def report_live_item(n, total, ref_name, ref_id, item, verbosity):
status = 'scanned %02.2f%%' % (n * 100.0 / total)
hex_id = hexstr(ref_id)
dirslash = b'/' if item.type == b'tree' else b''
chunk_path = item.chunk_path
if chunk_path:
if verbosity < 4:
return
ps = b'/'.join(item.path)
chunk_ps = b'/'.join(chunk_path)
log('%s %s:%s/%s%s\n' % (status, hex_id, path_msg(ps),
path_msg(chunk_ps), path_msg(dirslash)))
return
# Top commit, for example has none.
demangled = git.demangle_name(item.path[-1], item.mode)[0] if item.path \
else None
# Don't print mangled paths unless the verbosity is over 3.
if demangled:
ps = b'/'.join(item.path[:-1] + [demangled])
if verbosity == 1:
qprogress('%s %s:%s%s\r' %
(status, hex_id, path_msg(ps), path_msg(dirslash)))
elif (verbosity > 1 and item.type == b'tree') \
or (verbosity > 2 and item.type == b'blob'):
log('%s %s:%s%s\n' %
(status, hex_id, path_msg(ps), path_msg(dirslash)))
elif verbosity > 3:
ps = b'/'.join(item.path)
log('%s %s:%s%s\n' %
(status, hex_id, path_msg(ps), path_msg(dirslash)))
def prog(filenum, nbytes):
global total_bytes
total_bytes += nbytes
if filenum > 0:
qprogress("Splitting: file #%d, %d kbytes\r" % (filenum + 1, total_bytes / 1024))
else:
qprogress("Splitting: %d kbytes\r" % (total_bytes / 1024))
def report_live_item(n, total, ref_name, ref_id, item, verbosity):
status = 'scanned %02.2f%%' % (n * 100.0 / total)
hex_id = ref_id.encode('hex')
dirslash = '/' if item.type == 'tree' else ''
chunk_path = item.chunk_path
if chunk_path:
if verbosity < 4:
return
ps = '/'.join(item.path)
chunk_ps = '/'.join(chunk_path)
log('%s %s:%s/%s%s\n' % (status, hex_id, ps, chunk_ps, dirslash))
return
# Top commit, for example has none.
demangled = git.demangle_name(item.path[-1], item.mode)[0] if item.path \
else None
# Don't print mangled paths unless the verbosity is over 3.
if demangled:
ps = '/'.join(item.path[:-1] + [demangled])
if verbosity == 1:
qprogress('%s %s:%s%s\r' % (status, hex_id, ps, dirslash))
elif (verbosity > 1 and item.type == 'tree') \
or (verbosity > 2 and item.type == 'blob'):
log('%s %s:%s%s\n' % (status, hex_id, ps, dirslash))
elif verbosity > 3:
ps = '/'.join(item.path)
log('%s %s:%s%s\n' % (status, hex_id, ps, dirslash))
def report_live_item(n, total, ref_name, ref_id, item):
global opt
status = 'scanned %02.2f%%' % (n * 100.0 / total)
hex_id = ref_id.encode('hex')
dirslash = '/' if item.type == 'tree' else ''
chunk_path = item.chunk_path
if chunk_path:
if opt.verbose < 4:
return
ps = '/'.join(item.path)
chunk_ps = '/'.join(chunk_path)
log('%s %s:%s/%s%s\n' % (status, hex_id, ps, chunk_ps, dirslash))
return
# Top commit, for example has none.
demangled = git.demangle_name(item.path[-1], item.mode)[0] if item.path \
else None
# Don't print mangled paths unless the verbosity is over 3.
if demangled:
ps = '/'.join(item.path[:-1] + [demangled])
if opt.verbose == 1:
qprogress('%s %s:%s%s\r' % (status, hex_id, ps, dirslash))
elif (opt.verbose > 1 and item.type == 'tree') \
or (opt.verbose > 2 and item.type == 'blob'):
log('%s %s:%s%s\n' % (status, hex_id, ps, dirslash))
elif opt.verbose > 3:
ps = '/'.join(item.path)
log('%s %s:%s%s\n' % (status, hex_id, ps, dirslash))
def check_midx(name):
nicename = git.repo_rel(name)
log('Checking %s.\n' % path_msg(nicename))
try:
ix = git.open_idx(name)
except git.GitError as e:
add_error('%s: %s' % (path_msg(name), e))
return
for count,subname in enumerate(ix.idxnames):
sub = git.open_idx(os.path.join(os.path.dirname(name), subname))
for ecount,e in enumerate(sub):
if not (ecount % 1234):
qprogress(' %d/%d: %s %d/%d\r'
% (count, len(ix.idxnames),
git.shorten_hash(subname).decode('ascii'),
ecount, len(sub)))
if not sub.exists(e):
add_error("%s: %s: %s missing from idx"
% (path_msg(nicename),
git.shorten_hash(subname).decode('ascii'),
hexstr(e)))
if not ix.exists(e):
add_error("%s: %s: %s missing from midx"
% (path_msg(nicename),
git.shorten_hash(subname).decode('ascii'),
hexstr(e)))
prev = None
for ecount,e in enumerate(ix):
if not (ecount % 1234):
qprogress(' Ordering: %d/%d\r' % (ecount, len(ix)))
if e and prev and not e >= prev:
add_error('%s: ordering error: %s < %s'
% (nicename, hexstr(e), hexstr(prev)))
prev = e
def check_midx(name):
nicename = git.repo_rel(name)
log('Checking %s.\n' % nicename)
try:
ix = git.open_idx(name)
except git.GitError as e:
add_error('%s: %s' % (name, e))
return
for count,subname in enumerate(ix.idxnames):
sub = git.open_idx(os.path.join(os.path.dirname(name), subname))
for ecount,e in enumerate(sub):
if not (ecount % 1234):
qprogress(' %d/%d: %s %d/%d\r'
% (count, len(ix.idxnames),
git.shorten_hash(subname), ecount, len(sub)))
if not sub.exists(e):
add_error("%s: %s: %s missing from idx"
% (nicename, git.shorten_hash(subname),
str(e).encode('hex')))
if not ix.exists(e):
add_error("%s: %s: %s missing from midx"
% (nicename, git.shorten_hash(subname),
str(e).encode('hex')))
prev = None
for ecount,e in enumerate(ix):
if not (ecount % 1234):
qprogress(' Ordering: %d/%d\r' % (ecount, len(ix)))
if not e >= prev:
add_error('%s: ordering error: %s < %s'
% (nicename,
str(e).encode('hex'), str(prev).encode('hex')))
prev = e
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))
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))
def prog(filenum, nbytes):
global total_bytes
total_bytes += nbytes
if filenum > 0:
qprogress('Splitting: file #%d, %d kbytes\r' %
(filenum + 1, total_bytes / 1024))
else:
qprogress('Splitting: %d kbytes\r' % (total_bytes / 1024))
def main(argv):
o = options.Options(optspec)
opt, flags, extra = o.parse_bytes(argv[1:])
handle_ctrl_c()
opt.find = argv_bytes(opt.find) if opt.find else b''
if not extra:
o.fatal('you must provide at least one filename')
if len(opt.find) > 40:
o.fatal('--find parameter must be <= 40 chars long')
else:
if len(opt.find) % 2:
s = opt.find + b'0'
else:
s = opt.find
try:
bin = unhexlify(s)
except TypeError:
o.fatal('--find parameter is not a valid hex string')
sys.stdout.flush()
out = byte_stream(sys.stdout)
find = opt.find.lower()
count = 0
idxfiles = [argv_bytes(x) for x in extra]
for name in idxfiles:
try:
ix = git.open_idx(name)
except git.GitError as e:
add_error('%r: %s' % (name, e))
continue
if len(opt.find) == 40:
if ix.exists(bin):
out.write(b'%s %s\n' % (name, find))
else:
# slow, exhaustive search
for _i in ix:
i = hexlify(_i)
if i.startswith(find):
out.write(b'%s %s\n' % (name, i))
qprogress('Searching: %d\r' % count)
count += 1
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' %
len(saved_errors))
sys.exit(1)
def send_index(self, name, f, send_size):
self._require_command(b'send-index')
#debug1('requesting %r\n' % name)
self.check_busy()
self.conn.write(b'send-index %s\n' % name)
n = struct.unpack('!I', self.conn.read(4))[0]
assert(n)
send_size(n)
count = 0
progress('Receiving index from server: %d/%d\r' % (count, n))
for b in chunkyreader(self.conn, n):
f.write(b)
count += len(b)
qprogress('Receiving index from server: %d/%d\r' % (count, n))
progress('Receiving index from server: %d/%d, done.\n' % (count, n))
self.check_ok()
def sync_index(self, name):
#debug1('requesting %r\n' % name)
self.check_busy()
mkdirp(self.cachedir)
fn = os.path.join(self.cachedir, name)
if os.path.exists(fn):
msg = "won't request existing .idx, try `bup bloom --check %s`" % fn
raise ClientError(msg)
self.conn.write('send-index %s\n' % name)
n = struct.unpack('!I', self.conn.read(4))[0]
assert(n)
with atomically_replaced_file(fn, 'w') as f:
count = 0
for b in chunkyreader(self.conn, n):
f.write(b)
count += len(b)
qprogress('Receiving index from server: %d/%d\r' % (count, n))
self.check_ok()
def sync_index(self, name):
#debug1('requesting %r\n' % name)
self.check_busy()
mkdirp(self.cachedir)
fn = os.path.join(self.cachedir, name)
if os.path.exists(fn):
msg = "won't request existing .idx, try `bup bloom --check %s`" % fn
raise ClientError(msg)
self.conn.write('send-index %s\n' % name)
n = struct.unpack('!I', self.conn.read(4))[0]
assert (n)
with atomically_replaced_file(fn, 'w') as f:
count = 0
progress('Receiving index from server: %d/%d\r' % (count, n))
for b in chunkyreader(self.conn, n):
f.write(b)
count += len(b)
qprogress('Receiving index from server: %d/%d\r' % (count, n))
progress('Receiving index from server: %d/%d, done.\n' %
(count, n))
self.check_ok()
def progress_report(n):
global count, subcount, lastremain
subcount += n
cc = count + 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 (lastremain and (remain > lastremain)
and ((remain - lastremain)/lastremain < 0.05)):
remain = lastremain
else:
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))
def sweep(live_objects, existing_count, cat_pipe, threshold, compression,
verbosity):
# Traverse all the packs, saving the (probably) live data.
ns = Nonlocal()
ns.stale_files = []
def remove_stale_files(new_pack_prefix):
if verbosity and new_pack_prefix:
log('created ' + basename(new_pack_prefix) + '\n')
for p in ns.stale_files:
if new_pack_prefix and p.startswith(new_pack_prefix):
continue # Don't remove the new pack file
if verbosity:
log('removing ' + basename(p) + '\n')
os.unlink(p)
if ns.stale_files: # So git cat-pipe will close them
cat_pipe.restart()
ns.stale_files = []
writer = git.PackWriter(objcache_maker=None,
compression_level=compression,
run_midx=False,
on_pack_finish=remove_stale_files)
# FIXME: sanity check .idx names vs .pack names?
collect_count = 0
for idx_name in glob.glob(os.path.join(git.repo('objects/pack'), '*.idx')):
if verbosity:
qprogress('preserving live data (%d%% complete)\r'
% ((float(collect_count) / existing_count) * 100))
idx = git.open_idx(idx_name)
idx_live_count = 0
for i in xrange(0, len(idx)):
sha = idx.shatable[i * 20 : (i + 1) * 20]
if live_objects.exists(sha):
idx_live_count += 1
collect_count += idx_live_count
if idx_live_count == 0:
if verbosity:
log('deleting %s\n'
% git.repo_rel(basename(idx_name)))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + 'pack')
continue
live_frac = idx_live_count / float(len(idx))
if live_frac > ((100 - threshold) / 100.0):
if verbosity:
log('keeping %s (%d%% live)\n' % (git.repo_rel(basename(idx_name)),
live_frac * 100))
continue
if verbosity:
log('rewriting %s (%.2f%% live)\n' % (basename(idx_name),
live_frac * 100))
for i in xrange(0, len(idx)):
sha = idx.shatable[i * 20 : (i + 1) * 20]
if live_objects.exists(sha):
item_it = cat_pipe.get(sha.encode('hex'))
type = item_it.next()
writer.just_write(sha, type, ''.join(item_it))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + 'pack')
if verbosity:
progress('preserving live data (%d%% complete)\n'
% ((float(collect_count) / existing_count) * 100))
# Nothing should have recreated midx/bloom yet.
pack_dir = git.repo('objects/pack')
assert(not os.path.exists(os.path.join(pack_dir, 'bup.bloom')))
assert(not glob.glob(os.path.join(pack_dir, '*.midx')))
# try/catch should call writer.abort()?
# This will finally run midx.
writer.close() # Can only change refs (if needed) after this.
remove_stale_files(None) # In case we didn't write to the writer.
if verbosity:
log('discarded %d%% of objects\n'
% ((existing_count - count_objects(pack_dir, verbosity))
/ float(existing_count) * 100))
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()
def pfunc(count, total):
qprogress('bup: merging indexes (%d/%d)\r' % (count, total))
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".
def do_bloom(path, outfilename):
global _first
b = None
if os.path.exists(outfilename) and not opt.force:
b = bloom.ShaBloom(outfilename)
if not b.valid():
debug1("bloom: Existing invalid bloom found, regenerating.\n")
b = None
add = []
rest = []
add_count = 0
rest_count = 0
for i,name in enumerate(glob.glob('%s/*.idx' % path)):
progress('bloom: counting: %d\r' % i)
ix = git.open_idx(name)
ixbase = os.path.basename(name)
if b and (ixbase in b.idxnames):
rest.append(name)
rest_count += len(ix)
else:
add.append(name)
add_count += len(ix)
total = add_count + rest_count
if not add:
debug1("bloom: nothing to do.\n")
return
if b:
if len(b) != rest_count:
debug1("bloom: size %d != idx total %d, regenerating\n"
% (len(b), rest_count))
b = None
elif (b.bits < bloom.MAX_BLOOM_BITS and
b.pfalse_positive(add_count) > bloom.MAX_PFALSE_POSITIVE):
debug1("bloom: regenerating: adding %d entries gives "
"%.2f%% false positives.\n"
% (add_count, b.pfalse_positive(add_count)))
b = None
else:
b = bloom.ShaBloom(outfilename, readwrite=True, expected=add_count)
if not b: # Need all idxs to build from scratch
add += rest
add_count += rest_count
del rest
del rest_count
msg = b is None and 'creating from' or 'adding'
if not _first: _first = path
dirprefix = (_first != path) and git.repo_rel(path)+': ' or ''
progress('bloom: %s%s %d file%s (%d object%s).\n'
% (dirprefix, msg,
len(add), len(add)!=1 and 's' or '',
add_count, add_count!=1 and 's' or ''))
tfname = None
if b is None:
tfname = os.path.join(path, 'bup.tmp.bloom')
b = bloom.create(tfname, expected=add_count, k=opt.k)
count = 0
icount = 0
for name in add:
ix = git.open_idx(name)
qprogress('bloom: writing %.2f%% (%d/%d objects)\r'
% (icount*100.0/add_count, icount, add_count))
b.add_idx(ix)
count += 1
icount += len(ix)
# Currently, there's an open file object for tfname inside b.
# Make sure it's closed before rename.
b.close()
if tfname:
os.rename(tfname, outfilename)
def do_bloom(path, outfilename, k, force):
global _first
assert k in (None, 4, 5)
b = None
if os.path.exists(outfilename) and not force:
b = bloom.ShaBloom(outfilename)
if not b.valid():
debug1("bloom: Existing invalid bloom found, regenerating.\n")
b = None
add = []
rest = []
add_count = 0
rest_count = 0
for i, name in enumerate(glob.glob(b'%s/*.idx' % path)):
progress('bloom: counting: %d\r' % i)
ix = git.open_idx(name)
ixbase = os.path.basename(name)
if b and (ixbase in b.idxnames):
rest.append(name)
rest_count += len(ix)
else:
add.append(name)
add_count += len(ix)
if not add:
debug1("bloom: nothing to do.\n")
return
if b:
if len(b) != rest_count:
debug1("bloom: size %d != idx total %d, regenerating\n" %
(len(b), rest_count))
b = None
elif k is not None and k != b.k:
debug1("bloom: new k %d != existing k %d, regenerating\n" %
(k, b.k))
b = None
elif (b.bits < bloom.MAX_BLOOM_BITS[b.k]
and b.pfalse_positive(add_count) > bloom.MAX_PFALSE_POSITIVE):
debug1("bloom: regenerating: adding %d entries gives "
"%.2f%% false positives.\n" %
(add_count, b.pfalse_positive(add_count)))
b = None
else:
b = bloom.ShaBloom(outfilename, readwrite=True, expected=add_count)
if not b: # Need all idxs to build from scratch
add += rest
add_count += rest_count
del rest
del rest_count
msg = b is None and 'creating from' or 'adding'
if not _first: _first = path
dirprefix = (_first != path) and git.repo_rel(path) + b': ' or b''
progress('bloom: %s%s %d file%s (%d object%s).\r' %
(path_msg(dirprefix), msg, len(add), len(add) != 1 and 's'
or '', add_count, add_count != 1 and 's' or ''))
tfname = None
if b is None:
tfname = os.path.join(path, b'bup.tmp.bloom')
b = bloom.create(tfname, expected=add_count, k=k)
count = 0
icount = 0
for name in add:
ix = git.open_idx(name)
qprogress('bloom: writing %.2f%% (%d/%d objects)\r' %
(icount * 100.0 / add_count, icount, add_count))
b.add_idx(ix)
count += 1
icount += len(ix)
# Currently, there's an open file object for tfname inside b.
# Make sure it's closed before rename.
b.close()
if tfname:
os.rename(tfname, outfilename)
def sweep(live_objects, existing_count, cat_pipe, opt):
# Traverse all the packs, saving the (probably) live data.
ns = Nonlocal()
ns.stale_files = []
def remove_stale_files(new_pack_prefix):
if opt.verbose and new_pack_prefix:
log('created ' + basename(new_pack_prefix) + '\n')
for p in ns.stale_files:
if opt.verbose:
log('removing ' + basename(p) + '\n')
os.unlink(p)
ns.stale_files = []
writer = git.PackWriter(objcache_maker=None,
compression_level=opt.compress,
run_midx=False,
on_pack_finish=remove_stale_files)
# FIXME: sanity check .idx names vs .pack names?
collect_count = 0
for idx_name in glob.glob(os.path.join(git.repo('objects/pack'), '*.idx')):
if opt.verbose:
qprogress('preserving live data (%d%% complete)\r' %
((float(collect_count) / existing_count) * 100))
idx = git.open_idx(idx_name)
idx_live_count = 0
for i in xrange(0, len(idx)):
sha = idx.shatable[i * 20:(i + 1) * 20]
if live_objects.exists(sha):
idx_live_count += 1
collect_count += idx_live_count
if idx_live_count == 0:
if opt.verbose:
log('deleting %s\n' % git.repo_rel(basename(idx_name)))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + 'pack')
continue
live_frac = idx_live_count / float(len(idx))
if live_frac > ((100 - opt.threshold) / 100.0):
if opt.verbose:
log('keeping %s (%d%% live)\n' %
(git.repo_rel(basename(idx_name)), live_frac * 100))
continue
if opt.verbose:
log('rewriting %s (%.2f%% live)\n' %
(basename(idx_name), live_frac * 100))
for i in xrange(0, len(idx)):
sha = idx.shatable[i * 20:(i + 1) * 20]
if live_objects.exists(sha):
item_it = cat_pipe.get(sha.encode('hex'))
type = item_it.next()
writer.write(sha, type, ''.join(item_it))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + 'pack')
if opt.verbose:
progress('preserving live data (%d%% complete)\n' %
((float(collect_count) / existing_count) * 100))
# Nothing should have recreated midx/bloom yet.
pack_dir = git.repo('objects/pack')
assert (not os.path.exists(os.path.join(pack_dir, 'bup.bloom')))
assert (not glob.glob(os.path.join(pack_dir, '*.midx')))
# try/catch should call writer.abort()?
# This will finally run midx.
writer.close() # Can only change refs (if needed) after this.
remove_stale_files(None) # In case we didn't write to the writer.
if opt.verbose:
log('discarded %d%% of objects\n' %
((existing_count - count_objects(pack_dir)) /
float(existing_count) * 100))
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 pfunc(count, total):
qprogress('bup: merging indexes (%d/%d)\r' % (count, total))
def restore(repo, parent_path, name, item, top, sparse, numeric_ids, owner_map,
exclude_rxs, verbosity, hardlinks):
global total_restored
mode = vfs.item_mode(item)
treeish = S_ISDIR(mode)
fullname = parent_path + b'/' + name
# Match behavior of index --exclude-rx with respect to paths.
if should_rx_exclude_path(fullname + (b'/' if treeish else b''),
exclude_rxs):
return
if not treeish:
# Do this now so we'll have meta.symlink_target for verbose output
item = vfs.augment_item_meta(repo, item, include_size=True)
meta = item.meta
assert (meta.mode == mode)
if stat.S_ISDIR(mode):
if verbosity >= 1:
out.write(b'%s/\n' % fullname)
elif stat.S_ISLNK(mode):
assert (meta.symlink_target)
if verbosity >= 2:
out.write(b'%s@ -> %s\n' % (fullname, meta.symlink_target))
else:
if verbosity >= 2:
out.write(fullname + b'\n')
orig_cwd = os.getcwd()
try:
if treeish:
# Assumes contents() returns '.' with the full metadata first
sub_items = vfs.contents(repo, item, want_meta=True)
dot, item = next(sub_items, None)
assert (dot == b'.')
item = vfs.augment_item_meta(repo, item, include_size=True)
meta = item.meta
meta.create_path(name)
os.chdir(name)
total_restored += 1
if verbosity >= 0:
qprogress('Restoring: %d\r' % total_restored)
for sub_name, sub_item in sub_items:
restore(repo, fullname, sub_name, sub_item, top, sparse,
numeric_ids, owner_map, exclude_rxs, verbosity,
hardlinks)
os.chdir(b'..')
apply_metadata(meta, name, numeric_ids, owner_map)
else:
created_hardlink = False
if meta.hardlink_target:
created_hardlink = hardlink_if_possible(
fullname, item, top, hardlinks)
if not created_hardlink:
meta.create_path(name)
if stat.S_ISREG(meta.mode):
if sparse:
write_file_content_sparsely(repo, name, item)
else:
write_file_content(repo, name, item)
total_restored += 1
if verbosity >= 0:
qprogress('Restoring: %d\r' % total_restored)
if not created_hardlink:
apply_metadata(meta, name, numeric_ids, owner_map)
finally:
os.chdir(orig_cwd)
def sweep(live_objects, existing_count, cat_pipe, threshold, compression,
verbosity):
# Traverse all the packs, saving the (probably) live data.
ns = Nonlocal()
ns.stale_files = []
def remove_stale_files(new_pack_prefix):
if verbosity and new_pack_prefix:
log('created ' + path_msg(basename(new_pack_prefix)) + '\n')
for p in ns.stale_files:
if new_pack_prefix and p.startswith(new_pack_prefix):
continue # Don't remove the new pack file
if verbosity:
log('removing ' + path_msg(basename(p)) + '\n')
os.unlink(p)
if ns.stale_files: # So git cat-pipe will close them
cat_pipe.restart()
ns.stale_files = []
writer = git.PackWriter(objcache_maker=None,
compression_level=compression,
run_midx=False,
on_pack_finish=remove_stale_files)
try:
# FIXME: sanity check .idx names vs .pack names?
collect_count = 0
for idx_name in glob.glob(
os.path.join(git.repo(b'objects/pack'), b'*.idx')):
if verbosity:
qprogress('preserving live data (%d%% complete)\r' %
((float(collect_count) / existing_count) * 100))
with git.open_idx(idx_name) as idx:
idx_live_count = 0
for sha in idx:
if live_objects.exists(sha):
idx_live_count += 1
collect_count += idx_live_count
if idx_live_count == 0:
if verbosity:
log('deleting %s\n' %
path_msg(git.repo_rel(basename(idx_name))))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + b'pack')
continue
live_frac = idx_live_count / float(len(idx))
if live_frac > ((100 - threshold) / 100.0):
if verbosity:
log('keeping %s (%d%% live)\n' % (git.repo_rel(
basename(idx_name)), live_frac * 100))
continue
if verbosity:
log('rewriting %s (%.2f%% live)\n' %
(basename(idx_name), live_frac * 100))
for sha in idx:
if live_objects.exists(sha):
item_it = cat_pipe.get(hexlify(sha))
_, typ, _ = next(item_it)
writer.just_write(sha, typ, b''.join(item_it))
ns.stale_files.append(idx_name)
ns.stale_files.append(idx_name[:-3] + b'pack')
if verbosity:
progress('preserving live data (%d%% complete)\n' %
((float(collect_count) / existing_count) * 100))
# Nothing should have recreated midx/bloom yet.
pack_dir = git.repo(b'objects/pack')
assert (not os.path.exists(os.path.join(pack_dir, b'bup.bloom')))
assert (not glob.glob(os.path.join(pack_dir, b'*.midx')))
except BaseException as ex:
with pending_raise(ex):
writer.abort()
# This will finally run midx.
# Can only change refs (if needed) after this.
writer.close()
remove_stale_files(None) # In case we didn't write to the writer.
if verbosity:
log('discarded %d%% of objects\n' %
((existing_count - count_objects(pack_dir, verbosity)) /
float(existing_count) * 100))
def restore(repo, parent_path, name, item, top, sparse, numeric_ids, owner_map,
exclude_rxs, verbosity, hardlinks):
global total_restored
mode = vfs.item_mode(item)
treeish = S_ISDIR(mode)
fullname = parent_path + '/' + name
# Match behavior of index --exclude-rx with respect to paths.
if should_rx_exclude_path(fullname + ('/' if treeish else ''),
exclude_rxs):
return
if not treeish:
# Do this now so we'll have meta.symlink_target for verbose output
item = vfs.augment_item_meta(repo, item, include_size=True)
meta = item.meta
assert(meta.mode == mode)
if stat.S_ISDIR(mode):
if verbosity >= 1:
print('%s/' % fullname)
elif stat.S_ISLNK(mode):
assert(meta.symlink_target)
if verbosity >= 2:
print('%s@ -> %s' % (fullname, meta.symlink_target))
else:
if verbosity >= 2:
print(fullname)
orig_cwd = os.getcwd()
try:
if treeish:
# Assumes contents() returns '.' with the full metadata first
sub_items = vfs.contents(repo, item, want_meta=True)
dot, item = next(sub_items, None)
assert(dot == '.')
item = vfs.augment_item_meta(repo, item, include_size=True)
meta = item.meta
meta.create_path(name)
os.chdir(name)
total_restored += 1
if verbosity >= 0:
qprogress('Restoring: %d\r' % total_restored)
for sub_name, sub_item in sub_items:
restore(repo, fullname, sub_name, sub_item, top, sparse,
numeric_ids, owner_map, exclude_rxs, verbosity,
hardlinks)
os.chdir('..')
apply_metadata(meta, name, numeric_ids, owner_map)
else:
created_hardlink = False
if meta.hardlink_target:
created_hardlink = hardlink_if_possible(fullname, item, top,
hardlinks)
if not created_hardlink:
meta.create_path(name)
if stat.S_ISREG(meta.mode):
if sparse:
write_file_content_sparsely(repo, name, item)
else:
write_file_content(repo, name, item)
total_restored += 1
if verbosity >= 0:
qprogress('Restoring: %d\r' % total_restored)
if not created_hardlink:
apply_metadata(meta, name, numeric_ids, owner_map)
finally:
os.chdir(orig_cwd)
def pfunc(count, total):
qprogress('Reading indexes: %.2f%% (%d/%d)\r'
% (count*100.0/total, count, total))
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()
def plog(s):
if opt.quiet:
return
qprogress(s)
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".
def save_tree(opt, reader, hlink_db, msr, w):
# 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
# [StackDir(name=''), StackDir(name='home'), StackDir(name='someuser')],
# and the accumulated content and metadata for files in the dirs is stored
# in the .items member of the StackDir.
stack = []
def _push(part, metadata):
# Enter a new archive directory -- make it the current directory.
item = StackDir(part, metadata)
stack.append(item)
def _pop(force_tree=None, dir_metadata=None):
# Leave the current archive directory and add its tree to its parent.
item = stack.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 = []
for x in item.items:
name = x.name
if name in names_seen:
parent_path = b'/'.join(x.name for x in stack) + b'/'
add_error('error: ignoring duplicate path %s in %s'
% (path_msg(name), path_msg(parent_path)))
else:
names_seen.add(name)
clean_list.append(x)
# if set, overrides the original metadata pushed for this dir.
if dir_metadata is None:
dir_metadata = item.meta
metalist = [(b'', dir_metadata)]
metalist += [(git.shalist_item_sort_key((entry.mode, entry.name, None)),
entry.meta)
for entry in clean_list if entry.mode != GIT_MODE_TREE]
metalist.sort(key = lambda x: x[0])
metadata = BytesIO(b''.join(m[1].encode() for m in metalist))
mode, id = hashsplit.split_to_blob_or_tree(w.new_blob, w.new_tree,
[metadata],
keep_boundaries=False)
shalist = [(mode, b'.bupm', id)]
shalist += [(entry.gitmode,
git.mangle_name(entry.name, entry.mode, entry.gitmode),
entry.oid)
for entry in clean_list]
tree = w.new_tree(shalist)
if stack:
stack[-1].append(item.name, GIT_MODE_TREE, GIT_MODE_TREE, tree, None)
return tree
# Hack around lack of nonlocal vars in python 2
_nonlocal = {}
_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))
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]
return None
total = ftotal = 0
if opt.progress:
for transname, ent in reader.filter(opt.sources,
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 reader.filter(opt.sources,
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, opt.sources)
elif opt.strip_path:
dirp = stripped_path_components(dir, [opt.strip_path])
elif opt.grafts:
dirp = grafted_path_components(opt.grafts, 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 [x.name for x in stack] > [x[0] for x in dirp]:
_pop()
# If switching to a new sub-tree, start a new sub-tree.
for path_component in dirp[len(stack):]:
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(stack) == 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:
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)
stack[-1].append(file, ent.mode, ent.gitmode, ent.sha, 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()
stack[-1].append(file, ent.mode, ent.gitmode, id, 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(stack) > 1: # _pop() all the parts above the root
_pop()
# Finish the root directory.
# When there's a collision, use empty metadata for the root.
tree = _pop(dir_metadata = metadata.Metadata() if root_collision else None)
return tree
def pfunc(count, total):
qprogress('Reading indexes: %.2f%% (%d/%d)\r' %
(count * 100.0 / total, count, total))
s = opt.find
try:
bin = s.decode('hex')
except TypeError:
o.fatal('--find parameter is not a valid hex string')
find = opt.find.lower()
count = 0
for name in extra:
try:
ix = git.open_idx(name)
except git.GitError as e:
add_error('%s: %s' % (name, e))
continue
if len(opt.find) == 40:
if ix.exists(bin):
print name, find
else:
# slow, exhaustive search
for _i in ix:
i = str(_i).encode('hex')
if i.startswith(find):
print name, i
qprogress('Searching: %d\r' % count)
count += 1
if saved_errors:
log('WARNING: %d errors encountered while saving.\n' % len(saved_errors))
sys.exit(1)
