def test_cli_integrity_missing_file(test_dir, testname, monkeypatch):
monkeypatch.chdir(str(test_dir))
name = archive_name(tags=[testname])
base = Path("base")
missing = base / "data" / "not-present"
with missing.open("wt") as f:
f.write("Hello!")
manifest = Manifest(paths=[base])
with open("manifest.yaml", "wb") as f:
manifest.write(f)
mtime_parent = os.stat(str(missing.parent)).st_mtime
missing.unlink()
os.utime(str(missing.parent), times=(mtime_parent, mtime_parent))
with tarfile.open(name, "w") as tarf:
with open("manifest.yaml", "rb") as f:
manifest_info = tarf.gettarinfo(arcname="base/.manifest.yaml",
fileobj=f)
manifest_info.mode = stat.S_IFREG | 0o444
tarf.addfile(manifest_info, f)
tarf.add("base")
with TemporaryFile(mode="w+t", dir=str(test_dir)) as f:
args = ["verify", name]
callscript("archive-tool.py", args, returncode=3, stderr=f)
f.seek(0)
line = f.readline()
assert "%s:%s: missing" % (name, missing) in line
def open(self, path):
self.path = path
try:
self._file = tarfile.open(str(self.path), 'r')
except OSError as e:
raise ArchiveReadError(str(e))
md = self.get_metadata(".manifest.yaml")
self.basedir = md.path.parent
self.manifest = Manifest(fileobj=md.fileobj)
if not self.manifest.metadata:
# Legacy: Manifest version 1.0 did not have metadata.
self.manifest.add_metadata(self.basedir / ".manifest.yaml")
return self
def create(self,
path,
compression=None,
paths=None,
fileinfos=None,
basedir=None,
workdir=None,
excludes=None,
dedup=DedupMode.LINK,
tags=None):
if compression is None:
try:
compression = compression_map["".join(path.suffixes)]
except KeyError:
# Last ressort default
compression = 'gz'
mode = 'x:' + compression
save_wd = None
try:
if workdir:
save_wd = os.getcwd()
os.chdir(workdir)
self.path = path.resolve()
self._dedup = dedup
self._dupindex = {}
if fileinfos is not None:
if not isinstance(fileinfos, Sequence):
fileinfos = list(fileinfos)
self._check_paths([fi.path for fi in fileinfos], basedir)
try:
self.manifest = Manifest(fileinfos=fileinfos, tags=tags)
except ValueError as e:
raise ArchiveCreateError("invalid fileinfos: %s" % e)
else:
self._check_paths(paths, basedir, excludes)
self.manifest = Manifest(paths=paths,
excludes=excludes,
tags=tags)
bd_fi = self.manifest.find(self.basedir)
if bd_fi and not bd_fi.is_dir():
raise ArchiveCreateError("base directory %s must "
"be a directory" % self.basedir)
self.manifest.add_metadata(self.basedir / ".manifest.yaml")
for md in self._metadata:
md.set_path(self.basedir)
self.manifest.add_metadata(md.path)
self._create(mode)
finally:
if save_wd:
os.chdir(save_wd)
return self
def test_mnifest_sort(test_dir, monkeypatch):
"""Test the Manifest.sort() method.
"""
monkeypatch.chdir(str(test_dir))
manifest = Manifest(paths=[Path("base")])
check_manifest(manifest, testdata)
fileinfos = set(manifest)
manifest.sort(key=lambda fi: getattr(fi, "size", 0), reverse=True)
assert set(manifest) == fileinfos
prev = None
for fi in manifest:
k = getattr(fi, "size", 0)
if prev is not None:
assert k <= prev
prev = k
manifest.sort(key=lambda fi: (fi.type, fi.path))
assert set(manifest) == fileinfos
prev = None
for fi in manifest:
if prev is not None:
assert fi.type >= prev.type
if fi.type == prev.type:
assert fi.path >= prev.path
prev = fi
manifest.sort()
assert set(manifest) == fileinfos
prev = None
for fi in manifest:
if prev is not None:
assert fi.path >= prev.path
prev = fi
def test_manifest_exclude_samelevel(test_dir, monkeypatch):
"""Test excludes: exclude things explictely named in paths.
"""
monkeypatch.chdir(str(test_dir))
paths = [Path("base", "data"), Path("base", "empty")]
excludes = [paths[1]]
manifest = Manifest(paths=paths, excludes=excludes)
data = sub_testdata(testdata, Path("base"), paths[0])
check_manifest(manifest, data)
def test_manifest_exclude_subdir(test_dir, monkeypatch):
"""Test excludes: excluding a subdirectory.
"""
monkeypatch.chdir(str(test_dir))
paths = [Path("base")]
excludes = [Path("base", "data")]
manifest = Manifest(paths=paths, excludes=excludes)
data = sub_testdata(testdata, excludes[0])
check_manifest(manifest, data)
def test_manifest_exclude_file(test_dir, monkeypatch):
"""Test excludes: excluding one single file.
"""
monkeypatch.chdir(str(test_dir))
paths = [Path("base")]
excludes = [Path("base", "msg.txt")]
manifest = Manifest(paths=paths, excludes=excludes)
data = sub_testdata(testdata, excludes[0])
check_manifest(manifest, data)
def test_manifest_exclude_explicit_include(test_dir, monkeypatch):
"""Test excludes: it is possible to explicitely include files, even if
their parent directory is excluded.
"""
monkeypatch.chdir(str(test_dir))
paths = [Path("base"), Path("base", "data", "rnd.dat")]
excludes = [Path("base", "data")]
manifest = Manifest(paths=paths, excludes=excludes)
data = sub_testdata(testdata, excludes[0], paths[1])
check_manifest(manifest, data)
def test_create_fileinfos_manifest(test_dir, monkeypatch):
"""Create the archive from a Manifest.
A Manifest is an iterable of FileInfo objects.
"""
monkeypatch.chdir(test_dir)
manifest = Manifest(paths=[Path("base")])
archive_path = Path("archive-fi-manifest.tar")
Archive().create(archive_path, "", fileinfos=manifest)
with Archive().open(archive_path) as archive:
check_manifest(archive.manifest, testdata)
archive.verify()
def test_manifest_exclude_nonexistent(test_dir, monkeypatch):
"""Test excludes argument to Manifest: excluding a nonexistent file.
This is legal, but should have no effect.
"""
monkeypatch.chdir(str(test_dir))
paths = [Path("base")]
excludes = [Path("base", "non-existent.dat")]
manifest = Manifest(paths=paths, excludes=excludes)
data = sub_testdata(testdata, excludes[0])
check_manifest(manifest, data)
def get_fileinfos(config, schedule):
fileinfos = Manifest(paths=config.dirs, excludes=config.excludes)
try:
base_archives = schedule.get_base_archives(get_prev_backups(config))
except NoFullBackupError:
raise ArchiveCreateError("No previous full backup found, can not "
"create %s archive" % schedule.name)
for p in [i.path for i in base_archives]:
log.debug("considering %s to create differential archive", p)
with Archive().open(p) as base:
fileinfos = filter_fileinfos(base.manifest, fileinfos)
return fileinfos
def test_manifest_from_fileobj():
"""Read a manifest from a YAML file.
"""
with gettestdata("manifest.yaml").open("rt") as f:
manifest = Manifest(fileobj=f)
head = manifest.head
assert set(head.keys()) == {
"Checksums", "Date", "Generator", "Metadata", "Version"
}
assert manifest.version == "1.1"
assert isinstance(manifest.date, datetime.datetime)
assert manifest.checksums == ("sha256", )
assert manifest.tags == ()
check_manifest(manifest, testdata)
def test_manifest_from_paths(test_dir, monkeypatch):
"""Create a manifest reading the files in test_dir.
"""
monkeypatch.chdir(str(test_dir))
manifest = Manifest(paths=[Path("base")])
head = manifest.head
assert set(head.keys()) == {
"Checksums", "Date", "Generator", "Metadata", "Version"
}
assert manifest.version == Manifest.Version
assert isinstance(manifest.date, datetime.datetime)
assert manifest.checksums == tuple(FileInfo.Checksums)
assert manifest.tags == ()
check_manifest(manifest, testdata)
def test_data(tmpdir, monkeypatch):
monkeypatch.chdir(str(tmpdir))
shutil.rmtree("base", ignore_errors=True)
setup_testdata(tmpdir, testdata)
manifest = Manifest(paths=[Path("base")])
manifest.add_metadata(Path("base", ".manifest.yaml"))
with open("manifest.yaml", "wb") as f:
manifest.write(f)
return tmpdir
def test_manifest_from_fileinfos(test_dir, monkeypatch):
"""Create a manifest providing an iterable of fileinfos.
"""
monkeypatch.chdir(test_dir)
fileinfos = FileInfo.iterpaths([Path("base")], set())
manifest = Manifest(fileinfos=fileinfos)
head = manifest.head
assert set(head.keys()) == {
"Checksums", "Date", "Generator", "Metadata", "Version"
}
assert manifest.version == Manifest.Version
assert isinstance(manifest.date, datetime.datetime)
assert manifest.checksums == tuple(FileInfo.Checksums)
assert manifest.tags == ()
check_manifest(manifest, testdata)
def _create(self, path, mode, paths, basedir, excludes, dedup, tags):
self.path = path
self._check_paths(paths, basedir, excludes)
self.manifest = Manifest(paths=paths, excludes=excludes, tags=tags)
self.manifest.add_metadata(self.basedir / ".manifest.yaml")
for md in self._metadata:
md.set_path(self.basedir)
self.manifest.add_metadata(md.path)
with tarfile.open(str(self.path), mode) as tarf:
with tempfile.TemporaryFile() as tmpf:
self.manifest.write(tmpf)
tmpf.seek(0)
self.add_metadata(".manifest.yaml", tmpf)
md_names = self._add_metadata_files(tarf)
dupindex = {}
for fi in self.manifest:
p = fi.path
name = self._arcname(p)
if name in md_names:
raise ArchiveCreateError("cannot add %s: "
"this filename is reserved" % p)
if fi.is_file():
ti = tarf.gettarinfo(str(p), arcname=name)
dup = self._check_duplicate(fi, name, dedup, dupindex)
if dup:
ti.type = tarfile.LNKTYPE
ti.linkname = dup
tarf.addfile(ti)
else:
ti.size = fi.size
ti.type = tarfile.REGTYPE
ti.linkname = ''
with p.open("rb") as f:
tarf.addfile(ti, fileobj=f)
else:
tarf.add(str(p), arcname=name, recursive=False)
def test_verify_missing_metadata_item(test_data, testname):
name = archive_name(tags=[testname])
manifest = Manifest(paths=[Path("base")])
manifest.add_metadata(Path("base", ".manifest.yaml"))
manifest.add_metadata(Path("base", ".msg.txt"))
with tarfile.open(name, "w") as tarf:
with tempfile.TemporaryFile(dir=str(test_data)) as tmpf:
manifest.write(tmpf)
tmpf.seek(0)
ti = tarf.gettarinfo(arcname="base/.manifest.yaml",
fileobj=tmpf)
ti.mode = stat.S_IFREG | stat.S_IMODE(0o444)
tarf.addfile(ti, tmpf)
tarf.add("base")
with Archive().open(Path(name)) as archive:
with pytest.raises(ArchiveIntegrityError) as err:
archive.verify()
assert "'base/.msg.txt' not found" in str(err.value)
def test_create_manifest_missing_checksum(test_dir, testname, monkeypatch):
"""Same as last test, but now creating the archive from a precompiled
manifest.
"""
monkeypatch.chdir(test_dir)
name = archive_name(tags=[testname])
with monkeypatch.context() as m:
m.setattr(FileInfo, "Checksums", ['md5'])
manifest = Manifest(paths=[Path("base")])
# Checksums are calculated lazily, we must explicitely access
# the attribute while monkeypatching FileInfo.Checksums is
# active.
for fi in manifest:
if fi.is_file():
assert set(fi.checksum.keys()) == {'md5'}
with pytest.raises(ArchiveCreateError) as err:
Archive().create(Path(name), "", fileinfos=manifest)
assert "Missing checksum" in str(err.value)
def test_manifest_tags(test_dir, monkeypatch, tags, expected):
"""Set tags in a manifest reading the files in test_dir.
"""
monkeypatch.chdir(str(test_dir))
manifest = Manifest(paths=[Path("base")], tags=tags)
assert manifest.tags == expected
class Archive:
def __init__(self):
self.path = None
self.basedir = None
self.manifest = None
self._file = None
self._metadata = []
def create(self,
path,
compression,
paths,
basedir=None,
workdir=None,
excludes=None,
dedup=DedupMode.LINK,
tags=None):
if sys.version_info < (3, 5):
# The 'x' (exclusive creation) mode was added to tarfile
# in Python 3.5.
mode = 'w:' + compression
else:
mode = 'x:' + compression
if workdir:
with tmp_chdir(workdir):
self._create(workdir / path, mode, paths, basedir, excludes,
dedup, tags)
else:
self._create(path, mode, paths, basedir, excludes, dedup, tags)
return self
def _create(self, path, mode, paths, basedir, excludes, dedup, tags):
self.path = path
self._check_paths(paths, basedir, excludes)
self.manifest = Manifest(paths=paths, excludes=excludes, tags=tags)
self.manifest.add_metadata(self.basedir / ".manifest.yaml")
for md in self._metadata:
md.set_path(self.basedir)
self.manifest.add_metadata(md.path)
with tarfile.open(str(self.path), mode) as tarf:
with tempfile.TemporaryFile() as tmpf:
self.manifest.write(tmpf)
tmpf.seek(0)
self.add_metadata(".manifest.yaml", tmpf)
md_names = self._add_metadata_files(tarf)
dupindex = {}
for fi in self.manifest:
p = fi.path
name = self._arcname(p)
if name in md_names:
raise ArchiveCreateError("cannot add %s: "
"this filename is reserved" % p)
if fi.is_file():
ti = tarf.gettarinfo(str(p), arcname=name)
dup = self._check_duplicate(fi, name, dedup, dupindex)
if dup:
ti.type = tarfile.LNKTYPE
ti.linkname = dup
tarf.addfile(ti)
else:
ti.size = fi.size
ti.type = tarfile.REGTYPE
ti.linkname = ''
with p.open("rb") as f:
tarf.addfile(ti, fileobj=f)
else:
tarf.add(str(p), arcname=name, recursive=False)
def _check_paths(self, paths, basedir, excludes):
"""Check the paths to be added to an archive for several error
conditions. Accept a list of either strings or path-like
objects. Convert them to a list of Path objects. Also sets
self.basedir.
"""
if not paths:
raise ArchiveCreateError("refusing to create an empty archive")
if not basedir:
p = paths[0]
if p.is_absolute():
self.basedir = Path(self.path.name.split('.')[0])
else:
self.basedir = Path(p.parts[0])
else:
self.basedir = basedir
if self.basedir.is_absolute():
raise ArchiveCreateError("basedir must be relative")
# We allow two different cases: either
# - all paths are absolute, or
# - all paths are relative and start with basedir.
# The same rules for paths also apply to excludes, if
# provided. So we may just iterate over the chain of both
# lists.
abspath = None
for p in itertools.chain(paths, excludes or ()):
if not _is_normalized(p):
raise ArchiveCreateError(
"invalid path %s: must be normalized" % p)
if abspath is None:
abspath = p.is_absolute()
else:
if abspath != p.is_absolute():
raise ArchiveCreateError("mixing of absolute and relative "
"paths is not allowed")
if not p.is_absolute():
try:
# This will raise ValueError if p does not start
# with basedir:
p.relative_to(self.basedir)
except ValueError as e:
raise ArchiveCreateError(str(e))
if not abspath:
if self.basedir.is_symlink() or not self.basedir.is_dir():
raise ArchiveCreateError("basedir must be a directory")
def _add_metadata_files(self, tarf):
"""Add the metadata files to the tar file.
"""
md_names = set()
for md in self._metadata:
name = str(md.path)
if name in md_names:
raise ArchiveCreateError("duplicate metadata %s" % name)
md_names.add(name)
ti = tarf.gettarinfo(arcname=name, fileobj=md.fileobj)
ti.mode = stat.S_IFREG | stat.S_IMODE(md.mode)
tarf.addfile(ti, md.fileobj)
return md_names
def _check_duplicate(self, fileinfo, name, dedup, dupindex):
"""Check if the archive item fileinfo should be linked
to another item already added to the archive.
"""
assert fileinfo.is_file()
if dedup == DedupMode.LINK:
st = fileinfo.path.stat()
if st.st_nlink == 1:
return None
idxkey = (st.st_dev, st.st_ino)
elif dedup == DedupMode.CONTENT:
try:
hashalg = fileinfo.Checksums[0]
except IndexError:
return None
idxkey = fileinfo.checksum[hashalg]
else:
return None
if idxkey in dupindex:
return dupindex[idxkey]
else:
dupindex[idxkey] = name
return None
def add_metadata(self, name, fileobj, mode=0o444):
path = self.basedir / name if self.basedir else None
md = MetadataItem(name=name, path=path, fileobj=fileobj, mode=mode)
self._metadata.insert(0, md)
def open(self, path):
self.path = path
try:
self._file = tarfile.open(str(self.path), 'r')
except OSError as e:
raise ArchiveReadError(str(e))
md = self.get_metadata(".manifest.yaml")
self.basedir = md.path.parent
self.manifest = Manifest(fileobj=md.fileobj)
if not self.manifest.metadata:
# Legacy: Manifest version 1.0 did not have metadata.
self.manifest.add_metadata(self.basedir / ".manifest.yaml")
return self
def get_metadata(self, name):
ti = self._file.next()
path = Path(ti.path)
if path.name != name:
raise ArchiveIntegrityError("%s not found" % name)
fileobj = self._file.extractfile(ti)
md = MetadataItem(path=path, tarinfo=ti, fileobj=fileobj)
self._metadata.append(md)
return md
def close(self):
if self._file:
self._file.close()
self._file = None
def __enter__(self):
return self
def __exit__(self, type, value, tb):
self.close()
def __del__(self):
self.close()
def _arcname(self, p):
if p.is_absolute():
return str(self.basedir / p.relative_to(p.root))
else:
return str(p)
def verify(self):
if not self._file:
raise ValueError("archive is closed.")
# Verify that all metadata items are present in the proper
# order at the beginning of the tar file. Start iterating for
# TarInfo objects in the tarfile from the beginning,
# regardless of what has already been read:
tarf_it = iter(self._file)
for md in self.manifest.metadata:
ti = next(tarf_it)
if ti.name != md:
raise ArchiveIntegrityError("Expected metadata item '%s' "
"not found" % (md))
# Check the content of the archive.
for fileinfo in self.manifest:
self._verify_item(fileinfo)
def _verify_item(self, fileinfo):
def _check_condition(cond, item, message):
if not cond:
raise ArchiveIntegrityError("%s: %s" % (item, message))
itemname = "%s:%s" % (self.path, fileinfo.path)
try:
tarinfo = self._file.getmember(self._arcname(fileinfo.path))
except KeyError:
raise ArchiveIntegrityError("%s: missing" % itemname)
_check_condition(tarinfo.mode == fileinfo.mode, itemname, "wrong mode")
_check_condition(
int(tarinfo.mtime) == int(fileinfo.mtime), itemname,
"wrong modification time")
if fileinfo.is_dir():
_check_condition(tarinfo.isdir(), itemname,
"wrong type, expected directory")
elif fileinfo.is_file():
_check_condition(tarinfo.isfile() or tarinfo.islnk(), itemname,
"wrong type, expected regular file")
if tarinfo.isfile():
_check_condition(tarinfo.size == fileinfo.size, itemname,
"wrong size")
with self._file.extractfile(tarinfo) as f:
cs = checksum(f, fileinfo.checksum.keys())
_check_condition(cs == fileinfo.checksum, itemname,
"checksum does not match")
elif fileinfo.is_symlink():
_check_condition(tarinfo.issym(), itemname,
"wrong type, expected symbolic link")
_check_condition(tarinfo.linkname == str(fileinfo.target),
itemname, "wrong link target")
else:
raise ArchiveIntegrityError("%s: invalid type" % (itemname))
def extract(self, targetdir, inclmeta=False):
# We extract the directories last in reverse order. This way,
# the directory attributes, in particular the file modification
# time, is set correctly after the file content is written into
# the directory.
dirstack = []
if inclmeta:
for mi in self.manifest.metadata:
self._file.extract(mi, path=str(targetdir))
for fi in self.manifest:
if fi.is_dir():
dirstack.append(fi.path)
else:
self._file.extract(self._arcname(fi.path), path=str(targetdir))
while True:
try:
p = dirstack.pop()
except IndexError:
break
self._file.extract(self._arcname(p), path=str(targetdir))
