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
83
from ..trace import warning
86
class TransportRefsContainer(RefsContainer):
87
"""Refs container that reads refs from a transport."""
89
def __init__(self, transport, worktree_transport=None):
90
self.transport = transport
91
if worktree_transport is None:
92
worktree_transport = transport
93
self.worktree_transport = worktree_transport
94
self._packed_refs = None
95
self._peeled_refs = None
98
return "%s(%r)" % (self.__class__.__name__, self.transport)
100
def _ensure_dir_exists(self, path):
101
for n in range(path.count("/")):
102
dirname = "/".join(path.split("/")[:n + 1])
104
self.transport.mkdir(dirname)
108
def subkeys(self, base):
109
"""Refs present in this container under a base.
111
:param base: The base to return refs under.
112
:return: A set of valid refs in this container under the base; the base
113
prefix is stripped from the ref names returned.
116
base_len = len(base) + 1
117
for refname in self.allkeys():
118
if refname.startswith(base):
119
keys.add(refname[base_len:])
125
self.worktree_transport.get_bytes("HEAD")
131
iter_files = list(self.transport.clone(
132
"refs").iter_files_recursive())
133
for filename in iter_files:
134
unquoted_filename = urlutils.unquote_to_bytes(filename)
135
refname = osutils.pathjoin(b"refs", unquoted_filename)
136
if check_ref_format(refname):
138
except (TransportNotPossible, NoSuchFile):
140
keys.update(self.get_packed_refs())
143
def get_packed_refs(self):
144
"""Get contents of the packed-refs file.
146
:return: Dictionary mapping ref names to SHA1s
148
:note: Will return an empty dictionary when no packed-refs file is
151
# TODO: invalidate the cache on repacking
152
if self._packed_refs is None:
153
# set both to empty because we want _peeled_refs to be
154
# None if and only if _packed_refs is also None.
155
self._packed_refs = {}
156
self._peeled_refs = {}
158
f = self.transport.get("packed-refs")
162
first_line = next(iter(f)).rstrip()
163
if (first_line.startswith(b"# pack-refs") and b" peeled" in
165
for sha, name, peeled in read_packed_refs_with_peeled(f):
166
self._packed_refs[name] = sha
168
self._peeled_refs[name] = peeled
171
for sha, name in read_packed_refs(f):
172
self._packed_refs[name] = sha
175
return self._packed_refs
177
def get_peeled(self, name):
178
"""Return the cached peeled value of a ref, if available.
180
:param name: Name of the ref to peel
181
:return: The peeled value of the ref. If the ref is known not point to
182
a tag, this will be the SHA the ref refers to. If the ref may point
183
to a tag, but no cached information is available, None is returned.
185
self.get_packed_refs()
186
if self._peeled_refs is None or name not in self._packed_refs:
187
# No cache: no peeled refs were read, or this ref is loose
189
if name in self._peeled_refs:
190
return self._peeled_refs[name]
195
def read_loose_ref(self, name):
196
"""Read a reference file and return its contents.
198
If the reference file a symbolic reference, only read the first line of
199
the file. Otherwise, only read the first 40 bytes.
201
:param name: the refname to read, relative to refpath
202
:return: The contents of the ref file, or None if the file does not
204
:raises IOError: if any other error occurs
207
transport = self.worktree_transport
209
transport = self.transport
211
f = transport.get(urlutils.quote_from_bytes(name))
216
header = f.read(len(SYMREF))
218
# probably a directory
221
# Read only the first line
222
return header + next(iter(f)).rstrip(b"\r\n")
224
# Read only the first 40 bytes
225
return header + f.read(40 - len(SYMREF))
227
def _remove_packed_ref(self, name):
228
if self._packed_refs is None:
230
# reread cached refs from disk, while holding the lock
232
self._packed_refs = None
233
self.get_packed_refs()
235
if name not in self._packed_refs:
238
del self._packed_refs[name]
239
if name in self._peeled_refs:
240
del self._peeled_refs[name]
241
with self.transport.open_write_stream("packed-refs") as f:
242
write_packed_refs(f, self._packed_refs, self._peeled_refs)
244
def set_symbolic_ref(self, name, other):
245
"""Make a ref point at another ref.
247
:param name: Name of the ref to set
248
:param other: Name of the ref to point at
250
self._check_refname(name)
251
self._check_refname(other)
253
transport = self.transport
254
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
256
transport = self.worktree_transport
257
transport.put_bytes(urlutils.quote_from_bytes(
258
name), SYMREF + other + b'\n')
260
def set_if_equals(self, name, old_ref, new_ref):
261
"""Set a refname to new_ref only if it currently equals old_ref.
263
This method follows all symbolic references, and can be used to perform
264
an atomic compare-and-swap operation.
266
:param name: The refname to set.
267
:param old_ref: The old sha the refname must refer to, or None to set
269
:param new_ref: The new sha the refname will refer to.
270
:return: True if the set was successful, False otherwise.
273
realnames, _ = self.follow(name)
274
realname = realnames[-1]
275
except (KeyError, IndexError):
277
if realname == b'HEAD':
278
transport = self.worktree_transport
280
transport = self.transport
281
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
282
transport.put_bytes(urlutils.quote_from_bytes(
283
realname), new_ref + b"\n")
286
def add_if_new(self, name, ref):
287
"""Add a new reference only if it does not already exist.
289
This method follows symrefs, and only ensures that the last ref in the
290
chain does not exist.
292
:param name: The refname to set.
293
:param ref: The new sha the refname will refer to.
294
:return: True if the add was successful, False otherwise.
297
realnames, contents = self.follow(name)
298
if contents is not None:
300
realname = realnames[-1]
301
except (KeyError, IndexError):
303
self._check_refname(realname)
304
if realname == b'HEAD':
305
transport = self.worktree_transport
307
transport = self.transport
308
self._ensure_dir_exists(urlutils.quote_from_bytes(realname))
309
transport.put_bytes(urlutils.quote_from_bytes(realname), ref + b"\n")
312
def remove_if_equals(self, name, old_ref):
313
"""Remove a refname only if it currently equals old_ref.
315
This method does not follow symbolic references. It can be used to
316
perform an atomic compare-and-delete operation.
318
:param name: The refname to delete.
319
:param old_ref: The old sha the refname must refer to, or None to
320
delete unconditionally.
321
:return: True if the delete was successful, False otherwise.
323
self._check_refname(name)
326
transport = self.worktree_transport
328
transport = self.transport
330
transport.delete(urlutils.quote_from_bytes(name))
333
self._remove_packed_ref(name)
336
def get(self, name, default=None):
342
def unlock_ref(self, name):
344
transport = self.worktree_transport
346
transport = self.transport
347
lockname = name + b".lock"
349
transport.delete(urlutils.quote_from_bytes(lockname))
353
def lock_ref(self, name):
355
transport = self.worktree_transport
357
transport = self.transport
358
self._ensure_dir_exists(urlutils.quote_from_bytes(name))
359
lockname = urlutils.quote_from_bytes(name + b".lock")
361
local_path = transport.local_abspath(
362
urlutils.quote_from_bytes(name))
364
# This is racy, but what can we do?
365
if transport.has(lockname):
366
raise LockContention(name)
367
transport.put_bytes(lockname, b'Locked by brz-git')
368
return LogicalLockResult(lambda: transport.delete(lockname))
371
gf = GitFile(local_path, 'wb')
372
except FileLocked as e:
373
raise LockContention(name, e)
377
transport.delete(lockname)
379
raise LockBroken(lockname)
380
# GitFile.abort doesn't care if the lock has already
383
return LogicalLockResult(unlock)
386
# TODO(jelmer): Use upstream read_gitfile; unfortunately that expects strings
387
# rather than bytes..
389
"""Read a ``.git`` file.
391
The first line of the file should start with "gitdir: "
393
:param f: File-like object to read from
397
if not cs.startswith(b"gitdir: "):
398
raise ValueError("Expected file to start with 'gitdir: '")
399
return cs[len(b"gitdir: "):].rstrip(b"\n")
402
class TransportRepo(BaseRepo):
404
def __init__(self, transport, bare, refs_text=None):
405
self.transport = transport
408
with transport.get(CONTROLDIR) as f:
409
path = read_gitfile(f)
410
except (ReadError, NoSuchFile):
412
self._controltransport = self.transport
414
self._controltransport = self.transport.clone('.git')
416
self._controltransport = self.transport.clone(
417
urlutils.quote_from_bytes(path))
418
commondir = self.get_named_file(COMMONDIR)
419
if commondir is not None:
421
commondir = os.path.join(
423
commondir.read().rstrip(b"\r\n").decode(
424
sys.getfilesystemencoding()))
425
self._commontransport = \
426
_mod_transport.get_transport_from_path(commondir)
428
self._commontransport = self._controltransport
429
config = self.get_config()
430
object_store = TransportObjectStore.from_config(
431
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(b"HEAD")
441
refs_container._refs[b"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
loose_compression_level=-1, pack_compression_level=-1):
553
"""Open an object store.
555
:param transport: Transport to open data from
557
super(TransportObjectStore, self).__init__()
558
self.pack_compression_level = pack_compression_level
559
self.loose_compression_level = loose_compression_level
560
self.transport = transport
561
self.pack_transport = self.transport.clone(PACKDIR)
562
self._alternates = None
565
def from_config(cls, path, config):
567
default_compression_level = int(config.get(
568
(b'core', ), b'compression').decode())
570
default_compression_level = -1
572
loose_compression_level = int(config.get(
573
(b'core', ), b'looseCompression').decode())
575
loose_compression_level = default_compression_level
577
pack_compression_level = int(config.get(
578
(b'core', ), 'packCompression').decode())
580
pack_compression_level = default_compression_level
581
return cls(path, loose_compression_level, pack_compression_level)
583
def __eq__(self, other):
584
if not isinstance(other, TransportObjectStore):
586
return self.transport == other.transport
589
return "%s(%r)" % (self.__class__.__name__, self.transport)
592
def alternates(self):
593
if self._alternates is not None:
594
return self._alternates
595
self._alternates = []
596
for path in self._read_alternate_paths():
598
t = _mod_transport.get_transport_from_path(path)
599
self._alternates.append(self.__class__(t))
600
return self._alternates
602
def _read_alternate_paths(self):
604
f = self.transport.get("info/alternates")
609
for l in f.read().splitlines():
617
def _update_pack_cache(self):
618
pack_files = set(self._pack_names())
620
for basename in pack_files:
621
pack_name = basename + ".pack"
622
if basename not in self._pack_cache:
624
size = self.pack_transport.stat(pack_name).st_size
625
except TransportNotPossible:
626
f = self.pack_transport.get(pack_name)
627
# TODO(jelmer): Don't read entire file into memory?
628
f = BytesIO(f.read())
629
pd = PackData(pack_name, f)
632
pack_name, self.pack_transport.get(pack_name),
634
idxname = basename + ".idx"
635
idx = load_pack_index_file(
636
idxname, self.pack_transport.get(idxname))
637
pack = Pack.from_objects(pd, idx)
638
pack._basename = basename
639
self._pack_cache[basename] = pack
640
new_packs.append(pack)
641
# Remove disappeared pack files
642
for f in set(self._pack_cache) - pack_files:
643
self._pack_cache.pop(f).close()
646
def _pack_names(self):
649
dir_contents = self.pack_transport.list_dir(".")
650
for name in dir_contents:
651
if name.startswith("pack-") and name.endswith(".pack"):
652
# verify that idx exists first (otherwise the pack was not yet
654
idx_name = os.path.splitext(name)[0] + ".idx"
655
if idx_name in dir_contents:
656
pack_files.append(os.path.splitext(name)[0])
657
except TransportNotPossible:
659
f = self.transport.get('info/packs')
661
warning('No info/packs on remote host;'
662
'run \'git update-server-info\' on remote.')
666
os.path.splitext(name)[0]
667
for name in read_packs_file(f)]
672
def _remove_pack(self, pack):
673
self.pack_transport.delete(os.path.basename(pack.index.path))
674
self.pack_transport.delete(pack.data.filename)
676
del self._pack_cache[os.path.basename(pack._basename)]
680
def _iter_loose_objects(self):
681
for base in self.transport.list_dir('.'):
684
for rest in self.transport.list_dir(base):
685
yield (base + rest).encode(sys.getfilesystemencoding())
687
def _split_loose_object(self, sha):
688
return (sha[:2], sha[2:])
690
def _remove_loose_object(self, sha):
691
path = osutils.joinpath(self._split_loose_object(sha))
692
self.transport.delete(urlutils.quote_from_bytes(path))
694
def _get_loose_object(self, sha):
695
path = osutils.joinpath(self._split_loose_object(sha))
697
with self.transport.get(urlutils.quote_from_bytes(path)) as f:
698
return ShaFile.from_file(f)
702
def add_object(self, obj):
703
"""Add a single object to this object store.
705
:param obj: Object to add
707
(dir, file) = self._split_loose_object(obj.id)
709
self.transport.mkdir(urlutils.quote_from_bytes(dir))
712
path = urlutils.quote_from_bytes(osutils.pathjoin(dir, file))
713
if self.transport.has(path):
714
return # Already there, no need to write again
715
# Backwards compatibility with Dulwich < 0.20, which doesn't support
716
# the compression_level parameter.
717
if self.loose_compression_level not in (-1, None):
718
raw_string = obj.as_legacy_object(
719
compression_level=self.loose_compression_level)
721
raw_string = obj.as_legacy_object()
722
self.transport.put_bytes(path, raw_string)
724
def move_in_pack(self, f):
725
"""Move a specific file containing a pack into the pack directory.
727
:note: The file should be on the same file system as the
730
:param path: Path to the pack file.
733
p = PackData("", f, len(f.getvalue()))
734
entries = p.sorted_entries()
735
basename = "pack-%s" % iter_sha1(entry[0]
736
for entry in entries).decode('ascii')
737
p._filename = basename + ".pack"
739
self.pack_transport.put_file(basename + ".pack", f)
740
with self.pack_transport.open_write_stream(basename + ".idx") as idxfile:
741
write_pack_index_v2(idxfile, entries, p.get_stored_checksum())
742
idxfile = self.pack_transport.get(basename + ".idx")
743
idx = load_pack_index_file(basename + ".idx", idxfile)
744
final_pack = Pack.from_objects(p, idx)
745
final_pack._basename = basename
746
self._add_cached_pack(basename, final_pack)
749
def move_in_thin_pack(self, f):
750
"""Move a specific file containing a pack into the pack directory.
752
:note: The file should be on the same file system as the
755
:param path: Path to the pack file.
758
p = Pack('', resolve_ext_ref=self.get_raw)
759
p._data = PackData.from_file(f, len(f.getvalue()))
761
p._idx_load = lambda: MemoryPackIndex(
762
p.data.sorted_entries(), p.data.get_stored_checksum())
764
pack_sha = p.index.objects_sha1()
766
with self.pack_transport.open_write_stream(
767
"pack-%s.pack" % pack_sha.decode('ascii')) as datafile:
768
entries, data_sum = write_pack_objects(datafile, p.pack_tuples())
769
entries = sorted([(k, v[0], v[1]) for (k, v) in entries.items()])
770
with self.pack_transport.open_write_stream(
771
"pack-%s.idx" % pack_sha.decode('ascii')) as idxfile:
772
write_pack_index_v2(idxfile, entries, data_sum)
775
"""Add a new pack to this object store.
777
:return: Fileobject to write to and a commit function to
778
call when the pack is finished.
783
if len(f.getvalue()) > 0:
784
return self.move_in_pack(f)
790
return f, commit, abort
793
def init(cls, transport):
795
transport.mkdir('info')
799
transport.mkdir(PACKDIR)
802
return cls(transport)