1
# Copyright (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2008 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
18
## XXX: Can we do any better about making interrupted commits change
21
## XXX: If we merged two versions of a file then we still need to
22
## create a new version representing that merge, even if it didn't
23
## change from the parent.
25
## TODO: Read back the just-generated changeset, and make sure it
26
## applies and recreates the right state.
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.
37
from binascii import hexlify
38
57
from cStringIO import StringIO
40
from bzrlib.osutils import (local_time_offset, username,
41
rand_bytes, compact_date, user_email,
42
kind_marker, is_inside_any, quotefn,
43
sha_string, sha_file, isdir, isfile)
44
from bzrlib.branch import gen_file_id
45
from bzrlib.errors import BzrError, PointlessCommit
46
from bzrlib.revision import Revision, RevisionReference
47
from bzrlib.trace import mutter, note
48
from bzrlib.xml5 import serializer_v5
49
from bzrlib.inventory import Inventory
50
from bzrlib.delta import compare_trees
51
from bzrlib.weave import Weave
52
from bzrlib.weavefile import read_weave, write_weave_v5
53
from bzrlib.atomicfile import AtomicFile
56
def commit(*args, **kwargs):
57
"""Commit a new revision to a branch.
59
Function-style interface for convenience of old callers.
61
New code should use the Commit class instead.
63
Commit().commit(*args, **kwargs)
67
from bzrlib.branch import Branch
69
from bzrlib.errors import (BzrError, PointlessCommit,
73
from bzrlib.osutils import (get_user_encoding,
74
kind_marker, isdir,isfile, is_inside_any,
75
is_inside_or_parent_of_any,
76
minimum_path_selection,
77
quotefn, sha_file, split_lines,
80
from bzrlib.testament import Testament
81
from bzrlib.trace import mutter, note, warning, is_quiet
82
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
83
from bzrlib import symbol_versioning
84
from bzrlib.symbol_versioning import (deprecated_passed,
87
from bzrlib.workingtree import WorkingTree
88
from bzrlib.urlutils import unescape_for_display
66
92
class NullCommitReporter(object):
67
93
"""I report on progress of a commit."""
68
def added(self, path):
71
def removed(self, path):
74
def renamed(self, old_path, new_path):
95
def started(self, revno, revid, location=None):
97
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
98
"to started.", DeprecationWarning,
102
def snapshot_change(self, change, path):
105
def completed(self, revno, rev_id):
108
def deleted(self, path):
111
def missing(self, path):
114
def renamed(self, change, old_path, new_path):
117
def is_verbose(self):
78
121
class ReportCommitToLog(NullCommitReporter):
79
def added(self, path):
80
note('added %s', path)
82
def removed(self, path):
83
note('removed %s', path)
85
def renamed(self, old_path, new_path):
86
note('renamed %s => %s', old_path, new_path)
123
def _note(self, format, *args):
126
Subclasses may choose to override this method.
130
def snapshot_change(self, change, path):
131
if path == '' and change in ('added', 'modified'):
133
self._note("%s %s", change, path)
135
def started(self, revno, rev_id, location=None):
136
if location is not None:
137
location = ' to: ' + unescape_for_display(location, 'utf-8')
139
# When started was added, location was only made optional by
140
# accident. Matt Nordhoff 20071129
141
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
142
"to started.", DeprecationWarning,
145
self._note('Committing%s', location)
147
def completed(self, revno, rev_id):
148
self._note('Committed revision %d.', revno)
150
def deleted(self, path):
151
self._note('deleted %s', path)
153
def missing(self, path):
154
self._note('missing %s', path)
156
def renamed(self, change, old_path, new_path):
157
self._note('%s %s => %s', change, old_path, new_path)
159
def is_verbose(self):
89
163
class Commit(object):
99
173
working inventory.
101
175
def __init__(self,
103
if reporter is not None:
104
self.reporter = reporter
106
self.reporter = NullCommitReporter()
178
"""Create a Commit object.
180
:param reporter: the default reporter to use or None to decide later
182
self.reporter = reporter
114
190
specific_files=None,
116
allow_pointless=True):
192
allow_pointless=True,
200
message_callback=None,
203
possible_master_transports=None):
117
204
"""Commit working copy as a new revision.
119
The basic approach is to add all the file texts into the
120
store, then the inventory, then make a new revision pointing
121
to that inventory and store that.
123
This is not quite safe if the working copy changes during the
124
commit; for the moment that is simply not allowed. A better
125
approach is to make a temporary copy of the files before
126
computing their hashes, and then add those hashes in turn to
127
the inventory. This should mean at least that there are no
128
broken hash pointers. There is no way we can get a snapshot
129
of the whole directory at an instant. This would also have to
130
be robust against files disappearing, moving, etc. So the
131
whole thing is a bit hard.
133
This raises PointlessCommit if there are no changes, no new merges,
134
and allow_pointless is false.
136
timestamp -- if not None, seconds-since-epoch for a
137
postdated/predated commit.
140
If true, commit only those files.
143
If set, use this as the new revision id.
206
:param message: the commit message (it or message_callback is required)
208
:param timestamp: if not None, seconds-since-epoch for a
209
postdated/predated commit.
211
:param specific_files: If true, commit only those files.
213
:param rev_id: If set, use this as the new revision id.
144
214
Useful for test or import commands that need to tightly
145
215
control what revisions are assigned. If you duplicate
146
216
a revision id that exists elsewhere it is your own fault.
147
217
If null (default), a time/random revision id is generated.
219
:param allow_pointless: If true (default), commit even if nothing
220
has changed and no merges are recorded.
222
:param strict: If true, don't allow a commit if the working tree
223
contains unknown files.
225
:param revprops: Properties for new revision
226
:param local: Perform a local only commit.
227
:param reporter: the reporter to use or None for the default
228
:param verbose: if True and the reporter is not None, report everything
229
:param recursive: If set to 'down', commit in any subtrees that have
230
pending changes of any sort during this commit.
231
:param exclude: None or a list of relative paths to exclude from the
232
commit. Pending changes to excluded files will be ignored by the
151
self.branch.lock_write()
153
self.specific_files = specific_files
235
mutter('preparing to commit')
237
if working_tree is None:
238
raise BzrError("working_tree must be passed into commit().")
240
self.work_tree = working_tree
241
self.branch = self.work_tree.branch
242
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
243
if not self.branch.repository.supports_rich_root():
244
raise errors.RootNotRich()
245
if message_callback is None:
246
if message is not None:
247
if isinstance(message, str):
248
message = message.decode(get_user_encoding())
249
message_callback = lambda x: message
251
raise BzrError("The message or message_callback keyword"
252
" parameter is required for commit().")
254
self.bound_branch = None
255
self.any_entries_deleted = False
256
if exclude is not None:
257
self.exclude = sorted(
258
minimum_path_selection(exclude))
262
self.master_branch = None
263
self.master_locked = False
264
self.recursive = recursive
266
if specific_files is not None:
267
self.specific_files = sorted(
268
minimum_path_selection(specific_files))
270
self.specific_files = None
154
272
self.allow_pointless = allow_pointless
156
if timestamp is None:
157
self.timestamp = time.time()
159
self.timestamp = long(timestamp)
161
if committer is None:
162
self.committer = username(self.branch)
164
assert isinstance(committer, basestring), type(committer)
165
self.committer = committer
168
self.timezone = local_time_offset()
170
self.timezone = int(timezone)
172
assert isinstance(message, basestring), type(message)
173
self.message = message
176
# First walk over the working inventory; and both update that
177
# and also build a new revision inventory. The revision
178
# inventory needs to hold the text-id, sha1 and size of the
179
# actual file versions committed in the revision. (These are
180
# not present in the working inventory.) We also need to
181
# detect missing/deleted files, and remove them from the
184
self.work_tree = self.branch.working_tree()
185
self.work_inv = self.work_tree.inventory
186
self.basis_tree = self.branch.basis_tree()
187
self.basis_inv = self.basis_tree.inventory
189
self.pending_merges = self.branch.pending_merges()
190
if self.pending_merges:
191
raise NotImplementedError("sorry, can't commit merges to the weave format yet")
193
if self.rev_id is None:
194
self.rev_id = _gen_revision_id(self.branch, time.time())
196
# todo: update hashcache
197
self.delta = compare_trees(self.basis_tree, self.work_tree,
198
specific_files=self.specific_files)
200
if not (self.delta.has_changed()
201
or self.allow_pointless
202
or self.pending_merges):
203
raise PointlessCommit()
205
self.new_inv = self.basis_inv.copy()
207
self.delta.show(sys.stdout)
209
self._remove_deleted()
212
self.branch._write_inventory(self.work_inv)
213
self._record_inventory()
215
self._make_revision()
216
note('committted r%d {%s}', (self.branch.revno() + 1),
218
self.branch.append_revision(self.rev_id)
219
self.branch.set_pending_merges([])
224
def _record_inventory(self):
226
serializer_v5.write_inventory(self.new_inv, inv_tmp)
227
self.inv_sha1 = sha_string(inv_tmp.getvalue())
229
self.branch.inventory_store.add(inv_tmp, self.rev_id)
232
def _make_revision(self):
233
"""Record a new revision object for this commit."""
234
self.rev = Revision(timestamp=self.timestamp,
235
timezone=self.timezone,
236
committer=self.committer,
237
message=self.message,
238
inventory_sha1=self.inv_sha1,
239
revision_id=self.rev_id)
241
self.rev.parents = []
242
precursor_id = self.branch.last_patch()
244
self.rev.parents.append(RevisionReference(precursor_id))
245
for merge_rev in self.pending_merges:
246
rev.parents.append(RevisionReference(merge_rev))
248
rev_tmp = tempfile.TemporaryFile()
249
serializer_v5.write_revision(self.rev, rev_tmp)
251
self.branch.revision_store.add(rev_tmp, self.rev_id)
252
mutter('new revision_id is {%s}', self.rev_id)
255
def _remove_deleted(self):
256
"""Remove deleted files from the working and stored inventories."""
257
for path, id, kind in self.delta.removed:
258
if self.work_inv.has_id(id):
259
del self.work_inv[id]
260
if self.new_inv.has_id(id):
265
def _store_files(self):
266
"""Store new texts of modified/added files."""
267
# We must make sure that directories are added before anything
268
# inside them is added. the files within the delta report are
269
# sorted by path so we know the directory will come before its
271
for path, file_id, kind in self.delta.added:
273
ie = self.work_inv[file_id].copy()
276
self._store_file_text(file_id)
278
for path, file_id, kind in self.delta.modified:
281
self._store_file_text(file_id)
283
for old_path, new_path, file_id, kind, text_modified in self.delta.renamed:
286
if not text_modified:
288
self._store_file_text(file_id)
291
def _store_file_text(self, file_id):
292
"""Store updated text for one modified or added file."""
293
note('store new text for {%s} in revision {%s}',
294
file_id, self.rev_id)
295
new_lines = self.work_tree.get_file(file_id).readlines()
296
if file_id in self.new_inv: # was in basis inventory
297
ie = self.new_inv[file_id]
298
assert ie.file_id == file_id
299
assert file_id in self.basis_inv
300
assert self.basis_inv[file_id].kind == 'file'
301
old_version = self.basis_inv[file_id].text_version
302
file_parents = [old_version]
303
else: # new in this revision
304
ie = self.work_inv[file_id].copy()
306
assert file_id not in self.basis_inv
308
assert ie.kind == 'file'
309
self._add_text_to_weave(file_id, new_lines, file_parents)
310
# make a new inventory entry for this file, using whatever
311
# it had in the working copy, plus details on the new text
312
ie.text_sha1 = _sha_strings(new_lines)
313
ie.text_size = sum(map(len, new_lines))
314
ie.text_version = self.rev_id
315
ie.entry_version = self.rev_id
318
def _add_text_to_weave(self, file_id, new_lines, parents):
319
weave_fn = self.branch.controlfilename(['weaves', file_id+'.weave'])
320
if os.path.exists(weave_fn):
321
w = read_weave(file(weave_fn, 'rb'))
324
# XXX: Should set the appropriate parents by looking for this file_id
325
# in all revision parents
326
parent_idxs = map(w.lookup, parents)
327
w.add(self.rev_id, parent_idxs, new_lines)
328
af = AtomicFile(weave_fn)
330
write_weave_v5(w, af)
336
def _gen_revision_id(branch, when):
337
"""Return new revision-id."""
338
s = '%s-%s-' % (user_email(branch), compact_date(when))
339
s += hexlify(rand_bytes(8))
343
def _sha_strings(strings):
344
"""Return the sha-1 of concatenation of strings"""
346
map(s.update, strings)
273
self.revprops = revprops
274
self.message_callback = message_callback
275
self.timestamp = timestamp
276
self.timezone = timezone
277
self.committer = committer
279
self.verbose = verbose
281
self.work_tree.lock_write()
282
self.parents = self.work_tree.get_parent_ids()
283
# We can use record_iter_changes IFF iter_changes is compatible with
284
# the command line parameters, and the repository has fast delta
285
# generation. See bug 347649.
286
self.use_record_iter_changes = (
287
not self.specific_files and
289
not self.branch.repository._format.supports_tree_reference and
290
(self.branch.repository._format.fast_deltas or
291
len(self.parents) < 2))
292
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
293
self.basis_revid = self.work_tree.last_revision()
294
self.basis_tree = self.work_tree.basis_tree()
295
self.basis_tree.lock_read()
297
# Cannot commit with conflicts present.
298
if len(self.work_tree.conflicts()) > 0:
299
raise ConflictsInTree
301
# Setup the bound branch variables as needed.
302
self._check_bound_branch(possible_master_transports)
304
# Check that the working tree is up to date
305
old_revno, new_revno = self._check_out_of_date_tree()
307
# Complete configuration setup
308
if reporter is not None:
309
self.reporter = reporter
310
elif self.reporter is None:
311
self.reporter = self._select_reporter()
312
if self.config is None:
313
self.config = self.branch.get_config()
315
self._set_specific_file_ids()
317
# Setup the progress bar. As the number of files that need to be
318
# committed in unknown, progress is reported as stages.
319
# We keep track of entries separately though and include that
320
# information in the progress bar during the relevant stages.
321
self.pb_stage_name = ""
322
self.pb_stage_count = 0
323
self.pb_stage_total = 5
324
if self.bound_branch:
325
self.pb_stage_total += 1
326
self.pb.show_pct = False
327
self.pb.show_spinner = False
328
self.pb.show_eta = False
329
self.pb.show_count = True
330
self.pb.show_bar = True
332
self._gather_parents()
333
# After a merge, a selected file commit is not supported.
334
# See 'bzr help merge' for an explanation as to why.
335
if len(self.parents) > 1 and self.specific_files:
336
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
337
# Excludes are a form of selected file commit.
338
if len(self.parents) > 1 and self.exclude:
339
raise errors.CannotCommitSelectedFileMerge(self.exclude)
341
# Collect the changes
342
self._set_progress_stage("Collecting changes", counter=True)
343
self.builder = self.branch.get_commit_builder(self.parents,
344
self.config, timestamp, timezone, committer, revprops, rev_id)
347
self.builder.will_record_deletes()
348
# find the location being committed to
349
if self.bound_branch:
350
master_location = self.master_branch.base
352
master_location = self.branch.base
354
# report the start of the commit
355
self.reporter.started(new_revno, self.rev_id, master_location)
357
self._update_builder_with_changes()
358
self._check_pointless()
360
# TODO: Now the new inventory is known, check for conflicts.
361
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
362
# weave lines, because nothing should be recorded until it is known
363
# that commit will succeed.
364
self._set_progress_stage("Saving data locally")
365
self.builder.finish_inventory()
367
# Prompt the user for a commit message if none provided
368
message = message_callback(self)
369
self.message = message
371
# Add revision data to the local branch
372
self.rev_id = self.builder.commit(self.message)
375
mutter("aborting commit write group because of exception:")
376
trace.log_exception_quietly()
377
note("aborting commit write group: %r" % (e,))
381
self._process_pre_hooks(old_revno, new_revno)
383
# Upload revision data to the master.
384
# this will propagate merged revisions too if needed.
385
if self.bound_branch:
386
self._set_progress_stage("Uploading data to master branch")
387
# 'commit' to the master first so a timeout here causes the
388
# local branch to be out of date
389
self.master_branch.import_last_revision_info(
390
self.branch.repository, new_revno, self.rev_id)
392
# and now do the commit locally.
393
self.branch.set_last_revision_info(new_revno, self.rev_id)
395
# Make the working tree up to date with the branch
396
self._set_progress_stage("Updating the working tree")
397
self.work_tree.update_basis_by_delta(self.rev_id,
398
self.builder.get_basis_delta())
399
self.reporter.completed(new_revno, self.rev_id)
400
self._process_post_hooks(old_revno, new_revno)
405
def _select_reporter(self):
406
"""Select the CommitReporter to use."""
408
return NullCommitReporter()
409
return ReportCommitToLog()
411
def _check_pointless(self):
412
if self.allow_pointless:
414
# A merge with no effect on files
415
if len(self.parents) > 1:
417
# TODO: we could simplify this by using self.builder.basis_delta.
419
# The initial commit adds a root directory, but this in itself is not
420
# a worthwhile commit.
421
if (self.basis_revid == revision.NULL_REVISION and
422
((self.builder.new_inventory is not None and
423
len(self.builder.new_inventory) == 1) or
424
len(self.builder._basis_delta) == 1)):
425
raise PointlessCommit()
426
if self.builder.any_changes():
428
raise PointlessCommit()
430
def _check_bound_branch(self, possible_master_transports=None):
431
"""Check to see if the local branch is bound.
433
If it is bound, then most of the commit will actually be
434
done using the remote branch as the target branch.
435
Only at the end will the local branch be updated.
437
if self.local and not self.branch.get_bound_location():
438
raise errors.LocalRequiresBoundBranch()
441
self.master_branch = self.branch.get_master_branch(
442
possible_master_transports)
444
if not self.master_branch:
445
# make this branch the reference branch for out of date checks.
446
self.master_branch = self.branch
449
# If the master branch is bound, we must fail
450
master_bound_location = self.master_branch.get_bound_location()
451
if master_bound_location:
452
raise errors.CommitToDoubleBoundBranch(self.branch,
453
self.master_branch, master_bound_location)
455
# TODO: jam 20051230 We could automatically push local
456
# commits to the remote branch if they would fit.
457
# But for now, just require remote to be identical
460
# Make sure the local branch is identical to the master
461
master_info = self.master_branch.last_revision_info()
462
local_info = self.branch.last_revision_info()
463
if local_info != master_info:
464
raise errors.BoundBranchOutOfDate(self.branch,
467
# Now things are ready to change the master branch
469
self.bound_branch = self.branch
470
self.master_branch.lock_write()
471
self.master_locked = True
473
def _check_out_of_date_tree(self):
474
"""Check that the working tree is up to date.
476
:return: old_revision_number,new_revision_number tuple
479
first_tree_parent = self.work_tree.get_parent_ids()[0]
481
# if there are no parents, treat our parent as 'None'
482
# this is so that we still consider the master branch
483
# - in a checkout scenario the tree may have no
484
# parents but the branch may do.
485
first_tree_parent = bzrlib.revision.NULL_REVISION
486
old_revno, master_last = self.master_branch.last_revision_info()
487
if master_last != first_tree_parent:
488
if master_last != bzrlib.revision.NULL_REVISION:
489
raise errors.OutOfDateTree(self.work_tree)
490
if self.branch.repository.has_revision(first_tree_parent):
491
new_revno = old_revno + 1
493
# ghost parents never appear in revision history.
495
return old_revno,new_revno
497
def _process_pre_hooks(self, old_revno, new_revno):
498
"""Process any registered pre commit hooks."""
499
self._set_progress_stage("Running pre_commit hooks")
500
self._process_hooks("pre_commit", old_revno, new_revno)
502
def _process_post_hooks(self, old_revno, new_revno):
503
"""Process any registered post commit hooks."""
504
# Process the post commit hooks, if any
505
self._set_progress_stage("Running post_commit hooks")
506
# old style commit hooks - should be deprecated ? (obsoleted in
508
if self.config.post_commit() is not None:
509
hooks = self.config.post_commit().split(' ')
510
# this would be nicer with twisted.python.reflect.namedAny
512
result = eval(hook + '(branch, rev_id)',
513
{'branch':self.branch,
515
'rev_id':self.rev_id})
516
# process new style post commit hooks
517
self._process_hooks("post_commit", old_revno, new_revno)
519
def _process_hooks(self, hook_name, old_revno, new_revno):
520
if not Branch.hooks[hook_name]:
523
# new style commit hooks:
524
if not self.bound_branch:
525
hook_master = self.branch
528
hook_master = self.master_branch
529
hook_local = self.branch
530
# With bound branches, when the master is behind the local branch,
531
# the 'old_revno' and old_revid values here are incorrect.
532
# XXX: FIXME ^. RBC 20060206
534
old_revid = self.parents[0]
536
old_revid = bzrlib.revision.NULL_REVISION
538
if hook_name == "pre_commit":
539
future_tree = self.builder.revision_tree()
540
tree_delta = future_tree.changes_from(self.basis_tree,
543
for hook in Branch.hooks[hook_name]:
544
# show the running hook in the progress bar. As hooks may
545
# end up doing nothing (e.g. because they are not configured by
546
# the user) this is still showing progress, not showing overall
547
# actions - its up to each plugin to show a UI if it want's to
548
# (such as 'Emailing diff to foo@example.com').
549
self.pb_stage_name = "Running %s hooks [%s]" % \
550
(hook_name, Branch.hooks.get_hook_name(hook))
551
self._emit_progress()
552
if 'hooks' in debug.debug_flags:
553
mutter("Invoking commit hook: %r", hook)
554
if hook_name == "post_commit":
555
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
557
elif hook_name == "pre_commit":
558
hook(hook_local, hook_master,
559
old_revno, old_revid, new_revno, self.rev_id,
560
tree_delta, future_tree)
563
"""Cleanup any open locks, progress bars etc."""
564
cleanups = [self._cleanup_bound_branch,
565
self.basis_tree.unlock,
566
self.work_tree.unlock,
568
found_exception = None
569
for cleanup in cleanups:
572
# we want every cleanup to run no matter what.
573
# so we have a catchall here, but we will raise the
574
# last encountered exception up the stack: and
575
# typically this will be useful enough.
578
if found_exception is not None:
579
# don't do a plan raise, because the last exception may have been
580
# trashed, e is our sure-to-work exception even though it loses the
581
# full traceback. XXX: RBC 20060421 perhaps we could check the
582
# exc_info and if its the same one do a plain raise otherwise
583
# 'raise e' as we do now.
586
def _cleanup_bound_branch(self):
587
"""Executed at the end of a try/finally to cleanup a bound branch.
589
If the branch wasn't bound, this is a no-op.
590
If it was, it resents self.branch to the local branch, instead
593
if not self.bound_branch:
595
if self.master_locked:
596
self.master_branch.unlock()
598
def _gather_parents(self):
599
"""Record the parents of a merge for merge detection."""
600
# TODO: Make sure that this list doesn't contain duplicate
601
# entries and the order is preserved when doing this.
602
if self.use_record_iter_changes:
604
self.basis_inv = self.basis_tree.inventory
605
self.parent_invs = [self.basis_inv]
606
for revision in self.parents[1:]:
607
if self.branch.repository.has_revision(revision):
608
mutter('commit parent revision {%s}', revision)
609
inventory = self.branch.repository.get_inventory(revision)
610
self.parent_invs.append(inventory)
612
mutter('commit parent ghost revision {%s}', revision)
614
def _update_builder_with_changes(self):
615
"""Update the commit builder with the data about what has changed.
617
exclude = self.exclude
618
specific_files = self.specific_files or []
619
mutter("Selecting files for commit with filter %s", specific_files)
622
if self.use_record_iter_changes:
623
iter_changes = self.work_tree.iter_changes(self.basis_tree)
624
iter_changes = self._filter_iter_changes(iter_changes)
625
for file_id, path, fs_hash in self.builder.record_iter_changes(
626
self.work_tree, self.basis_revid, iter_changes):
627
self.work_tree._observed_sha1(file_id, path, fs_hash)
629
# Build the new inventory
630
self._populate_from_inventory()
631
self._record_unselected()
632
self._report_and_accumulate_deletes()
634
def _filter_iter_changes(self, iter_changes):
635
"""Process iter_changes.
637
This method reports on the changes in iter_changes to the user, and
638
converts 'missing' entries in the iter_changes iterator to 'deleted'
639
entries. 'missing' entries have their
641
:param iter_changes: An iter_changes to process.
642
:return: A generator of changes.
644
reporter = self.reporter
645
report_changes = reporter.is_verbose()
647
for change in iter_changes:
649
old_path = change[1][0]
650
new_path = change[1][1]
651
versioned = change[3][1]
653
versioned = change[3][1]
654
if kind is None and versioned:
657
reporter.missing(new_path)
658
deleted_ids.append(change[0])
659
# Reset the new path (None) and new versioned flag (False)
660
change = (change[0], (change[1][0], None), change[2],
661
(change[3][0], False)) + change[4:]
662
elif kind == 'tree-reference':
663
if self.recursive == 'down':
664
self._commit_nested_tree(change[0], change[1][1])
665
if change[3][0] or change[3][1]:
669
reporter.deleted(old_path)
670
elif old_path is None:
671
reporter.snapshot_change('added', new_path)
672
elif old_path != new_path:
673
reporter.renamed('renamed', old_path, new_path)
676
self.work_tree.branch.repository._format.rich_root_data):
677
# Don't report on changes to '' in non rich root
679
reporter.snapshot_change('modified', new_path)
680
self._next_progress_entry()
681
# Unversion IDs that were found to be deleted
682
self.work_tree.unversion(deleted_ids)
684
def _record_unselected(self):
685
# If specific files are selected, then all un-selected files must be
686
# recorded in their previous state. For more details, see
687
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
688
if self.specific_files or self.exclude:
689
specific_files = self.specific_files or []
690
for path, old_ie in self.basis_inv.iter_entries():
691
if old_ie.file_id in self.builder.new_inventory:
692
# already added - skip.
694
if (is_inside_any(specific_files, path)
695
and not is_inside_any(self.exclude, path)):
696
# was inside the selected path, and not excluded - if not
697
# present it has been deleted so skip.
699
# From here down it was either not selected, or was excluded:
700
# We preserve the entry unaltered.
702
# Note: specific file commits after a merge are currently
703
# prohibited. This test is for sanity/safety in case it's
704
# required after that changes.
705
if len(self.parents) > 1:
707
self.builder.record_entry_contents(ie, self.parent_invs, path,
708
self.basis_tree, None)
710
def _report_and_accumulate_deletes(self):
711
if (isinstance(self.basis_inv, Inventory)
712
and isinstance(self.builder.new_inventory, Inventory)):
713
# the older Inventory classes provide a _byid dict, and building a
714
# set from the keys of this dict is substantially faster than even
715
# getting a set of ids from the inventory
717
# <lifeless> set(dict) is roughly the same speed as
718
# set(iter(dict)) and both are significantly slower than
720
deleted_ids = set(self.basis_inv._byid.keys()) - \
721
set(self.builder.new_inventory._byid.keys())
723
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
725
self.any_entries_deleted = True
726
deleted = [(self.basis_tree.id2path(file_id), file_id)
727
for file_id in deleted_ids]
729
# XXX: this is not quite directory-order sorting
730
for path, file_id in deleted:
731
self.builder.record_delete(path, file_id)
732
self.reporter.deleted(path)
734
def _check_strict(self):
735
# XXX: when we use iter_changes this would likely be faster if
736
# iter_changes would check for us (even in the presence of
739
# raise an exception as soon as we find a single unknown.
740
for unknown in self.work_tree.unknowns():
741
raise StrictCommitFailed()
743
def _populate_from_inventory(self):
744
"""Populate the CommitBuilder by walking the working tree inventory."""
745
# Build the revision inventory.
747
# This starts by creating a new empty inventory. Depending on
748
# which files are selected for commit, and what is present in the
749
# current tree, the new inventory is populated. inventory entries
750
# which are candidates for modification have their revision set to
751
# None; inventory entries that are carried over untouched have their
752
# revision set to their prior value.
754
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
755
# results to create a new inventory at the same time, which results
756
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
759
specific_files = self.specific_files
760
exclude = self.exclude
761
report_changes = self.reporter.is_verbose()
763
# A tree of paths that have been deleted. E.g. if foo/bar has been
764
# deleted, then we have {'foo':{'bar':{}}}
766
# XXX: Note that entries may have the wrong kind because the entry does
767
# not reflect the status on disk.
768
work_inv = self.work_tree.inventory
769
# NB: entries will include entries within the excluded ids/paths
770
# because iter_entries_by_dir has no 'exclude' facility today.
771
entries = work_inv.iter_entries_by_dir(
772
specific_file_ids=self.specific_file_ids, yield_parents=True)
773
for path, existing_ie in entries:
774
file_id = existing_ie.file_id
775
name = existing_ie.name
776
parent_id = existing_ie.parent_id
777
kind = existing_ie.kind
778
# Skip files that have been deleted from the working tree.
779
# The deleted path ids are also recorded so they can be explicitly
782
path_segments = splitpath(path)
783
deleted_dict = deleted_paths
784
for segment in path_segments:
785
deleted_dict = deleted_dict.get(segment, None)
787
# We either took a path not present in the dict
788
# (deleted_dict was None), or we've reached an empty
789
# child dir in the dict, so are now a sub-path.
793
if deleted_dict is not None:
794
# the path has a deleted parent, do not add it.
796
if exclude and is_inside_any(exclude, path):
797
# Skip excluded paths. Excluded paths are processed by
798
# _update_builder_with_changes.
800
content_summary = self.work_tree.path_content_summary(path)
801
# Note that when a filter of specific files is given, we must only
802
# skip/record deleted files matching that filter.
803
if not specific_files or is_inside_any(specific_files, path):
804
if content_summary[0] == 'missing':
805
if not deleted_paths:
806
# path won't have been split yet.
807
path_segments = splitpath(path)
808
deleted_dict = deleted_paths
809
for segment in path_segments:
810
deleted_dict = deleted_dict.setdefault(segment, {})
811
self.reporter.missing(path)
812
self._next_progress_entry()
813
deleted_ids.append(file_id)
815
# TODO: have the builder do the nested commit just-in-time IF and
817
if content_summary[0] == 'tree-reference':
818
# enforce repository nested tree policy.
819
if (not self.work_tree.supports_tree_reference() or
820
# repository does not support it either.
821
not self.branch.repository._format.supports_tree_reference):
822
content_summary = ('directory',) + content_summary[1:]
823
kind = content_summary[0]
824
# TODO: specific_files filtering before nested tree processing
825
if kind == 'tree-reference':
826
if self.recursive == 'down':
827
nested_revision_id = self._commit_nested_tree(
829
content_summary = content_summary[:3] + (
832
content_summary = content_summary[:3] + (
833
self.work_tree.get_reference_revision(file_id),)
835
# Record an entry for this item
836
# Note: I don't particularly want to have the existing_ie
837
# parameter but the test suite currently (28-Jun-07) breaks
838
# without it thanks to a unicode normalisation issue. :-(
839
definitely_changed = kind != existing_ie.kind
840
self._record_entry(path, file_id, specific_files, kind, name,
841
parent_id, definitely_changed, existing_ie, report_changes,
844
# Unversion IDs that were found to be deleted
845
self.work_tree.unversion(deleted_ids)
847
def _commit_nested_tree(self, file_id, path):
848
"Commit a nested tree."
849
sub_tree = self.work_tree.get_nested_tree(file_id, path)
850
# FIXME: be more comprehensive here:
851
# this works when both trees are in --trees repository,
852
# but when both are bound to a different repository,
853
# it fails; a better way of approaching this is to
854
# finally implement the explicit-caches approach design
855
# a while back - RBC 20070306.
856
if sub_tree.branch.repository.has_same_location(
857
self.work_tree.branch.repository):
858
sub_tree.branch.repository = \
859
self.work_tree.branch.repository
861
return sub_tree.commit(message=None, revprops=self.revprops,
862
recursive=self.recursive,
863
message_callback=self.message_callback,
864
timestamp=self.timestamp, timezone=self.timezone,
865
committer=self.committer,
866
allow_pointless=self.allow_pointless,
867
strict=self.strict, verbose=self.verbose,
868
local=self.local, reporter=self.reporter)
869
except errors.PointlessCommit:
870
return self.work_tree.get_reference_revision(file_id)
872
def _record_entry(self, path, file_id, specific_files, kind, name,
873
parent_id, definitely_changed, existing_ie, report_changes,
875
"Record the new inventory entry for a path if any."
876
# mutter('check %s {%s}', path, file_id)
877
# mutter('%s selected for commit', path)
878
if definitely_changed or existing_ie is None:
879
ie = make_entry(kind, name, parent_id, file_id)
881
ie = existing_ie.copy()
883
# For carried over entries we don't care about the fs hash - the repo
884
# isn't generating a sha, so we're not saving computation time.
885
_, _, fs_hash = self.builder.record_entry_contents(
886
ie, self.parent_invs, path, self.work_tree, content_summary)
888
self._report_change(ie, path)
890
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
893
def _report_change(self, ie, path):
894
"""Report a change to the user.
896
The change that has occurred is described relative to the basis
899
if (self.basis_inv.has_id(ie.file_id)):
900
basis_ie = self.basis_inv[ie.file_id]
903
change = ie.describe_change(basis_ie, ie)
904
if change in (InventoryEntry.RENAMED,
905
InventoryEntry.MODIFIED_AND_RENAMED):
906
old_path = self.basis_inv.id2path(ie.file_id)
907
self.reporter.renamed(change, old_path, path)
908
self._next_progress_entry()
910
if change == 'unchanged':
912
self.reporter.snapshot_change(change, path)
913
self._next_progress_entry()
915
def _set_progress_stage(self, name, counter=False):
916
"""Set the progress stage and emit an update to the progress bar."""
917
self.pb_stage_name = name
918
self.pb_stage_count += 1
920
self.pb_entries_count = 0
922
self.pb_entries_count = None
923
self._emit_progress()
925
def _next_progress_entry(self):
926
"""Emit an update to the progress bar and increment the entry count."""
927
self.pb_entries_count += 1
928
self._emit_progress()
930
def _emit_progress(self):
931
if self.pb_entries_count is not None:
932
text = "%s [%d] - Stage" % (self.pb_stage_name,
933
self.pb_entries_count)
935
text = "%s - Stage" % (self.pb_stage_name, )
936
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
938
def _set_specific_file_ids(self):
939
"""populate self.specific_file_ids if we will use it."""
940
if not self.use_record_iter_changes:
941
# If provided, ensure the specified files are versioned
942
if self.specific_files is not None:
943
# Note: This routine is being called because it raises
944
# PathNotVersionedError as a side effect of finding the IDs. We
945
# later use the ids we found as input to the working tree
946
# inventory iterator, so we only consider those ids rather than
947
# examining the whole tree again.
948
# XXX: Dont we have filter_unversioned to do this more
950
self.specific_file_ids = tree.find_ids_across_trees(
951
self.specific_files, [self.basis_tree, self.work_tree])
953
self.specific_file_ids = None
added
removed
bzrlib/commit.py
