bup repositories are in Git format. This library allows us to
interact with the Git data structures.
"""
-import os, sys, zlib, time, subprocess, struct, stat, re, tempfile, glob
+
+import errno, os, sys, zlib, time, subprocess, struct, stat, re, tempfile, glob
from collections import namedtuple
+from itertools import islice
+
+from bup import _helpers, hashsplit, path, midx, bloom, xstat
+from bup.helpers import (Sha1, add_error, chunkyreader, debug1, debug2,
+ fdatasync,
+ hostname, localtime, log, merge_iter,
+ mmap_read, mmap_readwrite,
+ progress, qprogress, unlink, username, userfullname,
+ utc_offset_str)
-from bup.helpers import *
-from bup import _helpers, path, midx, bloom, xstat
max_pack_size = 1000*1000*1000 # larger packs will slow down pruning
max_pack_objects = 200*1000 # cache memory usage is about 83 bytes per object
return parse_commit(commit_content)
-def repo(sub = ''):
+def _local_git_date_str(epoch_sec):
+ return '%d %s' % (epoch_sec, utc_offset_str(epoch_sec))
+
+
+def _git_date_str(epoch_sec, tz_offset_sec):
+ offs = tz_offset_sec // 60
+ return '%d %s%02d%02d' \
+ % (epoch_sec,
+ '+' if offs >= 0 else '-',
+ abs(offs) // 60,
+ abs(offs) % 60)
+
+
+def repo(sub = '', repo_dir=None):
"""Get the path to the git repository or one of its subdirectories."""
global repodir
- if not repodir:
+ repo_dir = repo_dir or repodir
+ if not repo_dir:
raise GitError('You should call check_repo_or_die()')
# If there's a .git subdirectory, then the actual repo is in there.
- gd = os.path.join(repodir, '.git')
+ gd = os.path.join(repo_dir, '.git')
if os.path.exists(gd):
repodir = gd
- return os.path.join(repodir, sub)
+ return os.path.join(repo_dir, sub)
def shorten_hash(s):
args = [path.exe(), 'midx', '--auto', '--dir', objdir]
try:
rv = subprocess.call(args, stdout=open('/dev/null', 'w'))
- except OSError, e:
+ except OSError as e:
# make sure 'args' gets printed to help with debugging
add_error('%r: exception: %s' % (args, e))
raise
args = [path.exe(), 'bloom', '--dir', objdir]
try:
rv = subprocess.call(args, stdout=open('/dev/null', 'w'))
- except OSError, e:
+ except OSError as e:
# make sure 'args' gets printed to help with debugging
add_error('%r: exception: %s' % (args, e))
raise
disambiguate normal files from segmented ones.
"""
if stat.S_ISREG(mode) and not stat.S_ISREG(gitmode):
+ assert(stat.S_ISDIR(gitmode))
return name + '.bup'
elif name.endswith('.bup') or name[:-1].endswith('.bup'):
return name + '.bupl'
(BUP_NORMAL, BUP_CHUNKED) = (0,1)
-def demangle_name(name):
+def demangle_name(name, mode):
"""Remove name mangling from a file name, if necessary.
The return value is a tuple (demangled_filename,mode), where mode is one of
return (name[:-5], BUP_NORMAL)
elif name.endswith('.bup'):
return (name[:-4], BUP_CHUNKED)
+ elif name.endswith('.bupm'):
+ return (name[:-5],
+ BUP_CHUNKED if stat.S_ISDIR(mode) else BUP_NORMAL)
else:
return (name, BUP_NORMAL)
def _encode_packobj(type, content, compression_level=1):
+ if compression_level not in (0, 1, 2, 3, 4, 5, 6, 7, 8, 9):
+ raise ValueError('invalid compression level %s' % compression_level)
szout = ''
sz = len(content)
szbits = (sz & 0x0f) | (_typemap[type]<<4)
break
szbits = sz & 0x7f
sz >>= 7
- if compression_level > 9:
- compression_level = 9
- elif compression_level < 0:
- compression_level = 0
z = zlib.compressobj(compression_level)
yield szout
yield z.compress(content)
if not d.get(full):
try:
ix = open_idx(full)
- except GitError, e:
+ except GitError as e:
add_error(e)
continue
d[full] = ix
def _make_objcache():
return PackIdxList(repo('objects/pack'))
+# bup-gc assumes that it can disable all PackWriter activities
+# (bloom/midx/cache) via the constructor and close() arguments.
+
class PackWriter:
"""Writes Git objects inside a pack file."""
- def __init__(self, objcache_maker=_make_objcache, compression_level=1):
+ def __init__(self, objcache_maker=_make_objcache, compression_level=1,
+ run_midx=True, on_pack_finish=None):
+ self.file = None
+ self.parentfd = None
self.count = 0
self.outbytes = 0
self.filename = None
- self.file = None
self.idx = None
self.objcache_maker = objcache_maker
self.objcache = None
self.compression_level = compression_level
+ self.run_midx=run_midx
+ self.on_pack_finish = on_pack_finish
def __del__(self):
self.close()
def _open(self):
if not self.file:
- (fd,name) = tempfile.mkstemp(suffix='.pack', dir=repo('objects'))
- self.file = os.fdopen(fd, 'w+b')
+ objdir = dir=repo('objects')
+ fd, name = tempfile.mkstemp(suffix='.pack', dir=objdir)
+ try:
+ self.file = os.fdopen(fd, 'w+b')
+ except:
+ os.close(fd)
+ raise
+ try:
+ self.parentfd = os.open(objdir, os.O_RDONLY)
+ except:
+ f = self.file
+ self.file = None
+ f.close()
+ raise
assert(name.endswith('.pack'))
self.filename = name[:-5]
self.file.write('PACK\0\0\0\2\0\0\0\0')
oneblob = ''.join(datalist)
try:
f.write(oneblob)
- except IOError, e:
+ except IOError as e:
raise GitError, e, sys.exc_info()[2]
nw = len(oneblob)
crc = zlib.crc32(oneblob) & 0xffffffff
def breakpoint(self):
"""Clear byte and object counts and return the last processed id."""
- id = self._end()
+ id = self._end(self.run_midx)
self.outbytes = self.count = 0
return id
self._require_objcache()
return self.objcache.exists(id, want_source=want_source)
+ def write(self, sha, type, content):
+ """Write an object to the pack file. Fails if sha exists()."""
+ self._write(sha, type, content)
+
def maybe_write(self, type, content):
"""Write an object to the pack file if not present and return its id."""
sha = calc_hash(type, content)
if not self.exists(sha):
- self._write(sha, type, content)
+ self.write(sha, type, content)
self._require_objcache()
self.objcache.add(sha)
return sha
content = tree_encode(shalist)
return self.maybe_write('tree', content)
- def _new_commit(self, tree, parent, author, adate, committer, cdate, msg):
+ def new_commit(self, tree, parent,
+ author, adate_sec, adate_tz,
+ committer, cdate_sec, cdate_tz,
+ msg):
+ """Create a commit object in the pack. The date_sec values must be
+ epoch-seconds, and if a tz is None, the local timezone is assumed."""
+ if adate_tz:
+ adate_str = _git_date_str(adate_sec, adate_tz)
+ else:
+ adate_str = _local_git_date_str(adate_sec)
+ if cdate_tz:
+ cdate_str = _git_date_str(cdate_sec, cdate_tz)
+ else:
+ cdate_str = _local_git_date_str(cdate_sec)
l = []
if tree: l.append('tree %s' % tree.encode('hex'))
if parent: l.append('parent %s' % parent.encode('hex'))
- if author: l.append('author %s %s' % (author, _git_date(adate)))
- if committer: l.append('committer %s %s' % (committer, _git_date(cdate)))
+ if author: l.append('author %s %s' % (author, adate_str))
+ if committer: l.append('committer %s %s' % (committer, cdate_str))
l.append('')
l.append(msg)
return self.maybe_write('commit', '\n'.join(l))
- def new_commit(self, parent, tree, date, msg):
- """Create a commit object in the pack."""
- userline = '%s <%s@%s>' % (userfullname(), username(), hostname())
- commit = self._new_commit(tree, parent,
- userline, date, userline, date,
- msg)
- return commit
-
def abort(self):
"""Remove the pack file from disk."""
f = self.file
if f:
- self.idx = None
+ pfd = self.parentfd
self.file = None
- f.close()
- os.unlink(self.filename + '.pack')
+ self.parentfd = None
+ self.idx = None
+ try:
+ try:
+ os.unlink(self.filename + '.pack')
+ finally:
+ f.close()
+ finally:
+ if pfd is not None:
+ os.close(pfd)
def _end(self, run_midx=True):
f = self.file
if not f: return None
self.file = None
- self.objcache = None
- idx = self.idx
- self.idx = None
+ try:
+ self.objcache = None
+ idx = self.idx
+ self.idx = None
- # update object count
- f.seek(8)
- cp = struct.pack('!i', self.count)
- assert(len(cp) == 4)
- f.write(cp)
-
- # calculate the pack sha1sum
- f.seek(0)
- sum = Sha1()
- for b in chunkyreader(f):
- sum.update(b)
- packbin = sum.digest()
- f.write(packbin)
- f.close()
+ # update object count
+ f.seek(8)
+ cp = struct.pack('!i', self.count)
+ assert(len(cp) == 4)
+ f.write(cp)
+
+ # calculate the pack sha1sum
+ f.seek(0)
+ sum = Sha1()
+ for b in chunkyreader(f):
+ sum.update(b)
+ packbin = sum.digest()
+ f.write(packbin)
+ fdatasync(f.fileno())
+ finally:
+ f.close()
obj_list_sha = self._write_pack_idx_v2(self.filename + '.idx', idx, packbin)
os.unlink(self.filename + '.map')
os.rename(self.filename + '.pack', nameprefix + '.pack')
os.rename(self.filename + '.idx', nameprefix + '.idx')
+ try:
+ os.fsync(self.parentfd)
+ finally:
+ os.close(self.parentfd)
if run_midx:
auto_midx(repo('objects/pack'))
+
+ if self.on_pack_finish:
+ self.on_pack_finish(nameprefix)
+
return nameprefix
def close(self, run_midx=True):
idx_f = open(filename, 'w+b')
try:
idx_f.truncate(index_len)
+ fdatasync(idx_f.fileno())
idx_map = mmap_readwrite(idx_f, close=False)
- count = _helpers.write_idx(filename, idx_map, idx, self.count)
- assert(count == self.count)
+ try:
+ count = _helpers.write_idx(filename, idx_map, idx, self.count)
+ assert(count == self.count)
+ idx_map.flush()
+ finally:
+ idx_map.close()
finally:
- if idx_map: idx_map.close()
idx_f.close()
idx_f = open(filename, 'a+b')
for b in chunkyreader(idx_f):
idx_sum.update(b)
idx_f.write(idx_sum.digest())
+ fdatasync(idx_f.fileno())
return namebase
finally:
idx_f.close()
-def _git_date(date):
- return '%d %s' % (date, time.strftime('%z', time.localtime(date)))
+def _gitenv(repo_dir = None):
+ if not repo_dir:
+ repo_dir = repo()
+ def env():
+ os.environ['GIT_DIR'] = os.path.abspath(repo_dir)
+ return env
-def _gitenv():
- os.environ['GIT_DIR'] = os.path.abspath(repo())
+def list_refs(refname=None, repo_dir=None,
+ limit_to_heads=False, limit_to_tags=False):
+ """Yield (refname, hash) tuples for all repository refs unless a ref
+ name is specified. Given a ref name, only include tuples for that
+ particular ref. The limits restrict the result items to
+ refs/heads or refs/tags. If both limits are specified, items from
+ both sources will be included.
-
-def list_refs(refname = None):
- """Generate a list of tuples in the form (refname,hash).
- If a ref name is specified, list only this particular ref.
"""
- argv = ['git', 'show-ref', '--']
+ argv = ['git', 'show-ref']
+ if limit_to_heads:
+ argv.append('--heads')
+ if limit_to_tags:
+ argv.append('--tags')
+ argv.append('--')
if refname:
argv += [refname]
- p = subprocess.Popen(argv, preexec_fn = _gitenv, stdout = subprocess.PIPE)
+ p = subprocess.Popen(argv,
+ preexec_fn = _gitenv(repo_dir),
+ stdout = subprocess.PIPE)
out = p.stdout.read().strip()
rv = p.wait() # not fatal
if rv:
yield (name, sha.decode('hex'))
-def read_ref(refname):
+def read_ref(refname, repo_dir = None):
"""Get the commit id of the most recent commit made on a given ref."""
- l = list(list_refs(refname))
+ refs = list_refs(refname, repo_dir=repo_dir, limit_to_heads=True)
+ l = tuple(islice(refs, 2))
if l:
assert(len(l) == 1)
return l[0][1]
return None
-def rev_list(ref, count=None):
+def rev_list(ref, count=None, repo_dir=None):
"""Generate a list of reachable commits in reverse chronological order.
This generator walks through commits, from child to parent, that are
if count:
opts += ['-n', str(atoi(count))]
argv = ['git', 'rev-list', '--pretty=format:%at'] + opts + [ref, '--']
- p = subprocess.Popen(argv, preexec_fn = _gitenv, stdout = subprocess.PIPE)
+ p = subprocess.Popen(argv,
+ preexec_fn = _gitenv(repo_dir),
+ stdout = subprocess.PIPE)
commit = None
for row in p.stdout:
s = row.strip()
raise GitError, 'git rev-list returned error %d' % rv
-def get_commit_dates(refs):
+def get_commit_dates(refs, repo_dir=None):
"""Get the dates for the specified commit refs. For now, every unique
string in refs must resolve to a different commit or this
function will fail."""
result = []
for ref in refs:
- commit = get_commit_items(ref, cp())
+ commit = get_commit_items(ref, cp(repo_dir))
result.append(commit.author_sec)
return result
-def rev_parse(committish):
+def rev_parse(committish, repo_dir=None):
"""Resolve the full hash for 'committish', if it exists.
Should be roughly equivalent to 'git rev-parse'.
Returns the hex value of the hash if it is found, None if 'committish' does
not correspond to anything.
"""
- head = read_ref(committish)
+ head = read_ref(committish, repo_dir=repo_dir)
if head:
debug2("resolved from ref: commit = %s\n" % head.encode('hex'))
return head
- pL = PackIdxList(repo('objects/pack'))
+ pL = PackIdxList(repo('objects/pack', repo_dir=repo_dir))
if len(committish) == 40:
try:
return None
-def update_ref(refname, newval, oldval):
- """Change the commit pointed to by a branch."""
+def update_ref(refname, newval, oldval, repo_dir=None):
+ """Update a repository reference."""
if not oldval:
oldval = ''
- assert(refname.startswith('refs/heads/'))
+ assert(refname.startswith('refs/heads/') \
+ or refname.startswith('refs/tags/'))
p = subprocess.Popen(['git', 'update-ref', refname,
newval.encode('hex'), oldval.encode('hex')],
- preexec_fn = _gitenv)
+ preexec_fn = _gitenv(repo_dir))
+ _git_wait('git update-ref', p)
+
+
+def delete_ref(refname, oldvalue=None):
+ """Delete a repository reference (see git update-ref(1))."""
+ assert(refname.startswith('refs/'))
+ oldvalue = [] if not oldvalue else [oldvalue]
+ p = subprocess.Popen(['git', 'update-ref', '-d', refname] + oldvalue,
+ preexec_fn = _gitenv())
_git_wait('git update-ref', p)
if os.path.exists(d) and not os.path.isdir(os.path.join(d, '.')):
raise GitError('"%s" exists but is not a directory\n' % d)
p = subprocess.Popen(['git', '--bare', 'init'], stdout=sys.stderr,
- preexec_fn = _gitenv)
+ preexec_fn = _gitenv())
_git_wait('git init', p)
# Force the index version configuration in order to ensure bup works
# regardless of the version of the installed Git binary.
p = subprocess.Popen(['git', 'config', 'pack.indexVersion', '2'],
- stdout=sys.stderr, preexec_fn = _gitenv)
+ stdout=sys.stderr, preexec_fn = _gitenv())
_git_wait('git config', p)
# Enable the reflog
p = subprocess.Popen(['git', 'config', 'core.logAllRefUpdates', 'true'],
- stdout=sys.stderr, preexec_fn = _gitenv)
+ stdout=sys.stderr, preexec_fn = _gitenv())
_git_wait('git config', p)
guess_repo(path)
try:
os.stat(repo('objects/pack/.'))
- except OSError, e:
+ except OSError as e:
if e.errno == errno.ENOENT:
log('error: %r is not a bup repository; run "bup init"\n'
% repo())
def _git_capture(argv):
- p = subprocess.Popen(argv, stdout=subprocess.PIPE, preexec_fn = _gitenv)
+ p = subprocess.Popen(argv, stdout=subprocess.PIPE, preexec_fn = _gitenv())
r = p.stdout.read()
_git_wait(repr(argv), p)
return r
def next(self):
try:
return self.it.next()
- except StopIteration, e:
+ except StopIteration as e:
self.done = True
raise
except:
self.abort()
+class MissingObject(KeyError):
+ def __init__(self, id):
+ self.id = id
+ KeyError.__init__(self, 'object %r is missing' % id.encode('hex'))
+
+
_ver_warned = 0
class CatPipe:
"""Link to 'git cat-file' that is used to retrieve blob data."""
- def __init__(self):
+ def __init__(self, repo_dir = None):
global _ver_warned
+ self.repo_dir = repo_dir
wanted = ('1','5','6')
if ver() < wanted:
if not _ver_warned:
stdout=subprocess.PIPE,
close_fds = True,
bufsize = 4096,
- preexec_fn = _gitenv)
+ preexec_fn = _gitenv(self.repo_dir))
def _fast_get(self, id):
if not self.p or self.p.poll() != None:
hdr = self.p.stdout.readline()
if hdr.endswith(' missing\n'):
self.inprogress = None
- raise KeyError('blob %r is missing' % id)
+ raise MissingObject(id.decode('hex'))
spl = hdr.split(' ')
if len(spl) != 3 or len(spl[0]) != 40:
raise GitError('expected blob, got %r' % spl)
readline_result = self.p.stdout.readline()
assert(readline_result == '\n')
self.inprogress = None
- except Exception, e:
+ except Exception as e:
it.abort()
raise
p = subprocess.Popen(['git', 'cat-file', type, id],
stdout=subprocess.PIPE,
- preexec_fn = _gitenv)
+ preexec_fn = _gitenv(self.repo_dir))
for blob in chunkyreader(p.stdout):
yield blob
_git_wait('git cat-file', p)
log('booger!\n')
-_cp = (None, None)
+_cp = {}
-def cp():
- """Create a CatPipe object or reuse an already existing one."""
+def cp(repo_dir=None):
+ """Create a CatPipe object or reuse the already existing one."""
global _cp
- cp_dir, cp = _cp
- cur_dir = os.path.realpath(repo())
- if cur_dir != cp_dir:
- cp = CatPipe()
- _cp = (cur_dir, cp)
+ if not repo_dir:
+ repo_dir = repo()
+ repo_dir = os.path.abspath(repo_dir)
+ cp = _cp.get(repo_dir)
+ if not cp:
+ cp = CatPipe(repo_dir)
+ _cp[repo_dir] = cp
return cp
-def tags():
+def tags(repo_dir = None):
"""Return a dictionary of all tags in the form {hash: [tag_names, ...]}."""
tags = {}
- for (n,c) in list_refs():
- if n.startswith('refs/tags/'):
- name = n[10:]
- if not c in tags:
- tags[c] = []
+ for n, c in list_refs(repo_dir = repo_dir, limit_to_tags=True):
+ assert(n.startswith('refs/tags/'))
+ name = n[10:]
+ if not c in tags:
+ tags[c] = []
+ tags[c].append(name) # more than one tag can point at 'c'
+ return tags
- tags[c].append(name) # more than one tag can point at 'c'
- return tags
+WalkItem = namedtuple('WalkItem', ['id', 'type', 'mode',
+ 'path', 'chunk_path', 'data'])
+# The path is the mangled path, and if an item represents a fragment
+# of a chunked file, the chunk_path will be the chunked subtree path
+# for the chunk, i.e. ['', '2d3115e', ...]. The top-level path for a
+# chunked file will have a chunk_path of ['']. So some chunk subtree
+# of the file '/foo/bar/baz' might look like this:
+#
+# item.path = ['foo', 'bar', 'baz.bup']
+# item.chunk_path = ['', '2d3115e', '016b097']
+# item.type = 'tree'
+# ...
+
+
+def walk_object(cat_pipe, id,
+ stop_at=None,
+ include_data=None):
+ """Yield everything reachable from id via cat_pipe as a WalkItem,
+ stopping whenever stop_at(id) returns true. Throw MissingObject
+ if a hash encountered is missing from the repository.
+
+ """
+ # Maintain the pending stack on the heap to avoid stack overflow
+ pending = [(id, [], [], None)]
+ while len(pending):
+ id, parent_path, chunk_path, mode = pending.pop()
+ if stop_at and stop_at(id):
+ continue
+
+ item_it = cat_pipe.get(id) # FIXME: use include_data
+ type = item_it.next()
+ if type not in ('blob', 'commit', 'tree'):
+ raise Exception('unexpected repository object type %r' % type)
+
+ # FIXME: set the mode based on the type when the mode is None
+ if type == 'blob' and not include_data:
+ # Dump data until we can ask cat_pipe not to fetch it
+ for ignored in item_it:
+ pass
+ data = None
+ else:
+ data = ''.join(item_it)
+
+ yield WalkItem(id=id, type=type,
+ chunk_path=chunk_path, path=parent_path,
+ mode=mode,
+ data=(data if include_data else None))
+
+ if type == 'commit':
+ commit_items = parse_commit(data)
+ for pid in commit_items.parents:
+ pending.append((pid, parent_path, chunk_path, mode))
+ pending.append((commit_items.tree, parent_path, chunk_path,
+ hashsplit.GIT_MODE_TREE))
+ elif type == 'tree':
+ for mode, name, ent_id in tree_decode(data):
+ demangled, bup_type = demangle_name(name, mode)
+ if chunk_path:
+ sub_path = parent_path
+ sub_chunk_path = chunk_path + [name]
+ else:
+ sub_path = parent_path + [name]
+ if bup_type == BUP_CHUNKED:
+ sub_chunk_path = ['']
+ else:
+ sub_chunk_path = chunk_path
+ pending.append((ent_id.encode('hex'), sub_path, sub_chunk_path,
+ mode))
projects / bup.git / blobdiff