#!/usr/bin/python """This tool reads a Debian package from stdin and emits a yaml stream on stdout. It does not access a database. Therefore it can be run in parallel and on multiple machines. The generated yaml contains multiple documents. The first document contains package metadata. Then a document is emitted for each file. And finally a document consisting of the string "commit" is emitted.""" import binascii import hashlib import optparse import sys import tarfile import zlib import lzma import yaml from dedup.arreader import ArReader from dedup.debpkg import process_control, get_tar_hashes from dedup.hashing import DecompressedHash, SuppressingHash, HashedStream, \ HashBlacklistContent from dedup.compression import GzipDecompressor, DecompressedStream from dedup.image import GIFHash, PNGHash boring_content = set(("", "\n")) def decompress_tar(filelike, extension): if extension in (".lzma", ".xz"): filelike = DecompressedStream(filelike, lzma.LZMADecompressor()) extension = "" if extension not in ("", ".gz", ".bz2"): raise ValueError("unknown compression format with extension %r" % extension) return tarfile.open(fileobj=filelike, mode="r|" + extension[1:]) def process_package(filelike): af = ArReader(filelike) af.read_magic() state = "start" record = None while True: try: name = af.read_entry() except EOFError: raise ValueError("control.tar not found") if name.startswith("control.tar"): if state != "start": raise ValueError("unexpected control.tar") state = "control" tf = decompress_tar(af, name[11:]) controldata = {} for elem in tf: if not elem.isreg(): continue elemname = elem.name if elemname.startswith("./"): elemname = elemname[2:] content = tf.extractfile(elem).read() if elemname in controldata: raise ValueError("duplicate entry %r in control.tar" % elemname) controldata[elemname] = binascii.b2a_base64(content).strip() if elemname == "control": record = process_control(content) if record is None: raise ValueError("control file missing from control.tar") record["data"] = controldata yield record yield "commit" break def process_package_with_hash(filelike, sha256hash): hstream = HashedStream(filelike, hashlib.sha256()) for elem in process_package(hstream): if elem == "commit": while hstream.read(4096): pass if hstream.hexdigest() != sha256hash: raise ValueError("hash sum mismatch") yield elem break yield elem def main(): parser = optparse.OptionParser() parser.add_option("-H", "--hash", action="store", help="verify that stdin hash given sha256 hash") options, args = parser.parse_args() if options.hash: gen = process_package_with_hash(sys.stdin, options.hash) else: gen = process_package(sys.stdin) yaml.safe_dump_all(gen, sys.stdout) if __name__ == "__main__": main()