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))
215
header = f.read(len(SYMREF))
217
# probably a directory
220
# Read only the first line
221
return header + next(iter(f)).rstrip(b"\r\n")
223
# Read only the first 40 bytes
224
return header + f.read(40 - len(SYMREF))
226
def _remove_packed_ref(self, name):
227
if self._packed_refs is None:
229
# reread cached refs from disk, while holding the lock
231
self._packed_refs = None
232
self.get_packed_refs()
234
if name not in self._packed_refs:
237
del self._packed_refs[name]
238
if name in self._peeled_refs:
239
del self._peeled_refs[name]
240
f = self.transport.open_write_stream("packed-refs")
242
write_packed_refs(f, self._packed_refs, self._peeled_refs)
246
def set_symbolic_ref(self, name, other):
247
"""Make a ref point at another ref.
249
:param name: Name of the ref to set
250
:param other: Name of the ref to point at
252
self._check_refname(name)
253
self._check_refname(other)
255
transport = self.transport
256
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
258
transport = self.worktree_transport
259
transport.put_bytes(urlutils.quote_from_bytes(
260
name), SYMREF + other + b'\n')
262
def set_if_equals(self, name, old_ref, new_ref):
263
"""Set a refname to new_ref only if it currently equals old_ref.
265
This method follows all symbolic references, and can be used to perform
266
an atomic compare-and-swap operation.
268
:param name: The refname to set.
269
:param old_ref: The old sha the refname must refer to, or None to set
271
:param new_ref: The new sha the refname will refer to.
272
:return: True if the set was successful, False otherwise.
275
realnames, _ = self.follow(name)
276
realname = realnames[-1]
277
except (KeyError, IndexError):
279
if realname == b'HEAD':
280
transport = self.worktree_transport
282
transport = self.transport
283
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
284
transport.put_bytes(urlutils.quote_from_bytes(
285
realname), new_ref + b"\n")
288
def add_if_new(self, name, ref):
289
"""Add a new reference only if it does not already exist.
291
This method follows symrefs, and only ensures that the last ref in the
292
chain does not exist.
294
:param name: The refname to set.
295
:param ref: The new sha the refname will refer to.
296
:return: True if the add was successful, False otherwise.
299
realnames, contents = self.follow(name)
300
if contents is not None:
302
realname = realnames[-1]
303
except (KeyError, IndexError):
305
self._check_refname(realname)
306
if realname == b'HEAD':
307
transport = self.worktree_transport
309
transport = self.transport
310
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
311
transport.put_bytes(urlutils.quote_from_bytes(realname), ref + b"\n")
314
def remove_if_equals(self, name, old_ref):
315
"""Remove a refname only if it currently equals old_ref.
317
This method does not follow symbolic references. It can be used to
318
perform an atomic compare-and-delete operation.
320
:param name: The refname to delete.
321
:param old_ref: The old sha the refname must refer to, or None to
322
delete unconditionally.
323
:return: True if the delete was successful, False otherwise.
325
self._check_refname(name)
328
transport = self.worktree_transport
330
transport = self.transport
332
transport.delete(urlutils.quote_from_bytes(name))
335
self._remove_packed_ref(name)
338
def get(self, name, default=None):
344
def unlock_ref(self, name):
346
transport = self.worktree_transport
348
transport = self.transport
349
lockname = name + b".lock"
351
transport.delete(urlutils.quote_from_bytes(lockname))
355
def lock_ref(self, name):
357
transport = self.worktree_transport
359
transport = self.transport
360
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
361
lockname = urlutils.quote_from_bytes(name + b".lock")
363
local_path = transport.local_abspath(
364
urlutils.quote_from_bytes(name))
366
# This is racy, but what can we do?
367
if transport.has(lockname):
368
raise LockContention(name)
369
transport.put_bytes(lockname, b'Locked by brz-git')
370
return LogicalLockResult(lambda: transport.delete(lockname))
373
gf = GitFile(local_path, 'wb')
374
except FileLocked as e:
375
raise LockContention(name, e)
379
transport.delete(lockname)
381
raise LockBroken(lockname)
382
# GitFile.abort doesn't care if the lock has already
385
return LogicalLockResult(unlock)
388
# TODO(jelmer): Use upstream read_gitfile; unfortunately that expects strings
389
# rather than bytes..
391
"""Read a ``.git`` file.
393
The first line of the file should start with "gitdir: "
395
:param f: File-like object to read from
399
if not cs.startswith(b"gitdir: "):
400
raise ValueError("Expected file to start with 'gitdir: '")
401
return cs[len(b"gitdir: "):].rstrip(b"\n")
404
class TransportRepo(BaseRepo):
406
def __init__(self, transport, bare, refs_text=None):
407
self.transport = transport
410
with transport.get(CONTROLDIR) as f:
411
path = read_gitfile(f)
412
except (ReadError, NoSuchFile):
414
self._controltransport = self.transport
416
self._controltransport = self.transport.clone('.git')
418
self._controltransport = self.transport.clone(
419
urlutils.quote_from_bytes(path))
420
commondir = self.get_named_file(COMMONDIR)
421
if commondir is not None:
423
commondir = os.path.join(
425
commondir.read().rstrip(b"\r\n").decode(
426
sys.getfilesystemencoding()))
427
self._commontransport = \
428
_mod_transport.get_transport_from_path(commondir)
430
self._commontransport = self._controltransport
431
object_store = TransportObjectStore(
432
self._commontransport.clone(OBJECTDIR))
433
if refs_text is not None:
434
refs_container = InfoRefsContainer(BytesIO(refs_text))
436
head = TransportRefsContainer(
437
self._commontransport).read_loose_ref("HEAD")
441
refs_container._refs["HEAD"] = head
443
refs_container = TransportRefsContainer(
444
self._commontransport, self._controltransport)
445
super(TransportRepo, self).__init__(object_store,
448
def controldir(self):
449
return self._controltransport.local_abspath('.')
452
return self._commontransport.local_abspath('.')
456
return self.transport.local_abspath('.')
458
def _determine_file_mode(self):
459
# Be consistent with bzr
460
if sys.platform == 'win32':
464
def get_named_file(self, path):
465
"""Get a file from the control dir with a specific name.
467
Although the filename should be interpreted as a filename relative to
468
the control dir in a disk-baked Repo, the object returned need not be
469
pointing to a file in that location.
471
:param path: The path to the file, relative to the control dir.
472
:return: An open file object, or None if the file does not exist.
475
return self._controltransport.get(path.lstrip('/'))
479
def _put_named_file(self, relpath, contents):
480
self._controltransport.put_bytes(relpath, contents)
482
def index_path(self):
483
"""Return the path to the index file."""
484
return self._controltransport.local_abspath(INDEX_FILENAME)
486
def open_index(self):
487
"""Open the index for this repository."""
488
from dulwich.index import Index
489
if not self.has_index():
490
raise NoIndexPresent()
491
return Index(self.index_path())
494
"""Check if an index is present."""
495
# Bare repos must never have index files; non-bare repos may have a
496
# missing index file, which is treated as empty.
499
def get_config(self):
500
from dulwich.config import ConfigFile
502
with self._controltransport.get('config') as f:
503
return ConfigFile.from_file(f)
507
def get_config_stack(self):
508
from dulwich.config import StackedConfig
510
p = self.get_config()
516
backends.extend(StackedConfig.default_backends())
517
return StackedConfig(backends, writable=writable)
520
return "<%s for %r>" % (self.__class__.__name__, self.transport)
523
def init(cls, transport, bare=False):
526
transport.mkdir(".git")
528
raise AlreadyControlDirError(transport.base)
529
control_transport = transport.clone(".git")
531
control_transport = transport
532
for d in BASE_DIRECTORIES:
534
control_transport.mkdir("/".join(d))
538
control_transport.mkdir(OBJECTDIR)
540
raise AlreadyControlDirError(transport.base)
541
TransportObjectStore.init(control_transport.clone(OBJECTDIR))
542
ret = cls(transport, bare)
543
ret.refs.set_symbolic_ref(b"HEAD", b"refs/heads/master")
544
ret._init_files(bare)
548
class TransportObjectStore(PackBasedObjectStore):
549
"""Git-style object store that exists on disk."""
551
def __init__(self, transport):
552
"""Open an object store.
554
:param transport: Transport to open data from
556
super(TransportObjectStore, self).__init__()
557
self.transport = transport
558
self.pack_transport = self.transport.clone(PACKDIR)
559
self._alternates = None
561
def __eq__(self, other):
562
if not isinstance(other, TransportObjectStore):
564
return self.transport == other.transport
567
return "%s(%r)" % (self.__class__.__name__, self.transport)
570
def alternates(self):
571
if self._alternates is not None:
572
return self._alternates
573
self._alternates = []
574
for path in self._read_alternate_paths():
576
t = _mod_transport.get_transport_from_path(path)
577
self._alternates.append(self.__class__(t))
578
return self._alternates
580
def _read_alternate_paths(self):
582
f = self.transport.get("info/alternates")
587
for l in f.read().splitlines():
595
def _update_pack_cache(self):
597
pack_dir_contents = self._pack_names()
598
for name in pack_dir_contents:
599
if name.startswith("pack-") and name.endswith(".pack"):
600
# verify that idx exists first (otherwise the pack was not yet
602
idx_name = os.path.splitext(name)[0] + ".idx"
603
if idx_name in pack_dir_contents:
604
pack_files.add(os.path.splitext(name)[0])
607
for basename in pack_files:
608
pack_name = basename + ".pack"
609
if basename not in self._pack_cache:
611
size = self.pack_transport.stat(pack_name).st_size
612
except TransportNotPossible:
613
f = self.pack_transport.get(pack_name)
614
pd = PackData(pack_name, f)
617
pack_name, self.pack_transport.get(pack_name),
619
idxname = basename + ".idx"
620
idx = load_pack_index_file(
621
idxname, self.pack_transport.get(idxname))
622
pack = Pack.from_objects(pd, idx)
623
pack._basename = basename
624
self._pack_cache[basename] = pack
625
new_packs.append(pack)
626
# Remove disappeared pack files
627
for f in set(self._pack_cache) - pack_files:
628
self._pack_cache.pop(f).close()
631
def _pack_names(self):
633
return self.pack_transport.list_dir(".")
634
except TransportNotPossible:
636
f = self.transport.get('info/packs')
638
# Hmm, warn about running 'git update-server-info' ?
642
return read_packs_file(f)
646
def _remove_pack(self, pack):
647
self.pack_transport.delete(os.path.basename(pack.index.path))
648
self.pack_transport.delete(pack.data.filename)
650
del self._pack_cache[os.path.basename(pack._basename)]
654
def _iter_loose_objects(self):
655
for base in self.transport.list_dir('.'):
658
for rest in self.transport.list_dir(base):
659
yield (base + rest).encode(sys.getfilesystemencoding())
661
def _split_loose_object(self, sha):
662
return (sha[:2], sha[2:])
664
def _remove_loose_object(self, sha):
665
path = osutils.joinpath(self._split_loose_object(sha))
666
self.transport.delete(urlutils.quote_from_bytes(path))
668
def _get_loose_object(self, sha):
669
path = osutils.joinpath(self._split_loose_object(sha))
671
with self.transport.get(urlutils.quote_from_bytes(path)) as f:
672
return ShaFile.from_file(f)
676
def add_object(self, obj):
677
"""Add a single object to this object store.
679
:param obj: Object to add
681
(dir, file) = self._split_loose_object(obj.id)
683
self.transport.mkdir(urlutils.quote_from_bytes(dir))
686
path = urlutils.quote_from_bytes(osutils.pathjoin(dir, file))
687
if self.transport.has(path):
688
return # Already there, no need to write again
689
self.transport.put_bytes(path, obj.as_legacy_object())
691
def move_in_pack(self, f):
692
"""Move a specific file containing a pack into the pack directory.
694
:note: The file should be on the same file system as the
697
:param path: Path to the pack file.
700
p = PackData("", f, len(f.getvalue()))
701
entries = p.sorted_entries()
702
basename = "pack-%s" % iter_sha1(entry[0]
703
for entry in entries).decode('ascii')
704
p._filename = basename + ".pack"
706
self.pack_transport.put_file(basename + ".pack", f)
707
idxfile = self.pack_transport.open_write_stream(basename + ".idx")
709
write_pack_index_v2(idxfile, entries, p.get_stored_checksum())
712
idxfile = self.pack_transport.get(basename + ".idx")
713
idx = load_pack_index_file(basename + ".idx", idxfile)
714
final_pack = Pack.from_objects(p, idx)
715
final_pack._basename = basename
716
self._add_cached_pack(basename, final_pack)
719
def move_in_thin_pack(self, f):
720
"""Move a specific file containing a pack into the pack directory.
722
:note: The file should be on the same file system as the
725
:param path: Path to the pack file.
728
p = Pack('', resolve_ext_ref=self.get_raw)
729
p._data = PackData.from_file(f, len(f.getvalue()))
731
p._idx_load = lambda: MemoryPackIndex(
732
p.data.sorted_entries(), p.data.get_stored_checksum())
734
pack_sha = p.index.objects_sha1()
736
datafile = self.pack_transport.open_write_stream(
737
"pack-%s.pack" % pack_sha.decode('ascii'))
739
entries, data_sum = write_pack_objects(datafile, p.pack_tuples())
742
entries = sorted([(k, v[0], v[1]) for (k, v) in entries.items()])
743
idxfile = self.pack_transport.open_write_stream(
744
"pack-%s.idx" % pack_sha.decode('ascii'))
746
write_pack_index_v2(idxfile, entries, data_sum)
751
"""Add a new pack to this object store.
753
:return: Fileobject to write to and a commit function to
754
call when the pack is finished.
759
if len(f.getvalue()) > 0:
760
return self.move_in_pack(f)
766
return f, commit, abort
769
def init(cls, transport):
771
transport.mkdir('info')
775
transport.mkdir(PACKDIR)
778
return cls(transport)