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.
60
from bzrlib.branch import Branch
61
from bzrlib.cleanup import OperationWithCleanups
63
from bzrlib.errors import (BzrError, PointlessCommit,
67
from bzrlib.osutils import (get_user_encoding,
69
minimum_path_selection,
72
from bzrlib.trace import mutter, note, is_quiet
73
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
74
from bzrlib import symbol_versioning
75
from bzrlib.urlutils import unescape_for_display
78
class NullCommitReporter(object):
79
"""I report on progress of a commit."""
81
def started(self, revno, revid, location=None):
83
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
84
"to started.", DeprecationWarning,
88
def snapshot_change(self, change, path):
91
def completed(self, revno, rev_id):
94
def deleted(self, path):
97
def missing(self, path):
100
def renamed(self, change, old_path, new_path):
103
def is_verbose(self):
107
class ReportCommitToLog(NullCommitReporter):
109
def _note(self, format, *args):
112
Subclasses may choose to override this method.
116
def snapshot_change(self, change, path):
117
if path == '' and change in ('added', 'modified'):
119
self._note("%s %s", change, path)
121
def started(self, revno, rev_id, location=None):
122
if location is not None:
123
location = ' to: ' + unescape_for_display(location, 'utf-8')
125
# When started was added, location was only made optional by
126
# accident. Matt Nordhoff 20071129
127
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
128
"to started.", DeprecationWarning,
131
self._note('Committing%s', location)
133
def completed(self, revno, rev_id):
134
self._note('Committed revision %d.', revno)
135
# self._note goes to the console too; so while we want to log the
136
# rev_id, we can't trivially only log it. (See bug 526425). Long
137
# term we should rearrange the reporting structure, but for now
138
# we just mutter seperately. We mutter the revid and revno together
139
# so that concurrent bzr invocations won't lead to confusion.
140
mutter('Committed revid %s as revno %d.', rev_id, revno)
142
def deleted(self, path):
143
self._note('deleted %s', path)
145
def missing(self, path):
146
self._note('missing %s', path)
148
def renamed(self, change, old_path, new_path):
149
self._note('%s %s => %s', change, old_path, new_path)
151
def is_verbose(self):
155
class Commit(object):
156
"""Task of committing a new revision.
158
This is a MethodObject: it accumulates state as the commit is
159
prepared, and then it is discarded. It doesn't represent
160
historical revisions, just the act of recording a new one.
163
Modified to hold a list of files that have been deleted from
164
the working directory; these should be removed from the
170
"""Create a Commit object.
172
:param reporter: the default reporter to use or None to decide later
174
self.reporter = reporter
178
def update_revprops(revprops, branch, authors=None, author=None,
179
local=False, possible_master_transports=None):
182
if possible_master_transports is None:
183
possible_master_transports = []
184
if not 'branch-nick' in revprops:
185
revprops['branch-nick'] = branch._get_nick(
187
possible_master_transports)
188
if authors is not None:
189
if author is not None:
190
raise AssertionError('Specifying both author and authors '
191
'is not allowed. Specify just authors instead')
192
if 'author' in revprops or 'authors' in revprops:
193
# XXX: maybe we should just accept one of them?
194
raise AssertionError('author property given twice')
196
for individual in authors:
197
if '\n' in individual:
198
raise AssertionError('\\n is not a valid character '
199
'in an author identity')
200
revprops['authors'] = '\n'.join(authors)
201
if author is not None:
202
symbol_versioning.warn('The parameter author was deprecated'
203
' in version 1.13. Use authors instead',
205
if 'author' in revprops or 'authors' in revprops:
206
# XXX: maybe we should just accept one of them?
207
raise AssertionError('author property given twice')
209
raise AssertionError('\\n is not a valid character '
210
'in an author identity')
211
revprops['authors'] = author
221
allow_pointless=True,
229
message_callback=None,
232
possible_master_transports=None,
234
"""Commit working copy as a new revision.
236
:param message: the commit message (it or message_callback is required)
237
:param message_callback: A callback: message = message_callback(cmt_obj)
239
:param timestamp: if not None, seconds-since-epoch for a
240
postdated/predated commit.
242
:param specific_files: If not None, commit only those files. An empty
243
list means 'commit no files'.
245
:param rev_id: If set, use this as the new revision id.
246
Useful for test or import commands that need to tightly
247
control what revisions are assigned. If you duplicate
248
a revision id that exists elsewhere it is your own fault.
249
If null (default), a time/random revision id is generated.
251
:param allow_pointless: If true (default), commit even if nothing
252
has changed and no merges are recorded.
254
:param strict: If true, don't allow a commit if the working tree
255
contains unknown files.
257
:param revprops: Properties for new revision
258
:param local: Perform a local only commit.
259
:param reporter: the reporter to use or None for the default
260
:param verbose: if True and the reporter is not None, report everything
261
:param recursive: If set to 'down', commit in any subtrees that have
262
pending changes of any sort during this commit.
263
:param exclude: None or a list of relative paths to exclude from the
264
commit. Pending changes to excluded files will be ignored by the
266
:param lossy: When committing to a foreign VCS, ignore any
267
data that can not be natively represented.
269
operation = OperationWithCleanups(self._commit)
270
self.revprops = revprops or {}
271
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
272
self.config = config or self.config
273
return operation.run(
278
specific_files=specific_files,
280
allow_pointless=allow_pointless,
283
working_tree=working_tree,
286
message_callback=message_callback,
289
possible_master_transports=possible_master_transports,
292
def _commit(self, operation, message, timestamp, timezone, committer,
293
specific_files, rev_id, allow_pointless, strict, verbose,
294
working_tree, local, reporter, message_callback, recursive,
295
exclude, possible_master_transports, lossy):
296
mutter('preparing to commit')
298
if working_tree is None:
299
raise BzrError("working_tree must be passed into commit().")
301
self.work_tree = working_tree
302
self.branch = self.work_tree.branch
303
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
304
if not self.branch.repository.supports_rich_root():
305
raise errors.RootNotRich()
306
if message_callback is None:
307
if message is not None:
308
if isinstance(message, str):
309
message = message.decode(get_user_encoding())
310
message_callback = lambda x: message
312
raise BzrError("The message or message_callback keyword"
313
" parameter is required for commit().")
315
self.bound_branch = None
316
self.any_entries_deleted = False
317
if exclude is not None:
318
self.exclude = sorted(
319
minimum_path_selection(exclude))
323
self.master_branch = None
324
self.recursive = recursive
326
# self.specific_files is None to indicate no filter, or any iterable to
327
# indicate a filter - [] means no files at all, as per iter_changes.
328
if specific_files is not None:
329
self.specific_files = sorted(
330
minimum_path_selection(specific_files))
332
self.specific_files = None
334
self.allow_pointless = allow_pointless
335
self.message_callback = message_callback
336
self.timestamp = timestamp
337
self.timezone = timezone
338
self.committer = committer
340
self.verbose = verbose
342
self.work_tree.lock_write()
343
operation.add_cleanup(self.work_tree.unlock)
344
self.parents = self.work_tree.get_parent_ids()
345
# We can use record_iter_changes IFF iter_changes is compatible with
346
# the command line parameters, and the repository has fast delta
347
# generation. See bug 347649.
348
self.use_record_iter_changes = (
350
not self.branch.repository._format.supports_tree_reference and
351
(self.branch.repository._format.fast_deltas or
352
len(self.parents) < 2))
353
self.pb = ui.ui_factory.nested_progress_bar()
354
operation.add_cleanup(self.pb.finished)
355
self.basis_revid = self.work_tree.last_revision()
356
self.basis_tree = self.work_tree.basis_tree()
357
self.basis_tree.lock_read()
358
operation.add_cleanup(self.basis_tree.unlock)
359
# Cannot commit with conflicts present.
360
if len(self.work_tree.conflicts()) > 0:
361
raise ConflictsInTree
363
# Setup the bound branch variables as needed.
364
self._check_bound_branch(operation, possible_master_transports)
366
# Check that the working tree is up to date
367
old_revno, new_revno = self._check_out_of_date_tree()
369
# Complete configuration setup
370
if reporter is not None:
371
self.reporter = reporter
372
elif self.reporter is None:
373
self.reporter = self._select_reporter()
374
if self.config is None:
375
self.config = self.branch.get_config()
377
self._set_specific_file_ids()
379
# Setup the progress bar. As the number of files that need to be
380
# committed in unknown, progress is reported as stages.
381
# We keep track of entries separately though and include that
382
# information in the progress bar during the relevant stages.
383
self.pb_stage_name = ""
384
self.pb_stage_count = 0
385
self.pb_stage_total = 5
386
if self.bound_branch:
387
# 2 extra stages: "Uploading data to master branch" and "Merging
388
# tags to master branch"
389
self.pb_stage_total += 2
390
self.pb.show_pct = False
391
self.pb.show_spinner = False
392
self.pb.show_eta = False
393
self.pb.show_count = True
394
self.pb.show_bar = True
396
self._gather_parents()
397
# After a merge, a selected file commit is not supported.
398
# See 'bzr help merge' for an explanation as to why.
399
if len(self.parents) > 1 and self.specific_files is not None:
400
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
401
# Excludes are a form of selected file commit.
402
if len(self.parents) > 1 and self.exclude:
403
raise errors.CannotCommitSelectedFileMerge(self.exclude)
405
# Collect the changes
406
self._set_progress_stage("Collecting changes", counter=True)
408
self.builder = self.branch.get_commit_builder(self.parents,
409
self.config, timestamp, timezone, committer, self.revprops,
411
if not self.builder.supports_record_entry_contents and self.exclude:
413
raise errors.ExcludesUnsupported(self.branch.repository)
416
self.builder.will_record_deletes()
417
# find the location being committed to
418
if self.bound_branch:
419
master_location = self.master_branch.base
421
master_location = self.branch.base
423
# report the start of the commit
424
self.reporter.started(new_revno, self.rev_id, master_location)
426
self._update_builder_with_changes()
427
self._check_pointless()
429
# TODO: Now the new inventory is known, check for conflicts.
430
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
431
# weave lines, because nothing should be recorded until it is known
432
# that commit will succeed.
433
self._set_progress_stage("Saving data locally")
434
self.builder.finish_inventory()
436
# Prompt the user for a commit message if none provided
437
message = message_callback(self)
438
self.message = message
440
# Add revision data to the local branch
441
self.rev_id = self.builder.commit(self.message)
444
mutter("aborting commit write group because of exception:")
445
trace.log_exception_quietly()
449
self._process_pre_hooks(old_revno, new_revno)
451
# Upload revision data to the master.
452
# this will propagate merged revisions too if needed.
453
if self.bound_branch:
454
self._set_progress_stage("Uploading data to master branch")
455
# 'commit' to the master first so a timeout here causes the
456
# local branch to be out of date
457
(new_revno, self.rev_id) = self.master_branch.import_last_revision_info_and_tags(
458
self.branch, new_revno, self.rev_id, lossy=lossy)
460
self.branch.fetch(self.master_branch, self.rev_id)
462
# and now do the commit locally.
463
self.branch.set_last_revision_info(new_revno, self.rev_id)
465
# Merge local tags to remote
466
if self.bound_branch:
467
self._set_progress_stage("Merging tags to master branch")
468
tag_conflicts = self.branch.tags.merge_to(self.master_branch.tags)
470
warning_lines = [' ' + name for name, _, _ in tag_conflicts]
471
note("Conflicting tags in bound branch:\n" +
472
"\n".join(warning_lines))
474
# Make the working tree be up to date with the branch. This
475
# includes automatic changes scheduled to be made to the tree, such
476
# as updating its basis and unversioning paths that were missing.
477
self.work_tree.unversion(self.deleted_ids)
478
self._set_progress_stage("Updating the working tree")
479
self.work_tree.update_basis_by_delta(self.rev_id,
480
self.builder.get_basis_delta())
481
self.reporter.completed(new_revno, self.rev_id)
482
self._process_post_hooks(old_revno, new_revno)
485
def _select_reporter(self):
486
"""Select the CommitReporter to use."""
488
return NullCommitReporter()
489
return ReportCommitToLog()
491
def _check_pointless(self):
492
if self.allow_pointless:
494
# A merge with no effect on files
495
if len(self.parents) > 1:
497
if self.builder.any_changes():
499
raise PointlessCommit()
501
def _check_bound_branch(self, operation, possible_master_transports=None):
502
"""Check to see if the local branch is bound.
504
If it is bound, then most of the commit will actually be
505
done using the remote branch as the target branch.
506
Only at the end will the local branch be updated.
508
if self.local and not self.branch.get_bound_location():
509
raise errors.LocalRequiresBoundBranch()
512
self.master_branch = self.branch.get_master_branch(
513
possible_master_transports)
515
if not self.master_branch:
516
# make this branch the reference branch for out of date checks.
517
self.master_branch = self.branch
520
# If the master branch is bound, we must fail
521
master_bound_location = self.master_branch.get_bound_location()
522
if master_bound_location:
523
raise errors.CommitToDoubleBoundBranch(self.branch,
524
self.master_branch, master_bound_location)
526
# TODO: jam 20051230 We could automatically push local
527
# commits to the remote branch if they would fit.
528
# But for now, just require remote to be identical
531
# Make sure the local branch is identical to the master
532
master_info = self.master_branch.last_revision_info()
533
local_info = self.branch.last_revision_info()
534
if local_info != master_info:
535
raise errors.BoundBranchOutOfDate(self.branch,
538
# Now things are ready to change the master branch
540
self.bound_branch = self.branch
541
self.master_branch.lock_write()
542
operation.add_cleanup(self.master_branch.unlock)
544
def _check_out_of_date_tree(self):
545
"""Check that the working tree is up to date.
547
:return: old_revision_number,new_revision_number tuple
550
first_tree_parent = self.work_tree.get_parent_ids()[0]
552
# if there are no parents, treat our parent as 'None'
553
# this is so that we still consider the master branch
554
# - in a checkout scenario the tree may have no
555
# parents but the branch may do.
556
first_tree_parent = bzrlib.revision.NULL_REVISION
557
old_revno, master_last = self.master_branch.last_revision_info()
558
if master_last != first_tree_parent:
559
if master_last != bzrlib.revision.NULL_REVISION:
560
raise errors.OutOfDateTree(self.work_tree)
561
if self.branch.repository.has_revision(first_tree_parent):
562
new_revno = old_revno + 1
564
# ghost parents never appear in revision history.
566
return old_revno,new_revno
568
def _process_pre_hooks(self, old_revno, new_revno):
569
"""Process any registered pre commit hooks."""
570
self._set_progress_stage("Running pre_commit hooks")
571
self._process_hooks("pre_commit", old_revno, new_revno)
573
def _process_post_hooks(self, old_revno, new_revno):
574
"""Process any registered post commit hooks."""
575
# Process the post commit hooks, if any
576
self._set_progress_stage("Running post_commit hooks")
577
# old style commit hooks - should be deprecated ? (obsoleted in
579
if self.config.post_commit() is not None:
580
hooks = self.config.post_commit().split(' ')
581
# this would be nicer with twisted.python.reflect.namedAny
583
result = eval(hook + '(branch, rev_id)',
584
{'branch':self.branch,
586
'rev_id':self.rev_id})
587
# process new style post commit hooks
588
self._process_hooks("post_commit", old_revno, new_revno)
590
def _process_hooks(self, hook_name, old_revno, new_revno):
591
if not Branch.hooks[hook_name]:
594
# new style commit hooks:
595
if not self.bound_branch:
596
hook_master = self.branch
599
hook_master = self.master_branch
600
hook_local = self.branch
601
# With bound branches, when the master is behind the local branch,
602
# the 'old_revno' and old_revid values here are incorrect.
603
# XXX: FIXME ^. RBC 20060206
605
old_revid = self.parents[0]
607
old_revid = bzrlib.revision.NULL_REVISION
609
if hook_name == "pre_commit":
610
future_tree = self.builder.revision_tree()
611
tree_delta = future_tree.changes_from(self.basis_tree,
614
for hook in Branch.hooks[hook_name]:
615
# show the running hook in the progress bar. As hooks may
616
# end up doing nothing (e.g. because they are not configured by
617
# the user) this is still showing progress, not showing overall
618
# actions - its up to each plugin to show a UI if it want's to
619
# (such as 'Emailing diff to foo@example.com').
620
self.pb_stage_name = "Running %s hooks [%s]" % \
621
(hook_name, Branch.hooks.get_hook_name(hook))
622
self._emit_progress()
623
if 'hooks' in debug.debug_flags:
624
mutter("Invoking commit hook: %r", hook)
625
if hook_name == "post_commit":
626
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
628
elif hook_name == "pre_commit":
629
hook(hook_local, hook_master,
630
old_revno, old_revid, new_revno, self.rev_id,
631
tree_delta, future_tree)
633
def _gather_parents(self):
634
"""Record the parents of a merge for merge detection."""
635
# TODO: Make sure that this list doesn't contain duplicate
636
# entries and the order is preserved when doing this.
637
if self.use_record_iter_changes:
639
self.basis_inv = self.basis_tree.inventory
640
self.parent_invs = [self.basis_inv]
641
for revision in self.parents[1:]:
642
if self.branch.repository.has_revision(revision):
643
mutter('commit parent revision {%s}', revision)
644
inventory = self.branch.repository.get_inventory(revision)
645
self.parent_invs.append(inventory)
647
mutter('commit parent ghost revision {%s}', revision)
649
def _update_builder_with_changes(self):
650
"""Update the commit builder with the data about what has changed.
652
exclude = self.exclude
653
specific_files = self.specific_files
654
mutter("Selecting files for commit with filter %s", specific_files)
657
if self.use_record_iter_changes:
658
iter_changes = self.work_tree.iter_changes(self.basis_tree,
659
specific_files=specific_files)
660
iter_changes = self._filter_iter_changes(iter_changes)
661
for file_id, path, fs_hash in self.builder.record_iter_changes(
662
self.work_tree, self.basis_revid, iter_changes):
663
self.work_tree._observed_sha1(file_id, path, fs_hash)
665
# Build the new inventory
666
self._populate_from_inventory()
667
self._record_unselected()
668
self._report_and_accumulate_deletes()
670
def _filter_iter_changes(self, iter_changes):
671
"""Process iter_changes.
673
This method reports on the changes in iter_changes to the user, and
674
converts 'missing' entries in the iter_changes iterator to 'deleted'
675
entries. 'missing' entries have their
677
:param iter_changes: An iter_changes to process.
678
:return: A generator of changes.
680
reporter = self.reporter
681
report_changes = reporter.is_verbose()
683
for change in iter_changes:
685
old_path = change[1][0]
686
new_path = change[1][1]
687
versioned = change[3][1]
689
versioned = change[3][1]
690
if kind is None and versioned:
693
reporter.missing(new_path)
694
deleted_ids.append(change[0])
695
# Reset the new path (None) and new versioned flag (False)
696
change = (change[0], (change[1][0], None), change[2],
697
(change[3][0], False)) + change[4:]
698
elif kind == 'tree-reference':
699
if self.recursive == 'down':
700
self._commit_nested_tree(change[0], change[1][1])
701
if change[3][0] or change[3][1]:
705
reporter.deleted(old_path)
706
elif old_path is None:
707
reporter.snapshot_change('added', new_path)
708
elif old_path != new_path:
709
reporter.renamed('renamed', old_path, new_path)
712
self.work_tree.branch.repository._format.rich_root_data):
713
# Don't report on changes to '' in non rich root
715
reporter.snapshot_change('modified', new_path)
716
self._next_progress_entry()
717
# Unversion IDs that were found to be deleted
718
self.deleted_ids = deleted_ids
720
def _record_unselected(self):
721
# If specific files are selected, then all un-selected files must be
722
# recorded in their previous state. For more details, see
723
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
724
if self.specific_files or self.exclude:
725
specific_files = self.specific_files or []
726
for path, old_ie in self.basis_inv.iter_entries():
727
if old_ie.file_id in self.builder.new_inventory:
728
# already added - skip.
730
if (is_inside_any(specific_files, path)
731
and not is_inside_any(self.exclude, path)):
732
# was inside the selected path, and not excluded - if not
733
# present it has been deleted so skip.
735
# From here down it was either not selected, or was excluded:
736
# We preserve the entry unaltered.
738
# Note: specific file commits after a merge are currently
739
# prohibited. This test is for sanity/safety in case it's
740
# required after that changes.
741
if len(self.parents) > 1:
743
self.builder.record_entry_contents(ie, self.parent_invs, path,
744
self.basis_tree, None)
746
def _report_and_accumulate_deletes(self):
747
if (isinstance(self.basis_inv, Inventory)
748
and isinstance(self.builder.new_inventory, Inventory)):
749
# the older Inventory classes provide a _byid dict, and building a
750
# set from the keys of this dict is substantially faster than even
751
# getting a set of ids from the inventory
753
# <lifeless> set(dict) is roughly the same speed as
754
# set(iter(dict)) and both are significantly slower than
756
deleted_ids = set(self.basis_inv._byid.keys()) - \
757
set(self.builder.new_inventory._byid.keys())
759
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
761
self.any_entries_deleted = True
762
deleted = [(self.basis_tree.id2path(file_id), file_id)
763
for file_id in deleted_ids]
765
# XXX: this is not quite directory-order sorting
766
for path, file_id in deleted:
767
self.builder.record_delete(path, file_id)
768
self.reporter.deleted(path)
770
def _check_strict(self):
771
# XXX: when we use iter_changes this would likely be faster if
772
# iter_changes would check for us (even in the presence of
775
# raise an exception as soon as we find a single unknown.
776
for unknown in self.work_tree.unknowns():
777
raise StrictCommitFailed()
779
def _populate_from_inventory(self):
780
"""Populate the CommitBuilder by walking the working tree inventory."""
781
# Build the revision inventory.
783
# This starts by creating a new empty inventory. Depending on
784
# which files are selected for commit, and what is present in the
785
# current tree, the new inventory is populated. inventory entries
786
# which are candidates for modification have their revision set to
787
# None; inventory entries that are carried over untouched have their
788
# revision set to their prior value.
790
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
791
# results to create a new inventory at the same time, which results
792
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
795
specific_files = self.specific_files
796
exclude = self.exclude
797
report_changes = self.reporter.is_verbose()
799
# A tree of paths that have been deleted. E.g. if foo/bar has been
800
# deleted, then we have {'foo':{'bar':{}}}
802
# XXX: Note that entries may have the wrong kind because the entry does
803
# not reflect the status on disk.
804
work_inv = self.work_tree.inventory
805
# NB: entries will include entries within the excluded ids/paths
806
# because iter_entries_by_dir has no 'exclude' facility today.
807
entries = work_inv.iter_entries_by_dir(
808
specific_file_ids=self.specific_file_ids, yield_parents=True)
809
for path, existing_ie in entries:
810
file_id = existing_ie.file_id
811
name = existing_ie.name
812
parent_id = existing_ie.parent_id
813
kind = existing_ie.kind
814
# Skip files that have been deleted from the working tree.
815
# The deleted path ids are also recorded so they can be explicitly
818
path_segments = splitpath(path)
819
deleted_dict = deleted_paths
820
for segment in path_segments:
821
deleted_dict = deleted_dict.get(segment, None)
823
# We either took a path not present in the dict
824
# (deleted_dict was None), or we've reached an empty
825
# child dir in the dict, so are now a sub-path.
829
if deleted_dict is not None:
830
# the path has a deleted parent, do not add it.
832
if exclude and is_inside_any(exclude, path):
833
# Skip excluded paths. Excluded paths are processed by
834
# _update_builder_with_changes.
836
content_summary = self.work_tree.path_content_summary(path)
837
kind = content_summary[0]
838
# Note that when a filter of specific files is given, we must only
839
# skip/record deleted files matching that filter.
840
if not specific_files or is_inside_any(specific_files, path):
841
if kind == 'missing':
842
if not deleted_paths:
843
# path won't have been split yet.
844
path_segments = splitpath(path)
845
deleted_dict = deleted_paths
846
for segment in path_segments:
847
deleted_dict = deleted_dict.setdefault(segment, {})
848
self.reporter.missing(path)
849
self._next_progress_entry()
850
deleted_ids.append(file_id)
852
# TODO: have the builder do the nested commit just-in-time IF and
854
if kind == 'tree-reference':
855
# enforce repository nested tree policy.
856
if (not self.work_tree.supports_tree_reference() or
857
# repository does not support it either.
858
not self.branch.repository._format.supports_tree_reference):
860
content_summary = (kind, None, None, None)
861
elif self.recursive == 'down':
862
nested_revision_id = self._commit_nested_tree(
864
content_summary = (kind, None, None, nested_revision_id)
866
nested_revision_id = self.work_tree.get_reference_revision(file_id)
867
content_summary = (kind, None, None, nested_revision_id)
869
# Record an entry for this item
870
# Note: I don't particularly want to have the existing_ie
871
# parameter but the test suite currently (28-Jun-07) breaks
872
# without it thanks to a unicode normalisation issue. :-(
873
definitely_changed = kind != existing_ie.kind
874
self._record_entry(path, file_id, specific_files, kind, name,
875
parent_id, definitely_changed, existing_ie, report_changes,
878
# Unversion IDs that were found to be deleted
879
self.deleted_ids = deleted_ids
881
def _commit_nested_tree(self, file_id, path):
882
"Commit a nested tree."
883
sub_tree = self.work_tree.get_nested_tree(file_id, path)
884
# FIXME: be more comprehensive here:
885
# this works when both trees are in --trees repository,
886
# but when both are bound to a different repository,
887
# it fails; a better way of approaching this is to
888
# finally implement the explicit-caches approach design
889
# a while back - RBC 20070306.
890
if sub_tree.branch.repository.has_same_location(
891
self.work_tree.branch.repository):
892
sub_tree.branch.repository = \
893
self.work_tree.branch.repository
895
return sub_tree.commit(message=None, revprops=self.revprops,
896
recursive=self.recursive,
897
message_callback=self.message_callback,
898
timestamp=self.timestamp, timezone=self.timezone,
899
committer=self.committer,
900
allow_pointless=self.allow_pointless,
901
strict=self.strict, verbose=self.verbose,
902
local=self.local, reporter=self.reporter)
903
except errors.PointlessCommit:
904
return self.work_tree.get_reference_revision(file_id)
906
def _record_entry(self, path, file_id, specific_files, kind, name,
907
parent_id, definitely_changed, existing_ie, report_changes,
909
"Record the new inventory entry for a path if any."
910
# mutter('check %s {%s}', path, file_id)
911
# mutter('%s selected for commit', path)
912
if definitely_changed or existing_ie is None:
913
ie = make_entry(kind, name, parent_id, file_id)
915
ie = existing_ie.copy()
917
# For carried over entries we don't care about the fs hash - the repo
918
# isn't generating a sha, so we're not saving computation time.
919
_, _, fs_hash = self.builder.record_entry_contents(
920
ie, self.parent_invs, path, self.work_tree, content_summary)
922
self._report_change(ie, path)
924
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
927
def _report_change(self, ie, path):
928
"""Report a change to the user.
930
The change that has occurred is described relative to the basis
933
if (self.basis_inv.has_id(ie.file_id)):
934
basis_ie = self.basis_inv[ie.file_id]
937
change = ie.describe_change(basis_ie, ie)
938
if change in (InventoryEntry.RENAMED,
939
InventoryEntry.MODIFIED_AND_RENAMED):
940
old_path = self.basis_inv.id2path(ie.file_id)
941
self.reporter.renamed(change, old_path, path)
942
self._next_progress_entry()
944
if change == 'unchanged':
946
self.reporter.snapshot_change(change, path)
947
self._next_progress_entry()
949
def _set_progress_stage(self, name, counter=False):
950
"""Set the progress stage and emit an update to the progress bar."""
951
self.pb_stage_name = name
952
self.pb_stage_count += 1
954
self.pb_entries_count = 0
956
self.pb_entries_count = None
957
self._emit_progress()
959
def _next_progress_entry(self):
960
"""Emit an update to the progress bar and increment the entry count."""
961
self.pb_entries_count += 1
962
self._emit_progress()
964
def _emit_progress(self):
965
if self.pb_entries_count is not None:
966
text = "%s [%d] - Stage" % (self.pb_stage_name,
967
self.pb_entries_count)
969
text = "%s - Stage" % (self.pb_stage_name, )
970
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
972
def _set_specific_file_ids(self):
973
"""populate self.specific_file_ids if we will use it."""
974
if not self.use_record_iter_changes:
975
# If provided, ensure the specified files are versioned
976
if self.specific_files is not None:
977
# Note: This routine is being called because it raises
978
# PathNotVersionedError as a side effect of finding the IDs. We
979
# later use the ids we found as input to the working tree
980
# inventory iterator, so we only consider those ids rather than
981
# examining the whole tree again.
982
# XXX: Dont we have filter_unversioned to do this more
984
self.specific_file_ids = tree.find_ids_across_trees(
985
self.specific_files, [self.basis_tree, self.work_tree])
987
self.specific_file_ids = None
added
removed
bzrlib/commit.py
