def update_index(top, excluded_paths, exclude_rxs, xdev_exceptions):
# tmax and start must be epoch nanoseconds.
tmax = (time.time() - 1) * 10**9
ri = index.Reader(indexfile)
msw = index.MetaStoreWriter(indexfile + '.meta')
wi = index.Writer(indexfile, msw, tmax)
rig = IterHelper(ri.iter(name=top))
tstart = int(time.time()) * 10**9
hlinks = hlinkdb.HLinkDB(indexfile + '.hlink')
fake_hash = None
if opt.fake_valid:
def fake_hash(name):
return (GIT_MODE_FILE, index.FAKE_SHA)
total = 0
bup_dir = os.path.abspath(git.repo())
index_start = time.time()
for path, pst in recursive_dirlist([top],
xdev=opt.xdev,
bup_dir=bup_dir,
excluded_paths=excluded_paths,
exclude_rxs=exclude_rxs,
xdev_exceptions=xdev_exceptions):
if opt.verbose >= 2 or (opt.verbose == 1
and stat.S_ISDIR(pst.st_mode)):
sys.stdout.write('%s\n' % path)
sys.stdout.flush()
elapsed = time.time() - index_start
paths_per_sec = total / elapsed if elapsed else 0
qprogress('Indexing: %d (%d paths/s)\r' % (total, paths_per_sec))
elif not (total % 128):
elapsed = time.time() - index_start
paths_per_sec = total / elapsed if elapsed else 0
qprogress('Indexing: %d (%d paths/s)\r' % (total, paths_per_sec))
total += 1
while rig.cur and rig.cur.name > path: # deleted paths
if rig.cur.exists():
rig.cur.set_deleted()
rig.cur.repack()
if rig.cur.nlink > 1 and not stat.S_ISDIR(rig.cur.mode):
hlinks.del_path(rig.cur.name)
rig.next()
if rig.cur and rig.cur.name == path: # paths that already existed
need_repack = False
if (rig.cur.stale(pst, tstart, check_device=opt.check_device)):
try:
meta = metadata.from_path(path, statinfo=pst)
except (OSError, IOError) as e:
add_error(e)
rig.next()
continue
if not stat.S_ISDIR(rig.cur.mode) and rig.cur.nlink > 1:
hlinks.del_path(rig.cur.name)
if not stat.S_ISDIR(pst.st_mode) and pst.st_nlink > 1:
hlinks.add_path(path, pst.st_dev, pst.st_ino)
# Clear these so they don't bloat the store -- they're
# already in the index (since they vary a lot and they're
# fixed length). If you've noticed "tmax", you might
# wonder why it's OK to do this, since that code may
# adjust (mangle) the index mtime and ctime -- producing
# fake values which must not end up in a .bupm. However,
# it looks like that shouldn't be possible: (1) When
# "save" validates the index entry, it always reads the
# metadata from the filesytem. (2) Metadata is only
# read/used from the index if hashvalid is true. (3)
# "faked" entries will be stale(), and so we'll invalidate
# them below.
meta.ctime = meta.mtime = meta.atime = 0
meta_ofs = msw.store(meta)
rig.cur.update_from_stat(pst, meta_ofs)
rig.cur.invalidate()
need_repack = True
if not (rig.cur.flags & index.IX_HASHVALID):
if fake_hash:
rig.cur.gitmode, rig.cur.sha = fake_hash(path)
rig.cur.flags |= index.IX_HASHVALID
need_repack = True
if opt.fake_invalid:
rig.cur.invalidate()
need_repack = True
if need_repack:
rig.cur.repack()
rig.next()
else: # new paths
try:
meta = metadata.from_path(path, statinfo=pst)
except (OSError, IOError) as e:
add_error(e)
continue
# See same assignment to 0, above, for rationale.
meta.atime = meta.mtime = meta.ctime = 0
meta_ofs = msw.store(meta)
wi.add(path, pst, meta_ofs, hashgen=fake_hash)
if not stat.S_ISDIR(pst.st_mode) and pst.st_nlink > 1:
hlinks.add_path(path, pst.st_dev, pst.st_ino)
elapsed = time.time() - index_start
paths_per_sec = total / elapsed if elapsed else 0
progress('Indexing: %d, done (%d paths/s).\n' % (total, paths_per_sec))
hlinks.prepare_save()
if ri.exists():
ri.save()
wi.flush()
if wi.count:
wr = wi.new_reader()
if opt.check:
log('check: before merging: oldfile\n')
check_index(ri)
log('check: before merging: newfile\n')
check_index(wr)
mi = index.Writer(indexfile, msw, tmax)
for e in index.merge(ri, wr):
# FIXME: shouldn't we remove deleted entries eventually? When?
mi.add_ixentry(e)
ri.close()
mi.close()
wr.close()
wi.abort()
else:
wi.close()
msw.close()
hlinks.commit_save()
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)
git.check_repo_or_die()
handle_ctrl_c()
if opt.verbose is None:
opt.verbose = 0
if opt.indexfile:
indexfile = argv_bytes(opt.indexfile)
else:
indexfile = git.repo(b'bupindex')
if opt.check:
log('check: starting initial check.\n')
check_index(index.Reader(indexfile))
if opt.clear:
log('clear: clearing index.\n')
clear_index(indexfile)
sys.stdout.flush()
out = byte_stream(sys.stdout)
if opt.update:
if not extra:
o.fatal('update mode (-u) requested but no paths given')
extra = [argv_bytes(x) for x in extra]
excluded_paths = parse_excludes(flags, o.fatal)
exclude_rxs = parse_rx_excludes(flags, o.fatal)
xexcept = index.unique_resolved_paths(extra)
# FIXME: remove this once we account for timestamp races, i.e. index;
# touch new-file; index. It's possible for this to happen quickly
# enough that new-file ends up with the same timestamp as the first
# index, and then bup will ignore it.
tick_start = time.time()
time.sleep(1 - (tick_start - int(tick_start)))
git.check_repo_or_die()
indexfile = opt.indexfile or git.repo('bupindex')
handle_ctrl_c()
if opt.check:
log('check: starting initial check.\n')
check_index(index.Reader(indexfile))
if opt.clear:
log('clear: clearing index.\n')
clear_index(indexfile)
if opt.update:
if not extra:
o.fatal('update mode (-u) requested but no paths given')
excluded_paths = parse_excludes(flags, o.fatal)
exclude_rxs = parse_rx_excludes(flags, o.fatal)
xexcept = index.unique_resolved_paths(extra)
for rp, path in index.reduce_paths(extra):
update_index(rp, excluded_paths, exclude_rxs, xdev_exceptions=xexcept)
if opt['print'] or opt.status or opt.modified:
def main(argv):
o = options.Options(optspec)
opt, flags, extra = o.parse_bytes(argv[1:])
if not (opt.modified or \
opt['print'] or \
opt.status or \
opt.update or \
opt.check or \
opt.clear):
opt.update = 1
if (opt.fake_valid or opt.fake_invalid) and not opt.update:
o.fatal('--fake-{in,}valid are meaningless without -u')
if opt.fake_valid and opt.fake_invalid:
o.fatal('--fake-valid is incompatible with --fake-invalid')
if opt.clear and opt.indexfile:
o.fatal('cannot clear an external index (via -f)')
# FIXME: remove this once we account for timestamp races, i.e. index;
# touch new-file; index. It's possible for this to happen quickly
# enough that new-file ends up with the same timestamp as the first
# index, and then bup will ignore it.
tick_start = time.time()
time.sleep(1 - (tick_start - int(tick_start)))
git.check_repo_or_die()
handle_ctrl_c()
if opt.verbose is None:
opt.verbose = 0
if opt.indexfile:
indexfile = argv_bytes(opt.indexfile)
else:
indexfile = git.repo(b'bupindex')
if opt.check:
log('check: starting initial check.\n')
check_index(index.Reader(indexfile), opt.verbose)
if opt.clear:
log('clear: clearing index.\n')
clear_index(indexfile, opt.verbose)
sys.stdout.flush()
out = byte_stream(sys.stdout)
if opt.update:
if not extra:
o.fatal('update mode (-u) requested but no paths given')
extra = [argv_bytes(x) for x in extra]
excluded_paths = parse_excludes(flags, o.fatal)
exclude_rxs = parse_rx_excludes(flags, o.fatal)
xexcept = index.unique_resolved_paths(extra)
for rp, path in index.reduce_paths(extra):
update_index(rp,
excluded_paths,
exclude_rxs,
indexfile,
check=opt.check,
check_device=opt.check_device,
xdev=opt.xdev,
xdev_exceptions=xexcept,
fake_valid=opt.fake_valid,
fake_invalid=opt.fake_invalid,
out=out,
verbose=opt.verbose)
if opt['print'] or opt.status or opt.modified:
extra = [argv_bytes(x) for x in extra]
for name, ent in index.Reader(indexfile).filter(extra or [b'']):
if (opt.modified
and (ent.is_valid() or ent.is_deleted() or not ent.mode)):
continue
line = b''
if opt.status:
if ent.is_deleted():
line += b'D '
elif not ent.is_valid():
if ent.sha == index.EMPTY_SHA:
line += b'A '
else:
line += b'M '
else:
line += b' '
if opt.hash:
line += hexlify(ent.sha) + b' '
if opt.long:
line += b'%7s %7s ' % (oct(ent.mode).encode('ascii'),
oct(ent.gitmode).encode('ascii'))
out.write(line + (name or b'./') + b'\n')
if opt.check and (opt['print'] or opt.status or opt.modified
or opt.update):
log('check: starting final check.\n')
check_index(index.Reader(indexfile), opt.verbose)
if saved_errors:
log('WARNING: %d errors encountered.\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 update_index(top, excluded_paths):
tmax = time.time() - 1
ri = index.Reader(indexfile)
wi = index.Writer(indexfile, tmax)
rig = IterHelper(ri.iter(name=top))
tstart = int(time.time())
hashgen = None
if opt.fake_valid:
def hashgen(name):
return (GIT_MODE_FILE, index.FAKE_SHA)
total = 0
bup_dir = os.path.abspath(git.repo())
for (path,
pst) in drecurse.recursive_dirlist([top],
xdev=opt.xdev,
bup_dir=bup_dir,
excluded_paths=excluded_paths):
if opt.verbose >= 2 or (opt.verbose == 1
and stat.S_ISDIR(pst.st_mode)):
sys.stdout.write('%s\n' % path)
sys.stdout.flush()
qprogress('Indexing: %d\r' % total)
elif not (total % 128):
qprogress('Indexing: %d\r' % total)
total += 1
while rig.cur and rig.cur.name > path: # deleted paths
if rig.cur.exists():
rig.cur.set_deleted()
rig.cur.repack()
rig.next()
if rig.cur and rig.cur.name == path: # paths that already existed
if pst:
rig.cur.from_stat(pst, tstart)
if not (rig.cur.flags & index.IX_HASHVALID):
if hashgen:
(rig.cur.gitmode, rig.cur.sha) = hashgen(path)
rig.cur.flags |= index.IX_HASHVALID
if opt.fake_invalid:
rig.cur.invalidate()
rig.cur.repack()
rig.next()
else: # new paths
wi.add(path, pst, hashgen=hashgen)
progress('Indexing: %d, done.\n' % total)
if ri.exists():
ri.save()
wi.flush()
if wi.count:
wr = wi.new_reader()
if opt.check:
log('check: before merging: oldfile\n')
check_index(ri)
log('check: before merging: newfile\n')
check_index(wr)
mi = index.Writer(indexfile, tmax)
for e in index.merge(ri, wr):
# FIXME: shouldn't we remove deleted entries eventually? When?
mi.add_ixentry(e)
ri.close()
mi.close()
wr.close()
wi.abort()
else:
wi.close()
def update_index(top, excluded_paths, exclude_rxs):
# tmax and start must be epoch nanoseconds.
tmax = (time.time() - 1) * 10**9
ri = index.Reader(indexfile)
msw = index.MetaStoreWriter(indexfile + '.meta')
wi = index.Writer(indexfile, msw, tmax)
rig = IterHelper(ri.iter(name=top))
tstart = int(time.time()) * 10**9
hlinks = hlinkdb.HLinkDB(indexfile + '.hlink')
hashgen = None
if opt.fake_valid:
def hashgen(name):
return (GIT_MODE_FILE, index.FAKE_SHA)
total = 0
bup_dir = os.path.abspath(git.repo())
index_start = time.time()
for (path,
pst) in drecurse.recursive_dirlist([top],
xdev=opt.xdev,
bup_dir=bup_dir,
excluded_paths=excluded_paths,
exclude_rxs=exclude_rxs):
if opt.verbose >= 2 or (opt.verbose == 1
and stat.S_ISDIR(pst.st_mode)):
sys.stdout.write('%s\n' % path)
sys.stdout.flush()
elapsed = time.time() - index_start
paths_per_sec = total / elapsed if elapsed else 0
qprogress('Indexing: %d (%d paths/s)\r' % (total, paths_per_sec))
elif not (total % 128):
elapsed = time.time() - index_start
paths_per_sec = total / elapsed if elapsed else 0
qprogress('Indexing: %d (%d paths/s)\r' % (total, paths_per_sec))
total += 1
while rig.cur and rig.cur.name > path: # deleted paths
if rig.cur.exists():
rig.cur.set_deleted()
rig.cur.repack()
if rig.cur.nlink > 1 and not stat.S_ISDIR(rig.cur.mode):
hlinks.del_path(rig.cur.name)
rig.next()
if rig.cur and rig.cur.name == path: # paths that already existed
try:
meta = metadata.from_path(path, statinfo=pst)
except (OSError, IOError), e:
add_error(e)
rig.next()
continue
if not stat.S_ISDIR(rig.cur.mode) and rig.cur.nlink > 1:
hlinks.del_path(rig.cur.name)
if not stat.S_ISDIR(pst.st_mode) and pst.st_nlink > 1:
hlinks.add_path(path, pst.st_dev, pst.st_ino)
# Clear these so they don't bloat the store -- they're
# already in the index (since they vary a lot and they're
# fixed length). If you've noticed "tmax", you might
# wonder why it's OK to do this, since that code may
# adjust (mangle) the index mtime and ctime -- producing
# fake values which must not end up in a .bupm. However,
# it looks like that shouldn't be possible: (1) When
# "save" validates the index entry, it always reads the
# metadata from the filesytem. (2) Metadata is only
# read/used from the index if hashvalid is true. (3) index
# always invalidates "faked" entries, because "old != new"
# in from_stat().
meta.ctime = meta.mtime = meta.atime = 0
meta_ofs = msw.store(meta)
rig.cur.from_stat(pst,
meta_ofs,
tstart,
check_device=opt.check_device)
if not (rig.cur.flags & index.IX_HASHVALID):
if hashgen:
(rig.cur.gitmode, rig.cur.sha) = hashgen(path)
rig.cur.flags |= index.IX_HASHVALID
if opt.fake_invalid:
rig.cur.invalidate()
rig.cur.repack()
rig.next()
else: # new paths
try:
meta = metadata.from_path(path, statinfo=pst)
except (OSError, IOError), e:
add_error(e)
continue
# See same assignment to 0, above, for rationale.
meta.atime = meta.mtime = meta.ctime = 0
meta_ofs = msw.store(meta)
wi.add(path, pst, meta_ofs, hashgen=hashgen)
if not stat.S_ISDIR(pst.st_mode) and pst.st_nlink > 1:
hlinks.add_path(path, pst.st_dev, pst.st_ino)
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)