def test_multiple_suggestions():
with no_lingering_errors():
with test_tempdir(b'bup-tclient-') as tmpdir:
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
lw = git.PackWriter()
lw.new_blob(s2)
lw.close()
WVPASSEQ(len(glob.glob(git.repo(b'objects/pack' + IDX_PAT))), 2)
c = client.Client(bupdir, create=True)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 0)
rw = c.new_packwriter()
s1sha = rw.new_blob(s1)
WVPASS(rw.exists(s1sha))
s2sha = rw.new_blob(s2)
# This is a little hacky, but ensures that we test the
# code under test
while (len(glob.glob(c.cachedir + IDX_PAT)) < 2
and not c.conn.has_input()):
pass
rw.new_blob(s2)
WVPASS(rw.objcache.exists(s1sha))
WVPASS(rw.objcache.exists(s2sha))
rw.new_blob(s3)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 2)
rw.close()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 3)
def __init__(self, address):
self.address = address
self.client = client.Client(address)
self.new_packwriter = self.client.new_packwriter
self.update_ref = self.client.update_ref
self.rev_list = self.client.rev_list
self._id = _repo_id(self.address)
def test_midx_refreshing(tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
rw.new_blob(s1)
p1base = rw.breakpoint()
p1name = os.path.join(c.cachedir, p1base)
s1sha = rw.new_blob(s1) # should not be written; it's already in p1
s2sha = rw.new_blob(s2)
p2base = rw.close()
p2name = os.path.join(c.cachedir, p2base)
del rw
pi = git.PackIdxList(bupdir + b'/objects/pack')
assert len(pi.packs) == 2
pi.refresh()
assert len(pi.packs) == 2
assert sorted([os.path.basename(i.name)
for i in pi.packs]) == sorted([p1base, p2base])
p1 = git.open_idx(p1name)
assert p1.exists(s1sha)
p2 = git.open_idx(p2name)
assert not p2.exists(s1sha)
assert p2.exists(s2sha)
subprocess.call([path.exe(), b'midx', b'-f'])
pi.refresh()
assert len(pi.packs) == 1
pi.refresh(skip_midx=True)
assert len(pi.packs) == 2
pi.refresh(skip_midx=False)
assert len(pi.packs) == 1
def test_multiple_suggestions():
initial_failures = wvfailure_count()
tmpdir = tempfile.mkdtemp(dir=bup_tmp, prefix='bup-tclient-')
os.environ['BUP_MAIN_EXE'] = '../../../bup'
os.environ['BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
lw = git.PackWriter()
lw.new_blob(s2)
lw.close()
WVPASSEQ(len(glob.glob(git.repo('objects/pack' + IDX_PAT))), 2)
c = client.Client(bupdir, create=True)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 0)
rw = c.new_packwriter()
s1sha = rw.new_blob(s1)
WVPASS(rw.exists(s1sha))
s2sha = rw.new_blob(s2)
# This is a little hacky, but ensures that we test the code under test
while (len(glob.glob(c.cachedir + IDX_PAT)) < 2
and not c.conn.has_input()):
pass
rw.new_blob(s2)
WVPASS(rw.objcache.exists(s1sha))
WVPASS(rw.objcache.exists(s2sha))
rw.new_blob(s3)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 2)
rw.close()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 3)
if wvfailure_count() == initial_failures:
subprocess.call(['rm', '-rf', tmpdir])
def test_multiple_suggestions(tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
lw = git.PackWriter()
lw.new_blob(s2)
lw.close()
assert len(glob.glob(git.repo(b'objects/pack' + IDX_PAT,
repo_dir=bupdir))) == 2
c = client.Client(bupdir, create=True)
assert len(glob.glob(c.cachedir + IDX_PAT)) == 0
rw = c.new_packwriter()
s1sha = rw.new_blob(s1)
assert rw.exists(s1sha)
s2sha = rw.new_blob(s2)
# This is a little hacky, but ensures that we test the
# code under test. First, flush to ensure that we've
# actually sent all the command ('receive-objects-v2')
# and their data to the server. This may be needed if
# the output buffer size is bigger than the data (both
# command and objects) we're writing. To see the need
# for this, change the object sizes at the beginning
# of this file to be very small (e.g. 10 instead of 10k)
c.conn.outp.flush()
# Then, check if we've already received the idx files.
# This may happen if we're preempted just after writing
# the data, then the server runs and suggests, and only
# then we continue in PackWriter_Remote::_raw_write()
# and check the has_input(), in that case we'll receive
# the idx still in the rw.new_blob() calls above.
#
# In most cases though, that doesn't happen, and we'll
# get past the has_input() check before the server has
# a chance to respond - it has to actually hash the new
# object here, so it takes some time. So also break out
# of the loop if the server has sent something on the
# connection.
#
# Finally, abort this after a little while (about one
# second) just in case something's actually broken.
n = 0
while (len(glob.glob(c.cachedir + IDX_PAT)) < 2 and not c.conn.has_input()
and n < 10):
time.sleep(0.1)
n += 1
assert len(glob.glob(c.cachedir + IDX_PAT)) == 2 or c.conn.has_input()
rw.new_blob(s2)
assert rw.objcache.exists(s1sha)
assert rw.objcache.exists(s2sha)
rw.new_blob(s3)
assert len(glob.glob(c.cachedir + IDX_PAT)) == 2
rw.close()
assert len(glob.glob(c.cachedir + IDX_PAT)) == 3
def test_server_split_with_indexes(tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
with git.PackWriter() as lw:
lw.new_blob(s1)
with client.Client(bupdir, create=True) as c, \
c.new_packwriter() as rw:
rw.new_blob(s2)
rw.breakpoint()
rw.new_blob(s1)
def test_server_split_with_indexes(tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
lw.new_blob(s1)
lw.close()
rw.new_blob(s2)
rw.breakpoint()
rw.new_blob(s1)
rw.close()
def test_server_split_with_indexes():
os.environ['BUP_MAIN_EXE'] = '../../../bup'
os.environ['BUP_DIR'] = bupdir = 'buptest_tclient.tmp'
subprocess.call(['rm', '-rf', bupdir])
git.init_repo(bupdir)
lw = git.PackWriter()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
lw.new_blob(s1)
lw.close()
rw.new_blob(s2)
rw.breakpoint()
rw.new_blob(s1)
def test_dumb_client_server(tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
open(git.repo(b'bup-dumb-server'), 'w').close()
with git.PackWriter() as lw:
lw.new_blob(s1)
with client.Client(bupdir, create=True) as c, \
c.new_packwriter() as rw:
assert len(glob.glob(c.cachedir + IDX_PAT)) == 1
rw.new_blob(s1)
assert len(glob.glob(c.cachedir + IDX_PAT)) == 1
rw.new_blob(s2)
assert len(glob.glob(c.cachedir + IDX_PAT)) == 2
def test_server_split_with_indexes():
with no_lingering_errors():
with test_tempdir(b'bup-tclient-') as tmpdir:
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
lw.new_blob(s1)
lw.close()
rw.new_blob(s2)
rw.breakpoint()
rw.new_blob(s1)
rw.close()
def main(argv):
o = options.Options(optspec)
opt, flags, extra = o.parse_bytes(argv[1:])
if extra:
o.fatal("no arguments expected")
try:
git.init_repo() # local repo
except git.GitError as e:
log("bup: error: could not init repository: %s" % e)
sys.exit(1)
if opt.remote:
git.check_repo_or_die()
cli = client.Client(argv_bytes(opt.remote), create=True)
cli.close()
def test_server_split_with_indexes():
initial_failures = wvfailure_count()
tmpdir = tempfile.mkdtemp(dir=bup_tmp, prefix='bup-tclient-')
os.environ['BUP_MAIN_EXE'] = '../../../bup'
os.environ['BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
lw = git.PackWriter()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
lw.new_blob(s1)
lw.close()
rw.new_blob(s2)
rw.breakpoint()
rw.new_blob(s1)
if wvfailure_count() == initial_failures:
subprocess.call(['rm', '-rf', tmpdir])
def test_dumb_client_server():
with no_lingering_errors():
with test_tempdir(b'bup-tclient-') as tmpdir:
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
open(git.repo(b'bup-dumb-server'), 'w').close()
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 1)
rw.new_blob(s1)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 1)
rw.new_blob(s2)
rw.close()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 2)
def test_dumb_client_server():
os.environ['BUP_MAIN_EXE'] = '../../../bup'
os.environ['BUP_DIR'] = bupdir = 'buptest_tclient.tmp'
subprocess.call(['rm', '-rf', bupdir])
git.init_repo(bupdir)
open(git.repo('bup-dumb-server'), 'w').close()
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
WVPASSEQ(len(glob.glob(c.cachedir+IDX_PAT)), 1)
rw.new_blob(s1)
WVPASSEQ(len(glob.glob(c.cachedir+IDX_PAT)), 1)
rw.new_blob(s2)
rw.close()
WVPASSEQ(len(glob.glob(c.cachedir+IDX_PAT)), 2)
def __init__(self, address, create=False, compression_level=None,
max_pack_size=None, max_pack_objects=None):
# if client.Client() raises an exception, have a client
# anyway to avoid follow-up exceptions from __del__
self.client = None
self.client = client.Client(address)
self.config = self.client.config
# init the superclass only afterwards so it can access self.config()
super(RemoteRepo, self).__init__(address,
compression_level=compression_level,
max_pack_size=max_pack_size,
max_pack_objects=max_pack_objects)
self.rev_list = self.client.rev_list
self.list_indexes = self.client.list_indexes
self.read_ref = self.client.read_ref
self.send_index = self.client.send_index
self.join = self.client.join
self.refs = self.client.refs
self.resolve = self.client.resolve
self._packwriter = None
def test_dumb_client_server():
initial_failures = wvfailure_count()
tmpdir = tempfile.mkdtemp(dir=bup_tmp, prefix='bup-tclient-')
os.environ['BUP_MAIN_EXE'] = '../../../bup'
os.environ['BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
open(git.repo('bup-dumb-server'), 'w').close()
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 1)
rw.new_blob(s1)
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 1)
rw.new_blob(s2)
rw.close()
WVPASSEQ(len(glob.glob(c.cachedir + IDX_PAT)), 2)
if wvfailure_count() == initial_failures:
subprocess.call(['rm', '-rf', tmpdir])
def test_dumb_client_server(dumb_mode, tmpdir):
environ[b'BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
if dumb_mode == 'file':
open(git.repo(b'bup-dumb-server', repo_dir=bupdir), 'w').close()
elif dumb_mode == 'config':
git.git_config_write(b'bup.dumb-server', b'true', repo_dir=bupdir)
else:
assert False
lw = git.PackWriter()
lw.new_blob(s1)
lw.close()
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
assert len(glob.glob(c.cachedir + IDX_PAT)) == 1
rw.new_blob(s1)
assert len(glob.glob(c.cachedir + IDX_PAT)) == 1
rw.new_blob(s2)
rw.close()
assert len(glob.glob(c.cachedir + IDX_PAT)) == 2
def test_midx_refreshing():
initial_failures = wvfailure_count()
tmpdir = tempfile.mkdtemp(dir=bup_tmp, prefix='bup-tclient-')
os.environ['BUP_MAIN_EXE'] = bupmain = '../../../bup'
os.environ['BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
rw.new_blob(s1)
p1base = rw.breakpoint()
p1name = os.path.join(c.cachedir, p1base)
s1sha = rw.new_blob(s1) # should not be written; it's already in p1
s2sha = rw.new_blob(s2)
p2base = rw.close()
p2name = os.path.join(c.cachedir, p2base)
del rw
pi = git.PackIdxList(bupdir + '/objects/pack')
WVPASSEQ(len(pi.packs), 2)
pi.refresh()
WVPASSEQ(len(pi.packs), 2)
WVPASSEQ(sorted([os.path.basename(i.name) for i in pi.packs]),
sorted([p1base, p2base]))
p1 = git.open_idx(p1name)
WVPASS(p1.exists(s1sha))
p2 = git.open_idx(p2name)
WVFAIL(p2.exists(s1sha))
WVPASS(p2.exists(s2sha))
subprocess.call([bupmain, 'midx', '-f'])
pi.refresh()
WVPASSEQ(len(pi.packs), 1)
pi.refresh(skip_midx=True)
WVPASSEQ(len(pi.packs), 2)
pi.refresh(skip_midx=False)
WVPASSEQ(len(pi.packs), 1)
if wvfailure_count() == initial_failures:
subprocess.call(['rm', '-rf', tmpdir])
def test_midx_refreshing():
with no_lingering_errors():
with test_tempdir('bup-tclient-') as tmpdir:
os.environ['BUP_MAIN_EXE'] = bupmain = '../../../bup'
os.environ['BUP_DIR'] = bupdir = tmpdir
git.init_repo(bupdir)
c = client.Client(bupdir, create=True)
rw = c.new_packwriter()
rw.new_blob(s1)
p1base = rw.breakpoint()
p1name = os.path.join(c.cachedir, p1base)
s1sha = rw.new_blob(
s1) # should not be written; it's already in p1
s2sha = rw.new_blob(s2)
p2base = rw.close()
p2name = os.path.join(c.cachedir, p2base)
del rw
pi = git.PackIdxList(bupdir + '/objects/pack')
WVPASSEQ(len(pi.packs), 2)
pi.refresh()
WVPASSEQ(len(pi.packs), 2)
WVPASSEQ(sorted([os.path.basename(i.name) for i in pi.packs]),
sorted([p1base, p2base]))
p1 = git.open_idx(p1name)
WVPASS(p1.exists(s1sha))
p2 = git.open_idx(p2name)
WVFAIL(p2.exists(s1sha))
WVPASS(p2.exists(s2sha))
subprocess.call([bupmain, 'midx', '-f'])
pi.refresh()
WVPASSEQ(len(pi.packs), 1)
pi.refresh(skip_midx=True)
WVPASSEQ(len(pi.packs), 2)
pi.refresh(skip_midx=False)
WVPASSEQ(len(pi.packs), 1)
def main(argv):
opt = opts_from_cmdline(argv)
if opt.verbose >= 2:
git.verbose = opt.verbose - 1
if opt.fanout:
hashsplit.fanout = opt.fanout
if opt.blobs:
hashsplit.fanout = 0
if opt.bwlimit:
client.bwlimit = opt.bwlimit
start_time = time.time()
sys.stdout.flush()
out = byte_stream(sys.stdout)
stdin = byte_stream(sys.stdin)
if opt.git_ids:
# the input is actually a series of git object ids that we should retrieve
# and split.
#
# This is a bit messy, but basically it converts from a series of
# CatPipe.get() iterators into a series of file-type objects.
# It would be less ugly if either CatPipe.get() returned a file-like object
# (not very efficient), or split_to_shalist() expected an iterator instead
# of a file.
cp = git.CatPipe()
class IterToFile:
def __init__(self, it):
self.it = iter(it)
def read(self, size):
v = next(self.it, None)
return v or b''
def read_ids():
while 1:
line = stdin.readline()
if not line:
break
if line:
line = line.strip()
try:
it = cp.get(line.strip())
next(it, None) # skip the file info
except KeyError as e:
add_error('error: %s' % e)
continue
yield IterToFile(it)
files = read_ids()
else:
# the input either comes from a series of files or from stdin.
if opt.sources:
files = (open(argv_bytes(fn), 'rb') for fn in opt.sources)
else:
files = [stdin]
writing = not (opt.noop or opt.copy)
remote_dest = opt.remote or opt.is_reverse
if writing:
git.check_repo_or_die()
if remote_dest and writing:
cli = repo = client.Client(opt.remote)
else:
cli = nullcontext()
repo = git
# cli creation must be last nontrivial command in each if clause above
with cli:
if opt.name and writing:
refname = opt.name and b'refs/heads/%s' % opt.name
oldref = repo.read_ref(refname)
else:
refname = oldref = None
if not writing:
pack_writer = NoOpPackWriter()
elif not remote_dest:
pack_writer = git.PackWriter(compression_level=opt.compress,
max_pack_size=opt.max_pack_size,
max_pack_objects=opt.max_pack_objects)
else:
pack_writer = cli.new_packwriter(
compression_level=opt.compress,
max_pack_size=opt.max_pack_size,
max_pack_objects=opt.max_pack_objects)
# packwriter creation must be last command in each if clause above
with pack_writer:
commit = split(opt, files, oldref, out, pack_writer)
# pack_writer must be closed before we can update the ref
if refname:
repo.update_ref(refname, commit, oldref)
secs = time.time() - start_time
size = hashsplit.total_split
if opt.bench:
log('bup: %.2f kbytes in %.2f secs = %.2f kbytes/sec\n' %
(size / 1024, secs, size / 1024 / secs))
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' %
len(saved_errors))
sys.exit(1)
if not (old_path and new_path):
o.fatal("a graft point cannot be empty")
graft_points.append((resolve_parent(old_path),
resolve_parent(new_path)))
is_reverse = environ.get(b'BUP_SERVER_REVERSE')
if is_reverse and opt.remote:
o.fatal("don't use -r in reverse mode; it's automatic")
name = opt.name
if name and not valid_save_name(name):
o.fatal("'%s' is not a valid branch name" % path_msg(name))
refname = name and b'refs/heads/%s' % name or None
if opt.remote or is_reverse:
try:
cli = client.Client(opt.remote)
except client.ClientError as e:
log('error: %s' % e)
sys.exit(1)
oldref = refname and cli.read_ref(refname) or None
w = cli.new_packwriter(compression_level=opt.compress)
else:
cli = None
oldref = refname and git.read_ref(refname) or None
w = git.PackWriter(compression_level=opt.compress)
handle_ctrl_c()
# Metadata is stored in a file named .bupm in each directory. The
# first metadata entry will be the metadata for the current directory.
#!/usr/bin/env python
import sys
from bup import git, options, client
from bup.helpers import *
optspec = """
[BUP_DIR=...] bup init [-r host:path]
--
r,remote= remote repository path
"""
o = options.Options(optspec)
(opt, flags, extra) = o.parse(sys.argv[1:])
if extra:
o.fatal("no arguments expected")
try:
git.init_repo() # local repo
except git.GitError, e:
log("bup: error: could not init repository: %s" % e)
sys.exit(1)
if opt.remote:
git.check_repo_or_die()
cli = client.Client(opt.remote, create=True)
cli.close()
def main(argv):
handle_ctrl_c()
opt = opts_from_cmdline(argv)
client.bwlimit = opt.bwlimit
git.check_repo_or_die()
remote_dest = opt.remote or opt.is_reverse
if not remote_dest:
repo = git
cli = nullcontext()
else:
try:
cli = repo = client.Client(opt.remote)
except client.ClientError as e:
log('error: %s' % e)
sys.exit(1)
# cli creation must be last nontrivial command in each if clause above
with cli:
if not remote_dest:
w = git.PackWriter(compression_level=opt.compress)
else:
w = cli.new_packwriter(compression_level=opt.compress)
with w:
sys.stdout.flush()
out = byte_stream(sys.stdout)
if opt.name:
refname = b'refs/heads/%s' % opt.name
parent = repo.read_ref(refname)
else:
refname = parent = None
indexfile = opt.indexfile or git.repo(b'bupindex')
try:
msr = index.MetaStoreReader(indexfile + b'.meta')
except IOError as ex:
if ex.errno != ENOENT:
raise
log('error: cannot access %r; have you run bup index?'
% path_msg(indexfile))
sys.exit(1)
with msr, \
hlinkdb.HLinkDB(indexfile + b'.hlink') as hlink_db, \
index.Reader(indexfile) as reader:
tree = save_tree(opt, reader, hlink_db, msr, w)
if opt.tree:
out.write(hexlify(tree))
out.write(b'\n')
if opt.commit or opt.name:
commit = commit_tree(tree, parent, opt.date, argv, w)
if opt.commit:
out.write(hexlify(commit))
out.write(b'\n')
# packwriter must be closed before we can update the ref
if opt.name:
repo.update_ref(refname, commit, parent)
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' % len(saved_errors))
sys.exit(1)
def main(argv):
o = options.Options(optspec)
opt, flags, extra = o.parse_bytes(argv[1:])
if opt.name: opt.name = argv_bytes(opt.name)
if opt.remote: opt.remote = argv_bytes(opt.remote)
if opt.verbose is None: opt.verbose = 0
if not (opt.blobs or opt.tree or opt.commit or opt.name or opt.noop
or opt.copy):
o.fatal("use one or more of -b, -t, -c, -n, --noop, --copy")
if opt.copy and (opt.blobs or opt.tree):
o.fatal('--copy is incompatible with -b, -t')
if (opt.noop or opt.copy) and (opt.commit or opt.name):
o.fatal('--noop and --copy are incompatible with -c, -n')
if opt.blobs and (opt.tree or opt.commit or opt.name):
o.fatal('-b is incompatible with -t, -c, -n')
if extra and opt.git_ids:
o.fatal("don't provide filenames when using --git-ids")
if opt.verbose >= 2:
git.verbose = opt.verbose - 1
opt.bench = 1
max_pack_size = None
if opt.max_pack_size:
max_pack_size = parse_num(opt.max_pack_size)
max_pack_objects = None
if opt.max_pack_objects:
max_pack_objects = parse_num(opt.max_pack_objects)
if opt.fanout:
hashsplit.fanout = parse_num(opt.fanout)
if opt.blobs:
hashsplit.fanout = 0
if opt.bwlimit:
client.bwlimit = parse_num(opt.bwlimit)
if opt.date:
date = parse_date_or_fatal(opt.date, o.fatal)
else:
date = time.time()
# Hack around lack of nonlocal vars in python 2
total_bytes = [0]
def prog(filenum, nbytes):
total_bytes[0] += nbytes
if filenum > 0:
qprogress('Splitting: file #%d, %d kbytes\r' %
(filenum + 1, total_bytes[0] // 1024))
else:
qprogress('Splitting: %d kbytes\r' % (total_bytes[0] // 1024))
is_reverse = environ.get(b'BUP_SERVER_REVERSE')
if is_reverse and opt.remote:
o.fatal("don't use -r in reverse mode; it's automatic")
start_time = time.time()
if opt.name and not valid_save_name(opt.name):
o.fatal("'%r' is not a valid branch name." % opt.name)
refname = opt.name and b'refs/heads/%s' % opt.name or None
if opt.noop or opt.copy:
cli = pack_writer = oldref = None
elif opt.remote or is_reverse:
git.check_repo_or_die()
cli = client.Client(opt.remote)
oldref = refname and cli.read_ref(refname) or None
pack_writer = cli.new_packwriter(compression_level=opt.compress,
max_pack_size=max_pack_size,
max_pack_objects=max_pack_objects)
else:
git.check_repo_or_die()
cli = None
oldref = refname and git.read_ref(refname) or None
pack_writer = git.PackWriter(compression_level=opt.compress,
max_pack_size=max_pack_size,
max_pack_objects=max_pack_objects)
input = byte_stream(sys.stdin)
if opt.git_ids:
# the input is actually a series of git object ids that we should retrieve
# and split.
#
# This is a bit messy, but basically it converts from a series of
# CatPipe.get() iterators into a series of file-type objects.
# It would be less ugly if either CatPipe.get() returned a file-like object
# (not very efficient), or split_to_shalist() expected an iterator instead
# of a file.
cp = git.CatPipe()
class IterToFile:
def __init__(self, it):
self.it = iter(it)
def read(self, size):
v = next(self.it, None)
return v or b''
def read_ids():
while 1:
line = input.readline()
if not line:
break
if line:
line = line.strip()
try:
it = cp.get(line.strip())
next(it, None) # skip the file info
except KeyError as e:
add_error('error: %s' % e)
continue
yield IterToFile(it)
files = read_ids()
else:
# the input either comes from a series of files or from stdin.
files = extra and (open(argv_bytes(fn), 'rb')
for fn in extra) or [input]
if pack_writer:
new_blob = pack_writer.new_blob
new_tree = pack_writer.new_tree
elif opt.blobs or opt.tree:
# --noop mode
new_blob = lambda content: git.calc_hash(b'blob', content)
new_tree = lambda shalist: git.calc_hash(b'tree',
git.tree_encode(shalist))
sys.stdout.flush()
out = byte_stream(sys.stdout)
if opt.blobs:
shalist = hashsplit.split_to_blobs(new_blob,
files,
keep_boundaries=opt.keep_boundaries,
progress=prog)
for (sha, size, level) in shalist:
out.write(hexlify(sha) + b'\n')
reprogress()
elif opt.tree or opt.commit or opt.name:
if opt.name: # insert dummy_name which may be used as a restore target
mode, sha = \
hashsplit.split_to_blob_or_tree(new_blob, new_tree, files,
keep_boundaries=opt.keep_boundaries,
progress=prog)
splitfile_name = git.mangle_name(b'data', hashsplit.GIT_MODE_FILE,
mode)
shalist = [(mode, splitfile_name, sha)]
else:
shalist = hashsplit.split_to_shalist(
new_blob,
new_tree,
files,
keep_boundaries=opt.keep_boundaries,
progress=prog)
tree = new_tree(shalist)
else:
last = 0
it = hashsplit.hashsplit_iter(files,
keep_boundaries=opt.keep_boundaries,
progress=prog)
for (blob, level) in it:
hashsplit.total_split += len(blob)
if opt.copy:
sys.stdout.write(str(blob))
megs = hashsplit.total_split // 1024 // 1024
if not opt.quiet and last != megs:
last = megs
if opt.verbose:
log('\n')
if opt.tree:
out.write(hexlify(tree) + b'\n')
if opt.commit or opt.name:
msg = b'bup split\n\nGenerated by command:\n%r\n' % compat.get_argvb()
ref = opt.name and (b'refs/heads/%s' % opt.name) or None
userline = b'%s <%s@%s>' % (userfullname(), username(), hostname())
commit = pack_writer.new_commit(tree, oldref, userline, date, None,
userline, date, None, msg)
if opt.commit:
out.write(hexlify(commit) + b'\n')
if pack_writer:
pack_writer.close() # must close before we can update the ref
if opt.name:
if cli:
cli.update_ref(refname, commit, oldref)
else:
git.update_ref(refname, commit, oldref)
if cli:
cli.close()
secs = time.time() - start_time
size = hashsplit.total_split
if opt.bench:
log('bup: %.2f kbytes in %.2f secs = %.2f kbytes/sec\n' %
(size / 1024, secs, size / 1024 / secs))
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' %
len(saved_errors))
sys.exit(1)
def main(argv):
# Hack around lack of nonlocal vars in python 2
_nonlocal = {}
o = options.Options(optspec)
opt, flags, extra = o.parse_bytes(argv[1:])
if opt.indexfile:
opt.indexfile = argv_bytes(opt.indexfile)
if opt.name:
opt.name = argv_bytes(opt.name)
if opt.remote:
opt.remote = argv_bytes(opt.remote)
if opt.strip_path:
opt.strip_path = argv_bytes(opt.strip_path)
git.check_repo_or_die()
if not (opt.tree or opt.commit or opt.name):
o.fatal("use one or more of -t, -c, -n")
if not extra:
o.fatal("no filenames given")
extra = [argv_bytes(x) for x in extra]
opt.progress = (istty2 and not opt.quiet)
opt.smaller = parse_num(opt.smaller or 0)
if opt.bwlimit:
client.bwlimit = parse_num(opt.bwlimit)
if opt.date:
date = parse_date_or_fatal(opt.date, o.fatal)
else:
date = time.time()
if opt.strip and opt.strip_path:
o.fatal("--strip is incompatible with --strip-path")
graft_points = []
if opt.graft:
if opt.strip:
o.fatal("--strip is incompatible with --graft")
if opt.strip_path:
o.fatal("--strip-path is incompatible with --graft")
for (option, parameter) in flags:
if option == "--graft":
parameter = argv_bytes(parameter)
splitted_parameter = parameter.split(b'=')
if len(splitted_parameter) != 2:
o.fatal(
"a graft point must be of the form old_path=new_path")
old_path, new_path = splitted_parameter
if not (old_path and new_path):
o.fatal("a graft point cannot be empty")
graft_points.append(
(resolve_parent(old_path), resolve_parent(new_path)))
is_reverse = environ.get(b'BUP_SERVER_REVERSE')
if is_reverse and opt.remote:
o.fatal("don't use -r in reverse mode; it's automatic")
name = opt.name
if name and not valid_save_name(name):
o.fatal("'%s' is not a valid branch name" % path_msg(name))
refname = name and b'refs/heads/%s' % name or None
if opt.remote or is_reverse:
try:
cli = client.Client(opt.remote)
except client.ClientError as e:
log('error: %s' % e)
sys.exit(1)
oldref = refname and cli.read_ref(refname) or None
w = cli.new_packwriter(compression_level=opt.compress)
else:
cli = None
oldref = refname and git.read_ref(refname) or None
w = git.PackWriter(compression_level=opt.compress)
handle_ctrl_c()
# Metadata is stored in a file named .bupm in each directory. The
# first metadata entry will be the metadata for the current directory.
# The remaining entries will be for each of the other directory
# elements, in the order they're listed in the index.
#
# Since the git tree elements are sorted according to
# git.shalist_item_sort_key, the metalist items are accumulated as
# (sort_key, metadata) tuples, and then sorted when the .bupm file is
# created. The sort_key should have been computed using the element's
# mangled name and git mode (after hashsplitting), but the code isn't
# actually doing that but rather uses the element's real name and mode.
# This makes things a bit more difficult when reading it back, see
# vfs.ordered_tree_entries().
# Maintain a stack of information representing the current location in
# the archive being constructed. The current path is recorded in
# parts, which will be something like ['', 'home', 'someuser'], and
# the accumulated content and metadata for of the dirs in parts is
# stored in parallel stacks in shalists and metalists.
parts = [] # Current archive position (stack of dir names).
shalists = [] # Hashes for each dir in paths.
metalists = [] # Metadata for each dir in paths.
def _push(part, metadata):
# Enter a new archive directory -- make it the current directory.
parts.append(part)
shalists.append([])
metalists.append([(b'', metadata)]) # This dir's metadata (no name).
def _pop(force_tree, dir_metadata=None):
# Leave the current archive directory and add its tree to its parent.
assert (len(parts) >= 1)
part = parts.pop()
shalist = shalists.pop()
metalist = metalists.pop()
# FIXME: only test if collision is possible (i.e. given --strip, etc.)?
if force_tree:
tree = force_tree
else:
names_seen = set()
clean_list = []
metaidx = 1 # entry at 0 is for the dir
for x in shalist:
name = x[1]
if name in names_seen:
parent_path = b'/'.join(parts) + b'/'
add_error('error: ignoring duplicate path %s in %s' %
(path_msg(name), path_msg(parent_path)))
if not stat.S_ISDIR(x[0]):
del metalist[metaidx]
else:
names_seen.add(name)
clean_list.append(x)
if not stat.S_ISDIR(x[0]):
metaidx += 1
if dir_metadata: # Override the original metadata pushed for this dir.
metalist = [(b'', dir_metadata)] + metalist[1:]
sorted_metalist = sorted(metalist, key=lambda x: x[0])
metadata = b''.join([m[1].encode() for m in sorted_metalist])
metadata_f = BytesIO(metadata)
mode, id = hashsplit.split_to_blob_or_tree(w.new_blob,
w.new_tree,
[metadata_f],
keep_boundaries=False)
clean_list.append((mode, b'.bupm', id))
tree = w.new_tree(clean_list)
if shalists:
shalists[-1].append((GIT_MODE_TREE,
git.mangle_name(part, GIT_MODE_TREE,
GIT_MODE_TREE), tree))
return tree
_nonlocal['count'] = 0
_nonlocal['subcount'] = 0
_nonlocal['lastremain'] = None
def progress_report(n):
_nonlocal['subcount'] += n
cc = _nonlocal['count'] + _nonlocal['subcount']
pct = total and (cc * 100.0 / total) or 0
now = time.time()
elapsed = now - tstart
kps = elapsed and int(cc / 1024. / elapsed)
kps_frac = 10**int(math.log(kps + 1, 10) - 1)
kps = int(kps / kps_frac) * kps_frac
if cc:
remain = elapsed * 1.0 / cc * (total - cc)
else:
remain = 0.0
if (_nonlocal['lastremain'] and (remain > _nonlocal['lastremain']) and
((remain - _nonlocal['lastremain']) / _nonlocal['lastremain'] <
0.05)):
remain = _nonlocal['lastremain']
else:
_nonlocal['lastremain'] = remain
hours = int(remain / 60 / 60)
mins = int(remain / 60 - hours * 60)
secs = int(remain - hours * 60 * 60 - mins * 60)
if elapsed < 30:
remainstr = ''
kpsstr = ''
else:
kpsstr = '%dk/s' % kps
if hours:
remainstr = '%dh%dm' % (hours, mins)
elif mins:
remainstr = '%dm%d' % (mins, secs)
else:
remainstr = '%ds' % secs
qprogress(
'Saving: %.2f%% (%d/%dk, %d/%d files) %s %s\r' %
(pct, cc / 1024, total / 1024, fcount, ftotal, remainstr, kpsstr))
indexfile = opt.indexfile or git.repo(b'bupindex')
r = index.Reader(indexfile)
try:
msr = index.MetaStoreReader(indexfile + b'.meta')
except IOError as ex:
if ex.errno != EACCES:
raise
log('error: cannot access %r; have you run bup index?' %
path_msg(indexfile))
sys.exit(1)
hlink_db = hlinkdb.HLinkDB(indexfile + b'.hlink')
def already_saved(ent):
return ent.is_valid() and w.exists(ent.sha) and ent.sha
def wantrecurse_pre(ent):
return not already_saved(ent)
def wantrecurse_during(ent):
return not already_saved(ent) or ent.sha_missing()
def find_hardlink_target(hlink_db, ent):
if hlink_db and not stat.S_ISDIR(ent.mode) and ent.nlink > 1:
link_paths = hlink_db.node_paths(ent.dev, ent.ino)
if link_paths:
return link_paths[0]
total = ftotal = 0
if opt.progress:
for (transname, ent) in r.filter(extra, wantrecurse=wantrecurse_pre):
if not (ftotal % 10024):
qprogress('Reading index: %d\r' % ftotal)
exists = ent.exists()
hashvalid = already_saved(ent)
ent.set_sha_missing(not hashvalid)
if not opt.smaller or ent.size < opt.smaller:
if exists and not hashvalid:
total += ent.size
ftotal += 1
progress('Reading index: %d, done.\n' % ftotal)
hashsplit.progress_callback = progress_report
# Root collisions occur when strip or graft options map more than one
# path to the same directory (paths which originally had separate
# parents). When that situation is detected, use empty metadata for
# the parent. Otherwise, use the metadata for the common parent.
# Collision example: "bup save ... --strip /foo /foo/bar /bar".
# FIXME: Add collision tests, or handle collisions some other way.
# FIXME: Detect/handle strip/graft name collisions (other than root),
# i.e. if '/foo/bar' and '/bar' both map to '/'.
first_root = None
root_collision = None
tstart = time.time()
fcount = 0
lastskip_name = None
lastdir = b''
for (transname, ent) in r.filter(extra, wantrecurse=wantrecurse_during):
(dir, file) = os.path.split(ent.name)
exists = (ent.flags & index.IX_EXISTS)
hashvalid = already_saved(ent)
wasmissing = ent.sha_missing()
oldsize = ent.size
if opt.verbose:
if not exists:
status = 'D'
elif not hashvalid:
if ent.sha == index.EMPTY_SHA:
status = 'A'
else:
status = 'M'
else:
status = ' '
if opt.verbose >= 2:
log('%s %-70s\n' % (status, path_msg(ent.name)))
elif not stat.S_ISDIR(ent.mode) and lastdir != dir:
if not lastdir.startswith(dir):
log('%s %-70s\n' %
(status, path_msg(os.path.join(dir, b''))))
lastdir = dir
if opt.progress:
progress_report(0)
fcount += 1
if not exists:
continue
if opt.smaller and ent.size >= opt.smaller:
if exists and not hashvalid:
if opt.verbose:
log('skipping large file "%s"\n' % path_msg(ent.name))
lastskip_name = ent.name
continue
assert (dir.startswith(b'/'))
if opt.strip:
dirp = stripped_path_components(dir, extra)
elif opt.strip_path:
dirp = stripped_path_components(dir, [opt.strip_path])
elif graft_points:
dirp = grafted_path_components(graft_points, dir)
else:
dirp = path_components(dir)
# At this point, dirp contains a representation of the archive
# path that looks like [(archive_dir_name, real_fs_path), ...].
# So given "bup save ... --strip /foo/bar /foo/bar/baz", dirp
# might look like this at some point:
# [('', '/foo/bar'), ('baz', '/foo/bar/baz'), ...].
# This dual representation supports stripping/grafting, where the
# archive path may not have a direct correspondence with the
# filesystem. The root directory is represented by an initial
# component named '', and any component that doesn't have a
# corresponding filesystem directory (due to grafting, for
# example) will have a real_fs_path of None, i.e. [('', None),
# ...].
if first_root == None:
first_root = dirp[0]
elif first_root != dirp[0]:
root_collision = True
# If switching to a new sub-tree, finish the current sub-tree.
while parts > [x[0] for x in dirp]:
_pop(force_tree=None)
# If switching to a new sub-tree, start a new sub-tree.
for path_component in dirp[len(parts):]:
dir_name, fs_path = path_component
# Not indexed, so just grab the FS metadata or use empty metadata.
try:
meta = metadata.from_path(fs_path, normalized=True) \
if fs_path else metadata.Metadata()
except (OSError, IOError) as e:
add_error(e)
lastskip_name = dir_name
meta = metadata.Metadata()
_push(dir_name, meta)
if not file:
if len(parts) == 1:
continue # We're at the top level -- keep the current root dir
# Since there's no filename, this is a subdir -- finish it.
oldtree = already_saved(ent) # may be None
newtree = _pop(force_tree=oldtree)
if not oldtree:
if lastskip_name and lastskip_name.startswith(ent.name):
ent.invalidate()
else:
ent.validate(GIT_MODE_TREE, newtree)
ent.repack()
if exists and wasmissing:
_nonlocal['count'] += oldsize
continue
# it's not a directory
if hashvalid:
id = ent.sha
git_name = git.mangle_name(file, ent.mode, ent.gitmode)
git_info = (ent.gitmode, git_name, id)
shalists[-1].append(git_info)
sort_key = git.shalist_item_sort_key((ent.mode, file, id))
meta = msr.metadata_at(ent.meta_ofs)
meta.hardlink_target = find_hardlink_target(hlink_db, ent)
# Restore the times that were cleared to 0 in the metastore.
(meta.atime, meta.mtime, meta.ctime) = (ent.atime, ent.mtime,
ent.ctime)
metalists[-1].append((sort_key, meta))
else:
id = None
hlink = find_hardlink_target(hlink_db, ent)
try:
meta = metadata.from_path(
ent.name,
hardlink_target=hlink,
normalized=True,
after_stat=after_nondir_metadata_stat)
except (OSError, IOError) as e:
add_error(e)
lastskip_name = ent.name
continue
if stat.S_IFMT(ent.mode) != stat.S_IFMT(meta.mode):
# The mode changed since we indexed the file, this is bad.
# This can cause two issues:
# 1) We e.g. think the file is a regular file, but now it's
# something else (a device, socket, FIFO or symlink, etc.)
# and _read_ from it when we shouldn't.
# 2) We then record it as valid, but don't update the index
# metadata, and on a subsequent save it has 'hashvalid'
# but is recorded as the file type from the index, when
# the content is something else ...
# Avoid all of these consistency issues by just skipping such
# things - it really ought to not happen anyway.
add_error("%s: mode changed since indexing, skipping." %
path_msg(ent.name))
lastskip_name = ent.name
continue
if stat.S_ISREG(ent.mode):
try:
# If the file changes while we're reading it, then our reading
# may stop at some point, but the stat() above may have gotten
# a different size already. Recalculate the meta size so that
# the repository records the accurate size in the metadata, even
# if the other stat() data might be slightly older than the file
# content (which we can't fix, this is inherently racy, but we
# can prevent the size mismatch.)
meta.size = 0
def new_blob(data):
meta.size += len(data)
return w.new_blob(data)
before_saving_regular_file(ent.name)
with hashsplit.open_noatime(ent.name) as f:
(mode, id) = hashsplit.split_to_blob_or_tree(
new_blob, w.new_tree, [f], keep_boundaries=False)
except (IOError, OSError) as e:
add_error('%s: %s' % (ent.name, e))
lastskip_name = ent.name
elif stat.S_ISDIR(ent.mode):
assert (0) # handled above
elif stat.S_ISLNK(ent.mode):
mode, id = (GIT_MODE_SYMLINK, w.new_blob(meta.symlink_target))
else:
# Everything else should be fully described by its
# metadata, so just record an empty blob, so the paths
# in the tree and .bupm will match up.
(mode, id) = (GIT_MODE_FILE, w.new_blob(b''))
if id:
ent.validate(mode, id)
ent.repack()
git_name = git.mangle_name(file, ent.mode, ent.gitmode)
git_info = (mode, git_name, id)
shalists[-1].append(git_info)
sort_key = git.shalist_item_sort_key((ent.mode, file, id))
metalists[-1].append((sort_key, meta))
if exists and wasmissing:
_nonlocal['count'] += oldsize
_nonlocal['subcount'] = 0
if opt.progress:
pct = total and _nonlocal['count'] * 100.0 / total or 100
progress(
'Saving: %.2f%% (%d/%dk, %d/%d files), done. \n' %
(pct, _nonlocal['count'] / 1024, total / 1024, fcount, ftotal))
while len(parts) > 1: # _pop() all the parts above the root
_pop(force_tree=None)
assert (len(shalists) == 1)
assert (len(metalists) == 1)
# Finish the root directory.
tree = _pop(
force_tree=None,
# When there's a collision, use empty metadata for the root.
dir_metadata=metadata.Metadata() if root_collision else None)
sys.stdout.flush()
out = byte_stream(sys.stdout)
if opt.tree:
out.write(hexlify(tree))
out.write(b'\n')
if opt.commit or name:
if compat.py_maj > 2:
# Strip b prefix from python 3 bytes reprs to preserve previous format
msgcmd = b'[%s]' % b', '.join(
[repr(argv_bytes(x))[1:].encode('ascii') for x in argv])
else:
msgcmd = repr(argv)
msg = b'bup save\n\nGenerated by command:\n%s\n' % msgcmd
userline = (b'%s <%s@%s>' % (userfullname(), username(), hostname()))
commit = w.new_commit(tree, oldref, userline, date, None, userline,
date, None, msg)
if opt.commit:
out.write(hexlify(commit))
out.write(b'\n')
msr.close()
w.close() # must close before we can update the ref
if opt.name:
if cli:
cli.update_ref(refname, commit, oldref)
else:
git.update_ref(refname, commit, oldref)
if cli:
cli.close()
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' %
len(saved_errors))
sys.exit(1)
def __init__(self, address):
self.address = address
self.client = client.Client(address)
self.rev_list = self.client.rev_list
sys.path[:0] = [os.path.dirname(os.path.realpath(__file__)) + '/..']
from bup import compat, git, options, client
from bup.helpers import log, saved_errors
from bup.compat import argv_bytes
optspec = """
[BUP_DIR=...] bup init [-r host:path]
--
r,remote= remote repository path
"""
o = options.Options(optspec)
opt, flags, extra = o.parse(compat.argv[1:])
if extra:
o.fatal("no arguments expected")
try:
git.init_repo() # local repo
except git.GitError as e:
log("bup: error: could not init repository: %s" % e)
sys.exit(1)
if opt.remote:
git.check_repo_or_die()
cli = client.Client(argv_bytes(opt.remote), create=True)
cli.close()
else:
qprogress('Splitting: %d kbytes\r' % (total_bytes / 1024))
is_reverse = os.environ.get('BUP_SERVER_REVERSE')
if is_reverse and opt.remote:
o.fatal("don't use -r in reverse mode; it's automatic")
start_time = time.time()
if opt.name and opt.name.startswith('.'):
o.fatal("'%s' is not a valid branch name." % opt.name)
refname = opt.name and 'refs/heads/%s' % opt.name or None
if opt.noop or opt.copy:
cli = pack_writer = oldref = None
elif opt.remote or is_reverse:
cli = client.Client(opt.remote, compression_level=opt.compress)
oldref = refname and cli.read_ref(refname) or None
pack_writer = cli.new_packwriter()
else:
cli = None
oldref = refname and git.read_ref(refname) or None
pack_writer = git.PackWriter(compression_level=opt.compress)
if opt.git_ids:
# the input is actually a series of git object ids that we should retrieve
# and split.
#
# This is a bit messy, but basically it converts from a series of
# CatPipe.get() iterators into a series of file-type objects.
# It would be less ugly if either CatPipe.get() returned a file-like object
# (not very efficient), or split_to_shalist() expected an iterator instead