that were missing.
459
self.work_tree.unversion(self.deleted_ids)
460
self._set_progress_stage("Updating the working tree")
461
self.work_tree.update_basis_by_delta(self.rev_id,
462
self.builder.get_basis_delta())
463
self.reporter.completed(new_revno, self.rev_id)
464
self._process_post_hooks(old_revno, new_revno)
467
def _update_branches(self, old_revno, old_revid, new_revno):
468
"""Update the master and local branch to the new revision.
470
This will try to make sure that the master branch is updated
471
before the local branch.
473
:param old_revno: Revision number of master branch before the
475
:param old_revid: Tip of master branch before the commit
476
:param new_revno: Revision number of the new commit
478
if not self.builder.updates_branch:
479
self._process_pre_hooks(old_revno, new_revno)
481
# Upload revision data to the master.
482
# this will propagate merged revisions too if needed.
483
if self.bound_branch:
484
self._set_progress_stage("Uploading data to master branch")
485
# 'commit' to the master first so a timeout here causes the
486
# local branch to be out of date
487
(new_revno, self.rev_id) = self.master_branch.import_last_revision_info_and_tags(
488
self.branch, new_revno, self.rev_id, lossy=self._lossy)
490
self.branch.fetch(self.master_branch, self.rev_id)
492
# and now do the commit locally.
493
self.branch.set_last_revision_info(new_revno, self.rev_id)
496
self._process_pre_hooks(old_revno, new_revno)
498
# The commit builder will already have updated the branch,
500
self.branch.set_last_revision_info(old_revno, old_revid)
503
# Merge local tags to remote
504
if self.bound_branch:
505
self._set_progress_stage("Merging tags to master branch")
506
tag_updates, tag_conflicts = self.branch.tags.merge_to(
507
self.master_branch.tags)
509
warning_lines = [' ' + name for name, _, _ in tag_conflicts]
510
note( gettext("Conflicting tags in bound branch:\n{0}".format(
511
"\n".join(warning_lines))) )
513
def _select_reporter(self):
514
"""Select the CommitReporter to use."""
516
return NullCommitReporter()
517
return ReportCommitToLog()
519
def _check_pointless(self):
520
if self.allow_pointless:
522
# A merge with no effect on files
523
if len(self.parents) > 1:
525
if self.builder.any_changes():
527
raise PointlessCommit()
529
def _check_bound_branch(self, operation, possible_master_transports=None):
530
"""Check to see if the local branch is bound.
532
If it is bound, then most of the commit will actually be
533
done using the remote branch as the target branch.
534
Only at the end will the local branch be updated.
536
if self.local and not self.branch.get_bound_location():
537
raise errors.LocalRequiresBoundBranch()
540
self.master_branch = self.branch.get_master_branch(
541
possible_master_transports)
543
if not self.master_branch:
544
# make this branch the reference branch for out of date checks.
545
self.master_branch = self.branch
548
# If the master branch is bound, we must fail
549
master_bound_location = self.master_branch.get_bound_location()
550
if master_bound_location:
551
raise errors.CommitToDoubleBoundBranch(self.branch,
552
self.master_branch, master_bound_location)
554
# TODO: jam 20051230 We could automatically push local
555
# commits to the remote branch if they would fit.
556
# But for now, just require remote to be identical
559
# Make sure the local branch is identical to the master
560
master_info = self.master_branch.last_revision_info()
561
local_info = self.branch.last_revision_info()
562
if local_info != master_info:
563
raise errors.BoundBranchOutOfDate(self.branch,
566
# Now things are ready to change the master branch
568
self.bound_branch = self.branch
569
self.master_branch.lock_write()
570
operation.add_cleanup(self.master_branch.unlock)
572
def _check_out_of_date_tree(self):
573
"""Check that the working tree is up to date.
575
:return: old_revision_number, old_revision_id, new_revision_number
579
first_tree_parent = self.work_tree.get_parent_ids()[0]
581
# if there are no parents, treat our parent as 'None'
582
# this is so that we still consider the master branch
583
# - in a checkout scenario the tree may have no
584
# parents but the branch may do.
585
first_tree_parent = bzrlib.revision.NULL_REVISION
586
old_revno, master_last = self.master_branch.last_revision_info()
587
if master_last != first_tree_parent:
588
if master_last != bzrlib.revision.NULL_REVISION:
589
raise errors.OutOfDateTree(self.work_tree)
590
if self.branch.repository.has_revision(first_tree_parent):
591
new_revno = old_revno + 1
593
# ghost parents never appear in revision history.
595
return old_revno, master_last, new_revno
597
def _process_pre_hooks(self, old_revno, new_revno):
598
"""Process any registered pre commit hooks."""
599
self._set_progress_stage("Running pre_commit hooks")
600
self._process_hooks("pre_commit", old_revno, new_revno)
602
def _process_post_hooks(self, old_revno, new_revno):
603
"""Process any registered post commit hooks."""
604
# Process the post commit hooks, if any
605
self._set_progress_stage("Running post_commit hooks")
606
# old style commit hooks - should be deprecated ? (obsoleted in
608
if self.config.post_commit() is not None:
609
hooks = self.config.post_commit().split(' ')
610
# this would be nicer with twisted.python.reflect.namedAny
612
result = eval(hook + '(branch, rev_id)',
613
{'branch':self.branch,
615
'rev_id':self.rev_id})
616
# process new style post commit hooks
617
self._process_hooks("post_commit", old_revno, new_revno)
619
def _process_hooks(self, hook_name, old_revno, new_revno):
620
if not Branch.hooks[hook_name]:
623
# new style commit hooks:
624
if not self.bound_branch:
625
hook_master = self.branch
628
hook_master = self.master_branch
629
hook_local = self.branch
630
# With bound branches, when the master is behind the local branch,
631
# the 'old_revno' and old_revid values here are incorrect.
632
# XXX: FIXME ^. RBC 20060206
634
old_revid = self.parents[0]
636
old_revid = bzrlib.revision.NULL_REVISION
638
if hook_name == "pre_commit":
639
future_tree = self.builder.revision_tree()
640
tree_delta = future_tree.changes_from(self.basis_tree,
643
for hook in Branch.hooks[hook_name]:
644
# show the running hook in the progress bar. As hooks may
645
# end up doing nothing (e.g. because they are not configured by
646
# the user) this is still showing progress, not showing overall
647
# actions - its up to each plugin to show a UI if it want's to
648
# (such as 'Emailing diff to foo@example.com').
649
self.pb_stage_name = "Running %s hooks [%s]" % \
650
(hook_name, Branch.hooks.get_hook_name(hook))
651
self._emit_progress()
652
if 'hooks' in debug.debug_flags:
653
mutter("Invoking commit hook: %r", hook)
654
if hook_name == "post_commit":
655
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
657
elif hook_name == "pre_commit":
658
hook(hook_local, hook_master,
659
old_revno, old_revid, new_revno, self.rev_id,
660
tree_delta, future_tree)
662
def _gather_parents(self):
663
"""Record the parents of a merge for merge detection."""
664
# TODO: Make sure that this list doesn't contain duplicate
665
# entries and the order is preserved when doing this.
666
if self.use_record_iter_changes:
668
self.basis_inv = self.basis_tree.inventory
669
self.parent_invs = [self.basis_inv]
670
for revision in self.parents[1:]:
671
if self.branch.repository.has_revision(revision):
672
mutter('commit parent revision {%s}', revision)
673
inventory = self.branch.repository.get_inventory(revision)
674
self.parent_invs.append(inventory)
676
mutter('commit parent ghost revision {%s}', revision)
678
def _update_builder_with_changes(self):
679
"""Update the commit builder with the data about what has changed.
681
exclude = self.exclude
682
specific_files = self.specific_files
683
mutter("Selecting files for commit with filter %s", specific_files)
686
if self.use_record_iter_changes:
687
iter_changes = self.work_tree.iter_changes(self.basis_tree,
688
specific_files=specific_files)
689
iter_changes = self._filter_iter_changes(iter_changes)
690
for file_id, path, fs_hash in self.builder.record_iter_changes(
691
self.work_tree, self.basis_revid, iter_changes):
692
self.work_tree._observed_sha1(file_id, path, fs_hash)
694
# Build the new inventory
695
self._populate_from_inventory()
696
self._record_unselected()
697
self._report_and_accumulate_deletes()
699
def _filter_iter_changes(self, iter_changes):
700
"""Process iter_changes.
702
This method reports on the changes in iter_changes to the user, and
703
converts 'missing' entries in the iter_changes iterator to 'deleted'
704
entries. 'missing' entries have their
706
:param iter_changes: An iter_changes to process.
707
:return: A generator of changes.
709
reporter = self.reporter
710
report_changes = reporter.is_verbose()
712
for change in iter_changes:
714
old_path = change[1][0]
715
new_path = change[1][1]
716
versioned = change[3][1]
718
versioned = change[3][1]
719
if kind is None and versioned:
722
reporter.missing(new_path)
723
deleted_ids.append(change[0])
724
# Reset the new path (None) and new versioned flag (False)
725
change = (change[0], (change[1][0], None), change[2],
726
(change[3][0], False)) + change[4:]
727
new_path = change[1][1]
729
elif kind == 'tree-reference':
730
if self.recursive == 'down':
731
self._commit_nested_tree(change[0], change[1][1])
732
if change[3][0] or change[3][1]:
736
reporter.deleted(old_path)
737
elif old_path is None:
738
reporter.snapshot_change(gettext('added'), new_path)
739
elif old_path != new_path:
740
reporter.renamed(gettext('renamed'), old_path, new_path)
743
self.work_tree.branch.repository._format.rich_root_data):
744
# Don't report on changes to '' in non rich root
746
reporter.snapshot_change(gettext('modified'), new_path)
747
self._next_progress_entry()
748
# Unversion IDs that were found to be deleted
749
self.deleted_ids = deleted_ids
751
def _record_unselected(self):
752
# If specific files are selected, then all un-selected files must be
753
# recorded in their previous state. For more details, see
754
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
755
if self.specific_files or self.exclude:
756
specific_files = self.specific_files or []
757
for path, old_ie in self.basis_inv.iter_entries():
758
if self.builder.new_inventory.has_id(old_ie.file_id):
759
# already added - skip.
761
if (is_inside_any(specific_files, path)
762
and not is_inside_any(self.exclude, path)):
763
# was inside the selected path, and not excluded - if not
764
# present it has been deleted so skip.
766
# From here down it was either not selected, or was excluded:
767
# We preserve the entry unaltered.
769
# Note: specific file commits after a merge are currently
770
# prohibited. This test is for sanity/safety in case it's
771
# required after that changes.
772
if len(self.parents) > 1:
774
self.builder.record_entry_contents(ie, self.parent_invs, path,
775
self.basis_tree, None)
777
def _report_and_accumulate_deletes(self):
778
if (isinstance(self.basis_inv, Inventory)
779
and isinstance(self.builder.new_inventory, Inventory)):
780
# the older Inventory classes provide a _byid dict, and building a
781
# set from the keys of this dict is substantially faster than even
782
# getting a set of ids from the inventory
784
# <lifeless> set(dict) is roughly the same speed as
785
# set(iter(dict)) and both are significantly slower than
787
deleted_ids = set(self.basis_inv._byid.keys()) - \
788
set(self.builder.new_inventory._byid.keys())
790
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
792
self.any_entries_deleted = True
793
deleted = [(self.basis_tree.id2path(file_id), file_id)
794
for file_id in deleted_ids]
796
# XXX: this is not quite directory-order sorting
797
for path, file_id in deleted:
798
self.builder.record_delete(path, file_id)
799
self.reporter.deleted(path)
801
def _check_strict(self):
802
# XXX: when we use iter_changes this would likely be faster if
803
# iter_changes would check for us (even in the presence of
806
# raise an exception as soon as we find a single unknown.
807
for unknown in self.work_tree.unknowns():
808
raise StrictCommitFailed()
810
def _populate_from_inventory(self):
811
"""Populate the CommitBuilder by walking the working tree inventory."""
812
# Build the revision inventory.
814
# This starts by creating a new empty inventory. Depending on
815
# which files are selected for commit, and what is present in the
816
# current tree, the new inventory is populated. inventory entries
817
# which are candidates for modification have their revision set to
818
# None; inventory entries that are carried over untouched have their
819
# revision set to their prior value.
821
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
822
# results to create a new inventory at the same time, which results
823
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
826
specific_files = self.specific_files
827
exclude = self.exclude
828
report_changes = self.reporter.is_verbose()
830
# A tree of paths that have been deleted. E.g. if foo/bar has been
831
# deleted, then we have {'foo':{'bar':{}}}
833
# XXX: Note that entries may have the wrong kind because the entry does
834
# not reflect the status on disk.
835
work_inv = self.work_tree.inventory
836
# NB: entries will include entries within the excluded ids/paths
837
# because iter_entries_by_dir has no 'exclude' facility today.
838
entries = work_inv.iter_entries_by_dir(
839
specific_file_ids=self.specific_file_ids, yield_parents=True)
840
for path, existing_ie in entries:
841
file_id = existing_ie.file_id
842
name = existing_ie.name
843
parent_id = existing_ie.parent_id
844
kind = existing_ie.kind
845
# Skip files that have been deleted from the working tree.
846
# The deleted path ids are also recorded so they can be explicitly
849
path_segments = splitpath(path)
850
deleted_dict = deleted_paths
851
for segment in path_segments:
852
deleted_dict = deleted_dict.get(segment, None)
854
# We either took a path not present in the dict
855
# (deleted_dict was None), or we've reached an empty
856
# child dir in the dict, so are now a sub-path.
860
if deleted_dict is not None:
861
# the path has a deleted parent, do not add it.
863
if exclude and is_inside_any(exclude, path):
864
# Skip excluded paths. Excluded paths are processed by
865
# _update_builder_with_changes.
867
content_summary = self.work_tree.path_content_summary(path)
868
kind = content_summary[0]
869
# Note that when a filter of specific files is given, we must only
870
# skip/record deleted files matching that filter.
871
if not specific_files or is_inside_any(specific_files, path):
872
if kind == 'missing':
873
if not deleted_paths:
874
# path won't have been split yet.
875
path_segments = splitpath(path)
876
deleted_dict = deleted_paths
877
for segment in path_segments:
878
deleted_dict = deleted_dict.setdefault(segment, {})
879
self.reporter.missing(path)
880
self._next_progress_entry()
881
deleted_ids.append(file_id)
883
# TODO: have the builder do the nested commit just-in-time IF and
885
if kind == 'tree-reference':
886
# enforce repository nested tree policy.
887
if (not self.work_tree.supports_tree_reference() or
888
# repository does not support it either.
889
not self.branch.repository._format.supports_tree_reference):
891
content_summary = (kind, None, None, None)
892
elif self.recursive == 'down':
893
nested_revision_id = self._commit_nested_tree(
895
content_summary = (kind, None, None, nested_revision_id)
897
nested_revision_id = self.work_tree.get_reference_revision(file_id)
898
content_summary = (kind, None, None, nested_revision_id)
900
# Record an entry for this item
901
# Note: I don't particularly want to have the existing_ie
902
# parameter but the test suite currently (28-Jun-07) breaks
903
# without it thanks to a unicode normalisation issue. :-(
904
definitely_changed = kind != existing_ie.kind
905
self._record_entry(path, file_id, specific_files, kind, name,
906
parent_id, definitely_changed, existing_ie, report_changes,
909
# Unversion IDs that were found to be deleted
910
self.deleted_ids = deleted_ids
912
def _commit_nested_tree(self, file_id, path):
913
"Commit a nested tree."
914
sub_tree = self.work_tree.get_nested_tree(file_id, path)
915
# FIXME: be more comprehensive here:
916
# this works when both trees are in --trees repository,
917
# but when both are bound to a different repository,
918
# it fails; a better way of approaching this is to
919
# finally implement the explicit-caches approach design
920
# a while back - RBC 20070306.
921
if sub_tree.branch.repository.has_same_location(
922
self.work_tree.branch.repository):
923
sub_tree.branch.repository = \
924
self.work_tree.branch.repository
926
return sub_tree.commit(message=None, revprops=self.revprops,
927
recursive=self.recursive,
928
message_callback=self.message_callback,
929
timestamp=self.timestamp, timezone=self.timezone,
930
committer=self.committer,
931
allow_pointless=self.allow_pointless,
932
strict=self.strict, verbose=self.verbose,
933
local=self.local, reporter=self.reporter)
934
except errors.PointlessCommit:
935
return self.work_tree.get_reference_revision(file_id)
937
def _record_entry(self, path, file_id, specific_files, kind, name,
938
parent_id, definitely_changed, existing_ie, report_changes,
940
"Record the new inventory entry for a path if any."
941
# mutter('check %s {%s}', path, file_id)
942
# mutter('%s selected for commit', path)
943
if definitely_changed or existing_ie is None:
944
ie = make_entry(kind, name, parent_id, file_id)
946
ie = existing_ie.copy()
948
# For carried over entries we don't care about the fs hash - the repo
949
# isn't generating a sha, so we're not saving computation time.
950
_, _, fs_hash = self.builder.record_entry_contents(
951
ie, self.parent_invs, path, self.work_tree, content_summary)
953
self._report_change(ie, path)
955
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
958
def _report_change(self, ie, path):
959
"""Report a change to the user.
961
The change that has occurred is described relative to the basis
964
if (self.basis_inv.has_id(ie.file_id)):
965
basis_ie = self.basis_inv[ie.file_id]
968
change = ie.describe_change(basis_ie, ie)
969
if change in (InventoryEntry.RENAMED,
970
InventoryEntry.MODIFIED_AND_RENAMED):
971
old_path = self.basis_inv.id2path(ie.file_id)
972
self.reporter.renamed(change, old_path, path)
973
self._next_progress_entry()
975
if change == gettext('unchanged'):
977
self.reporter.snapshot_change(change, path)
978
self._next_progress_entry()
980
def _set_progress_stage(self, name, counter=False):
981
"""Set the progress stage and emit an update to the progress bar."""
982
self.pb_stage_name = name
983
self.pb_stage_count += 1
985
self.pb_entries_count = 0
987
self.pb_entries_count = None
988
self._emit_progress()
990
def _next_progress_entry(self):
991
"""Emit an update to the progress bar and increment the entry count."""
992
self.pb_entries_count += 1
993
self._emit_progress()
995
def _emit_progress(self):
996
if self.pb_entries_count is not None:
997
text = gettext("{0} [{1}] - Stage").format(self.pb_stage_name,
998
self.pb_entries_count)
1000
text = gettext("%s - Stage") % (self.pb_stage_name, )
1001
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
1003
def _set_specific_file_ids(self):
1004
"""populate self.specific_file_ids if we will use it."""
1005
if not self.use_record_iter_changes:
1006
# If provided, ensure the specified files are versioned
1007
if self.specific_files is not None:
1008
# Note: This routine is being called because it raises
1009
# PathNotVersionedError as a side effect of finding the IDs. We
1010
# later use the ids we found as input to the working tree
1011
# inventory iterator, so we only consider those ids rather than
1012
# examining the whole tree again.
1013
# XXX: Dont we have filter_unversioned to do this more
1015
self.specific_file_ids = tree.find_ids_across_trees(
1016
self.specific_files, [self.basis_tree, self.work_tree])
1018
self.specific_file_ids = None
added
removed
bzrlib/commit.py
