1
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
19
# FIXME: "bzr commit doc/format" commits doc/format.txt!
21
def commit(branch, message,
28
"""Commit working copy as a new revision.
30
The basic approach is to add all the file texts into the
31
store, then the inventory, then make a new revision pointing
32
to that inventory and store that.
34
This is not quite safe if the working copy changes during the
35
commit; for the moment that is simply not allowed. A better
36
approach is to make a temporary copy of the files before
37
computing their hashes, and then add those hashes in turn to
38
the inventory. This should mean at least that there are no
39
broken hash pointers. There is no way we can get a snapshot
40
of the whole directory at an instant. This would also have to
41
be robust against files disappearing, moving, etc. So the
42
whole thing is a bit hard.
44
timestamp -- if not None, seconds-since-epoch for a
45
postdated/predated commit.
48
If true, commit only those files.
51
If set, use this as the new revision id.
52
Useful for test or import commands that need to tightly
53
control what revisions are assigned. If you duplicate
54
a revision id that exists elsewhere it is your own fault.
55
If null (default), a time/random revision id is generated.
60
from bzrlib.osutils import local_time_offset, username
61
from bzrlib.branch import gen_file_id
62
from bzrlib.errors import BzrError
63
from bzrlib.revision import Revision, RevisionReference
64
from bzrlib.trace import mutter, note
65
from bzrlib.xml import pack_xml
70
# First walk over the working inventory; and both update that
71
# and also build a new revision inventory. The revision
72
# inventory needs to hold the text-id, sha1 and size of the
73
# actual file versions committed in the revision. (These are
74
# not present in the working inventory.) We also need to
75
# detect missing/deleted files, and remove them from the
78
work_tree = branch.working_tree()
79
work_inv = work_tree.inventory
80
basis = branch.basis_tree()
81
basis_inv = basis.inventory
84
note('looking for changes...')
86
pending_merges = branch.pending_merges()
88
missing_ids, new_inv = _gather_commit(branch,
95
for file_id in missing_ids:
96
# Any files that have been deleted are now removed from the
97
# working inventory. Files that were not selected for commit
98
# are left as they were in the working inventory and ommitted
99
# from the revision inventory.
101
# have to do this later so we don't mess up the iterator.
102
# since parents may be removed before their children we
105
# FIXME: There's probably a better way to do this; perhaps
106
# the workingtree should know how to filter itbranch.
107
if work_inv.has_id(file_id):
108
del work_inv[file_id]
18
# XXX: Can we do any better about making interrupted commits change
19
# nothing? Perhaps the best approach is to integrate commit of
20
# AtomicFiles with releasing the lock on the Branch.
22
# TODO: Separate 'prepare' phase where we find a list of potentially
23
# committed files. We then can then pause the commit to prompt for a
24
# commit message, knowing the summary will be the same as what's
25
# actually used for the commit. (But perhaps simpler to simply get
26
# the tree status, then use that for a selective commit?)
28
# The newly committed revision is going to have a shape corresponding
29
# to that of the working inventory. Files that are not in the
30
# working tree and that were in the predecessor are reported as
31
# removed --- this can include files that were either removed from the
32
# inventory or deleted in the working tree. If they were only
33
# deleted from disk, they are removed from the working inventory.
35
# We then consider the remaining entries, which will be in the new
36
# version. Directory entries are simply copied across. File entries
37
# must be checked to see if a new version of the file should be
38
# recorded. For each parent revision inventory, we check to see what
39
# version of the file was present. If the file was present in at
40
# least one tree, and if it was the same version in all the trees,
41
# then we can just refer to that version. Otherwise, a new version
42
# representing the merger of the file versions must be added.
44
# TODO: Update hashcache before and after - or does the WorkingTree
47
# TODO: Rather than mashing together the ancestry and storing it back,
48
# perhaps the weave should have single method which does it all in one
49
# go, avoiding a lot of redundant work.
51
# TODO: Perhaps give a warning if one of the revisions marked as
52
# merged is already in the ancestry, and then don't record it as a
55
# TODO: If the file is newly merged but unchanged from the version it
56
# merges from, then it should still be reported as newly added
57
# relative to the basis revision.
59
# TODO: Do checks that the tree can be committed *before* running the
60
# editor; this should include checks for a pointless commit and for
61
# unknown or missing files.
63
# TODO: If commit fails, leave the message in a file somewhere.
72
from binascii import hexlify
73
from cStringIO import StringIO
75
from bzrlib.osutils import (local_time_offset,
76
rand_bytes, compact_date,
77
kind_marker, is_inside_any, quotefn,
78
sha_string, sha_strings, sha_file, isdir, isfile,
80
from bzrlib.branch import gen_file_id
82
from bzrlib.errors import (BzrError, PointlessCommit,
87
import bzrlib.gpg as gpg
88
from bzrlib.revision import Revision
89
from bzrlib.testament import Testament
90
from bzrlib.trace import mutter, note, warning
91
from bzrlib.xml5 import serializer_v5
92
from bzrlib.inventory import Inventory, ROOT_ID
93
from bzrlib.weave import Weave
94
from bzrlib.weavefile import read_weave, write_weave_v5
95
from bzrlib.atomicfile import AtomicFile
98
def commit(*args, **kwargs):
99
"""Commit a new revision to a branch.
101
Function-style interface for convenience of old callers.
103
New code should use the Commit class instead.
105
## XXX: Remove this in favor of Branch.commit?
106
Commit().commit(*args, **kwargs)
109
class NullCommitReporter(object):
110
"""I report on progress of a commit."""
112
def snapshot_change(self, change, path):
115
def completed(self, revno, rev_id):
118
def deleted(self, file_id):
121
def escaped(self, escape_count, message):
124
def missing(self, path):
127
class ReportCommitToLog(NullCommitReporter):
129
def snapshot_change(self, change, path):
130
note("%s %s", change, path)
132
def completed(self, revno, rev_id):
133
note('committed r%d {%s}', revno, rev_id)
135
def deleted(self, file_id):
136
note('deleted %s', file_id)
138
def escaped(self, escape_count, message):
139
note("replaced %d control characters in message", escape_count)
141
def missing(self, path):
142
note('missing %s', path)
144
class Commit(object):
145
"""Task of committing a new revision.
147
This is a MethodObject: it accumulates state as the commit is
148
prepared, and then it is discarded. It doesn't represent
149
historical revisions, just the act of recording a new one.
152
Modified to hold a list of files that have been deleted from
153
the working directory; these should be removed from the
159
if reporter is not None:
160
self.reporter = reporter
162
self.reporter = NullCommitReporter()
163
if config is not None:
175
allow_pointless=True,
179
"""Commit working copy as a new revision.
181
timestamp -- if not None, seconds-since-epoch for a
182
postdated/predated commit.
184
specific_files -- If true, commit only those files.
186
rev_id -- If set, use this as the new revision id.
187
Useful for test or import commands that need to tightly
188
control what revisions are assigned. If you duplicate
189
a revision id that exists elsewhere it is your own fault.
190
If null (default), a time/random revision id is generated.
192
allow_pointless -- If true (default), commit even if nothing
193
has changed and no merges are recorded.
195
strict -- If true, don't allow a commit if the working tree
196
contains unknown files.
198
revprops -- Properties for new revision
200
mutter('preparing to commit')
203
self.weave_store = branch.weave_store
205
self.specific_files = specific_files
206
self.allow_pointless = allow_pointless
207
self.revprops = revprops
210
# raise an exception as soon as we find a single unknown.
211
for unknown in branch.unknowns():
212
raise StrictCommitFailed()
214
if timestamp is None:
215
self.timestamp = time.time()
217
self.timestamp = long(timestamp)
219
if self.config is None:
220
self.config = bzrlib.config.BranchConfig(self.branch)
111
222
if rev_id is None:
112
rev_id = _gen_revision_id(time.time())
115
inv_tmp = tempfile.TemporaryFile()
116
pack_xml(new_inv, inv_tmp)
118
branch.inventory_store.add(inv_tmp, inv_id)
119
mutter('new inventory_id is {%s}' % inv_id)
121
# We could also just sha hash the inv_tmp file
122
# however, in the case that branch.inventory_store.add()
123
# ever actually does anything special
124
inv_sha1 = branch.get_inventory_sha1(inv_id)
126
branch._write_inventory(work_inv)
128
if timestamp == None:
129
timestamp = time.time()
131
if committer == None:
132
committer = username()
135
timezone = local_time_offset()
137
mutter("building commit log message")
138
rev = Revision(timestamp=timestamp,
143
inventory_sha1=inv_sha1,
147
precursor_id = branch.last_patch()
223
self.rev_id = _gen_revision_id(self.config, self.timestamp)
227
if committer is None:
228
self.committer = self.config.username()
230
assert isinstance(committer, basestring), type(committer)
231
self.committer = committer
234
self.timezone = local_time_offset()
236
self.timezone = int(timezone)
238
assert isinstance(message, basestring), type(message)
239
self.message = message
240
self._escape_commit_message()
242
self.branch.lock_write()
244
self.work_tree = self.branch.working_tree()
245
self.work_inv = self.work_tree.inventory
246
self.basis_tree = self.branch.basis_tree()
247
self.basis_inv = self.basis_tree.inventory
249
self._gather_parents()
250
if len(self.parents) > 1 and self.specific_files:
251
raise NotImplementedError('selected-file commit of merges is not supported yet')
252
self._check_parents_present()
254
self._remove_deleted()
255
self._populate_new_inv()
256
self._store_snapshot()
257
self._report_deletes()
259
if not (self.allow_pointless
260
or len(self.parents) > 1
261
or self.new_inv != self.basis_inv):
262
raise PointlessCommit()
264
if len(list(self.work_tree.iter_conflicts()))>0:
265
raise ConflictsInTree
267
self._record_inventory()
268
self._make_revision()
269
self.branch.append_revision(self.rev_id)
270
self.branch.set_pending_merges([])
271
self.reporter.completed(self.branch.revno()+1, self.rev_id)
272
if self.config.post_commit() is not None:
273
hooks = self.config.post_commit().split(' ')
274
# this would be nicer with twisted.python.reflect.namedAny
276
result = eval(hook + '(branch, rev_id)',
277
{'branch':self.branch,
279
'rev_id':self.rev_id})
283
def _record_inventory(self):
284
"""Store the inventory for the new revision."""
285
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
286
self.inv_sha1 = sha_string(inv_text)
287
s = self.branch.control_weaves
288
s.add_text('inventory', self.rev_id,
289
split_lines(inv_text), self.present_parents,
290
self.branch.get_transaction())
292
def _escape_commit_message(self):
293
"""Replace xml-incompatible control characters."""
294
# Python strings can include characters that can't be
295
# represented in well-formed XML; escape characters that
296
# aren't listed in the XML specification
297
# (http://www.w3.org/TR/REC-xml/#NT-Char).
298
if isinstance(self.message, unicode):
299
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
301
# Use a regular 'str' as pattern to avoid having re.subn
302
# return 'unicode' results.
303
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
304
self.message, escape_count = re.subn(
306
lambda match: match.group(0).encode('unicode_escape'),
309
self.reporter.escaped(escape_count, self.message)
311
def _gather_parents(self):
312
"""Record the parents of a merge for merge detection."""
313
pending_merges = self.branch.pending_merges()
315
self.parent_invs = []
316
self.present_parents = []
317
precursor_id = self.branch.last_revision()
149
precursor_sha1 = branch.get_revision_sha1(precursor_id)
150
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
151
for merge_rev in pending_merges:
152
rev.parents.append(RevisionReference(merge_rev))
319
self.parents.append(precursor_id)
320
self.parents += pending_merges
321
for revision in self.parents:
322
if self.branch.has_revision(revision):
323
self.parent_invs.append(self.branch.get_inventory(revision))
324
self.present_parents.append(revision)
154
rev_tmp = tempfile.TemporaryFile()
155
pack_xml(rev, rev_tmp)
326
def _check_parents_present(self):
327
for parent_id in self.parents:
328
mutter('commit parent revision {%s}', parent_id)
329
if not self.branch.has_revision(parent_id):
330
if parent_id == self.branch.last_revision():
331
warning("parent is missing %r", parent_id)
332
raise HistoryMissing(self.branch, 'revision', parent_id)
334
mutter("commit will ghost revision %r", parent_id)
336
def _make_revision(self):
337
"""Record a new revision object for this commit."""
338
self.rev = Revision(timestamp=self.timestamp,
339
timezone=self.timezone,
340
committer=self.committer,
341
message=self.message,
342
inventory_sha1=self.inv_sha1,
343
revision_id=self.rev_id,
344
properties=self.revprops)
345
self.rev.parent_ids = self.parents
347
serializer_v5.write_revision(self.rev, rev_tmp)
157
branch.revision_store.add(rev_tmp, rev_id)
158
mutter("new revision_id is {%s}" % rev_id)
160
## XXX: Everything up to here can simply be orphaned if we abort
161
## the commit; it will leave junk files behind but that doesn't
164
## TODO: Read back the just-generated changeset, and make sure it
165
## applies and recreates the right state.
167
## TODO: Also calculate and store the inventory SHA1
168
mutter("committing patch r%d" % (branch.revno() + 1))
170
branch.append_revision(rev_id)
172
branch.set_pending_merges([])
175
note("commited r%d" % branch.revno())
181
def _gen_revision_id(when):
349
if self.config.signature_needed():
350
plaintext = Testament(self.rev, self.new_inv).as_short_text()
351
self.branch.store_revision_signature(gpg.GPGStrategy(self.config),
352
plaintext, self.rev_id)
353
self.branch.revision_store.add(rev_tmp, self.rev_id)
354
mutter('new revision_id is {%s}', self.rev_id)
356
def _remove_deleted(self):
357
"""Remove deleted files from the working inventories.
359
This is done prior to taking the working inventory as the
360
basis for the new committed inventory.
362
This returns true if any files
363
*that existed in the basis inventory* were deleted.
364
Files that were added and deleted
365
in the working copy don't matter.
367
specific = self.specific_files
369
for path, ie in self.work_inv.iter_entries():
370
if specific and not is_inside_any(specific, path):
372
if not self.work_tree.has_filename(path):
373
self.reporter.missing(path)
374
deleted_ids.append((path, ie.file_id))
376
deleted_ids.sort(reverse=True)
377
for path, file_id in deleted_ids:
378
del self.work_inv[file_id]
379
self.branch._write_inventory(self.work_inv)
381
def _store_snapshot(self):
382
"""Pass over inventory and record a snapshot.
384
Entries get a new revision when they are modified in
385
any way, which includes a merge with a new set of
386
parents that have the same entry.
388
# XXX: Need to think more here about when the user has
389
# made a specific decision on a particular value -- c.f.
391
for path, ie in self.new_inv.iter_entries():
392
previous_entries = ie.find_previous_heads(
394
self.weave_store.get_weave_or_empty(ie.file_id,
395
self.branch.get_transaction()))
396
if ie.revision is None:
397
change = ie.snapshot(self.rev_id, path, previous_entries,
398
self.work_tree, self.weave_store,
399
self.branch.get_transaction())
402
self.reporter.snapshot_change(change, path)
404
def _populate_new_inv(self):
405
"""Build revision inventory.
407
This creates a new empty inventory. Depending on
408
which files are selected for commit, and what is present in the
409
current tree, the new inventory is populated. inventory entries
410
which are candidates for modification have their revision set to
411
None; inventory entries that are carried over untouched have their
412
revision set to their prior value.
414
mutter("Selecting files for commit with filter %s", self.specific_files)
415
self.new_inv = Inventory()
416
for path, new_ie in self.work_inv.iter_entries():
417
file_id = new_ie.file_id
418
mutter('check %s {%s}', path, new_ie.file_id)
419
if self.specific_files:
420
if not is_inside_any(self.specific_files, path):
421
mutter('%s not selected for commit', path)
422
self._carry_entry(file_id)
425
# this is selected, ensure its parents are too.
426
parent_id = new_ie.parent_id
427
while parent_id != ROOT_ID:
428
if not self.new_inv.has_id(parent_id):
429
ie = self._select_entry(self.work_inv[parent_id])
430
mutter('%s selected for commit because of %s',
431
self.new_inv.id2path(parent_id), path)
433
ie = self.new_inv[parent_id]
434
if ie.revision is not None:
436
mutter('%s selected for commit because of %s',
437
self.new_inv.id2path(parent_id), path)
438
parent_id = ie.parent_id
439
mutter('%s selected for commit', path)
440
self._select_entry(new_ie)
442
def _select_entry(self, new_ie):
443
"""Make new_ie be considered for committing."""
449
def _carry_entry(self, file_id):
450
"""Carry the file unchanged from the basis revision."""
451
if self.basis_inv.has_id(file_id):
452
self.new_inv.add(self.basis_inv[file_id].copy())
454
def _report_deletes(self):
455
for file_id in self.basis_inv:
456
if file_id not in self.new_inv:
457
self.reporter.deleted(self.basis_inv.id2path(file_id))
459
def _gen_revision_id(config, when):
182
460
"""Return new revision-id."""
183
from binascii import hexlify
184
from osutils import rand_bytes, compact_date, user_email
186
s = '%s-%s-' % (user_email(), compact_date(when))
461
s = '%s-%s-' % (config.user_email(), compact_date(when))
187
462
s += hexlify(rand_bytes(8))
191
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
193
"""Build inventory preparatory to commit.
195
This adds any changed files into the text store, and sets their
196
test-id, sha and size in the returned inventory appropriately.
199
Modified to hold a list of files that have been deleted from
200
the working directory; these should be removed from the
203
from bzrlib.inventory import Inventory
204
from osutils import isdir, isfile, sha_string, quotefn, \
205
local_time_offset, username, kind_marker, is_inside_any
207
from branch import gen_file_id
208
from errors import BzrError
209
from revision import Revision
210
from bzrlib.trace import mutter, note
215
for path, entry in work_inv.iter_entries():
216
## TODO: Check that the file kind has not changed from the previous
217
## revision of this file (if any).
219
p = branch.abspath(path)
220
file_id = entry.file_id
221
mutter('commit prep file %s, id %r ' % (p, file_id))
223
if specific_files and not is_inside_any(specific_files, path):
224
if basis_inv.has_id(file_id):
225
# carry over with previous state
226
inv.add(basis_inv[file_id].copy())
228
# omit this from committed inventory
232
if not work_tree.has_id(file_id):
234
print('deleted %s%s' % (path, kind_marker(entry.kind)))
235
mutter(" file is missing, removing from inventory")
236
missing_ids.append(file_id)
239
# this is present in the new inventory; may be new, modified or
241
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
247
old_kind = old_ie.kind
248
if old_kind != entry.kind:
249
raise BzrError("entry %r changed kind from %r to %r"
250
% (file_id, old_kind, entry.kind))
252
if entry.kind == 'directory':
254
raise BzrError("%s is entered as directory but not a directory"
256
elif entry.kind == 'file':
258
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
260
new_sha1 = work_tree.get_file_sha1(file_id)
263
and old_ie.text_sha1 == new_sha1):
264
## assert content == basis.get_file(file_id).read()
265
entry.text_id = old_ie.text_id
266
entry.text_sha1 = new_sha1
267
entry.text_size = old_ie.text_size
268
mutter(' unchanged from previous text_id {%s}' %
271
content = file(p, 'rb').read()
273
# calculate the sha again, just in case the file contents
274
# changed since we updated the cache
275
entry.text_sha1 = sha_string(content)
276
entry.text_size = len(content)
278
entry.text_id = gen_file_id(entry.name)
279
branch.text_store.add(content, entry.text_id)
280
mutter(' stored with text_id {%s}' % entry.text_id)
283
marked = path + kind_marker(entry.kind)
285
print 'added', marked
286
elif old_ie == entry:
288
elif (old_ie.name == entry.name
289
and old_ie.parent_id == entry.parent_id):
290
print 'modified', marked
292
print 'renamed', marked
294
return missing_ids, inv
added
removed
bzrlib/commit.py
