mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-14 13:25:11 -05:00
0d9fbc1ad7
This version of SickBeard uses both TVDB and TVRage to search and gather it's series data from allowing you to now have access to and download shows that you couldn't before because of being locked into only what TheTVDB had to offer. Also this edition is based off the code we used in our XEM editon so it does come with scene numbering support as well as all the other features our XEM edition has to offer. Please before using this with your existing database (sickbeard.db) please make a backup copy of it and delete any other database files such as cache.db and failed.db if present, we HIGHLY recommend starting out with no database files at all to make this a fresh start but the choice is at your own risk! Enjoy!
268 lines
8.4 KiB
Python
268 lines
8.4 KiB
Python
"""
|
|
RPM archive parser.
|
|
|
|
Author: Victor Stinner, 1st December 2005.
|
|
"""
|
|
|
|
from lib.hachoir_parser import Parser
|
|
from lib.hachoir_core.field import (FieldSet, ParserError,
|
|
UInt8, UInt16, UInt32, UInt64, Enum,
|
|
NullBytes, Bytes, RawBytes, SubFile,
|
|
Character, CString, String)
|
|
from lib.hachoir_core.endian import BIG_ENDIAN
|
|
from lib.hachoir_parser.archive.gzip_parser import GzipParser
|
|
from lib.hachoir_parser.archive.bzip2_parser import Bzip2Parser
|
|
|
|
class ItemContent(FieldSet):
|
|
format_type = {
|
|
0: UInt8,
|
|
1: Character,
|
|
2: UInt8,
|
|
3: UInt16,
|
|
4: UInt32,
|
|
5: UInt64,
|
|
6: CString,
|
|
7: RawBytes,
|
|
8: CString,
|
|
9: CString
|
|
}
|
|
|
|
def __init__(self, parent, name, item):
|
|
FieldSet.__init__(self, parent, name, item.description)
|
|
self.related_item = item
|
|
self._name = "content_%s" % item.name
|
|
|
|
def createFields(self):
|
|
item = self.related_item
|
|
type = item["type"].value
|
|
|
|
cls = self.format_type[type]
|
|
count = item["count"].value
|
|
if cls is RawBytes: # or type == 8:
|
|
if cls is RawBytes:
|
|
args = (self, "value", count)
|
|
else:
|
|
args = (self, "value") # cls is CString
|
|
count = 1
|
|
else:
|
|
if 1 < count:
|
|
args = (self, "value[]")
|
|
else:
|
|
args = (self, "value")
|
|
for index in xrange(count):
|
|
yield cls(*args)
|
|
|
|
class Item(FieldSet):
|
|
type_name = {
|
|
0: "NULL",
|
|
1: "CHAR",
|
|
2: "INT8",
|
|
3: "INT16",
|
|
4: "INT32",
|
|
5: "INT64",
|
|
6: "CSTRING",
|
|
7: "BIN",
|
|
8: "CSTRING_ARRAY",
|
|
9: "CSTRING?"
|
|
}
|
|
tag_name = {
|
|
1000: "File size",
|
|
1001: "(Broken) MD5 signature",
|
|
1002: "PGP 2.6.3 signature",
|
|
1003: "(Broken) MD5 signature",
|
|
1004: "MD5 signature",
|
|
1005: "GnuPG signature",
|
|
1006: "PGP5 signature",
|
|
1007: "Uncompressed payload size (bytes)",
|
|
256+8: "Broken SHA1 header digest",
|
|
256+9: "Broken SHA1 header digest",
|
|
256+13: "Broken SHA1 header digest",
|
|
256+11: "DSA header signature",
|
|
256+12: "RSA header signature"
|
|
}
|
|
|
|
def __init__(self, parent, name, description=None, tag_name_dict=None):
|
|
FieldSet.__init__(self, parent, name, description)
|
|
if tag_name_dict is None:
|
|
tag_name_dict = Item.tag_name
|
|
self.tag_name_dict = tag_name_dict
|
|
|
|
def createFields(self):
|
|
yield Enum(UInt32(self, "tag", "Tag"), self.tag_name_dict)
|
|
yield Enum(UInt32(self, "type", "Type"), Item.type_name)
|
|
yield UInt32(self, "offset", "Offset")
|
|
yield UInt32(self, "count", "Count")
|
|
|
|
def createDescription(self):
|
|
return "Item: %s (%s)" % (self["tag"].display, self["type"].display)
|
|
|
|
class ItemHeader(Item):
|
|
tag_name = {
|
|
61: "Current image",
|
|
62: "Signatures",
|
|
63: "Immutable",
|
|
64: "Regions",
|
|
100: "I18N string locales",
|
|
1000: "Name",
|
|
1001: "Version",
|
|
1002: "Release",
|
|
1003: "Epoch",
|
|
1004: "Summary",
|
|
1005: "Description",
|
|
1006: "Build time",
|
|
1007: "Build host",
|
|
1008: "Install time",
|
|
1009: "Size",
|
|
1010: "Distribution",
|
|
1011: "Vendor",
|
|
1012: "Gif",
|
|
1013: "Xpm",
|
|
1014: "Licence",
|
|
1015: "Packager",
|
|
1016: "Group",
|
|
1017: "Changelog",
|
|
1018: "Source",
|
|
1019: "Patch",
|
|
1020: "Url",
|
|
1021: "OS",
|
|
1022: "Arch",
|
|
1023: "Prein",
|
|
1024: "Postin",
|
|
1025: "Preun",
|
|
1026: "Postun",
|
|
1027: "Old filenames",
|
|
1028: "File sizes",
|
|
1029: "File states",
|
|
1030: "File modes",
|
|
1031: "File uids",
|
|
1032: "File gids",
|
|
1033: "File rdevs",
|
|
1034: "File mtimes",
|
|
1035: "File MD5s",
|
|
1036: "File link to's",
|
|
1037: "File flags",
|
|
1038: "Root",
|
|
1039: "File username",
|
|
1040: "File groupname",
|
|
1043: "Icon",
|
|
1044: "Source rpm",
|
|
1045: "File verify flags",
|
|
1046: "Archive size",
|
|
1047: "Provide name",
|
|
1048: "Require flags",
|
|
1049: "Require name",
|
|
1050: "Require version",
|
|
1051: "No source",
|
|
1052: "No patch",
|
|
1053: "Conflict flags",
|
|
1054: "Conflict name",
|
|
1055: "Conflict version",
|
|
1056: "Default prefix",
|
|
1057: "Build root",
|
|
1058: "Install prefix",
|
|
1059: "Exclude arch",
|
|
1060: "Exclude OS",
|
|
1061: "Exclusive arch",
|
|
1062: "Exclusive OS",
|
|
1064: "RPM version",
|
|
1065: "Trigger scripts",
|
|
1066: "Trigger name",
|
|
1067: "Trigger version",
|
|
1068: "Trigger flags",
|
|
1069: "Trigger index",
|
|
1079: "Verify script",
|
|
#TODO: Finish the list (id 1070..1162 using rpm library source code)
|
|
}
|
|
|
|
def __init__(self, parent, name, description=None):
|
|
Item.__init__(self, parent, name, description, self.tag_name)
|
|
|
|
def sortRpmItem(a,b):
|
|
return int( a["offset"].value - b["offset"].value )
|
|
|
|
class PropertySet(FieldSet):
|
|
def __init__(self, parent, name, *args):
|
|
FieldSet.__init__(self, parent, name, *args)
|
|
self._size = self["content_item[1]"].address + self["size"].value * 8
|
|
|
|
def createFields(self):
|
|
# Read chunk header
|
|
yield Bytes(self, "signature", 3, r"Property signature (\x8E\xAD\xE8)")
|
|
if self["signature"].value != "\x8E\xAD\xE8":
|
|
raise ParserError("Invalid property signature")
|
|
yield UInt8(self, "version", "Signature version")
|
|
yield NullBytes(self, "reserved", 4, "Reserved")
|
|
yield UInt32(self, "count", "Count")
|
|
yield UInt32(self, "size", "Size")
|
|
|
|
# Read item header
|
|
items = []
|
|
for i in range(0, self["count"].value):
|
|
item = ItemHeader(self, "item[]")
|
|
yield item
|
|
items.append(item)
|
|
|
|
# Sort items by their offset
|
|
items.sort( sortRpmItem )
|
|
|
|
# Read item content
|
|
start = self.current_size/8
|
|
for item in items:
|
|
offset = item["offset"].value
|
|
diff = offset - (self.current_size/8 - start)
|
|
if 0 < diff:
|
|
yield NullBytes(self, "padding[]", diff)
|
|
yield ItemContent(self, "content[]", item)
|
|
size = start + self["size"].value - self.current_size/8
|
|
if 0 < size:
|
|
yield NullBytes(self, "padding[]", size)
|
|
|
|
class RpmFile(Parser):
|
|
PARSER_TAGS = {
|
|
"id": "rpm",
|
|
"category": "archive",
|
|
"file_ext": ("rpm",),
|
|
"mime": (u"application/x-rpm",),
|
|
"min_size": (96 + 16 + 16)*8, # file header + checksum + content header
|
|
"magic": (('\xED\xAB\xEE\xDB', 0),),
|
|
"description": "RPM package"
|
|
}
|
|
TYPE_NAME = {
|
|
0: "Binary",
|
|
1: "Source"
|
|
}
|
|
endian = BIG_ENDIAN
|
|
|
|
def validate(self):
|
|
if self["signature"].value != '\xED\xAB\xEE\xDB':
|
|
return "Invalid signature"
|
|
if self["major_ver"].value != 3:
|
|
return "Unknown major version (%u)" % self["major_ver"].value
|
|
if self["type"].value not in self.TYPE_NAME:
|
|
return "Invalid RPM type"
|
|
return True
|
|
|
|
def createFields(self):
|
|
yield Bytes(self, "signature", 4, r"RPM file signature (\xED\xAB\xEE\xDB)")
|
|
yield UInt8(self, "major_ver", "Major version")
|
|
yield UInt8(self, "minor_ver", "Minor version")
|
|
yield Enum(UInt16(self, "type", "RPM type"), RpmFile.TYPE_NAME)
|
|
yield UInt16(self, "architecture", "Architecture")
|
|
yield String(self, "name", 66, "Archive name", strip="\0", charset="ASCII")
|
|
yield UInt16(self, "os", "OS")
|
|
yield UInt16(self, "signature_type", "Type of signature")
|
|
yield NullBytes(self, "reserved", 16, "Reserved")
|
|
yield PropertySet(self, "checksum", "Checksum (signature)")
|
|
yield PropertySet(self, "header", "Header")
|
|
|
|
if self._size is None: # TODO: is it possible to handle piped input?
|
|
raise NotImplementedError
|
|
|
|
size = (self._size - self.current_size) // 8
|
|
if size:
|
|
if 3 <= size and self.stream.readBytes(self.current_size, 3) == "BZh":
|
|
yield SubFile(self, "content", size, "bzip2 content", parser=Bzip2Parser)
|
|
else:
|
|
yield SubFile(self, "content", size, "gzip content", parser=GzipParser)
|
|
|