50
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
51
50
# the rest of the code; add a deprecation of the old name.
57
from cStringIO import StringIO
59
from .branch import Branch
60
from .cleanup import ExitStack
62
from .errors import (BzrError,
66
from .osutils import (get_user_encoding,
69
minimum_path_selection,
71
from .trace import mutter, note, is_quiet
72
from .tree import TreeChange
73
from .urlutils import unescape_for_display
74
from .i18n import gettext
77
class PointlessCommit(BzrError):
79
_fmt = "No changes to commit"
82
class CannotCommitSelectedFileMerge(BzrError):
84
_fmt = 'Selected-file commit of merges is not supported yet:'\
85
' files %(files_str)s'
87
def __init__(self, files):
88
files_str = ', '.join(files)
89
BzrError.__init__(self, files=files, files_str=files_str)
92
def filter_excluded(iter_changes, exclude):
93
"""Filter exclude filenames.
95
:param iter_changes: iter_changes function
96
:param exclude: List of paths to exclude
97
:return: iter_changes function
99
for change in iter_changes:
100
new_excluded = (change.path[1] is not None and
101
is_inside_any(exclude, change.path[1]))
103
old_excluded = (change.path[0] is not None and
104
is_inside_any(exclude, change.path[0]))
106
if old_excluded and new_excluded:
109
if old_excluded or new_excluded:
110
# TODO(jelmer): Perhaps raise an error here instead?
66
from bzrlib.branch import Branch
68
from bzrlib.errors import (BzrError, PointlessCommit,
72
from bzrlib.osutils import (get_user_encoding,
73
kind_marker, isdir,isfile, is_inside_any,
74
is_inside_or_parent_of_any,
75
minimum_path_selection,
76
quotefn, sha_file, split_lines,
79
from bzrlib.testament import Testament
80
from bzrlib.trace import mutter, note, warning, is_quiet
81
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
82
from bzrlib import symbol_versioning
83
from bzrlib.symbol_versioning import (deprecated_passed,
86
from bzrlib.workingtree import WorkingTree
87
from bzrlib.urlutils import unescape_for_display
116
91
class NullCommitReporter(object):
117
92
"""I report on progress of a commit."""
119
def started(self, revno, revid, location):
94
def started(self, revno, revid, location=None):
96
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
97
"to started.", DeprecationWarning,
122
101
def snapshot_change(self, change, path):
148
130
note(format, *args)
150
132
def snapshot_change(self, change, path):
151
if path == '' and change in (gettext('added'), gettext('modified')):
133
if change == 'unchanged':
135
if change == 'added' and path == '':
153
137
self._note("%s %s", change, path)
155
def started(self, revno, rev_id, location):
157
gettext('Committing to: %s'),
158
unescape_for_display(location, 'utf-8'))
139
def started(self, revno, rev_id, location=None):
140
if location is not None:
141
location = ' to: ' + unescape_for_display(location, 'utf-8')
143
# When started was added, location was only made optional by
144
# accident. Matt Nordhoff 20071129
145
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
146
"to started.", DeprecationWarning,
149
self._note('Committing%s', location)
160
151
def completed(self, revno, rev_id):
161
if revno is not None:
162
self._note(gettext('Committed revision %d.'), revno)
163
# self._note goes to the console too; so while we want to log the
164
# rev_id, we can't trivially only log it. (See bug 526425). Long
165
# term we should rearrange the reporting structure, but for now
166
# we just mutter seperately. We mutter the revid and revno together
167
# so that concurrent bzr invocations won't lead to confusion.
168
mutter('Committed revid %s as revno %d.', rev_id, revno)
170
self._note(gettext('Committed revid %s.'), rev_id)
172
def deleted(self, path):
173
self._note(gettext('deleted %s'), path)
152
self._note('Committed revision %d.', revno)
154
def deleted(self, file_id):
155
self._note('deleted %s', file_id)
157
def escaped(self, escape_count, message):
158
self._note("replaced %d control characters in message", escape_count)
175
160
def missing(self, path):
176
self._note(gettext('missing %s'), path)
161
self._note('missing %s', path)
178
163
def renamed(self, change, old_path, new_path):
179
164
self._note('%s %s => %s', change, old_path, new_path)
282
238
:param exclude: None or a list of relative paths to exclude from the
283
239
commit. Pending changes to excluded files will be ignored by the
285
:param lossy: When committing to a foreign VCS, ignore any
286
data that can not be natively represented.
288
with ExitStack() as stack:
289
self.revprops = revprops or {}
290
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
291
self.config_stack = config or self.config_stack
292
mutter('preparing to commit')
294
if working_tree is None:
295
raise BzrError("working_tree must be passed into commit().")
297
self.work_tree = working_tree
298
self.branch = self.work_tree.branch
299
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
300
if not self.branch.repository.supports_rich_root():
301
raise errors.RootNotRich()
302
if message_callback is None:
303
if message is not None:
304
if isinstance(message, bytes):
305
message = message.decode(get_user_encoding())
307
def message_callback(x):
310
raise BzrError("The message or message_callback keyword"
311
" parameter is required for commit().")
313
self.bound_branch = None
314
self.any_entries_deleted = False
315
if exclude is not None:
316
self.exclude = sorted(
317
minimum_path_selection(exclude))
321
self.master_branch = None
322
self.recursive = recursive
324
# self.specific_files is None to indicate no filter, or any iterable to
325
# indicate a filter - [] means no files at all, as per iter_changes.
326
if specific_files is not None:
327
self.specific_files = sorted(
328
minimum_path_selection(specific_files))
330
self.specific_files = None
332
self.allow_pointless = allow_pointless
333
self.message_callback = message_callback
334
self.timestamp = timestamp
335
self.timezone = timezone
336
self.committer = committer
338
self.verbose = verbose
340
stack.enter_context(self.work_tree.lock_write())
341
self.parents = self.work_tree.get_parent_ids()
342
self.pb = ui.ui_factory.nested_progress_bar()
343
stack.callback(self.pb.finished)
344
self.basis_revid = self.work_tree.last_revision()
345
self.basis_tree = self.work_tree.basis_tree()
346
stack.enter_context(self.basis_tree.lock_read())
242
mutter('preparing to commit')
244
if working_tree is None:
245
raise BzrError("working_tree must be passed into commit().")
247
self.work_tree = working_tree
248
self.branch = self.work_tree.branch
249
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
250
if not self.branch.repository.supports_rich_root():
251
raise errors.RootNotRich()
252
if message_callback is None:
253
if message is not None:
254
if isinstance(message, str):
255
message = message.decode(get_user_encoding())
256
message_callback = lambda x: message
258
raise BzrError("The message or message_callback keyword"
259
" parameter is required for commit().")
261
self.bound_branch = None
262
self.any_entries_deleted = False
263
if exclude is not None:
264
self.exclude = sorted(
265
minimum_path_selection(exclude))
269
self.master_branch = None
270
self.master_locked = False
271
self.recursive = recursive
273
if specific_files is not None:
274
self.specific_files = sorted(
275
minimum_path_selection(specific_files))
277
self.specific_files = None
278
self.specific_file_ids = None
279
self.allow_pointless = allow_pointless
280
self.revprops = revprops
281
self.message_callback = message_callback
282
self.timestamp = timestamp
283
self.timezone = timezone
284
self.committer = committer
286
self.verbose = verbose
288
self.work_tree.lock_write()
289
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
290
self.basis_revid = self.work_tree.last_revision()
291
self.basis_tree = self.work_tree.basis_tree()
292
self.basis_tree.lock_read()
347
294
# Cannot commit with conflicts present.
348
295
if len(self.work_tree.conflicts()) > 0:
349
296
raise ConflictsInTree
351
298
# Setup the bound branch variables as needed.
352
self._check_bound_branch(stack, possible_master_transports)
353
if self.config_stack is None:
354
self.config_stack = self.work_tree.get_config_stack()
299
self._check_bound_branch(possible_master_transports)
356
301
# Check that the working tree is up to date
357
old_revno, old_revid, new_revno = self._check_out_of_date_tree()
302
old_revno, new_revno = self._check_out_of_date_tree()
359
304
# Complete configuration setup
360
305
if reporter is not None:
361
306
self.reporter = reporter
362
307
elif self.reporter is None:
363
308
self.reporter = self._select_reporter()
309
if self.config is None:
310
self.config = self.branch.get_config()
312
# If provided, ensure the specified files are versioned
313
if self.specific_files is not None:
314
# Note: This routine is being called because it raises
315
# PathNotVersionedError as a side effect of finding the IDs. We
316
# later use the ids we found as input to the working tree
317
# inventory iterator, so we only consider those ids rather than
318
# examining the whole tree again.
319
# XXX: Dont we have filter_unversioned to do this more
321
self.specific_file_ids = tree.find_ids_across_trees(
322
specific_files, [self.basis_tree, self.work_tree])
365
324
# Setup the progress bar. As the number of files that need to be
366
325
# committed in unknown, progress is reported as stages.
654
573
old_revno, old_revid, new_revno, self.rev_id,
655
574
tree_delta, future_tree)
577
"""Cleanup any open locks, progress bars etc."""
578
cleanups = [self._cleanup_bound_branch,
579
self.basis_tree.unlock,
580
self.work_tree.unlock,
582
found_exception = None
583
for cleanup in cleanups:
586
# we want every cleanup to run no matter what.
587
# so we have a catchall here, but we will raise the
588
# last encountered exception up the stack: and
589
# typically this will be useful enough.
592
if found_exception is not None:
593
# don't do a plan raise, because the last exception may have been
594
# trashed, e is our sure-to-work exception even though it loses the
595
# full traceback. XXX: RBC 20060421 perhaps we could check the
596
# exc_info and if its the same one do a plain raise otherwise
597
# 'raise e' as we do now.
600
def _cleanup_bound_branch(self):
601
"""Executed at the end of a try/finally to cleanup a bound branch.
603
If the branch wasn't bound, this is a no-op.
604
If it was, it resents self.branch to the local branch, instead
607
if not self.bound_branch:
609
if self.master_locked:
610
self.master_branch.unlock()
612
def _escape_commit_message(self):
613
"""Replace xml-incompatible control characters."""
614
# FIXME: RBC 20060419 this should be done by the revision
615
# serialiser not by commit. Then we can also add an unescaper
616
# in the deserializer and start roundtripping revision messages
617
# precisely. See repository_implementations/test_repository.py
619
# Python strings can include characters that can't be
620
# represented in well-formed XML; escape characters that
621
# aren't listed in the XML specification
622
# (http://www.w3.org/TR/REC-xml/#NT-Char).
623
self.message, escape_count = re.subn(
624
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
625
lambda match: match.group(0).encode('unicode_escape'),
628
self.reporter.escaped(escape_count, self.message)
630
def _gather_parents(self):
631
"""Record the parents of a merge for merge detection."""
632
# TODO: Make sure that this list doesn't contain duplicate
633
# entries and the order is preserved when doing this.
634
self.parents = self.work_tree.get_parent_ids()
635
self.parent_invs = [self.basis_inv]
636
for revision in self.parents[1:]:
637
if self.branch.repository.has_revision(revision):
638
mutter('commit parent revision {%s}', revision)
639
inventory = self.branch.repository.get_inventory(revision)
640
self.parent_invs.append(inventory)
642
mutter('commit parent ghost revision {%s}', revision)
657
644
def _update_builder_with_changes(self):
658
645
"""Update the commit builder with the data about what has changed.
660
specific_files = self.specific_files
661
mutter("Selecting files for commit with filter %r", specific_files)
664
iter_changes = self.work_tree.iter_changes(
665
self.basis_tree, specific_files=specific_files)
667
iter_changes = filter_excluded(iter_changes, self.exclude)
668
iter_changes = self._filter_iter_changes(iter_changes)
669
for path, fs_hash in self.builder.record_iter_changes(
670
self.work_tree, self.basis_revid, iter_changes):
671
self.work_tree._observed_sha1(path, fs_hash)
673
def _filter_iter_changes(self, iter_changes):
674
"""Process iter_changes.
676
This method reports on the changes in iter_changes to the user, and
677
converts 'missing' entries in the iter_changes iterator to 'deleted'
678
entries. 'missing' entries have their
680
:param iter_changes: An iter_changes to process.
681
:return: A generator of changes.
683
reporter = self.reporter
684
report_changes = reporter.is_verbose()
686
for change in iter_changes:
688
old_path = change.path[0]
689
new_path = change.path[1]
690
versioned = change.versioned[1]
691
kind = change.kind[1]
692
versioned = change.versioned[1]
693
if kind is None and versioned:
696
reporter.missing(new_path)
697
if change.kind[0] == 'symlink' and not self.work_tree.supports_symlinks():
698
trace.warning('Ignoring "%s" as symlinks are not '
699
'supported on this filesystem.' % (change.path[0],))
701
deleted_paths.append(change.path[1])
702
# Reset the new path (None) and new versioned flag (False)
703
change = change.discard_new()
704
new_path = change.path[1]
706
elif kind == 'tree-reference':
707
if self.recursive == 'down':
708
self._commit_nested_tree(change.path[1])
709
if change.versioned[0] or change.versioned[1]:
713
reporter.deleted(old_path)
714
elif old_path is None:
715
reporter.snapshot_change(gettext('added'), new_path)
716
elif old_path != new_path:
717
reporter.renamed(gettext('renamed'),
721
or self.work_tree.branch.repository._format.rich_root_data):
722
# Don't report on changes to '' in non rich root
724
reporter.snapshot_change(
725
gettext('modified'), new_path)
726
self._next_progress_entry()
727
# Unversion files that were found to be deleted
728
self.deleted_paths = deleted_paths
730
def _check_strict(self):
731
# XXX: when we use iter_changes this would likely be faster if
732
# iter_changes would check for us (even in the presence of
647
# Build the revision inventory.
649
# This starts by creating a new empty inventory. Depending on
650
# which files are selected for commit, and what is present in the
651
# current tree, the new inventory is populated. inventory entries
652
# which are candidates for modification have their revision set to
653
# None; inventory entries that are carried over untouched have their
654
# revision set to their prior value.
656
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
657
# results to create a new inventory at the same time, which results
658
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
661
exclude = self.exclude
662
specific_files = self.specific_files or []
663
mutter("Selecting files for commit with filter %s", specific_files)
665
# Build the new inventory
666
self._populate_from_inventory()
668
# If specific files are selected, then all un-selected files must be
669
# recorded in their previous state. For more details, see
670
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
671
if specific_files or exclude:
672
for path, old_ie in self.basis_inv.iter_entries():
673
if old_ie.file_id in self.builder.new_inventory:
674
# already added - skip.
676
if (is_inside_any(specific_files, path)
677
and not is_inside_any(exclude, path)):
678
# was inside the selected path, and not excluded - if not
679
# present it has been deleted so skip.
681
# From here down it was either not selected, or was excluded:
682
if old_ie.kind == 'directory':
683
self._next_progress_entry()
684
# We preserve the entry unaltered.
686
# Note: specific file commits after a merge are currently
687
# prohibited. This test is for sanity/safety in case it's
688
# required after that changes.
689
if len(self.parents) > 1:
691
self.builder.record_entry_contents(ie, self.parent_invs, path,
692
self.basis_tree, None)
694
def _report_and_accumulate_deletes(self):
695
# XXX: Could the list of deleted paths and ids be instead taken from
696
# _populate_from_inventory?
697
if (isinstance(self.basis_inv, Inventory)
698
and isinstance(self.builder.new_inventory, Inventory)):
699
# the older Inventory classes provide a _byid dict, and building a
700
# set from the keys of this dict is substantially faster than even
701
# getting a set of ids from the inventory
703
# <lifeless> set(dict) is roughly the same speed as
704
# set(iter(dict)) and both are significantly slower than
706
deleted_ids = set(self.basis_inv._byid.keys()) - \
707
set(self.builder.new_inventory._byid.keys())
709
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
711
self.any_entries_deleted = True
712
deleted = [(self.basis_tree.id2path(file_id), file_id)
713
for file_id in deleted_ids]
715
# XXX: this is not quite directory-order sorting
716
for path, file_id in deleted:
717
self.builder.record_delete(path, file_id)
718
self.reporter.deleted(path)
720
def _populate_from_inventory(self):
721
"""Populate the CommitBuilder by walking the working tree inventory."""
735
723
# raise an exception as soon as we find a single unknown.
736
724
for unknown in self.work_tree.unknowns():
737
725
raise StrictCommitFailed()
739
def _commit_nested_tree(self, path):
727
specific_files = self.specific_files
728
exclude = self.exclude
729
report_changes = self.reporter.is_verbose()
731
# A tree of paths that have been deleted. E.g. if foo/bar has been
732
# deleted, then we have {'foo':{'bar':{}}}
734
# XXX: Note that entries may have the wrong kind because the entry does
735
# not reflect the status on disk.
736
work_inv = self.work_tree.inventory
737
# NB: entries will include entries within the excluded ids/paths
738
# because iter_entries_by_dir has no 'exclude' facility today.
739
entries = work_inv.iter_entries_by_dir(
740
specific_file_ids=self.specific_file_ids, yield_parents=True)
741
for path, existing_ie in entries:
742
file_id = existing_ie.file_id
743
name = existing_ie.name
744
parent_id = existing_ie.parent_id
745
kind = existing_ie.kind
746
if kind == 'directory':
747
self._next_progress_entry()
748
# Skip files that have been deleted from the working tree.
749
# The deleted path ids are also recorded so they can be explicitly
752
path_segments = splitpath(path)
753
deleted_dict = deleted_paths
754
for segment in path_segments:
755
deleted_dict = deleted_dict.get(segment, None)
757
# We either took a path not present in the dict
758
# (deleted_dict was None), or we've reached an empty
759
# child dir in the dict, so are now a sub-path.
763
if deleted_dict is not None:
764
# the path has a deleted parent, do not add it.
766
if exclude and is_inside_any(exclude, path):
767
# Skip excluded paths. Excluded paths are processed by
768
# _update_builder_with_changes.
770
content_summary = self.work_tree.path_content_summary(path)
771
# Note that when a filter of specific files is given, we must only
772
# skip/record deleted files matching that filter.
773
if not specific_files or is_inside_any(specific_files, path):
774
if content_summary[0] == 'missing':
775
if not deleted_paths:
776
# path won't have been split yet.
777
path_segments = splitpath(path)
778
deleted_dict = deleted_paths
779
for segment in path_segments:
780
deleted_dict = deleted_dict.setdefault(segment, {})
781
self.reporter.missing(path)
782
deleted_ids.append(file_id)
784
# TODO: have the builder do the nested commit just-in-time IF and
786
if content_summary[0] == 'tree-reference':
787
# enforce repository nested tree policy.
788
if (not self.work_tree.supports_tree_reference() or
789
# repository does not support it either.
790
not self.branch.repository._format.supports_tree_reference):
791
content_summary = ('directory',) + content_summary[1:]
792
kind = content_summary[0]
793
# TODO: specific_files filtering before nested tree processing
794
if kind == 'tree-reference':
795
if self.recursive == 'down':
796
nested_revision_id = self._commit_nested_tree(
798
content_summary = content_summary[:3] + (
801
content_summary = content_summary[:3] + (
802
self.work_tree.get_reference_revision(file_id),)
804
# Record an entry for this item
805
# Note: I don't particularly want to have the existing_ie
806
# parameter but the test suite currently (28-Jun-07) breaks
807
# without it thanks to a unicode normalisation issue. :-(
808
definitely_changed = kind != existing_ie.kind
809
self._record_entry(path, file_id, specific_files, kind, name,
810
parent_id, definitely_changed, existing_ie, report_changes,
813
# Unversion IDs that were found to be deleted
814
self.work_tree.unversion(deleted_ids)
816
def _commit_nested_tree(self, file_id, path):
740
817
"Commit a nested tree."
741
sub_tree = self.work_tree.get_nested_tree(path)
818
sub_tree = self.work_tree.get_nested_tree(file_id, path)
742
819
# FIXME: be more comprehensive here:
743
820
# this works when both trees are in --trees repository,
744
821
# but when both are bound to a different repository,
746
823
# finally implement the explicit-caches approach design
747
824
# a while back - RBC 20070306.
748
825
if sub_tree.branch.repository.has_same_location(
749
self.work_tree.branch.repository):
826
self.work_tree.branch.repository):
750
827
sub_tree.branch.repository = \
751
828
self.work_tree.branch.repository
753
830
return sub_tree.commit(message=None, revprops=self.revprops,
754
recursive=self.recursive,
755
message_callback=self.message_callback,
756
timestamp=self.timestamp,
757
timezone=self.timezone,
758
committer=self.committer,
759
allow_pointless=self.allow_pointless,
760
strict=self.strict, verbose=self.verbose,
761
local=self.local, reporter=self.reporter)
762
except PointlessCommit:
763
return self.work_tree.get_reference_revision(path)
765
def _set_progress_stage(self, name, counter=False):
831
recursive=self.recursive,
832
message_callback=self.message_callback,
833
timestamp=self.timestamp, timezone=self.timezone,
834
committer=self.committer,
835
allow_pointless=self.allow_pointless,
836
strict=self.strict, verbose=self.verbose,
837
local=self.local, reporter=self.reporter)
838
except errors.PointlessCommit:
839
return self.work_tree.get_reference_revision(file_id)
841
def _record_entry(self, path, file_id, specific_files, kind, name,
842
parent_id, definitely_changed, existing_ie, report_changes,
844
"Record the new inventory entry for a path if any."
845
# mutter('check %s {%s}', path, file_id)
846
# mutter('%s selected for commit', path)
847
if definitely_changed or existing_ie is None:
848
ie = make_entry(kind, name, parent_id, file_id)
850
ie = existing_ie.copy()
852
# For carried over entries we don't care about the fs hash - the repo
853
# isn't generating a sha, so we're not saving computation time.
854
_, _, fs_hash = self.builder.record_entry_contents(
855
ie, self.parent_invs, path, self.work_tree, content_summary)
857
self._report_change(ie, path)
859
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
862
def _report_change(self, ie, path):
863
"""Report a change to the user.
865
The change that has occurred is described relative to the basis
868
if (self.basis_inv.has_id(ie.file_id)):
869
basis_ie = self.basis_inv[ie.file_id]
872
change = ie.describe_change(basis_ie, ie)
873
if change in (InventoryEntry.RENAMED,
874
InventoryEntry.MODIFIED_AND_RENAMED):
875
old_path = self.basis_inv.id2path(ie.file_id)
876
self.reporter.renamed(change, old_path, path)
878
self.reporter.snapshot_change(change, path)
880
def _set_progress_stage(self, name, entries_title=None):
766
881
"""Set the progress stage and emit an update to the progress bar."""
767
882
self.pb_stage_name = name
768
883
self.pb_stage_count += 1
884
self.pb_entries_title = entries_title
885
if entries_title is not None:
770
886
self.pb_entries_count = 0
772
self.pb_entries_count = None
887
self.pb_entries_total = '?'
773
888
self._emit_progress()
775
890
def _next_progress_entry(self):
added
removed
bzrlib/commit.py
