1
# Copyright (C) 2005 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
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
20
import sys, os, os.path, random, time, sha, sets, types, re, shutil, tempfile
21
import traceback, socket, fnmatch, difflib, time
22
from binascii import hexlify
25
from inventory import Inventory
26
from trace import mutter, note
27
from tree import Tree, EmptyTree, RevisionTree
28
from inventory import InventoryEntry, Inventory
29
from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \
30
format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \
31
joinpath, sha_string, file_kind, local_time_offset, appendpath
32
from store import ImmutableStore
33
from revision import Revision
34
from errors import bailout, BzrError
35
from textui import show_status
37
BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"
38
## TODO: Maybe include checks for common corruption of newlines, etc?
42
def find_branch(f, **args):
43
if f and (f.startswith('http://') or f.startswith('https://')):
45
return remotebranch.RemoteBranch(f, **args)
47
return Branch(f, **args)
50
def find_branch_root(f=None):
51
"""Find the branch root enclosing f, or pwd.
53
f may be a filename or a URL.
55
It is not necessary that f exists.
57
Basically we keep looking up until we find the control directory or
61
elif hasattr(os.path, 'realpath'):
62
f = os.path.realpath(f)
64
f = os.path.abspath(f)
65
if not os.path.exists(f):
66
raise BzrError('%r does not exist' % f)
72
if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):
74
head, tail = os.path.split(f)
76
# reached the root, whatever that may be
77
raise BzrError('%r is not in a branch' % orig_f)
82
######################################################################
86
"""Branch holding a history of revisions.
89
Base directory of the branch.
93
def __init__(self, base, init=False, find_root=True, lock_mode='w'):
94
"""Create new branch object at a particular location.
96
base -- Base directory for the branch.
98
init -- If True, create new control files in a previously
99
unversioned directory. If False, the branch must already
102
find_root -- If true and init is false, find the root of the
103
existing branch containing base.
105
In the test suite, creation of new trees is tested using the
106
`ScratchBranch` class.
109
self.base = os.path.realpath(base)
112
self.base = find_branch_root(base)
114
self.base = os.path.realpath(base)
115
if not isdir(self.controlfilename('.')):
116
bailout("not a bzr branch: %s" % quotefn(base),
117
['use "bzr init" to initialize a new working tree',
118
'current bzr can only operate from top-of-tree'])
122
self.text_store = ImmutableStore(self.controlfilename('text-store'))
123
self.revision_store = ImmutableStore(self.controlfilename('revision-store'))
124
self.inventory_store = ImmutableStore(self.controlfilename('inventory-store'))
128
return '%s(%r)' % (self.__class__.__name__, self.base)
135
def lock(self, mode='w'):
136
"""Lock the on-disk branch, excluding other processes."""
142
om = os.O_WRONLY | os.O_CREAT
147
raise BzrError("invalid locking mode %r" % mode)
150
lockfile = os.open(self.controlfilename('branch-lock'), om)
152
if e.errno == errno.ENOENT:
153
# might not exist on branches from <0.0.4
154
self.controlfile('branch-lock', 'w').close()
155
lockfile = os.open(self.controlfilename('branch-lock'), om)
159
fcntl.lockf(lockfile, lm)
161
fcntl.lockf(lockfile, fcntl.LOCK_UN)
163
self._lockmode = None
165
self._lockmode = mode
167
warning("please write a locking method for platform %r" % sys.platform)
169
self._lockmode = None
171
self._lockmode = mode
174
def _need_readlock(self):
175
if self._lockmode not in ['r', 'w']:
176
raise BzrError('need read lock on branch, only have %r' % self._lockmode)
178
def _need_writelock(self):
179
if self._lockmode not in ['w']:
180
raise BzrError('need write lock on branch, only have %r' % self._lockmode)
183
def abspath(self, name):
184
"""Return absolute filename for something in the branch"""
185
return os.path.join(self.base, name)
188
def relpath(self, path):
189
"""Return path relative to this branch of something inside it.
191
Raises an error if path is not in this branch."""
192
rp = os.path.realpath(path)
194
if not rp.startswith(self.base):
195
bailout("path %r is not within branch %r" % (rp, self.base))
196
rp = rp[len(self.base):]
197
rp = rp.lstrip(os.sep)
201
def controlfilename(self, file_or_path):
202
"""Return location relative to branch."""
203
if isinstance(file_or_path, types.StringTypes):
204
file_or_path = [file_or_path]
205
return os.path.join(self.base, bzrlib.BZRDIR, *file_or_path)
208
def controlfile(self, file_or_path, mode='r'):
209
"""Open a control file for this branch.
211
There are two classes of file in the control directory: text
212
and binary. binary files are untranslated byte streams. Text
213
control files are stored with Unix newlines and in UTF-8, even
214
if the platform or locale defaults are different.
216
Controlfiles should almost never be opened in write mode but
217
rather should be atomically copied and replaced using atomicfile.
220
fn = self.controlfilename(file_or_path)
222
if mode == 'rb' or mode == 'wb':
223
return file(fn, mode)
224
elif mode == 'r' or mode == 'w':
225
# open in binary mode anyhow so there's no newline translation;
226
# codecs uses line buffering by default; don't want that.
228
return codecs.open(fn, mode + 'b', 'utf-8',
231
raise BzrError("invalid controlfile mode %r" % mode)
235
def _make_control(self):
236
os.mkdir(self.controlfilename([]))
237
self.controlfile('README', 'w').write(
238
"This is a Bazaar-NG control directory.\n"
239
"Do not change any files in this directory.")
240
self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT)
241
for d in ('text-store', 'inventory-store', 'revision-store'):
242
os.mkdir(self.controlfilename(d))
243
for f in ('revision-history', 'merged-patches',
244
'pending-merged-patches', 'branch-name',
246
self.controlfile(f, 'w').write('')
247
mutter('created control directory in ' + self.base)
248
Inventory().write_xml(self.controlfile('inventory','w'))
251
def _check_format(self):
252
"""Check this branch format is supported.
254
The current tool only supports the current unstable format.
256
In the future, we might need different in-memory Branch
257
classes to support downlevel branches. But not yet.
259
# This ignores newlines so that we can open branches created
260
# on Windows from Linux and so on. I think it might be better
261
# to always make all internal files in unix format.
262
fmt = self.controlfile('branch-format', 'r').read()
263
fmt.replace('\r\n', '')
264
if fmt != BZR_BRANCH_FORMAT:
265
bailout('sorry, branch format %r not supported' % fmt,
266
['use a different bzr version',
267
'or remove the .bzr directory and "bzr init" again'])
270
def read_working_inventory(self):
271
"""Read the working inventory."""
272
self._need_readlock()
274
# ElementTree does its own conversion from UTF-8, so open in
276
inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))
277
mutter("loaded inventory of %d items in %f"
278
% (len(inv), time.time() - before))
282
def _write_inventory(self, inv):
283
"""Update the working inventory.
285
That is to say, the inventory describing changes underway, that
286
will be committed to the next revision.
288
self._need_writelock()
289
## TODO: factor out to atomicfile? is rename safe on windows?
290
## TODO: Maybe some kind of clean/dirty marker on inventory?
291
tmpfname = self.controlfilename('inventory.tmp')
292
tmpf = file(tmpfname, 'wb')
295
inv_fname = self.controlfilename('inventory')
296
if sys.platform == 'win32':
298
os.rename(tmpfname, inv_fname)
299
mutter('wrote working inventory')
302
inventory = property(read_working_inventory, _write_inventory, None,
303
"""Inventory for the working copy.""")
306
def add(self, files, verbose=False):
307
"""Make files versioned.
309
Note that the command line normally calls smart_add instead.
311
This puts the files in the Added state, so that they will be
312
recorded by the next commit.
314
TODO: Perhaps have an option to add the ids even if the files do
317
TODO: Perhaps return the ids of the files? But then again it
318
is easy to retrieve them if they're needed.
320
TODO: Option to specify file id.
322
TODO: Adding a directory should optionally recurse down and
323
add all non-ignored children. Perhaps do that in a
326
self._need_writelock()
328
# TODO: Re-adding a file that is removed in the working copy
329
# should probably put it back with the previous ID.
330
if isinstance(files, types.StringTypes):
333
inv = self.read_working_inventory()
335
if is_control_file(f):
336
bailout("cannot add control file %s" % quotefn(f))
341
bailout("cannot add top-level %r" % f)
343
fullpath = os.path.normpath(self.abspath(f))
346
kind = file_kind(fullpath)
348
# maybe something better?
349
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
351
if kind != 'file' and kind != 'directory':
352
bailout('cannot add: not a regular file or directory: %s' % quotefn(f))
354
file_id = gen_file_id(f)
355
inv.add_path(f, kind=kind, file_id=file_id)
358
show_status('A', kind, quotefn(f))
360
mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))
362
self._write_inventory(inv)
365
def print_file(self, file, revno):
366
"""Print `file` to stdout."""
367
self._need_readlock()
368
tree = self.revision_tree(self.lookup_revision(revno))
369
# use inventory as it was in that revision
370
file_id = tree.inventory.path2id(file)
372
bailout("%r is not present in revision %d" % (file, revno))
373
tree.print_file(file_id)
376
def remove(self, files, verbose=False):
377
"""Mark nominated files for removal from the inventory.
379
This does not remove their text. This does not run on
381
TODO: Refuse to remove modified files unless --force is given?
383
TODO: Do something useful with directories.
385
TODO: Should this remove the text or not? Tough call; not
386
removing may be useful and the user can just use use rm, and
387
is the opposite of add. Removing it is consistent with most
388
other tools. Maybe an option.
390
## TODO: Normalize names
391
## TODO: Remove nested loops; better scalability
392
self._need_writelock()
394
if isinstance(files, types.StringTypes):
397
tree = self.working_tree()
400
# do this before any modifications
404
bailout("cannot remove unversioned file %s" % quotefn(f))
405
mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))
407
# having remove it, it must be either ignored or unknown
408
if tree.is_ignored(f):
412
show_status(new_status, inv[fid].kind, quotefn(f))
415
self._write_inventory(inv)
419
"""Return all unknown files.
421
These are files in the working directory that are not versioned or
422
control files or ignored.
424
>>> b = ScratchBranch(files=['foo', 'foo~'])
425
>>> list(b.unknowns())
428
>>> list(b.unknowns())
431
>>> list(b.unknowns())
434
return self.working_tree().unknowns()
437
def append_revision(self, revision_id):
438
mutter("add {%s} to revision-history" % revision_id)
439
rev_history = self.revision_history()
441
tmprhname = self.controlfilename('revision-history.tmp')
442
rhname = self.controlfilename('revision-history')
444
f = file(tmprhname, 'wt')
445
rev_history.append(revision_id)
446
f.write('\n'.join(rev_history))
450
if sys.platform == 'win32':
452
os.rename(tmprhname, rhname)
456
def get_revision(self, revision_id):
457
"""Return the Revision object for a named revision"""
458
self._need_readlock()
459
r = Revision.read_xml(self.revision_store[revision_id])
460
assert r.revision_id == revision_id
464
def get_inventory(self, inventory_id):
465
"""Get Inventory object by hash.
467
TODO: Perhaps for this and similar methods, take a revision
468
parameter which can be either an integer revno or a
470
self._need_readlock()
471
i = Inventory.read_xml(self.inventory_store[inventory_id])
475
def get_revision_inventory(self, revision_id):
476
"""Return inventory of a past revision."""
477
self._need_readlock()
478
if revision_id == None:
481
return self.get_inventory(self.get_revision(revision_id).inventory_id)
484
def revision_history(self):
485
"""Return sequence of revision hashes on to this branch.
487
>>> ScratchBranch().revision_history()
490
self._need_readlock()
491
return [l.rstrip('\r\n') for l in self.controlfile('revision-history', 'r').readlines()]
494
def enum_history(self, direction):
495
"""Return (revno, revision_id) for history of branch.
498
'forward' is from earliest to latest
499
'reverse' is from latest to earliest
501
rh = self.revision_history()
502
if direction == 'forward':
507
elif direction == 'reverse':
513
raise BzrError('invalid history direction %r' % direction)
517
"""Return current revision number for this branch.
519
That is equivalent to the number of revisions committed to
522
return len(self.revision_history())
525
def last_patch(self):
526
"""Return last patch hash, or None if no history.
528
ph = self.revision_history()
535
def commit(self, *args, **kw):
537
from bzrlib.commit import commit
538
commit(self, *args, **kw)
541
def lookup_revision(self, revno):
542
"""Return revision hash for revision number."""
547
# list is 0-based; revisions are 1-based
548
return self.revision_history()[revno-1]
550
raise BzrError("no such revision %s" % revno)
553
def revision_tree(self, revision_id):
554
"""Return Tree for a revision on this branch.
556
`revision_id` may be None for the null revision, in which case
557
an `EmptyTree` is returned."""
558
self._need_readlock()
559
if revision_id == None:
562
inv = self.get_revision_inventory(revision_id)
563
return RevisionTree(self.text_store, inv)
566
def working_tree(self):
567
"""Return a `Tree` for the working copy."""
568
from workingtree import WorkingTree
569
return WorkingTree(self.base, self.read_working_inventory())
572
def basis_tree(self):
573
"""Return `Tree` object for last revision.
575
If there are no revisions yet, return an `EmptyTree`.
577
r = self.last_patch()
581
return RevisionTree(self.text_store, self.get_revision_inventory(r))
585
def rename_one(self, from_rel, to_rel):
588
This can change the directory or the filename or both.
590
self._need_writelock()
591
tree = self.working_tree()
593
if not tree.has_filename(from_rel):
594
bailout("can't rename: old working file %r does not exist" % from_rel)
595
if tree.has_filename(to_rel):
596
bailout("can't rename: new working file %r already exists" % to_rel)
598
file_id = inv.path2id(from_rel)
600
bailout("can't rename: old name %r is not versioned" % from_rel)
602
if inv.path2id(to_rel):
603
bailout("can't rename: new name %r is already versioned" % to_rel)
605
to_dir, to_tail = os.path.split(to_rel)
606
to_dir_id = inv.path2id(to_dir)
607
if to_dir_id == None and to_dir != '':
608
bailout("can't determine destination directory id for %r" % to_dir)
610
mutter("rename_one:")
611
mutter(" file_id {%s}" % file_id)
612
mutter(" from_rel %r" % from_rel)
613
mutter(" to_rel %r" % to_rel)
614
mutter(" to_dir %r" % to_dir)
615
mutter(" to_dir_id {%s}" % to_dir_id)
617
inv.rename(file_id, to_dir_id, to_tail)
619
print "%s => %s" % (from_rel, to_rel)
621
from_abs = self.abspath(from_rel)
622
to_abs = self.abspath(to_rel)
624
os.rename(from_abs, to_abs)
626
bailout("failed to rename %r to %r: %s"
627
% (from_abs, to_abs, e[1]),
628
["rename rolled back"])
630
self._write_inventory(inv)
634
def move(self, from_paths, to_name):
637
to_name must exist as a versioned directory.
639
If to_name exists and is a directory, the files are moved into
640
it, keeping their old names. If it is a directory,
642
Note that to_name is only the last component of the new name;
643
this doesn't change the directory.
645
self._need_writelock()
646
## TODO: Option to move IDs only
647
assert not isinstance(from_paths, basestring)
648
tree = self.working_tree()
650
to_abs = self.abspath(to_name)
651
if not isdir(to_abs):
652
bailout("destination %r is not a directory" % to_abs)
653
if not tree.has_filename(to_name):
654
bailout("destination %r not in working directory" % to_abs)
655
to_dir_id = inv.path2id(to_name)
656
if to_dir_id == None and to_name != '':
657
bailout("destination %r is not a versioned directory" % to_name)
658
to_dir_ie = inv[to_dir_id]
659
if to_dir_ie.kind not in ('directory', 'root_directory'):
660
bailout("destination %r is not a directory" % to_abs)
662
to_idpath = Set(inv.get_idpath(to_dir_id))
665
if not tree.has_filename(f):
666
bailout("%r does not exist in working tree" % f)
667
f_id = inv.path2id(f)
669
bailout("%r is not versioned" % f)
670
name_tail = splitpath(f)[-1]
671
dest_path = appendpath(to_name, name_tail)
672
if tree.has_filename(dest_path):
673
bailout("destination %r already exists" % dest_path)
674
if f_id in to_idpath:
675
bailout("can't move %r to a subdirectory of itself" % f)
677
# OK, so there's a race here, it's possible that someone will
678
# create a file in this interval and then the rename might be
679
# left half-done. But we should have caught most problems.
682
name_tail = splitpath(f)[-1]
683
dest_path = appendpath(to_name, name_tail)
684
print "%s => %s" % (f, dest_path)
685
inv.rename(inv.path2id(f), to_dir_id, name_tail)
687
os.rename(self.abspath(f), self.abspath(dest_path))
689
bailout("failed to rename %r to %r: %s" % (f, dest_path, e[1]),
690
["rename rolled back"])
692
self._write_inventory(inv)
697
class ScratchBranch(Branch):
698
"""Special test class: a branch that cleans up after itself.
700
>>> b = ScratchBranch()
708
def __init__(self, files=[], dirs=[]):
709
"""Make a test branch.
711
This creates a temporary directory and runs init-tree in it.
713
If any files are listed, they are created in the working copy.
715
Branch.__init__(self, tempfile.mkdtemp(), init=True)
717
os.mkdir(self.abspath(d))
720
file(os.path.join(self.base, f), 'w').write('content of %s' % f)
727
"""Destroy the test branch, removing the scratch directory."""
729
mutter("delete ScratchBranch %s" % self.base)
730
shutil.rmtree(self.base)
732
# Work around for shutil.rmtree failing on Windows when
733
# readonly files are encountered
734
mutter("hit exception in destroying ScratchBranch: %s" % e)
735
for root, dirs, files in os.walk(self.base, topdown=False):
737
os.chmod(os.path.join(root, name), 0700)
738
shutil.rmtree(self.base)
743
######################################################################
747
def is_control_file(filename):
748
## FIXME: better check
749
filename = os.path.normpath(filename)
750
while filename != '':
751
head, tail = os.path.split(filename)
752
## mutter('check %r for control file' % ((head, tail), ))
753
if tail == bzrlib.BZRDIR:
762
def gen_file_id(name):
763
"""Return new file id.
765
This should probably generate proper UUIDs, but for the moment we
766
cope with just randomness because running uuidgen every time is
768
idx = name.rfind('/')
770
name = name[idx+1 : ]
771
idx = name.rfind('\\')
773
name = name[idx+1 : ]
775
name = name.lstrip('.')
777
s = hexlify(rand_bytes(8))
778
return '-'.join((name, compact_date(time.time()), s))