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
60
from .branch import Branch
61
from .cleanup import OperationWithCleanups
63
from .errors import (BzrError,
67
from .osutils import (get_user_encoding,
69
minimum_path_selection,
72
from .trace import mutter, note, is_quiet
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
old_path = change[1][0]
101
new_path = change[1][1]
103
new_excluded = (new_path is not None and
104
is_inside_any(exclude, new_path))
106
old_excluded = (old_path is not None and
107
is_inside_any(exclude, old_path))
109
if old_excluded and new_excluded:
112
if old_excluded or new_excluded:
113
# TODO(jelmer): Perhaps raise an error here instead?
67
from bzrlib.branch import Branch
68
from bzrlib.cleanup import OperationWithCleanups
70
from bzrlib.errors import (BzrError, PointlessCommit,
74
from bzrlib.osutils import (get_user_encoding,
75
kind_marker, isdir,isfile, is_inside_any,
76
is_inside_or_parent_of_any,
77
minimum_path_selection,
78
quotefn, sha_file, split_lines,
81
from bzrlib.testament import Testament
82
from bzrlib.trace import mutter, note, warning, is_quiet
83
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
84
from bzrlib import symbol_versioning
85
from bzrlib.symbol_versioning import (deprecated_passed,
88
from bzrlib.workingtree import WorkingTree
89
from bzrlib.urlutils import unescape_for_display
119
93
class NullCommitReporter(object):
120
94
"""I report on progress of a commit."""
122
def started(self, revno, revid, location):
96
def started(self, revno, revid, location=None):
98
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
99
"to started.", DeprecationWarning,
125
103
def snapshot_change(self, change, path):
151
129
note(format, *args)
153
131
def snapshot_change(self, change, path):
154
if path == '' and change in (gettext('added'), gettext('modified')):
132
if path == '' and change in ('added', 'modified'):
156
134
self._note("%s %s", change, path)
158
def started(self, revno, rev_id, location):
160
gettext('Committing to: %s'),
161
unescape_for_display(location, 'utf-8'))
136
def started(self, revno, rev_id, location=None):
137
if location is not None:
138
location = ' to: ' + unescape_for_display(location, 'utf-8')
140
# When started was added, location was only made optional by
141
# accident. Matt Nordhoff 20071129
142
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
143
"to started.", DeprecationWarning,
146
self._note('Committing%s', location)
163
148
def completed(self, revno, rev_id):
164
self._note(gettext('Committed revision %d.'), revno)
165
# self._note goes to the console too; so while we want to log the
166
# rev_id, we can't trivially only log it. (See bug 526425). Long
167
# term we should rearrange the reporting structure, but for now
168
# we just mutter seperately. We mutter the revid and revno together
169
# so that concurrent bzr invocations won't lead to confusion.
170
mutter('Committed revid %s as revno %d.', rev_id, revno)
149
self._note('Committed revision %d.', revno)
172
151
def deleted(self, path):
173
self._note(gettext('deleted %s'), path)
152
self._note('deleted %s', path)
175
154
def missing(self, path):
176
self._note(gettext('missing %s'), path)
155
self._note('missing %s', path)
178
157
def renamed(self, change, old_path, new_path):
179
158
self._note('%s %s => %s', change, old_path, new_path)
197
176
def __init__(self,
200
179
"""Create a Commit object.
202
181
:param reporter: the default reporter to use or None to decide later
204
183
self.reporter = reporter
205
self.config_stack = config_stack
208
def update_revprops(revprops, branch, authors=None,
209
local=False, possible_master_transports=None):
212
if possible_master_transports is None:
213
possible_master_transports = []
214
if (not u'branch-nick' in revprops and
215
branch.repository._format.supports_storing_branch_nick):
216
revprops[u'branch-nick'] = branch._get_nick(
218
possible_master_transports)
219
if authors is not None:
220
if u'author' in revprops or u'authors' in revprops:
221
# XXX: maybe we should just accept one of them?
222
raise AssertionError('author property given twice')
224
for individual in authors:
225
if '\n' in individual:
226
raise AssertionError('\\n is not a valid character '
227
'in an author identity')
228
revprops[u'authors'] = '\n'.join(authors)
297
246
allow_pointless=allow_pointless,
300
250
working_tree=working_tree,
302
252
reporter=reporter,
303
254
message_callback=message_callback,
304
255
recursive=recursive,
306
possible_master_transports=possible_master_transports,
257
possible_master_transports=possible_master_transports)
309
259
def _commit(self, operation, message, timestamp, timezone, committer,
310
specific_files, rev_id, allow_pointless, strict, verbose,
311
working_tree, local, reporter, message_callback, recursive,
312
exclude, possible_master_transports, lossy):
260
specific_files, rev_id, allow_pointless, strict, verbose, revprops,
261
working_tree, local, reporter, config, message_callback, recursive,
262
exclude, possible_master_transports):
313
263
mutter('preparing to commit')
315
265
if working_tree is None:
391
352
self.pb_stage_count = 0
392
353
self.pb_stage_total = 5
393
354
if self.bound_branch:
394
# 2 extra stages: "Uploading data to master branch" and "Merging
395
# tags to master branch"
396
self.pb_stage_total += 2
355
self.pb_stage_total += 1
397
356
self.pb.show_pct = False
398
357
self.pb.show_spinner = False
399
358
self.pb.show_eta = False
400
359
self.pb.show_count = True
401
360
self.pb.show_bar = True
362
self._gather_parents()
403
363
# After a merge, a selected file commit is not supported.
404
364
# See 'bzr help merge' for an explanation as to why.
405
365
if len(self.parents) > 1 and self.specific_files is not None:
406
raise CannotCommitSelectedFileMerge(self.specific_files)
366
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
407
367
# Excludes are a form of selected file commit.
408
368
if len(self.parents) > 1 and self.exclude:
409
raise CannotCommitSelectedFileMerge(self.exclude)
369
raise errors.CannotCommitSelectedFileMerge(self.exclude)
411
371
# Collect the changes
412
372
self._set_progress_stage("Collecting changes", counter=True)
414
373
self.builder = self.branch.get_commit_builder(self.parents,
415
self.config_stack, timestamp, timezone, committer, self.revprops,
418
if self.builder.updates_branch and self.bound_branch:
420
raise AssertionError(
421
"bound branches not supported for commit builders "
422
"that update the branch")
374
self.config, timestamp, timezone, committer, revprops, rev_id)
377
self.builder.will_record_deletes()
425
378
# find the location being committed to
426
379
if self.bound_branch:
427
380
master_location = self.master_branch.base
448
401
# Add revision data to the local branch
449
402
self.rev_id = self.builder.commit(self.message)
451
except Exception as e:
452
405
mutter("aborting commit write group because of exception:")
453
406
trace.log_exception_quietly()
407
note("aborting commit write group: %r" % (e,))
454
408
self.builder.abort()
457
self._update_branches(old_revno, old_revid, new_revno)
411
self._process_pre_hooks(old_revno, new_revno)
413
# Upload revision data to the master.
414
# this will propagate merged revisions too if needed.
415
if self.bound_branch:
416
self._set_progress_stage("Uploading data to master branch")
417
# 'commit' to the master first so a timeout here causes the
418
# local branch to be out of date
419
self.master_branch.import_last_revision_info(
420
self.branch.repository, new_revno, self.rev_id)
422
# and now do the commit locally.
423
self.branch.set_last_revision_info(new_revno, self.rev_id)
459
425
# Make the working tree be up to date with the branch. This
460
426
# includes automatic changes scheduled to be made to the tree, such
461
427
# as updating its basis and unversioning paths that were missing.
462
self.work_tree.unversion(self.deleted_paths)
428
self.work_tree.unversion(self.deleted_ids)
463
429
self._set_progress_stage("Updating the working tree")
464
430
self.work_tree.update_basis_by_delta(self.rev_id,
465
431
self.builder.get_basis_delta())
467
433
self._process_post_hooks(old_revno, new_revno)
468
434
return self.rev_id
470
def _update_branches(self, old_revno, old_revid, new_revno):
471
"""Update the master and local branch to the new revision.
473
This will try to make sure that the master branch is updated
474
before the local branch.
476
:param old_revno: Revision number of master branch before the
478
:param old_revid: Tip of master branch before the commit
479
:param new_revno: Revision number of the new commit
481
if not self.builder.updates_branch:
482
self._process_pre_hooks(old_revno, new_revno)
484
# Upload revision data to the master.
485
# this will propagate merged revisions too if needed.
486
if self.bound_branch:
487
self._set_progress_stage("Uploading data to master branch")
488
# 'commit' to the master first so a timeout here causes the
489
# local branch to be out of date
490
(new_revno, self.rev_id) = self.master_branch.import_last_revision_info_and_tags(
491
self.branch, new_revno, self.rev_id, lossy=self._lossy)
493
self.branch.fetch(self.master_branch, self.rev_id)
495
# and now do the commit locally.
496
self.branch.set_last_revision_info(new_revno, self.rev_id)
499
self._process_pre_hooks(old_revno, new_revno)
501
# The commit builder will already have updated the branch,
503
self.branch.set_last_revision_info(old_revno, old_revid)
506
# Merge local tags to remote
507
if self.bound_branch:
508
self._set_progress_stage("Merging tags to master branch")
509
tag_updates, tag_conflicts = self.branch.tags.merge_to(
510
self.master_branch.tags)
512
warning_lines = [' ' + name for name, _, _ in tag_conflicts]
513
note( gettext("Conflicting tags in bound branch:\n{0}".format(
514
"\n".join(warning_lines))) )
516
436
def _select_reporter(self):
517
437
"""Select the CommitReporter to use."""
585
513
# this is so that we still consider the master branch
586
514
# - in a checkout scenario the tree may have no
587
515
# parents but the branch may do.
588
first_tree_parent = breezy.revision.NULL_REVISION
516
first_tree_parent = bzrlib.revision.NULL_REVISION
589
517
old_revno, master_last = self.master_branch.last_revision_info()
590
518
if master_last != first_tree_parent:
591
if master_last != breezy.revision.NULL_REVISION:
519
if master_last != bzrlib.revision.NULL_REVISION:
592
520
raise errors.OutOfDateTree(self.work_tree)
593
521
if self.branch.repository.has_revision(first_tree_parent):
594
522
new_revno = old_revno + 1
596
524
# ghost parents never appear in revision history.
598
return old_revno, master_last, new_revno
526
return old_revno,new_revno
600
528
def _process_pre_hooks(self, old_revno, new_revno):
601
529
"""Process any registered pre commit hooks."""
663
590
old_revno, old_revid, new_revno, self.rev_id,
664
591
tree_delta, future_tree)
593
def _gather_parents(self):
594
"""Record the parents of a merge for merge detection."""
595
# TODO: Make sure that this list doesn't contain duplicate
596
# entries and the order is preserved when doing this.
597
if self.use_record_iter_changes:
599
self.basis_inv = self.basis_tree.inventory
600
self.parent_invs = [self.basis_inv]
601
for revision in self.parents[1:]:
602
if self.branch.repository.has_revision(revision):
603
mutter('commit parent revision {%s}', revision)
604
inventory = self.branch.repository.get_inventory(revision)
605
self.parent_invs.append(inventory)
607
mutter('commit parent ghost revision {%s}', revision)
666
609
def _update_builder_with_changes(self):
667
610
"""Update the commit builder with the data about what has changed.
612
exclude = self.exclude
669
613
specific_files = self.specific_files
670
mutter("Selecting files for commit with filter %r", specific_files)
614
mutter("Selecting files for commit with filter %s", specific_files)
672
616
self._check_strict()
673
iter_changes = self.work_tree.iter_changes(self.basis_tree,
674
specific_files=specific_files)
676
iter_changes = filter_excluded(iter_changes, self.exclude)
677
iter_changes = self._filter_iter_changes(iter_changes)
678
for file_id, path, fs_hash in self.builder.record_iter_changes(
679
self.work_tree, self.basis_revid, iter_changes):
680
self.work_tree._observed_sha1(file_id, path, fs_hash)
617
if self.use_record_iter_changes:
618
iter_changes = self.work_tree.iter_changes(self.basis_tree,
619
specific_files=specific_files)
620
iter_changes = self._filter_iter_changes(iter_changes)
621
for file_id, path, fs_hash in self.builder.record_iter_changes(
622
self.work_tree, self.basis_revid, iter_changes):
623
self.work_tree._observed_sha1(file_id, path, fs_hash)
625
# Build the new inventory
626
self._populate_from_inventory()
627
self._record_unselected()
628
self._report_and_accumulate_deletes()
682
630
def _filter_iter_changes(self, iter_changes):
683
631
"""Process iter_changes.
718
664
if new_path is None:
719
665
reporter.deleted(old_path)
720
666
elif old_path is None:
721
reporter.snapshot_change(gettext('added'), new_path)
667
reporter.snapshot_change('added', new_path)
722
668
elif old_path != new_path:
723
reporter.renamed(gettext('renamed'), old_path, new_path)
669
reporter.renamed('renamed', old_path, new_path)
726
672
self.work_tree.branch.repository._format.rich_root_data):
727
673
# Don't report on changes to '' in non rich root
729
reporter.snapshot_change(gettext('modified'), new_path)
675
reporter.snapshot_change('modified', new_path)
730
676
self._next_progress_entry()
731
# Unversion files that were found to be deleted
732
self.deleted_paths = deleted_paths
677
# Unversion IDs that were found to be deleted
678
self.deleted_ids = deleted_ids
680
def _record_unselected(self):
681
# If specific files are selected, then all un-selected files must be
682
# recorded in their previous state. For more details, see
683
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
684
if self.specific_files or self.exclude:
685
specific_files = self.specific_files or []
686
for path, old_ie in self.basis_inv.iter_entries():
687
if old_ie.file_id in self.builder.new_inventory:
688
# already added - skip.
690
if (is_inside_any(specific_files, path)
691
and not is_inside_any(self.exclude, path)):
692
# was inside the selected path, and not excluded - if not
693
# present it has been deleted so skip.
695
# From here down it was either not selected, or was excluded:
696
# We preserve the entry unaltered.
698
# Note: specific file commits after a merge are currently
699
# prohibited. This test is for sanity/safety in case it's
700
# required after that changes.
701
if len(self.parents) > 1:
703
self.builder.record_entry_contents(ie, self.parent_invs, path,
704
self.basis_tree, None)
706
def _report_and_accumulate_deletes(self):
707
if (isinstance(self.basis_inv, Inventory)
708
and isinstance(self.builder.new_inventory, Inventory)):
709
# the older Inventory classes provide a _byid dict, and building a
710
# set from the keys of this dict is substantially faster than even
711
# getting a set of ids from the inventory
713
# <lifeless> set(dict) is roughly the same speed as
714
# set(iter(dict)) and both are significantly slower than
716
deleted_ids = set(self.basis_inv._byid.keys()) - \
717
set(self.builder.new_inventory._byid.keys())
719
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
721
self.any_entries_deleted = True
722
deleted = [(self.basis_tree.id2path(file_id), file_id)
723
for file_id in deleted_ids]
725
# XXX: this is not quite directory-order sorting
726
for path, file_id in deleted:
727
self.builder.record_delete(path, file_id)
728
self.reporter.deleted(path)
734
730
def _check_strict(self):
735
731
# XXX: when we use iter_changes this would likely be faster if
740
736
for unknown in self.work_tree.unknowns():
741
737
raise StrictCommitFailed()
739
def _populate_from_inventory(self):
740
"""Populate the CommitBuilder by walking the working tree inventory."""
741
# Build the revision inventory.
743
# This starts by creating a new empty inventory. Depending on
744
# which files are selected for commit, and what is present in the
745
# current tree, the new inventory is populated. inventory entries
746
# which are candidates for modification have their revision set to
747
# None; inventory entries that are carried over untouched have their
748
# revision set to their prior value.
750
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
751
# results to create a new inventory at the same time, which results
752
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
755
specific_files = self.specific_files
756
exclude = self.exclude
757
report_changes = self.reporter.is_verbose()
759
# A tree of paths that have been deleted. E.g. if foo/bar has been
760
# deleted, then we have {'foo':{'bar':{}}}
762
# XXX: Note that entries may have the wrong kind because the entry does
763
# not reflect the status on disk.
764
work_inv = self.work_tree.inventory
765
# NB: entries will include entries within the excluded ids/paths
766
# because iter_entries_by_dir has no 'exclude' facility today.
767
entries = work_inv.iter_entries_by_dir(
768
specific_file_ids=self.specific_file_ids, yield_parents=True)
769
for path, existing_ie in entries:
770
file_id = existing_ie.file_id
771
name = existing_ie.name
772
parent_id = existing_ie.parent_id
773
kind = existing_ie.kind
774
# Skip files that have been deleted from the working tree.
775
# The deleted path ids are also recorded so they can be explicitly
778
path_segments = splitpath(path)
779
deleted_dict = deleted_paths
780
for segment in path_segments:
781
deleted_dict = deleted_dict.get(segment, None)
783
# We either took a path not present in the dict
784
# (deleted_dict was None), or we've reached an empty
785
# child dir in the dict, so are now a sub-path.
789
if deleted_dict is not None:
790
# the path has a deleted parent, do not add it.
792
if exclude and is_inside_any(exclude, path):
793
# Skip excluded paths. Excluded paths are processed by
794
# _update_builder_with_changes.
796
content_summary = self.work_tree.path_content_summary(path)
797
kind = content_summary[0]
798
# Note that when a filter of specific files is given, we must only
799
# skip/record deleted files matching that filter.
800
if not specific_files or is_inside_any(specific_files, path):
801
if kind == 'missing':
802
if not deleted_paths:
803
# path won't have been split yet.
804
path_segments = splitpath(path)
805
deleted_dict = deleted_paths
806
for segment in path_segments:
807
deleted_dict = deleted_dict.setdefault(segment, {})
808
self.reporter.missing(path)
809
self._next_progress_entry()
810
deleted_ids.append(file_id)
812
# TODO: have the builder do the nested commit just-in-time IF and
814
if kind == 'tree-reference':
815
# enforce repository nested tree policy.
816
if (not self.work_tree.supports_tree_reference() or
817
# repository does not support it either.
818
not self.branch.repository._format.supports_tree_reference):
820
content_summary = (kind, None, None, None)
821
elif self.recursive == 'down':
822
nested_revision_id = self._commit_nested_tree(
824
content_summary = (kind, None, None, nested_revision_id)
826
nested_revision_id = self.work_tree.get_reference_revision(file_id)
827
content_summary = (kind, None, None, nested_revision_id)
829
# Record an entry for this item
830
# Note: I don't particularly want to have the existing_ie
831
# parameter but the test suite currently (28-Jun-07) breaks
832
# without it thanks to a unicode normalisation issue. :-(
833
definitely_changed = kind != existing_ie.kind
834
self._record_entry(path, file_id, specific_files, kind, name,
835
parent_id, definitely_changed, existing_ie, report_changes,
838
# Unversion IDs that were found to be deleted
839
self.deleted_ids = deleted_ids
743
841
def _commit_nested_tree(self, file_id, path):
744
842
"Commit a nested tree."
745
sub_tree = self.work_tree.get_nested_tree(path, file_id)
843
sub_tree = self.work_tree.get_nested_tree(file_id, path)
746
844
# FIXME: be more comprehensive here:
747
845
# this works when both trees are in --trees repository,
748
846
# but when both are bound to a different repository,
762
860
allow_pointless=self.allow_pointless,
763
861
strict=self.strict, verbose=self.verbose,
764
862
local=self.local, reporter=self.reporter)
765
except PointlessCommit:
766
return self.work_tree.get_reference_revision(path, file_id)
863
except errors.PointlessCommit:
864
return self.work_tree.get_reference_revision(file_id)
866
def _record_entry(self, path, file_id, specific_files, kind, name,
867
parent_id, definitely_changed, existing_ie, report_changes,
869
"Record the new inventory entry for a path if any."
870
# mutter('check %s {%s}', path, file_id)
871
# mutter('%s selected for commit', path)
872
if definitely_changed or existing_ie is None:
873
ie = make_entry(kind, name, parent_id, file_id)
875
ie = existing_ie.copy()
877
# For carried over entries we don't care about the fs hash - the repo
878
# isn't generating a sha, so we're not saving computation time.
879
_, _, fs_hash = self.builder.record_entry_contents(
880
ie, self.parent_invs, path, self.work_tree, content_summary)
882
self._report_change(ie, path)
884
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
887
def _report_change(self, ie, path):
888
"""Report a change to the user.
890
The change that has occurred is described relative to the basis
893
if (self.basis_inv.has_id(ie.file_id)):
894
basis_ie = self.basis_inv[ie.file_id]
897
change = ie.describe_change(basis_ie, ie)
898
if change in (InventoryEntry.RENAMED,
899
InventoryEntry.MODIFIED_AND_RENAMED):
900
old_path = self.basis_inv.id2path(ie.file_id)
901
self.reporter.renamed(change, old_path, path)
902
self._next_progress_entry()
904
if change == 'unchanged':
906
self.reporter.snapshot_change(change, path)
907
self._next_progress_entry()
768
909
def _set_progress_stage(self, name, counter=False):
769
910
"""Set the progress stage and emit an update to the progress bar."""
783
924
def _emit_progress(self):
784
925
if self.pb_entries_count is not None:
785
text = gettext("{0} [{1}] - Stage").format(self.pb_stage_name,
926
text = "%s [%d] - Stage" % (self.pb_stage_name,
786
927
self.pb_entries_count)
788
text = gettext("%s - Stage") % (self.pb_stage_name, )
929
text = "%s - Stage" % (self.pb_stage_name, )
789
930
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
932
def _set_specific_file_ids(self):
933
"""populate self.specific_file_ids if we will use it."""
934
if not self.use_record_iter_changes:
935
# If provided, ensure the specified files are versioned
936
if self.specific_files is not None:
937
# Note: This routine is being called because it raises
938
# PathNotVersionedError as a side effect of finding the IDs. We
939
# later use the ids we found as input to the working tree
940
# inventory iterator, so we only consider those ids rather than
941
# examining the whole tree again.
942
# XXX: Dont we have filter_unversioned to do this more
944
self.specific_file_ids = tree.find_ids_across_trees(
945
self.specific_files, [self.basis_tree, self.work_tree])
947
self.specific_file_ids = None
added
removed
bzrlib/commit.py
