1
# Copyright (C) 2010-2018 Jelmer Vernooij <jelmer@jelmer.uk>
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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""A Git repository implementation that uses a Bazaar transport."""
19
from __future__ import absolute_import
21
from io import BytesIO
26
from dulwich.errors import (
29
from dulwich.file import (
33
from dulwich.objects import (
36
from dulwich.object_store import (
41
from dulwich.pack import (
50
from dulwich.repo import (
61
read_packed_refs_with_peeled,
68
transport as _mod_transport,
71
from ..errors import (
72
AlreadyControlDirError,
82
from ..lock import LogicalLockResult
85
class TransportRefsContainer(RefsContainer):
86
"""Refs container that reads refs from a transport."""
88
def __init__(self, transport, worktree_transport=None):
89
self.transport = transport
90
if worktree_transport is None:
91
worktree_transport = transport
92
self.worktree_transport = worktree_transport
93
self._packed_refs = None
94
self._peeled_refs = None
97
return "%s(%r)" % (self.__class__.__name__, self.transport)
99
def _ensure_dir_exists(self, path):
100
for n in range(path.count("/")):
101
dirname = "/".join(path.split("/")[:n + 1])
103
self.transport.mkdir(dirname)
107
def subkeys(self, base):
108
"""Refs present in this container under a base.
110
:param base: The base to return refs under.
111
:return: A set of valid refs in this container under the base; the base
112
prefix is stripped from the ref names returned.
115
base_len = len(base) + 1
116
for refname in self.allkeys():
117
if refname.startswith(base):
118
keys.add(refname[base_len:])
124
self.worktree_transport.get_bytes("HEAD")
130
iter_files = list(self.transport.clone(
131
"refs").iter_files_recursive())
132
for filename in iter_files:
133
unquoted_filename = urlutils.unquote_to_bytes(filename)
134
refname = osutils.pathjoin(b"refs", unquoted_filename)
135
if check_ref_format(refname):
137
except (TransportNotPossible, NoSuchFile):
139
keys.update(self.get_packed_refs())
142
def get_packed_refs(self):
143
"""Get contents of the packed-refs file.
145
:return: Dictionary mapping ref names to SHA1s
147
:note: Will return an empty dictionary when no packed-refs file is
150
# TODO: invalidate the cache on repacking
151
if self._packed_refs is None:
152
# set both to empty because we want _peeled_refs to be
153
# None if and only if _packed_refs is also None.
154
self._packed_refs = {}
155
self._peeled_refs = {}
157
f = self.transport.get("packed-refs")
161
first_line = next(iter(f)).rstrip()
162
if (first_line.startswith(b"# pack-refs") and b" peeled" in
164
for sha, name, peeled in read_packed_refs_with_peeled(f):
165
self._packed_refs[name] = sha
167
self._peeled_refs[name] = peeled
170
for sha, name in read_packed_refs(f):
171
self._packed_refs[name] = sha
174
return self._packed_refs
176
def get_peeled(self, name):
177
"""Return the cached peeled value of a ref, if available.
179
:param name: Name of the ref to peel
180
:return: The peeled value of the ref. If the ref is known not point to
181
a tag, this will be the SHA the ref refers to. If the ref may point
182
to a tag, but no cached information is available, None is returned.
184
self.get_packed_refs()
185
if self._peeled_refs is None or name not in self._packed_refs:
186
# No cache: no peeled refs were read, or this ref is loose
188
if name in self._peeled_refs:
189
return self._peeled_refs[name]
194
def read_loose_ref(self, name):
195
"""Read a reference file and return its contents.
197
If the reference file a symbolic reference, only read the first line of
198
the file. Otherwise, only read the first 40 bytes.
200
:param name: the refname to read, relative to refpath
201
:return: The contents of the ref file, or None if the file does not
203
:raises IOError: if any other error occurs
206
transport = self.worktree_transport
208
transport = self.transport
210
f = transport.get(urlutils.quote_from_bytes(name))
214
header = f.read(len(SYMREF))
216
# Read only the first line
217
return header + next(iter(f)).rstrip(b"\r\n")
219
# Read only the first 40 bytes
220
return header + f.read(40 - len(SYMREF))
222
def _remove_packed_ref(self, name):
223
if self._packed_refs is None:
225
# reread cached refs from disk, while holding the lock
227
self._packed_refs = None
228
self.get_packed_refs()
230
if name not in self._packed_refs:
233
del self._packed_refs[name]
234
if name in self._peeled_refs:
235
del self._peeled_refs[name]
236
with self.transport.open_write_stream("packed-refs") as f:
237
write_packed_refs(f, self._packed_refs, self._peeled_refs)
239
def set_symbolic_ref(self, name, other):
240
"""Make a ref point at another ref.
242
:param name: Name of the ref to set
243
:param other: Name of the ref to point at
245
self._check_refname(name)
246
self._check_refname(other)
248
transport = self.transport
249
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
251
transport = self.worktree_transport
252
transport.put_bytes(urlutils.quote_from_bytes(
253
name), SYMREF + other + b'\n')
255
def set_if_equals(self, name, old_ref, new_ref):
256
"""Set a refname to new_ref only if it currently equals old_ref.
258
This method follows all symbolic references, and can be used to perform
259
an atomic compare-and-swap operation.
261
:param name: The refname to set.
262
:param old_ref: The old sha the refname must refer to, or None to set
264
:param new_ref: The new sha the refname will refer to.
265
:return: True if the set was successful, False otherwise.
268
realnames, _ = self.follow(name)
269
realname = realnames[-1]
270
except (KeyError, IndexError):
272
if realname == b'HEAD':
273
transport = self.worktree_transport
275
transport = self.transport
276
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
277
transport.put_bytes(urlutils.quote_from_bytes(
278
realname), new_ref + b"\n")
281
def add_if_new(self, name, ref):
282
"""Add a new reference only if it does not already exist.
284
This method follows symrefs, and only ensures that the last ref in the
285
chain does not exist.
287
:param name: The refname to set.
288
:param ref: The new sha the refname will refer to.
289
:return: True if the add was successful, False otherwise.
292
realnames, contents = self.follow(name)
293
if contents is not None:
295
realname = realnames[-1]
296
except (KeyError, IndexError):
298
self._check_refname(realname)
299
if realname == b'HEAD':
300
transport = self.worktree_transport
302
transport = self.transport
303
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
304
transport.put_bytes(urlutils.quote_from_bytes(realname), ref + b"\n")
307
def remove_if_equals(self, name, old_ref):
308
"""Remove a refname only if it currently equals old_ref.
310
This method does not follow symbolic references. It can be used to
311
perform an atomic compare-and-delete operation.
313
:param name: The refname to delete.
314
:param old_ref: The old sha the refname must refer to, or None to
315
delete unconditionally.
316
:return: True if the delete was successful, False otherwise.
318
self._check_refname(name)
321
transport = self.worktree_transport
323
transport = self.transport
325
transport.delete(urlutils.quote_from_bytes(name))
328
self._remove_packed_ref(name)
331
def get(self, name, default=None):
337
def unlock_ref(self, name):
339
transport = self.worktree_transport
341
transport = self.transport
342
lockname = name + b".lock"
344
transport.delete(urlutils.quote_from_bytes(lockname))
348
def lock_ref(self, name):
350
transport = self.worktree_transport
352
transport = self.transport
353
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
354
lockname = urlutils.quote_from_bytes(name + b".lock")
356
local_path = transport.local_abspath(
357
urlutils.quote_from_bytes(name))
359
# This is racy, but what can we do?
360
if transport.has(lockname):
361
raise LockContention(name)
362
transport.put_bytes(lockname, b'Locked by brz-git')
363
return LogicalLockResult(lambda: transport.delete(lockname))
366
gf = GitFile(local_path, 'wb')
367
except FileLocked as e:
368
raise LockContention(name, e)
372
transport.delete(lockname)
374
raise LockBroken(lockname)
375
# GitFile.abort doesn't care if the lock has already
378
return LogicalLockResult(unlock)
381
# TODO(jelmer): Use upstream read_gitfile; unfortunately that expects strings
382
# rather than bytes..
384
"""Read a ``.git`` file.
386
The first line of the file should start with "gitdir: "
388
:param f: File-like object to read from
392
if not cs.startswith(b"gitdir: "):
393
raise ValueError("Expected file to start with 'gitdir: '")
394
return cs[len(b"gitdir: "):].rstrip(b"\n")
397
class TransportRepo(BaseRepo):
399
def __init__(self, transport, bare, refs_text=None):
400
self.transport = transport
403
with transport.get(CONTROLDIR) as f:
404
path = read_gitfile(f)
405
except (ReadError, NoSuchFile):
407
self._controltransport = self.transport
409
self._controltransport = self.transport.clone('.git')
411
self._controltransport = self.transport.clone(
412
urlutils.quote_from_bytes(path))
413
commondir = self.get_named_file(COMMONDIR)
414
if commondir is not None:
416
commondir = os.path.join(
418
commondir.read().rstrip(b"\r\n").decode(
419
sys.getfilesystemencoding()))
420
self._commontransport = \
421
_mod_transport.get_transport_from_path(commondir)
423
self._commontransport = self._controltransport
424
object_store = TransportObjectStore(
425
self._commontransport.clone(OBJECTDIR))
426
if refs_text is not None:
427
refs_container = InfoRefsContainer(BytesIO(refs_text))
429
head = TransportRefsContainer(
430
self._commontransport).read_loose_ref(b"HEAD")
434
refs_container._refs[b"HEAD"] = head
436
refs_container = TransportRefsContainer(
437
self._commontransport, self._controltransport)
438
super(TransportRepo, self).__init__(object_store,
441
def controldir(self):
442
return self._controltransport.local_abspath('.')
445
return self._commontransport.local_abspath('.')
449
return self.transport.local_abspath('.')
451
def _determine_file_mode(self):
452
# Be consistent with bzr
453
if sys.platform == 'win32':
457
def get_named_file(self, path):
458
"""Get a file from the control dir with a specific name.
460
Although the filename should be interpreted as a filename relative to
461
the control dir in a disk-baked Repo, the object returned need not be
462
pointing to a file in that location.
464
:param path: The path to the file, relative to the control dir.
465
:return: An open file object, or None if the file does not exist.
468
return self._controltransport.get(path.lstrip('/'))
472
def _put_named_file(self, relpath, contents):
473
self._controltransport.put_bytes(relpath, contents)
475
def index_path(self):
476
"""Return the path to the index file."""
477
return self._controltransport.local_abspath(INDEX_FILENAME)
479
def open_index(self):
480
"""Open the index for this repository."""
481
from dulwich.index import Index
482
if not self.has_index():
483
raise NoIndexPresent()
484
return Index(self.index_path())
487
"""Check if an index is present."""
488
# Bare repos must never have index files; non-bare repos may have a
489
# missing index file, which is treated as empty.
492
def get_config(self):
493
from dulwich.config import ConfigFile
495
with self._controltransport.get('config') as f:
496
return ConfigFile.from_file(f)
500
def get_config_stack(self):
501
from dulwich.config import StackedConfig
503
p = self.get_config()
509
backends.extend(StackedConfig.default_backends())
510
return StackedConfig(backends, writable=writable)
513
return "<%s for %r>" % (self.__class__.__name__, self.transport)
516
def init(cls, transport, bare=False):
519
transport.mkdir(".git")
521
raise AlreadyControlDirError(transport.base)
522
control_transport = transport.clone(".git")
524
control_transport = transport
525
for d in BASE_DIRECTORIES:
527
control_transport.mkdir("/".join(d))
531
control_transport.mkdir(OBJECTDIR)
533
raise AlreadyControlDirError(transport.base)
534
TransportObjectStore.init(control_transport.clone(OBJECTDIR))
535
ret = cls(transport, bare)
536
ret.refs.set_symbolic_ref(b"HEAD", b"refs/heads/master")
537
ret._init_files(bare)
541
class TransportObjectStore(PackBasedObjectStore):
542
"""Git-style object store that exists on disk."""
544
def __init__(self, transport):
545
"""Open an object store.
547
:param transport: Transport to open data from
549
super(TransportObjectStore, self).__init__()
550
self.transport = transport
551
self.pack_transport = self.transport.clone(PACKDIR)
552
self._alternates = None
554
def __eq__(self, other):
555
if not isinstance(other, TransportObjectStore):
557
return self.transport == other.transport
560
return "%s(%r)" % (self.__class__.__name__, self.transport)
563
def alternates(self):
564
if self._alternates is not None:
565
return self._alternates
566
self._alternates = []
567
for path in self._read_alternate_paths():
569
t = _mod_transport.get_transport_from_path(path)
570
self._alternates.append(self.__class__(t))
571
return self._alternates
573
def _read_alternate_paths(self):
575
f = self.transport.get("info/alternates")
580
for l in f.read().splitlines():
588
def _update_pack_cache(self):
590
pack_dir_contents = self._pack_names()
591
for name in pack_dir_contents:
592
if name.startswith("pack-") and name.endswith(".pack"):
593
# verify that idx exists first (otherwise the pack was not yet
595
idx_name = os.path.splitext(name)[0] + ".idx"
596
if idx_name in pack_dir_contents:
597
pack_files.add(os.path.splitext(name)[0])
600
for basename in pack_files:
601
pack_name = basename + ".pack"
602
if basename not in self._pack_cache:
604
size = self.pack_transport.stat(pack_name).st_size
605
except TransportNotPossible:
606
f = self.pack_transport.get(pack_name)
607
pd = PackData(pack_name, f)
610
pack_name, self.pack_transport.get(pack_name),
612
idxname = basename + ".idx"
613
idx = load_pack_index_file(
614
idxname, self.pack_transport.get(idxname))
615
pack = Pack.from_objects(pd, idx)
616
pack._basename = basename
617
self._pack_cache[basename] = pack
618
new_packs.append(pack)
619
# Remove disappeared pack files
620
for f in set(self._pack_cache) - pack_files:
621
self._pack_cache.pop(f).close()
624
def _pack_names(self):
626
return self.pack_transport.list_dir(".")
627
except TransportNotPossible:
629
f = self.transport.get('info/packs')
631
# Hmm, warn about running 'git update-server-info' ?
635
return read_packs_file(f)
639
def _remove_pack(self, pack):
640
self.pack_transport.delete(os.path.basename(pack.index.path))
641
self.pack_transport.delete(pack.data.filename)
643
del self._pack_cache[os.path.basename(pack._basename)]
647
def _iter_loose_objects(self):
648
for base in self.transport.list_dir('.'):
651
for rest in self.transport.list_dir(base):
652
yield (base + rest).encode(sys.getfilesystemencoding())
654
def _split_loose_object(self, sha):
655
return (sha[:2], sha[2:])
657
def _remove_loose_object(self, sha):
658
path = osutils.joinpath(self._split_loose_object(sha))
659
self.transport.delete(urlutils.quote_from_bytes(path))
661
def _get_loose_object(self, sha):
662
path = osutils.joinpath(self._split_loose_object(sha))
664
with self.transport.get(urlutils.quote_from_bytes(path)) as f:
665
return ShaFile.from_file(f)
669
def add_object(self, obj):
670
"""Add a single object to this object store.
672
:param obj: Object to add
674
(dir, file) = self._split_loose_object(obj.id)
676
self.transport.mkdir(urlutils.quote_from_bytes(dir))
679
path = urlutils.quote_from_bytes(osutils.pathjoin(dir, file))
680
if self.transport.has(path):
681
return # Already there, no need to write again
682
self.transport.put_bytes(path, obj.as_legacy_object())
684
def move_in_pack(self, f):
685
"""Move a specific file containing a pack into the pack directory.
687
:note: The file should be on the same file system as the
690
:param path: Path to the pack file.
693
p = PackData("", f, len(f.getvalue()))
694
entries = p.sorted_entries()
695
basename = "pack-%s" % iter_sha1(entry[0]
696
for entry in entries).decode('ascii')
697
p._filename = basename + ".pack"
699
self.pack_transport.put_file(basename + ".pack", f)
700
with self.pack_transport.open_write_stream(basename + ".idx") as idxfile:
701
write_pack_index_v2(idxfile, entries, p.get_stored_checksum())
702
idxfile = self.pack_transport.get(basename + ".idx")
703
idx = load_pack_index_file(basename + ".idx", idxfile)
704
final_pack = Pack.from_objects(p, idx)
705
final_pack._basename = basename
706
self._add_cached_pack(basename, final_pack)
709
def move_in_thin_pack(self, f):
710
"""Move a specific file containing a pack into the pack directory.
712
:note: The file should be on the same file system as the
715
:param path: Path to the pack file.
718
p = Pack('', resolve_ext_ref=self.get_raw)
719
p._data = PackData.from_file(f, len(f.getvalue()))
721
p._idx_load = lambda: MemoryPackIndex(
722
p.data.sorted_entries(), p.data.get_stored_checksum())
724
pack_sha = p.index.objects_sha1()
726
with self.pack_transport.open_write_stream(
727
"pack-%s.pack" % pack_sha.decode('ascii')) as datafile:
728
entries, data_sum = write_pack_objects(datafile, p.pack_tuples())
729
entries = sorted([(k, v[0], v[1]) for (k, v) in entries.items()])
730
with self.pack_transport.open_write_stream(
731
"pack-%s.idx" % pack_sha.decode('ascii')) as idxfile:
732
write_pack_index_v2(idxfile, entries, data_sum)
735
"""Add a new pack to this object store.
737
:return: Fileobject to write to and a commit function to
738
call when the pack is finished.
743
if len(f.getvalue()) > 0:
744
return self.move_in_pack(f)
750
return f, commit, abort
753
def init(cls, transport):
755
transport.mkdir('info')
759
transport.mkdir(PACKDIR)
762
return cls(transport)