X-Git-Url: https://arthur.barton.de/cgi-bin/gitweb.cgi?p=bup.git;a=blobdiff_plain;f=lib%2Fbup%2Fmetadata.py;h=d5b4b4ac30e902148c27fa45f27c64508daa1d7b;hp=81c234fa2e2dcec7123ffe64dc68e4b9a712d248;hb=f86e3d704af40bf98a557e58f634b1c917019821;hpb=b836113e503ab3f8e3c30c3283eb59f936fb0023 diff --git a/lib/bup/metadata.py b/lib/bup/metadata.py index 81c234f..d5b4b4a 100644 --- a/lib/bup/metadata.py +++ b/lib/bup/metadata.py @@ -4,30 +4,46 @@ # # This code is covered under the terms of the GNU Library General # Public License as described in the bup LICENSE file. -import errno, os, sys, stat, time, pwd, grp, socket -from cStringIO import StringIO -from bup import vint, xstat + +from __future__ import absolute_import, print_function +from binascii import hexlify +from copy import deepcopy +from errno import EACCES, EINVAL, ENOTTY, ENOSYS, EOPNOTSUPP +from io import BytesIO +from time import gmtime, strftime +import errno, os, sys, stat, time, pwd, grp, socket, struct + +from bup import compat, vint, xstat +from bup.compat import py_maj from bup.drecurse import recursive_dirlist -from bup.helpers import add_error, mkdirp, log, is_superuser -from bup.helpers import pwd_from_uid, pwd_from_name, grp_from_gid, grp_from_name +from bup.helpers import add_error, mkdirp, log, is_superuser, format_filesize +from bup.io import path_msg +from bup.pwdgrp import pwd_from_uid, pwd_from_name, grp_from_gid, grp_from_name from bup.xstat import utime, lutime xattr = None if sys.platform.startswith('linux'): + # prefer python-pyxattr (it's a lot faster), but fall back to python-xattr + # as the two are incompatible and only one can be installed on a system try: import xattr except ImportError: log('Warning: Linux xattr support missing; install python-pyxattr.\n') - if xattr: + if xattr and getattr(xattr, 'get_all', None) is None: try: - xattr.get_all - except AttributeError: - log('Warning: python-xattr module is too old; ' - 'install python-pyxattr instead.\n') + from xattr import pyxattr_compat as xattr + if not isinstance(xattr.NS_USER, bytes): + xattr = None + except ImportError: xattr = None + if xattr is None: + log('Warning: python-xattr module is too old; ' + 'upgrade or install python-pyxattr instead.\n') posix1e = None -if not (sys.platform.startswith('cygwin') or sys.platform.startswith('darwin')): +if not (sys.platform.startswith('cygwin') \ + or sys.platform.startswith('darwin') \ + or sys.platform.startswith('netbsd')): try: import posix1e except ImportError: @@ -40,7 +56,20 @@ except ImportError: # not on Linux, in which case files don't have any linux attrs anyway, so # lacking the functions isn't a problem. get_linux_file_attr = set_linux_file_attr = None - + + +# See the bup_get_linux_file_attr() comments. +_suppress_linux_file_attr = \ + sys.byteorder == 'big' and struct.calcsize('@l') > struct.calcsize('@i') + +def check_linux_file_attr_api(): + global get_linux_file_attr, set_linux_file_attr + if not (get_linux_file_attr or set_linux_file_attr): + return + if _suppress_linux_file_attr: + log('Warning: Linux attr support disabled (see "bup help index").\n') + get_linux_file_attr = set_linux_file_attr = None + # WARNING: the metadata encoding is *not* stable yet. Caveat emptor! @@ -56,7 +85,6 @@ except ImportError: # FIXME: Add nfsv4 acl handling - see nfs4-acl-tools. # FIXME: Consider other entries mentioned in stat(2) (S_IFDOOR, etc.). # FIXME: Consider pack('vvvvsss', ...) optimization. -# FIXME: Consider caching users/groups. ## FS notes: # @@ -100,59 +128,59 @@ def _clean_up_path_for_archive(p): result = p # Take everything after any '/../'. - pos = result.rfind('/../') + pos = result.rfind(b'/../') if pos != -1: - result = result[result.rfind('/../') + 4:] + result = result[result.rfind(b'/../') + 4:] # Take everything after any remaining '../'. - if result.startswith("../"): + if result.startswith(b"../"): result = result[3:] # Remove any '/./' sequences. - pos = result.find('/./') + pos = result.find(b'/./') while pos != -1: - result = result[0:pos] + '/' + result[pos + 3:] - pos = result.find('/./') + result = result[0:pos] + b'/' + result[pos + 3:] + pos = result.find(b'/./') # Remove any leading '/'s. - result = result.lstrip('/') + result = result.lstrip(b'/') # Replace '//' with '/' everywhere. - pos = result.find('//') + pos = result.find(b'//') while pos != -1: - result = result[0:pos] + '/' + result[pos + 2:] - pos = result.find('//') + result = result[0:pos] + b'/' + result[pos + 2:] + pos = result.find(b'//') # Take everything after any remaining './'. - if result.startswith('./'): + if result.startswith(b'./'): result = result[2:] # Take everything before any remaining '/.'. - if result.endswith('/.'): + if result.endswith(b'/.'): result = result[:-2] - if result == '' or result.endswith('/..'): - result = '.' + if result == b'' or result.endswith(b'/..'): + result = b'.' return result def _risky_path(p): - if p.startswith('/'): + if p.startswith(b'/'): return True - if p.find('/../') != -1: + if p.find(b'/../') != -1: return True - if p.startswith('../'): + if p.startswith(b'../'): return True - if p.endswith('/..'): + if p.endswith(b'/..'): return True return False def _clean_up_extract_path(p): - result = p.lstrip('/') - if result == '': - return '.' + result = p.lstrip(b'/') + if result == b'': + return b'.' elif _risky_path(result): return None else: @@ -163,13 +191,17 @@ def _clean_up_extract_path(p): # must be unique, and must *never* be changed. _rec_tag_end = 0 _rec_tag_path = 1 -_rec_tag_common = 2 # times, user, group, type, perms, etc. +_rec_tag_common_v1 = 2 # times, user, group, type, perms, etc. (legacy/broken) _rec_tag_symlink_target = 3 _rec_tag_posix1e_acl = 4 # getfacl(1), setfacl(1), etc. -_rec_tag_nfsv4_acl = 5 # intended to supplant posix1e acls? +_rec_tag_nfsv4_acl = 5 # intended to supplant posix1e? (unimplemented) _rec_tag_linux_attr = 6 # lsattr(1) chattr(1) _rec_tag_linux_xattr = 7 # getfattr(1) setfattr(1) _rec_tag_hardlink_target = 8 # hard link target path +_rec_tag_common_v2 = 9 # times, user, group, type, perms, etc. (current) +_rec_tag_common_v3 = 10 # adds optional size to v2 + +_warned_about_attr_einval = None class ApplyError(Exception): @@ -197,13 +229,15 @@ class Metadata: # must be non-negative and < 10**9. def _add_common(self, path, st): + assert(st.st_uid >= 0) + assert(st.st_gid >= 0) + self.size = st.st_size self.uid = st.st_uid self.gid = st.st_gid - self.rdev = st.st_rdev self.atime = st.st_atime self.mtime = st.st_mtime self.ctime = st.st_ctime - self.user = self.group = '' + self.user = self.group = b'' entry = pwd_from_uid(st.st_uid) if entry: self.user = entry.pw_name @@ -211,17 +245,25 @@ class Metadata: if entry: self.group = entry.gr_name self.mode = st.st_mode + # Only collect st_rdev if we might need it for a mknod() + # during restore. On some platforms (i.e. kFreeBSD), it isn't + # stable for other file types. For example "cp -a" will + # change it for a plain file. + if stat.S_ISCHR(st.st_mode) or stat.S_ISBLK(st.st_mode): + self.rdev = st.st_rdev + else: + self.rdev = 0 def _same_common(self, other): """Return true or false to indicate similarity in the hardlink sense.""" return self.uid == other.uid \ and self.gid == other.gid \ and self.rdev == other.rdev \ - and self.atime == other.atime \ and self.mtime == other.mtime \ and self.ctime == other.ctime \ and self.user == other.user \ - and self.group == other.group + and self.group == other.group \ + and self.size == other.size def _encode_common(self): if not self.mode: @@ -229,7 +271,7 @@ class Metadata: atime = xstat.nsecs_to_timespec(self.atime) mtime = xstat.nsecs_to_timespec(self.mtime) ctime = xstat.nsecs_to_timespec(self.ctime) - result = vint.pack('VVsVsVvVvVvV', + result = vint.pack('vvsvsvvVvVvVv', self.mode, self.uid, self.user, @@ -241,23 +283,36 @@ class Metadata: mtime[0], mtime[1], ctime[0], - ctime[1]) + ctime[1], + self.size if self.size is not None else -1) return result - def _load_common_rec(self, port): + def _load_common_rec(self, port, version=3): + if version == 3: + # Added trailing size to v2, negative when None. + unpack_fmt = 'vvsvsvvVvVvVv' + elif version == 2: + unpack_fmt = 'vvsvsvvVvVvV' + elif version == 1: + unpack_fmt = 'VVsVsVvVvVvV' + else: + raise Exception('unexpected common_rec version %d' % version) data = vint.read_bvec(port) - (self.mode, - self.uid, - self.user, - self.gid, - self.group, - self.rdev, - self.atime, - atime_ns, - self.mtime, - mtime_ns, - self.ctime, - ctime_ns) = vint.unpack('VVsVsVvVvVvV', data) + values = vint.unpack(unpack_fmt, data) + if version == 3: + (self.mode, self.uid, self.user, self.gid, self.group, + self.rdev, + self.atime, atime_ns, + self.mtime, mtime_ns, + self.ctime, ctime_ns, size) = values + if size >= 0: + self.size = size + else: + (self.mode, self.uid, self.user, self.gid, self.group, + self.rdev, + self.atime, atime_ns, + self.mtime, mtime_ns, + self.ctime, ctime_ns) = values self.atime = xstat.timespec_to_nsecs((self.atime, atime_ns)) self.mtime = xstat.timespec_to_nsecs((self.mtime, mtime_ns)) self.ctime = xstat.timespec_to_nsecs((self.ctime, ctime_ns)) @@ -273,49 +328,50 @@ class Metadata: def _create_via_common_rec(self, path, create_symlinks=True): if not self.mode: - raise ApplyError('no metadata - cannot create path ' + path) + raise ApplyError('no metadata - cannot create path ' + + path_msg(path)) # If the path already exists and is a dir, try rmdir. # If the path already exists and is anything else, try unlink. st = None try: st = xstat.lstat(path) - except OSError, e: + except OSError as e: if e.errno != errno.ENOENT: raise if st: if stat.S_ISDIR(st.st_mode): try: os.rmdir(path) - except OSError, e: + except OSError as e: if e.errno in (errno.ENOTEMPTY, errno.EEXIST): - msg = 'refusing to overwrite non-empty dir ' + path - raise Exception(msg) + raise Exception('refusing to overwrite non-empty dir ' + + path_msg(path)) raise else: os.unlink(path) if stat.S_ISREG(self.mode): assert(self._recognized_file_type()) - fd = os.open(path, os.O_CREAT|os.O_WRONLY|os.O_EXCL, 0600) + fd = os.open(path, os.O_CREAT|os.O_WRONLY|os.O_EXCL, 0o600) os.close(fd) elif stat.S_ISDIR(self.mode): assert(self._recognized_file_type()) - os.mkdir(path, 0700) + os.mkdir(path, 0o700) elif stat.S_ISCHR(self.mode): assert(self._recognized_file_type()) - os.mknod(path, 0600 | stat.S_IFCHR, self.rdev) + os.mknod(path, 0o600 | stat.S_IFCHR, self.rdev) elif stat.S_ISBLK(self.mode): assert(self._recognized_file_type()) - os.mknod(path, 0600 | stat.S_IFBLK, self.rdev) + os.mknod(path, 0o600 | stat.S_IFBLK, self.rdev) elif stat.S_ISFIFO(self.mode): assert(self._recognized_file_type()) - os.mknod(path, 0600 | stat.S_IFIFO) + os.mkfifo(path, 0o600 | stat.S_IFIFO) elif stat.S_ISSOCK(self.mode): try: - os.mknod(path, 0600 | stat.S_IFSOCK) - except OSError, e: - if e.errno == errno.EINVAL: + os.mknod(path, 0o600 | stat.S_IFSOCK) + except OSError as e: + if e.errno in (errno.EINVAL, errno.EPERM): s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) s.bind(path) else: @@ -326,7 +382,7 @@ class Metadata: # on MacOS, symlink() permissions depend on umask, and there's # no way to chown a symlink after creating it, so we have to # be careful here! - oldumask = os.umask((self.mode & 0777) ^ 0777) + oldumask = os.umask((self.mode & 0o777) ^ 0o777) try: os.symlink(self.symlink_target, path) finally: @@ -335,18 +391,18 @@ class Metadata: else: assert(not self._recognized_file_type()) add_error('not creating "%s" with unrecognized mode "0x%x"\n' - % (path, self.mode)) + % (path_msg(path), self.mode)) def _apply_common_rec(self, path, restore_numeric_ids=False): if not self.mode: - raise ApplyError('no metadata - cannot apply to ' + path) + raise ApplyError('no metadata - cannot apply to ' + path_msg(path)) # FIXME: S_ISDOOR, S_IFMPB, S_IFCMP, S_IFNWK, ... see stat(2). # EACCES errors at this stage are fatal for the current path. if lutime and stat.S_ISLNK(self.mode): try: lutime(path, (self.atime, self.mtime)) - except OSError, e: + except OSError as e: if e.errno == errno.EACCES: raise ApplyError('lutime: %s' % e) else: @@ -354,19 +410,18 @@ class Metadata: else: try: utime(path, (self.atime, self.mtime)) - except OSError, e: + except OSError as e: if e.errno == errno.EACCES: raise ApplyError('utime: %s' % e) else: raise - # Implement tar/rsync-like semantics; see bup-restore(1). - # FIXME: should we consider caching user/group name <-> id - # mappings, getgroups(), etc.? uid = gid = -1 # By default, do nothing. if is_superuser(): - uid = self.uid - gid = self.gid + if self.uid is not None: + uid = self.uid + if self.gid is not None: + gid = self.gid if not restore_numeric_ids: if self.uid != 0 and self.user: entry = pwd_from_name(self.user) @@ -391,18 +446,21 @@ class Metadata: if uid != -1 or gid != -1: try: os.lchown(path, uid, gid) - except OSError, e: + except OSError as e: if e.errno == errno.EPERM: add_error('lchown: %s' % e) elif sys.platform.startswith('cygwin') \ and e.errno == errno.EINVAL: add_error('lchown: unknown uid/gid (%d/%d) for %s' - % (uid, gid, path)) + % (uid, gid, path_msg(path))) else: raise if _have_lchmod: - os.lchmod(path, stat.S_IMODE(self.mode)) + try: + os.lchmod(path, stat.S_IMODE(self.mode)) + except errno.ENOSYS: # Function not implemented + pass elif not stat.S_ISLNK(self.mode): os.chmod(path, stat.S_IMODE(self.mode)) @@ -425,14 +483,19 @@ class Metadata: try: if stat.S_ISLNK(st.st_mode): self.symlink_target = os.readlink(path) - except OSError, e: - add_error('readlink: %s', e) + except OSError as e: + add_error('readlink: %s' % e) def _encode_symlink_target(self): return self.symlink_target def _load_symlink_target_rec(self, port): - self.symlink_target = vint.read_bvec(port) + target = vint.read_bvec(port) + self.symlink_target = target + if self.size is None: + self.size = len(target) + else: + assert(self.size == len(target)) ## Hardlink targets @@ -460,18 +523,31 @@ class Metadata: # The numeric/text distinction only matters when reading/restoring # a stored record. def _add_posix1e_acl(self, path, st): - if not posix1e: return + if not posix1e or not posix1e.HAS_EXTENDED_CHECK: + return if not stat.S_ISLNK(st.st_mode): + acls = None + def_acls = None try: if posix1e.has_extended(path): acl = posix1e.ACL(file=path) - self.posix1e_acl = [acl, acl] # txt and num are the same + acls = [acl, acl] # txt and num are the same if stat.S_ISDIR(st.st_mode): - acl = posix1e.ACL(filedef=path) - self.posix1e_acl.extend([acl, acl]) - except EnvironmentError, e: + def_acl = posix1e.ACL(filedef=(path if py_maj < 3 + else path.decode('iso-8859-1'))) + def_acls = [def_acl, def_acl] + except EnvironmentError as e: if e.errno not in (errno.EOPNOTSUPP, errno.ENOSYS): raise + if acls: + txt_flags = posix1e.TEXT_ABBREVIATE + num_flags = posix1e.TEXT_ABBREVIATE | posix1e.TEXT_NUMERIC_IDS + acl_rep = [acls[0].to_any_text('', b'\n', txt_flags), + acls[1].to_any_text('', b'\n', num_flags)] + if def_acls: + acl_rep.append(def_acls[0].to_any_text('', b'\n', txt_flags)) + acl_rep.append(def_acls[1].to_any_text('', b'\n', num_flags)) + self.posix1e_acl = acl_rep def _same_posix1e_acl(self, other): """Return true or false to indicate similarity in the hardlink sense.""" @@ -481,33 +557,34 @@ class Metadata: # Encode as two strings (w/default ACL string possibly empty). if self.posix1e_acl: acls = self.posix1e_acl - txt_flags = posix1e.TEXT_ABBREVIATE - num_flags = posix1e.TEXT_ABBREVIATE | posix1e.TEXT_NUMERIC_IDS - acl_reps = [acls[0].to_any_text('', '\n', txt_flags), - acls[1].to_any_text('', '\n', num_flags)] - if len(acls) < 3: - acl_reps += ['', ''] - else: - acl_reps.append(acls[2].to_any_text('', '\n', txt_flags)) - acl_reps.append(acls[3].to_any_text('', '\n', num_flags)) - return vint.pack('ssss', - acl_reps[0], acl_reps[1], acl_reps[2], acl_reps[3]) + if len(acls) == 2: + acls.extend([b'', b'']) + return vint.pack('ssss', acls[0], acls[1], acls[2], acls[3]) else: return None def _load_posix1e_acl_rec(self, port): - if not posix1e: return - data = vint.read_bvec(port) - acl_reps = vint.unpack('ssss', data) - if acl_reps[2] == '': - acl_reps = acl_reps[:2] - self.posix1e_acl = [posix1e.ACL(text=x) for x in acl_reps] + acl_rep = vint.unpack('ssss', vint.read_bvec(port)) + if acl_rep[2] == b'': + acl_rep = acl_rep[:2] + self.posix1e_acl = acl_rep def _apply_posix1e_acl_rec(self, path, restore_numeric_ids=False): - def apply_acl(acl, kind): + def apply_acl(acl_rep, kind): + try: + acl = posix1e.ACL(text=acl_rep.decode('ascii')) + except IOError as e: + if e.errno == 0: + # pylibacl appears to return an IOError with errno + # set to 0 if a group referred to by the ACL rep + # doesn't exist on the current system. + raise ApplyError("POSIX1e ACL: can't create %r for %r" + % (acl_rep, path_msg(path))) + else: + raise try: acl.applyto(path, kind) - except IOError, e: + except IOError as e: if e.errno == errno.EPERM or e.errno == errno.EOPNOTSUPP: raise ApplyError('POSIX1e ACL applyto: %s' % e) else: @@ -516,7 +593,7 @@ class Metadata: if not posix1e: if self.posix1e_acl: add_error("%s: can't restore ACLs; posix1e support missing.\n" - % path) + % path_msg(path)) return if self.posix1e_acl: acls = self.posix1e_acl @@ -534,20 +611,27 @@ class Metadata: ## Linux attributes (lsattr(1), chattr(1)) def _add_linux_attr(self, path, st): + check_linux_file_attr_api() if not get_linux_file_attr: return if stat.S_ISREG(st.st_mode) or stat.S_ISDIR(st.st_mode): try: attr = get_linux_file_attr(path) if attr != 0: self.linux_attr = attr - except OSError, e: + except OSError as e: if e.errno == errno.EACCES: add_error('read Linux attr: %s' % e) - elif e.errno == errno.ENOTTY or e.errno == errno.ENOSYS: - # ENOTTY: Function not implemented. - # ENOSYS: Inappropriate ioctl for device. + elif e.errno in (ENOTTY, ENOSYS, EOPNOTSUPP): # Assume filesystem doesn't support attrs. return + elif e.errno == EINVAL: + global _warned_about_attr_einval + if not _warned_about_attr_einval: + log("Ignoring attr EINVAL;" + + " if you're not using ntfs-3g, please report: " + + path_msg(path) + '\n') + _warned_about_attr_einval = True + return else: raise @@ -567,15 +651,21 @@ class Metadata: def _apply_linux_attr_rec(self, path, restore_numeric_ids=False): if self.linux_attr: + check_linux_file_attr_api() if not set_linux_file_attr: add_error("%s: can't restore linuxattrs: " - "linuxattr support missing.\n" % path) + "linuxattr support missing.\n" % path_msg(path)) return try: set_linux_file_attr(path, self.linux_attr) - except OSError, e: - if e.errno in (errno.ENOTTY, errno.EOPNOTSUPP, errno.ENOSYS): - raise ApplyError('Linux chattr: %s' % e) + except OSError as e: + if e.errno in (EACCES, ENOTTY, EOPNOTSUPP, ENOSYS): + raise ApplyError('Linux chattr: %s (0x%s)' + % (e, hex(self.linux_attr))) + elif e.errno == EINVAL: + msg = "if you're not using ntfs-3g, please report" + raise ApplyError('Linux chattr: %s (0x%s) (%s)' + % (e, hex(self.linux_attr), msg)) else: raise @@ -586,7 +676,7 @@ class Metadata: if not xattr: return try: self.linux_xattr = xattr.get_all(path, nofollow=True) - except EnvironmentError, e: + except EnvironmentError as e: if e.errno != errno.EOPNOTSUPP: raise @@ -605,7 +695,7 @@ class Metadata: def _load_linux_xattr_rec(self, file): data = vint.read_bvec(file) - memfile = StringIO(data) + memfile = BytesIO(data) result = [] for i in range(vint.read_vuint(memfile)): key = vint.read_bvec(memfile) @@ -617,33 +707,41 @@ class Metadata: if not xattr: if self.linux_xattr: add_error("%s: can't restore xattr; xattr support missing.\n" - % path) + % path_msg(path)) return - existing_xattrs = set(xattr.list(path, nofollow=True)) - if self.linux_xattr: - for k, v in self.linux_xattr: - if k not in existing_xattrs \ - or v != xattr.get(path, k, nofollow=True): - try: - xattr.set(path, k, v, nofollow=True) - except IOError, e: - if e.errno == errno.EPERM \ - or e.errno == errno.EOPNOTSUPP: - raise ApplyError('xattr.set: %s' % e) - else: - raise - existing_xattrs -= frozenset([k]) - for k in existing_xattrs: + if not self.linux_xattr: + return + try: + existing_xattrs = set(xattr.list(path, nofollow=True)) + except IOError as e: + if e.errno == errno.EACCES: + raise ApplyError('xattr.set %r: %s' % (path_msg(path), e)) + else: + raise + for k, v in self.linux_xattr: + if k not in existing_xattrs \ + or v != xattr.get(path, k, nofollow=True): try: - xattr.remove(path, k, nofollow=True) - except IOError, e: - if e.errno == errno.EPERM: - raise ApplyError('xattr.remove: %s' % e) + xattr.set(path, k, v, nofollow=True) + except IOError as e: + if e.errno == errno.EPERM \ + or e.errno == errno.EOPNOTSUPP: + raise ApplyError('xattr.set %r: %s' % (path_msg(path), e)) else: raise + existing_xattrs -= frozenset([k]) + for k in existing_xattrs: + try: + xattr.remove(path, k, nofollow=True) + except IOError as e: + if e.errno in (errno.EPERM, errno.EACCES): + raise ApplyError('xattr.remove %r: %s' % (path_msg(path), e)) + else: + raise def __init__(self): - self.mode = None + self.mode = self.uid = self.gid = self.user = self.group = None + self.atime = self.mtime = self.ctime = None # optional members self.path = None self.size = None @@ -652,11 +750,75 @@ class Metadata: self.linux_attr = None self.linux_xattr = None self.posix1e_acl = None - self.posix1e_acl_default = None + + def __eq__(self, other): + if not isinstance(other, Metadata): return False + if self.mode != other.mode: return False + if self.mtime != other.mtime: return False + if self.ctime != other.ctime: return False + if self.atime != other.atime: return False + if self.path != other.path: return False + if self.uid != other.uid: return False + if self.gid != other.gid: return False + if self.size != other.size: return False + if self.user != other.user: return False + if self.group != other.group: return False + if self.symlink_target != other.symlink_target: return False + if self.hardlink_target != other.hardlink_target: return False + if self.linux_attr != other.linux_attr: return False + if self.posix1e_acl != other.posix1e_acl: return False + return True + + def __ne__(self, other): + return not self.__eq__(other) + + def __hash__(self): + return hash((self.mode, + self.mtime, + self.ctime, + self.atime, + self.path, + self.uid, + self.gid, + self.size, + self.user, + self.group, + self.symlink_target, + self.hardlink_target, + self.linux_attr, + self.posix1e_acl)) + + def __repr__(self): + result = ['<%s instance at %s' % (self.__class__, hex(id(self)))] + if self.path is not None: + result += ' path:' + repr(self.path) + if self.mode is not None: + result += ' mode: %o (%s)' % (self.mode, xstat.mode_str(self.mode)) + if self.uid is not None: + result += ' uid:' + str(self.uid) + if self.gid is not None: + result += ' gid:' + str(self.gid) + if self.user is not None: + result += ' user:' + repr(self.user) + if self.group is not None: + result += ' group:' + repr(self.group) + if self.size is not None: + result += ' size:' + repr(self.size) + for name, val in (('atime', self.atime), + ('mtime', self.mtime), + ('ctime', self.ctime)): + if val is not None: + result += ' %s:%r (%d)' \ + % (name, + strftime('%Y-%m-%d %H:%M %z', + gmtime(xstat.fstime_floor_secs(val))), + val) + result += '>' + return ''.join(result) def write(self, port, include_path=True): records = include_path and [(_rec_tag_path, self._encode_path())] or [] - records.extend([(_rec_tag_common, self._encode_common()), + records.extend([(_rec_tag_common_v3, self._encode_common()), (_rec_tag_symlink_target, self._encode_symlink_target()), (_rec_tag_hardlink_target, @@ -671,10 +833,13 @@ class Metadata: vint.write_vuint(port, _rec_tag_end) def encode(self, include_path=True): - port = StringIO() + port = BytesIO() self.write(port, include_path) return port.getvalue() + def copy(self): + return deepcopy(self) + @staticmethod def read(port): # This method should either return a valid Metadata object, @@ -690,22 +855,24 @@ class Metadata: while True: # only exit is error (exception) or _rec_tag_end if tag == _rec_tag_path: result._load_path_rec(port) - elif tag == _rec_tag_common: - result._load_common_rec(port) + elif tag == _rec_tag_common_v3: + result._load_common_rec(port, version=3) + elif tag == _rec_tag_common_v2: + result._load_common_rec(port, version=2) elif tag == _rec_tag_symlink_target: result._load_symlink_target_rec(port) elif tag == _rec_tag_hardlink_target: result._load_hardlink_target_rec(port) elif tag == _rec_tag_posix1e_acl: result._load_posix1e_acl_rec(port) - elif tag == _rec_tag_nfsv4_acl: - result._load_nfsv4_acl_rec(port) elif tag == _rec_tag_linux_attr: result._load_linux_attr_rec(port) elif tag == _rec_tag_linux_xattr: result._load_linux_xattr_rec(port) elif tag == _rec_tag_end: return result + elif tag == _rec_tag_common_v1: # Should be very rare. + result._load_common_rec(port, version=1) else: # unknown record vint.skip_bvec(port) tag = vint.read_vuint(port) @@ -725,17 +892,18 @@ class Metadata: if not path: raise Exception('Metadata.apply_to_path() called with no path') if not self._recognized_file_type(): - add_error('not applying metadata to "%s"' % path + add_error('not applying metadata to "%s"' % path_msg(path) + ' with unrecognized mode "0x%x"\n' % self.mode) return num_ids = restore_numeric_ids - try: - self._apply_common_rec(path, restore_numeric_ids=num_ids) - self._apply_posix1e_acl_rec(path, restore_numeric_ids=num_ids) - self._apply_linux_attr_rec(path, restore_numeric_ids=num_ids) - self._apply_linux_xattr_rec(path, restore_numeric_ids=num_ids) - except ApplyError, e: - add_error(e) + for apply_metadata in (self._apply_common_rec, + self._apply_posix1e_acl_rec, + self._apply_linux_attr_rec, + self._apply_linux_xattr_rec): + try: + apply_metadata(path, restore_numeric_ids=num_ids) + except ApplyError as e: + add_error(e) def same_file(self, other): """Compare this to other for equivalency. Return true if @@ -750,11 +918,14 @@ class Metadata: def from_path(path, statinfo=None, archive_path=None, - save_symlinks=True, hardlink_target=None): + save_symlinks=True, hardlink_target=None, + normalized=False): + """Return the metadata associated with the path. When normalized is + true, return the metadata appropriate for a typical save, which + may or may not be all of it.""" result = Metadata() result.path = archive_path st = statinfo or xstat.lstat(path) - result.size = st.st_size result._add_common(path, st) if save_symlinks: result._add_symlink_target(path, st) @@ -762,6 +933,10 @@ def from_path(path, statinfo=None, archive_path=None, result._add_posix1e_acl(path, st) result._add_linux_attr(path, st) result._add_linux_xattr(path, st) + if normalized: + # Only store sizes for regular files and symlinks for now. + if not (stat.S_ISREG(result.mode) or stat.S_ISLNK(result.mode)): + result.size = None return result @@ -775,22 +950,35 @@ def save_tree(output_file, paths, for path in paths: safe_path = _clean_up_path_for_archive(path) if safe_path != path: - log('archiving "%s" as "%s"\n' % (path, safe_path)) + log('archiving "%s" as "%s"\n' + % (path_msg(path), path_msg(safe_path))) - start_dir = os.getcwd() - try: - for (p, st) in recursive_dirlist(paths, xdev=xdev): - dirlist_dir = os.getcwd() - os.chdir(start_dir) + if not recurse: + for p in paths: safe_path = _clean_up_path_for_archive(p) + st = xstat.lstat(p) + if stat.S_ISDIR(st.st_mode): + safe_path += b'/' m = from_path(p, statinfo=st, archive_path=safe_path, save_symlinks=save_symlinks) if verbose: - print >> sys.stderr, m.path + print(m.path, file=sys.stderr) m.write(output_file, include_path=write_paths) - os.chdir(dirlist_dir) - finally: - os.chdir(start_dir) + else: + start_dir = os.getcwd() + try: + for (p, st) in recursive_dirlist(paths, xdev=xdev): + dirlist_dir = os.getcwd() + os.chdir(start_dir) + safe_path = _clean_up_path_for_archive(p) + m = from_path(p, statinfo=st, archive_path=safe_path, + save_symlinks=save_symlinks) + if verbose: + print(m.path, file=sys.stderr) + m.write(output_file, include_path=write_paths) + os.chdir(dirlist_dir) + finally: + os.chdir(start_dir) def _set_up_path(meta, create_symlinks=True): @@ -822,32 +1010,61 @@ all_fields = frozenset(['path', 'posix1e-acl']) -def summary_str(meta): - mode_val = xstat.mode_str(meta.mode) - user_val = meta.user - if not user_val: - user_val = str(meta.uid) - group_val = meta.group - if not group_val: - group_val = str(meta.gid) - size_or_dev_val = '-' - if stat.S_ISCHR(meta.mode) or stat.S_ISBLK(meta.mode): - size_or_dev_val = '%d,%d' % (os.major(meta.rdev), os.minor(meta.rdev)) - elif meta.size: - size_or_dev_val = meta.size - mtime_secs = xstat.fstime_floor_secs(meta.mtime) - time_val = time.strftime('%Y-%m-%d %H:%M', time.localtime(mtime_secs)) - path_val = meta.path or '' - if stat.S_ISLNK(meta.mode): - path_val += ' -> ' + meta.symlink_target - return '%-10s %-11s %11s %16s %s' % (mode_val, - user_val + "/" + group_val, - size_or_dev_val, - time_val, - path_val) - - -def detailed_str(meta, fields = None): +def summary_bytes(meta, numeric_ids = False, classification = None, + human_readable = False): + """Return bytes containing the "ls -l" style listing for meta. + Classification may be "all", "type", or None.""" + user_str = group_str = size_or_dev_str = b'?' + symlink_target = None + if meta: + name = meta.path + mode_str = xstat.mode_str(meta.mode).encode('ascii') + symlink_target = meta.symlink_target + mtime_secs = xstat.fstime_floor_secs(meta.mtime) + mtime_str = strftime('%Y-%m-%d %H:%M', + time.localtime(mtime_secs)).encode('ascii') + if meta.user and not numeric_ids: + user_str = meta.user + elif meta.uid != None: + user_str = str(meta.uid).encode() + if meta.group and not numeric_ids: + group_str = meta.group + elif meta.gid != None: + group_str = str(meta.gid).encode() + if stat.S_ISCHR(meta.mode) or stat.S_ISBLK(meta.mode): + if meta.rdev: + size_or_dev_str = ('%d,%d' % (os.major(meta.rdev), + os.minor(meta.rdev))).encode() + elif meta.size != None: + if human_readable: + size_or_dev_str = format_filesize(meta.size).encode() + else: + size_or_dev_str = str(meta.size).encode() + else: + size_or_dev_str = b'-' + if classification: + classification_str = \ + xstat.classification_str(meta.mode, + classification == 'all').encode() + else: + mode_str = b'?' * 10 + mtime_str = b'????-??-?? ??:??' + classification_str = b'?' + + name = name or b'' + if classification: + name += classification_str + if symlink_target: + name += b' -> ' + meta.symlink_target + + return b'%-10s %-11s %11s %16s %s' % (mode_str, + user_str + b'/' + group_str, + size_or_dev_str, + mtime_str, + name) + + +def detailed_bytes(meta, fields = None): # FIXME: should optional fields be omitted, or empty i.e. "rdev: # 0", "link-target:", etc. if not fields: @@ -855,29 +1072,29 @@ def detailed_str(meta, fields = None): result = [] if 'path' in fields: - path = meta.path or '' - result.append('path: ' + path) + path = meta.path or b'' + result.append(b'path: ' + path) if 'mode' in fields: - result.append('mode: %s (%s)' % (oct(meta.mode), - xstat.mode_str(meta.mode))) + result.append(b'mode: %o (%s)' + % (meta.mode, xstat.mode_str(meta.mode).encode('ascii'))) if 'link-target' in fields and stat.S_ISLNK(meta.mode): - result.append('link-target: ' + meta.symlink_target) + result.append(b'link-target: ' + meta.symlink_target) if 'rdev' in fields: if meta.rdev: - result.append('rdev: %d,%d' % (os.major(meta.rdev), - os.minor(meta.rdev))) + result.append(b'rdev: %d,%d' % (os.major(meta.rdev), + os.minor(meta.rdev))) else: - result.append('rdev: 0') - if 'size' in fields and meta.size: - result.append('size: ' + str(meta.size)) + result.append(b'rdev: 0') + if 'size' in fields and meta.size is not None: + result.append(b'size: %d' % meta.size) if 'uid' in fields: - result.append('uid: ' + str(meta.uid)) + result.append(b'uid: %d' % meta.uid) if 'gid' in fields: - result.append('gid: ' + str(meta.gid)) + result.append(b'gid: %d' % meta.gid) if 'user' in fields: - result.append('user: ' + meta.user) + result.append(b'user: ' + meta.user) if 'group' in fields: - result.append('group: ' + meta.group) + result.append(b'group: ' + meta.group) if 'atime' in fields: # If we don't have xstat.lutime, that means we have to use # utime(), and utime() has no way to set the mtime/atime of a @@ -885,41 +1102,39 @@ def detailed_str(meta, fields = None): # so let's not report it. (That way scripts comparing # before/after won't trigger.) if xstat.lutime or not stat.S_ISLNK(meta.mode): - result.append('atime: ' + xstat.fstime_to_sec_str(meta.atime)) + result.append(b'atime: ' + xstat.fstime_to_sec_bytes(meta.atime)) else: - result.append('atime: 0') + result.append(b'atime: 0') if 'mtime' in fields: if xstat.lutime or not stat.S_ISLNK(meta.mode): - result.append('mtime: ' + xstat.fstime_to_sec_str(meta.mtime)) + result.append(b'mtime: ' + xstat.fstime_to_sec_bytes(meta.mtime)) else: - result.append('mtime: 0') + result.append(b'mtime: 0') if 'ctime' in fields: - result.append('ctime: ' + xstat.fstime_to_sec_str(meta.ctime)) + result.append(b'ctime: ' + xstat.fstime_to_sec_bytes(meta.ctime)) if 'linux-attr' in fields and meta.linux_attr: - result.append('linux-attr: ' + hex(meta.linux_attr)) + result.append(b'linux-attr: %x' % meta.linux_attr) if 'linux-xattr' in fields and meta.linux_xattr: for name, value in meta.linux_xattr: - result.append('linux-xattr: %s -> %s' % (name, repr(value))) - if 'posix1e-acl' in fields and meta.posix1e_acl and posix1e: - flags = posix1e.TEXT_ABBREVIATE + result.append(b'linux-xattr: %s -> %s' % (name, value)) + if 'posix1e-acl' in fields and meta.posix1e_acl: + acl = meta.posix1e_acl[0] + result.append(b'posix1e-acl: ' + acl + b'\n') if stat.S_ISDIR(meta.mode): - acl = meta.posix1e_acl[0] - default_acl = meta.posix1e_acl[2] - result.append(acl.to_any_text('posix1e-acl: ', '\n', flags)) - result.append(acl.to_any_text('posix1e-acl-default: ', '\n', flags)) - else: - acl = meta.posix1e_acl[0] - result.append(acl.to_any_text('posix1e-acl: ', '\n', flags)) - return '\n'.join(result) + def_acl = meta.posix1e_acl[2] + result.append(b'posix1e-acl-default: ' + def_acl + b'\n') + return b'\n'.join(result) class _ArchiveIterator: - def next(self): + def __next__(self): try: return Metadata.read(self._file) except EOFError: raise StopIteration() + next = __next__ + def __iter__(self): return self @@ -927,25 +1142,27 @@ class _ArchiveIterator: self._file = file -def display_archive(file): +def display_archive(file, out): if verbose > 1: first_item = True for meta in _ArchiveIterator(file): if not first_item: - print - print detailed_str(meta) + out.write(b'\n') + out.write(detailed_bytes(meta)) + out.write(b'\n') first_item = False elif verbose > 0: for meta in _ArchiveIterator(file): - print summary_str(meta) + out.write(summary_bytes(meta)) + out.write(b'\n') elif verbose == 0: for meta in _ArchiveIterator(file): if not meta.path: - print >> sys.stderr, \ - 'bup: no metadata path, but asked to only display path', \ - '(increase verbosity?)' + log('bup: no metadata path, but asked to only display path' + ' (increase verbosity?)') sys.exit(1) - print meta.path + out.write(meta.path) + out.write(b'\n') def start_extract(file, create_symlinks=True): @@ -953,10 +1170,11 @@ def start_extract(file, create_symlinks=True): if not meta: # Hit end record. break if verbose: - print >> sys.stderr, meta.path + print(path_msg(meta.path), file=sys.stderr) xpath = _clean_up_extract_path(meta.path) if not xpath: - add_error(Exception('skipping risky path "%s"' % meta.path)) + add_error(Exception('skipping risky path "%s"' + % path_msg(meta.path))) else: meta.path = xpath _set_up_path(meta, create_symlinks=create_symlinks) @@ -969,13 +1187,14 @@ def finish_extract(file, restore_numeric_ids=False): break xpath = _clean_up_extract_path(meta.path) if not xpath: - add_error(Exception('skipping risky path "%s"' % dir.path)) + add_error(Exception('skipping risky path "%s"' + % path_msg(meta.path))) else: if os.path.isdir(meta.path): all_dirs.append(meta) else: if verbose: - print >> sys.stderr, meta.path + print(path_msg(meta.path), file=sys.stderr) meta.apply_to_path(path=xpath, restore_numeric_ids=restore_numeric_ids) all_dirs.sort(key = lambda x : len(x.path), reverse=True) @@ -983,7 +1202,7 @@ def finish_extract(file, restore_numeric_ids=False): # Don't need to check xpath -- won't be in all_dirs if not OK. xpath = _clean_up_extract_path(dir.path) if verbose: - print >> sys.stderr, dir.path + print(path_msg(dir.path), file=sys.stderr) dir.apply_to_path(path=xpath, restore_numeric_ids=restore_numeric_ids) @@ -996,24 +1215,25 @@ def extract(file, restore_numeric_ids=False, create_symlinks=True): break xpath = _clean_up_extract_path(meta.path) if not xpath: - add_error(Exception('skipping risky path "%s"' % meta.path)) + add_error(Exception('skipping risky path "%s"' + % path_msg(meta.path))) else: meta.path = xpath if verbose: - print >> sys.stderr, '+', meta.path + print('+', path_msg(meta.path), file=sys.stderr) _set_up_path(meta, create_symlinks=create_symlinks) if os.path.isdir(meta.path): all_dirs.append(meta) else: if verbose: - print >> sys.stderr, '=', meta.path + print('=', path_msg(meta.path), file=sys.stderr) meta.apply_to_path(restore_numeric_ids=restore_numeric_ids) all_dirs.sort(key = lambda x : len(x.path), reverse=True) for dir in all_dirs: # Don't need to check xpath -- won't be in all_dirs if not OK. xpath = _clean_up_extract_path(dir.path) if verbose: - print >> sys.stderr, '=', xpath + print('=', path_msg(xpath), file=sys.stderr) # Shouldn't have to check for risky paths here (omitted above). dir.apply_to_path(path=dir.path, restore_numeric_ids=restore_numeric_ids)