1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
1
# Copyright (C) 2005 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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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.
57
from cStringIO import StringIO
65
from bzrlib.branch import Branch
67
from bzrlib.errors import (BzrError, PointlessCommit,
71
from bzrlib.osutils import (get_user_encoding,
72
kind_marker, isdir,isfile, is_inside_any,
73
is_inside_or_parent_of_any,
74
minimum_path_selection,
75
quotefn, sha_file, split_lines,
78
from bzrlib.testament import Testament
79
from bzrlib.trace import mutter, note, warning, is_quiet
80
from bzrlib.inventory import InventoryEntry, make_entry
81
from bzrlib import symbol_versioning
82
from bzrlib.symbol_versioning import (deprecated_passed,
85
from bzrlib.workingtree import WorkingTree
86
from bzrlib.urlutils import unescape_for_display
90
class NullCommitReporter(object):
91
"""I report on progress of a commit."""
93
def started(self, revno, revid, location=None):
95
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
96
"to started.", DeprecationWarning,
100
def snapshot_change(self, change, path):
103
def completed(self, revno, rev_id):
106
def deleted(self, file_id):
109
def escaped(self, escape_count, message):
112
def missing(self, path):
115
def renamed(self, change, old_path, new_path):
118
def is_verbose(self):
122
class ReportCommitToLog(NullCommitReporter):
124
def _note(self, format, *args):
127
Subclasses may choose to override this method.
131
def snapshot_change(self, change, path):
132
if change == 'unchanged':
134
if change == 'added' and path == '':
136
self._note("%s %s", change, path)
138
def started(self, revno, rev_id, location=None):
139
if location is not None:
140
location = ' to: ' + unescape_for_display(location, 'utf-8')
142
# When started was added, location was only made optional by
143
# accident. Matt Nordhoff 20071129
144
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
145
"to started.", DeprecationWarning,
148
self._note('Committing%s', location)
150
def completed(self, revno, rev_id):
151
self._note('Committed revision %d.', revno)
153
def deleted(self, file_id):
154
self._note('deleted %s', file_id)
156
def escaped(self, escape_count, message):
157
self._note("replaced %d control characters in message", escape_count)
159
def missing(self, path):
160
self._note('missing %s', path)
162
def renamed(self, change, old_path, new_path):
163
self._note('%s %s => %s', change, old_path, new_path)
165
def is_verbose(self):
169
class Commit(object):
170
"""Task of committing a new revision.
172
This is a MethodObject: it accumulates state as the commit is
173
prepared, and then it is discarded. It doesn't represent
174
historical revisions, just the act of recording a new one.
177
Modified to hold a list of files that have been deleted from
178
the working directory; these should be removed from the
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.
184
"""Create a Commit object.
186
:param reporter: the default reporter to use or None to decide later
188
self.reporter = reporter
198
allow_pointless=True,
206
message_callback=None,
209
"""Commit working copy as a new revision.
211
:param message: the commit message (it or message_callback is required)
213
:param timestamp: if not None, seconds-since-epoch for a
214
postdated/predated commit.
216
:param specific_files: If true, commit only those files.
218
:param rev_id: If set, use this as the new revision id.
219
Useful for test or import commands that need to tightly
220
control what revisions are assigned. If you duplicate
221
a revision id that exists elsewhere it is your own fault.
222
If null (default), a time/random revision id is generated.
224
:param allow_pointless: If true (default), commit even if nothing
225
has changed and no merges are recorded.
227
:param strict: If true, don't allow a commit if the working tree
228
contains unknown files.
230
:param revprops: Properties for new revision
231
:param local: Perform a local only commit.
232
:param reporter: the reporter to use or None for the default
233
:param verbose: if True and the reporter is not None, report everything
234
:param recursive: If set to 'down', commit in any subtrees that have
235
pending changes of any sort during this commit.
236
:param exclude: None or a list of relative paths to exclude from the
237
commit. Pending changes to excluded files will be ignored by the
240
mutter('preparing to commit')
242
if working_tree is None:
243
raise BzrError("working_tree must be passed into commit().")
245
self.work_tree = working_tree
246
self.branch = self.work_tree.branch
247
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
248
if not self.branch.repository.supports_rich_root():
249
raise errors.RootNotRich()
250
if message_callback is None:
251
if message is not None:
252
if isinstance(message, str):
253
message = message.decode(get_user_encoding())
254
message_callback = lambda x: message
256
raise BzrError("The message or message_callback keyword"
257
" parameter is required for commit().")
259
self.bound_branch = None
260
self.any_entries_changed = False
261
self.any_entries_deleted = False
262
if exclude is not None:
263
self.exclude = sorted(
264
minimum_path_selection(exclude))
268
self.master_branch = None
269
self.master_locked = False
270
self.recursive = recursive
272
if specific_files is not None:
273
self.specific_files = sorted(
274
minimum_path_selection(specific_files))
276
self.specific_files = None
277
self.specific_file_ids = None
278
self.allow_pointless = allow_pointless
279
self.revprops = revprops
280
self.message_callback = message_callback
281
self.timestamp = timestamp
282
self.timezone = timezone
283
self.committer = committer
285
self.verbose = verbose
286
# accumulates an inventory delta to the basis entry, so we can make
287
# just the necessary updates to the workingtree's cached basis.
288
self._basis_delta = []
290
self.work_tree.lock_write()
291
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
292
self.basis_revid = self.work_tree.last_revision()
293
self.basis_tree = self.work_tree.basis_tree()
294
self.basis_tree.lock_read()
296
# Cannot commit with conflicts present.
297
if len(self.work_tree.conflicts()) > 0:
298
raise ConflictsInTree
300
# Setup the bound branch variables as needed.
301
self._check_bound_branch()
303
# Check that the working tree is up to date
304
old_revno, new_revno = self._check_out_of_date_tree()
306
# Complete configuration setup
307
if reporter is not None:
308
self.reporter = reporter
309
elif self.reporter is None:
310
self.reporter = self._select_reporter()
311
if self.config is None:
312
self.config = self.branch.get_config()
314
# If provided, ensure the specified files are versioned
315
if self.specific_files is not None:
316
# Note: This routine is being called because it raises
317
# PathNotVersionedError as a side effect of finding the IDs. We
318
# later use the ids we found as input to the working tree
319
# inventory iterator, so we only consider those ids rather than
320
# examining the whole tree again.
321
# XXX: Dont we have filter_unversioned to do this more
323
self.specific_file_ids = tree.find_ids_across_trees(
324
specific_files, [self.basis_tree, self.work_tree])
326
# Setup the progress bar. As the number of files that need to be
327
# committed in unknown, progress is reported as stages.
328
# We keep track of entries separately though and include that
329
# information in the progress bar during the relevant stages.
330
self.pb_stage_name = ""
331
self.pb_stage_count = 0
332
self.pb_stage_total = 5
333
if self.bound_branch:
334
self.pb_stage_total += 1
335
self.pb.show_pct = False
336
self.pb.show_spinner = False
337
self.pb.show_eta = False
338
self.pb.show_count = True
339
self.pb.show_bar = True
341
self.basis_inv = self.basis_tree.inventory
342
self._gather_parents()
343
# After a merge, a selected file commit is not supported.
344
# See 'bzr help merge' for an explanation as to why.
345
if len(self.parents) > 1 and self.specific_files:
346
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
347
# Excludes are a form of selected file commit.
348
if len(self.parents) > 1 and self.exclude:
349
raise errors.CannotCommitSelectedFileMerge(self.exclude)
351
# Collect the changes
352
self._set_progress_stage("Collecting changes",
353
entries_title="Directory")
354
self.builder = self.branch.get_commit_builder(self.parents,
355
self.config, timestamp, timezone, committer, revprops, rev_id)
358
# find the location being committed to
359
if self.bound_branch:
360
master_location = self.master_branch.base
362
master_location = self.branch.base
364
# report the start of the commit
365
self.reporter.started(new_revno, self.rev_id, master_location)
367
self._update_builder_with_changes()
368
self._report_and_accumulate_deletes()
369
self._check_pointless()
371
# TODO: Now the new inventory is known, check for conflicts.
372
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
373
# weave lines, because nothing should be recorded until it is known
374
# that commit will succeed.
375
self._set_progress_stage("Saving data locally")
376
self.builder.finish_inventory()
378
# Prompt the user for a commit message if none provided
379
message = message_callback(self)
380
self.message = message
381
self._escape_commit_message()
383
# Add revision data to the local branch
384
self.rev_id = self.builder.commit(self.message)
390
self._process_pre_hooks(old_revno, new_revno)
392
# Upload revision data to the master.
393
# this will propagate merged revisions too if needed.
394
if self.bound_branch:
395
if not self.master_branch.repository.has_same_location(
396
self.branch.repository):
397
self._set_progress_stage("Uploading data to master branch")
398
self.master_branch.repository.fetch(self.branch.repository,
399
revision_id=self.rev_id)
400
# now the master has the revision data
401
# 'commit' to the master first so a timeout here causes the
402
# local branch to be out of date
403
self.master_branch.set_last_revision_info(new_revno,
406
# and now do the commit locally.
407
self.branch.set_last_revision_info(new_revno, self.rev_id)
409
# Make the working tree up to date with the branch
410
self._set_progress_stage("Updating the working tree")
411
self.work_tree.update_basis_by_delta(self.rev_id,
413
self.reporter.completed(new_revno, self.rev_id)
414
self._process_post_hooks(old_revno, new_revno)
419
def _select_reporter(self):
420
"""Select the CommitReporter to use."""
422
return NullCommitReporter()
423
return ReportCommitToLog()
425
def _check_pointless(self):
426
if self.allow_pointless:
428
# A merge with no effect on files
429
if len(self.parents) > 1:
431
# TODO: we could simplify this by using self._basis_delta.
433
# The initial commit adds a root directory, but this in itself is not
434
# a worthwhile commit.
435
if (self.basis_revid == revision.NULL_REVISION and
436
len(self.builder.new_inventory) == 1):
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):
437
101
raise PointlessCommit()
438
# If length == 1, then we only have the root entry. Which means
439
# that there is no real difference (only the root could be different)
440
# unless deletes occured, in which case the length is irrelevant.
441
if (self.any_entries_deleted or
442
(len(self.builder.new_inventory) != 1 and
443
self.any_entries_changed)):
445
raise PointlessCommit()
447
def _check_bound_branch(self):
448
"""Check to see if the local branch is bound.
450
If it is bound, then most of the commit will actually be
451
done using the remote branch as the target branch.
452
Only at the end will the local branch be updated.
454
if self.local and not self.branch.get_bound_location():
455
raise errors.LocalRequiresBoundBranch()
458
self.master_branch = self.branch.get_master_branch()
460
if not self.master_branch:
461
# make this branch the reference branch for out of date checks.
462
self.master_branch = self.branch
465
# If the master branch is bound, we must fail
466
master_bound_location = self.master_branch.get_bound_location()
467
if master_bound_location:
468
raise errors.CommitToDoubleBoundBranch(self.branch,
469
self.master_branch, master_bound_location)
471
# TODO: jam 20051230 We could automatically push local
472
# commits to the remote branch if they would fit.
473
# But for now, just require remote to be identical
476
# Make sure the local branch is identical to the master
477
master_info = self.master_branch.last_revision_info()
478
local_info = self.branch.last_revision_info()
479
if local_info != master_info:
480
raise errors.BoundBranchOutOfDate(self.branch,
483
# Now things are ready to change the master branch
485
self.bound_branch = self.branch
486
self.master_branch.lock_write()
487
self.master_locked = True
489
def _check_out_of_date_tree(self):
490
"""Check that the working tree is up to date.
492
:return: old_revision_number,new_revision_number tuple
495
first_tree_parent = self.work_tree.get_parent_ids()[0]
497
# if there are no parents, treat our parent as 'None'
498
# this is so that we still consider the master branch
499
# - in a checkout scenario the tree may have no
500
# parents but the branch may do.
501
first_tree_parent = bzrlib.revision.NULL_REVISION
502
old_revno, master_last = self.master_branch.last_revision_info()
503
if master_last != first_tree_parent:
504
if master_last != bzrlib.revision.NULL_REVISION:
505
raise errors.OutOfDateTree(self.work_tree)
506
if self.branch.repository.has_revision(first_tree_parent):
507
new_revno = old_revno + 1
509
# ghost parents never appear in revision history.
511
return old_revno,new_revno
513
def _process_pre_hooks(self, old_revno, new_revno):
514
"""Process any registered pre commit hooks."""
515
self._set_progress_stage("Running pre_commit hooks")
516
self._process_hooks("pre_commit", old_revno, new_revno)
518
def _process_post_hooks(self, old_revno, new_revno):
519
"""Process any registered post commit hooks."""
520
# Process the post commit hooks, if any
521
self._set_progress_stage("Running post_commit hooks")
522
# old style commit hooks - should be deprecated ? (obsoleted in
524
if self.config.post_commit() is not None:
525
hooks = self.config.post_commit().split(' ')
526
# this would be nicer with twisted.python.reflect.namedAny
528
result = eval(hook + '(branch, rev_id)',
529
{'branch':self.branch,
531
'rev_id':self.rev_id})
532
# process new style post commit hooks
533
self._process_hooks("post_commit", old_revno, new_revno)
535
def _process_hooks(self, hook_name, old_revno, new_revno):
536
if not Branch.hooks[hook_name]:
539
# new style commit hooks:
540
if not self.bound_branch:
541
hook_master = self.branch
544
hook_master = self.master_branch
545
hook_local = self.branch
546
# With bound branches, when the master is behind the local branch,
547
# the 'old_revno' and old_revid values here are incorrect.
548
# XXX: FIXME ^. RBC 20060206
550
old_revid = self.parents[0]
552
old_revid = bzrlib.revision.NULL_REVISION
554
if hook_name == "pre_commit":
555
future_tree = self.builder.revision_tree()
556
tree_delta = future_tree.changes_from(self.basis_tree,
559
for hook in Branch.hooks[hook_name]:
560
# show the running hook in the progress bar. As hooks may
561
# end up doing nothing (e.g. because they are not configured by
562
# the user) this is still showing progress, not showing overall
563
# actions - its up to each plugin to show a UI if it want's to
564
# (such as 'Emailing diff to foo@example.com').
565
self.pb_stage_name = "Running %s hooks [%s]" % \
566
(hook_name, Branch.hooks.get_hook_name(hook))
567
self._emit_progress()
568
if 'hooks' in debug.debug_flags:
569
mutter("Invoking commit hook: %r", hook)
570
if hook_name == "post_commit":
571
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
573
elif hook_name == "pre_commit":
574
hook(hook_local, hook_master,
575
old_revno, old_revid, new_revno, self.rev_id,
576
tree_delta, future_tree)
579
"""Cleanup any open locks, progress bars etc."""
580
cleanups = [self._cleanup_bound_branch,
581
self.basis_tree.unlock,
582
self.work_tree.unlock,
584
found_exception = None
585
for cleanup in cleanups:
588
# we want every cleanup to run no matter what.
589
# so we have a catchall here, but we will raise the
590
# last encountered exception up the stack: and
591
# typically this will be useful enough.
594
if found_exception is not None:
595
# don't do a plan raise, because the last exception may have been
596
# trashed, e is our sure-to-work exception even though it loses the
597
# full traceback. XXX: RBC 20060421 perhaps we could check the
598
# exc_info and if its the same one do a plain raise otherwise
599
# 'raise e' as we do now.
602
def _cleanup_bound_branch(self):
603
"""Executed at the end of a try/finally to cleanup a bound branch.
605
If the branch wasn't bound, this is a no-op.
606
If it was, it resents self.branch to the local branch, instead
609
if not self.bound_branch:
611
if self.master_locked:
612
self.master_branch.unlock()
614
def _escape_commit_message(self):
615
"""Replace xml-incompatible control characters."""
616
# FIXME: RBC 20060419 this should be done by the revision
617
# serialiser not by commit. Then we can also add an unescaper
618
# in the deserializer and start roundtripping revision messages
619
# precisely. See repository_implementations/test_repository.py
621
# Python strings can include characters that can't be
622
# represented in well-formed XML; escape characters that
623
# aren't listed in the XML specification
624
# (http://www.w3.org/TR/REC-xml/#NT-Char).
625
self.message, escape_count = re.subn(
626
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
627
lambda match: match.group(0).encode('unicode_escape'),
630
self.reporter.escaped(escape_count, self.message)
632
def _gather_parents(self):
633
"""Record the parents of a merge for merge detection."""
634
# TODO: Make sure that this list doesn't contain duplicate
635
# entries and the order is preserved when doing this.
636
self.parents = self.work_tree.get_parent_ids()
637
self.parent_invs = [self.basis_inv]
638
for revision in self.parents[1:]:
639
if self.branch.repository.has_revision(revision):
640
mutter('commit parent revision {%s}', revision)
641
inventory = self.branch.repository.get_inventory(revision)
642
self.parent_invs.append(inventory)
644
mutter('commit parent ghost revision {%s}', revision)
646
def _update_builder_with_changes(self):
647
"""Update the commit builder with the data about what has changed.
649
# Build the revision inventory.
651
# This starts by creating a new empty inventory. Depending on
652
# which files are selected for commit, and what is present in the
653
# current tree, the new inventory is populated. inventory entries
654
# which are candidates for modification have their revision set to
655
# None; inventory entries that are carried over untouched have their
656
# revision set to their prior value.
658
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
659
# results to create a new inventory at the same time, which results
660
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
663
exclude = self.exclude
664
specific_files = self.specific_files or []
665
mutter("Selecting files for commit with filter %s", specific_files)
667
# Build the new inventory
668
self._populate_from_inventory()
670
# If specific files are selected, then all un-selected files must be
671
# recorded in their previous state. For more details, see
672
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
673
if specific_files or exclude:
674
for path, old_ie in self.basis_inv.iter_entries():
675
if old_ie.file_id in self.builder.new_inventory:
676
# already added - skip.
678
if (is_inside_any(specific_files, path)
679
and not is_inside_any(exclude, path)):
680
# was inside the selected path, and not excluded - if not
681
# present it has been deleted so skip.
683
# From here down it was either not selected, or was excluded:
684
if old_ie.kind == 'directory':
685
self._next_progress_entry()
686
# We preserve the entry unaltered.
688
# Note: specific file commits after a merge are currently
689
# prohibited. This test is for sanity/safety in case it's
690
# required after that changes.
691
if len(self.parents) > 1:
693
delta, version_recorded, _ = self.builder.record_entry_contents(
694
ie, self.parent_invs, path, self.basis_tree, None)
696
self.any_entries_changed = True
698
self._basis_delta.append(delta)
700
def _report_and_accumulate_deletes(self):
701
# XXX: Could the list of deleted paths and ids be instead taken from
702
# _populate_from_inventory?
703
deleted_ids = set(self.basis_inv._byid.keys()) - \
704
set(self.builder.new_inventory._byid.keys())
706
self.any_entries_deleted = True
707
deleted = [(self.basis_tree.id2path(file_id), file_id)
708
for file_id in deleted_ids]
710
# XXX: this is not quite directory-order sorting
711
for path, file_id in deleted:
712
self._basis_delta.append((path, None, file_id, None))
713
self.reporter.deleted(path)
715
def _populate_from_inventory(self):
716
"""Populate the CommitBuilder by walking the working tree inventory."""
718
# raise an exception as soon as we find a single unknown.
719
for unknown in self.work_tree.unknowns():
720
raise StrictCommitFailed()
722
specific_files = self.specific_files
723
exclude = self.exclude
724
report_changes = self.reporter.is_verbose()
726
# A tree of paths that have been deleted. E.g. if foo/bar has been
727
# deleted, then we have {'foo':{'bar':{}}}
729
# XXX: Note that entries may have the wrong kind because the entry does
730
# not reflect the status on disk.
731
work_inv = self.work_tree.inventory
732
# NB: entries will include entries within the excluded ids/paths
733
# because iter_entries_by_dir has no 'exclude' facility today.
734
entries = work_inv.iter_entries_by_dir(
735
specific_file_ids=self.specific_file_ids, yield_parents=True)
736
for path, existing_ie in entries:
737
file_id = existing_ie.file_id
738
name = existing_ie.name
739
parent_id = existing_ie.parent_id
740
kind = existing_ie.kind
741
if kind == 'directory':
742
self._next_progress_entry()
743
# Skip files that have been deleted from the working tree.
744
# The deleted path ids are also recorded so they can be explicitly
747
path_segments = splitpath(path)
748
deleted_dict = deleted_paths
749
for segment in path_segments:
750
deleted_dict = deleted_dict.get(segment, None)
752
# We either took a path not present in the dict
753
# (deleted_dict was None), or we've reached an empty
754
# child dir in the dict, so are now a sub-path.
758
if deleted_dict is not None:
759
# the path has a deleted parent, do not add it.
761
if exclude and is_inside_any(exclude, path):
762
# Skip excluded paths. Excluded paths are processed by
763
# _update_builder_with_changes.
765
content_summary = self.work_tree.path_content_summary(path)
766
# Note that when a filter of specific files is given, we must only
767
# skip/record deleted files matching that filter.
768
if not specific_files or is_inside_any(specific_files, path):
769
if content_summary[0] == 'missing':
770
if not deleted_paths:
771
# path won't have been split yet.
772
path_segments = splitpath(path)
773
deleted_dict = deleted_paths
774
for segment in path_segments:
775
deleted_dict = deleted_dict.setdefault(segment, {})
776
self.reporter.missing(path)
777
deleted_ids.append(file_id)
779
# TODO: have the builder do the nested commit just-in-time IF and
781
if content_summary[0] == 'tree-reference':
782
# enforce repository nested tree policy.
783
if (not self.work_tree.supports_tree_reference() or
784
# repository does not support it either.
785
not self.branch.repository._format.supports_tree_reference):
786
content_summary = ('directory',) + content_summary[1:]
787
kind = content_summary[0]
788
# TODO: specific_files filtering before nested tree processing
789
if kind == 'tree-reference':
790
if self.recursive == 'down':
791
nested_revision_id = self._commit_nested_tree(
793
content_summary = content_summary[:3] + (
796
content_summary = content_summary[:3] + (
797
self.work_tree.get_reference_revision(file_id),)
799
# Record an entry for this item
800
# Note: I don't particularly want to have the existing_ie
801
# parameter but the test suite currently (28-Jun-07) breaks
802
# without it thanks to a unicode normalisation issue. :-(
803
definitely_changed = kind != existing_ie.kind
804
self._record_entry(path, file_id, specific_files, kind, name,
805
parent_id, definitely_changed, existing_ie, report_changes,
808
# Unversion IDs that were found to be deleted
809
self.work_tree.unversion(deleted_ids)
811
def _commit_nested_tree(self, file_id, path):
812
"Commit a nested tree."
813
sub_tree = self.work_tree.get_nested_tree(file_id, path)
814
# FIXME: be more comprehensive here:
815
# this works when both trees are in --trees repository,
816
# but when both are bound to a different repository,
817
# it fails; a better way of approaching this is to
818
# finally implement the explicit-caches approach design
819
# a while back - RBC 20070306.
820
if sub_tree.branch.repository.has_same_location(
821
self.work_tree.branch.repository):
822
sub_tree.branch.repository = \
823
self.work_tree.branch.repository
825
return sub_tree.commit(message=None, revprops=self.revprops,
826
recursive=self.recursive,
827
message_callback=self.message_callback,
828
timestamp=self.timestamp, timezone=self.timezone,
829
committer=self.committer,
830
allow_pointless=self.allow_pointless,
831
strict=self.strict, verbose=self.verbose,
832
local=self.local, reporter=self.reporter)
833
except errors.PointlessCommit:
834
return self.work_tree.get_reference_revision(file_id)
836
def _record_entry(self, path, file_id, specific_files, kind, name,
837
parent_id, definitely_changed, existing_ie, report_changes,
839
"Record the new inventory entry for a path if any."
840
# mutter('check %s {%s}', path, file_id)
841
# mutter('%s selected for commit', path)
842
if definitely_changed or existing_ie is None:
843
ie = make_entry(kind, name, parent_id, file_id)
845
ie = existing_ie.copy()
847
# For carried over entries we don't care about the fs hash - the repo
848
# isn't generating a sha, so we're not saving computation time.
849
delta, version_recorded, fs_hash = self.builder.record_entry_contents(
850
ie, self.parent_invs, path, self.work_tree, content_summary)
852
self._basis_delta.append(delta)
854
self.any_entries_changed = True
856
self._report_change(ie, path)
858
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
861
def _report_change(self, ie, path):
862
"""Report a change to the user.
864
The change that has occurred is described relative to the basis
867
if (self.basis_inv.has_id(ie.file_id)):
868
basis_ie = self.basis_inv[ie.file_id]
871
change = ie.describe_change(basis_ie, ie)
872
if change in (InventoryEntry.RENAMED,
873
InventoryEntry.MODIFIED_AND_RENAMED):
874
old_path = self.basis_inv.id2path(ie.file_id)
875
self.reporter.renamed(change, old_path, path)
877
self.reporter.snapshot_change(change, path)
879
def _set_progress_stage(self, name, entries_title=None):
880
"""Set the progress stage and emit an update to the progress bar."""
881
self.pb_stage_name = name
882
self.pb_stage_count += 1
883
self.pb_entries_title = entries_title
884
if entries_title is not None:
885
self.pb_entries_count = 0
886
self.pb_entries_total = '?'
887
self._emit_progress()
889
def _next_progress_entry(self):
890
"""Emit an update to the progress bar and increment the entry count."""
891
self.pb_entries_count += 1
892
self._emit_progress()
894
def _emit_progress(self):
895
if self.pb_entries_title:
896
if self.pb_entries_total == '?':
897
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
898
self.pb_entries_title, self.pb_entries_count)
900
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
901
self.pb_entries_title, self.pb_entries_count,
902
str(self.pb_entries_total))
904
text = "%s - Stage" % (self.pb_stage_name)
905
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
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 osutils import isdir, isfile, sha_string, quotefn, \
215
local_time_offset, username, kind_marker, is_inside_any
217
from branch import gen_file_id
218
from errors import BzrError
219
from 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
added
removed
bzrlib/commit.py
