diff options
Diffstat (limited to 'source4/scripting/bin')
-rwxr-xr-x | source4/scripting/bin/reorgldb.py | 60 | ||||
-rwxr-xr-x | source4/scripting/bin/setup_dns.sh | 1 | ||||
-rwxr-xr-x | source4/scripting/bin/w32err_code.py | 361 |
3 files changed, 362 insertions, 60 deletions
diff --git a/source4/scripting/bin/reorgldb.py b/source4/scripting/bin/reorgldb.py deleted file mode 100755 index 571363fdc7..0000000000 --- a/source4/scripting/bin/reorgldb.py +++ /dev/null @@ -1,60 +0,0 @@ -#!/usr/bin/python -# -# Copyright (C) Matthieu Patou <mat@matws.net> 2009 -# This script realize an offline reorganisation of an LDB -# file it helps to reduce (sometime drastically) the -# size of LDB files. -import sys -import optparse -import os -sys.path.insert(0, "bin/python") - -import samba -from samba.credentials import DONT_USE_KERBEROS -from samba.auth import system_session -from samba import Ldb, substitute_var, valid_netbios_name, check_all_substituted -from ldb import SCOPE_SUBTREE, SCOPE_ONELEVEL, SCOPE_BASE, LdbError -import ldb -import samba.getopt as options -from samba.samdb import SamDB -from samba import param -from samba.provision import ProvisionPaths, ProvisionNames,provision_paths_from_lp, Schema - -parser = optparse.OptionParser("provision [options]") -sambaopts = options.SambaOptions(parser) -parser.add_option_group(sambaopts) -parser.add_option_group(options.VersionOptions(parser)) -credopts = options.CredentialsOptions(parser) -parser.add_option_group(credopts) -parser.add_option("--database", type="string", metavar="FILE", - help="LDB to reorganize") -opts = parser.parse_args()[0] -lp = sambaopts.get_loadparm() -smbconf = lp.configfile - -if not opts.database: - print "Parameter database is mandatory" - sys.exit(1) -creds = credopts.get_credentials(lp) -creds.set_kerberos_state(DONT_USE_KERBEROS) -session = system_session() -empty = ldb.Message() -newname="%s.new"%(opts.database) -if os.path.exists(newname): - os.remove(newname) -old_ldb = Ldb(opts.database, session_info=session, credentials=creds,lp=lp) -new_ldb = Ldb(newname,session_info=session, credentials=creds,lp=lp) - -new_ldb.transaction_start() -res = old_ldb.search(expression="(dn=*)",base="", scope=SCOPE_SUBTREE) -for i in range(0,len(res)): - if str(res[i].dn) == "@BASEINFO": - continue - if str(res[i].dn).startswith("@INDEX:"): - continue - delta = new_ldb.msg_diff(empty,res[i]) - delta.dn = res[i].dn - delta.remove("distinguishedName") - new_ldb.add(delta) - -new_ldb.transaction_commit() diff --git a/source4/scripting/bin/setup_dns.sh b/source4/scripting/bin/setup_dns.sh index f296de2b8e..f20ad145c1 100755 --- a/source4/scripting/bin/setup_dns.sh +++ b/source4/scripting/bin/setup_dns.sh @@ -33,5 +33,6 @@ scripting/bin/nsupdate-gss --realm=$DOMAIN --noverify --ntype="CNAME" $OBJECTGUI exit 1 } echo "Checking" +rndc flush host $HOSTNAME.$DOMAIN host $OBJECTGUID._msdcs.$DOMAIN diff --git a/source4/scripting/bin/w32err_code.py b/source4/scripting/bin/w32err_code.py new file mode 100755 index 0000000000..cad6f6ecc9 --- /dev/null +++ b/source4/scripting/bin/w32err_code.py @@ -0,0 +1,361 @@ +#!/usr/bin/python + +# Unix SMB/CIFS implementation. +# Copyright (C) Kamen Mazdrashki <kamen.mazdrashki@postpath.com> 2009 +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +# + +"""Import generete werror.h/doserr.c files from WSPP HTML""" + +import re +import os +import sys +import urllib +import pprint +from xml.dom import minidom +from optparse import OptionParser, OptionGroup + +_wspp_werror_url = 'http://msdn.microsoft.com/en-us/library/cc231199%28PROT.10%29.aspx' + +class WerrorHtmlParser(object): + """ + Parses HTML from WSPP documentation generating dictionary of + dictionaries with following keys: + - "err_hex" - hex number (as string) + - "err_name" - error name + - "err_desc" - error long description + For the key of returned dictionary err_hex is used, + i.e. "hex-error-code-str" => {error dictionary object} + """ + + ERROR_PREFIX = ['ERROR_', 'NERR_', 'FRS_', 'RPC_', 'EPT_', 'OR_', 'WAIT_TIMEOUT'] + ERROR_REPLACE = ['ERROR_'] + + def __init__(self, opt): + self.opt = opt + self._errors_skipped = [] + pass + + def _is_error_code_name(self, err_name): + for pref in self.ERROR_PREFIX: + if err_name.startswith(pref): + return True + return False + + def _make_werr_name(self, err_name): + err_name = err_name.upper() + for pref in self.ERROR_REPLACE: + if err_name.startswith(pref): + return err_name.replace(pref, 'WERR_', 1) + return 'WERR_' + err_name + + def parse_url(self, url): + errors = {} + html = self._load_url(url) + + # let minidom to parse the tree, should be: + # table -> tr -> td + # p -> [hex code, br, error code] + # p -> [description] + table_node = minidom.parseString(html) + for row_node in table_node.getElementsByTagName("tr"): + # verify we got right number of td elements + td_nodes = row_node.getElementsByTagName('td') + if len(td_nodes) != 2: + continue + # now get the real data + p_nodes = row_node.getElementsByTagName('p') + if len(p_nodes) != 2: continue + if len(p_nodes[0].childNodes) != 3: continue + if len(p_nodes[1].childNodes) != 1: continue + err_hex = str(p_nodes[0].childNodes[0].nodeValue) + err_name = str(p_nodes[0].childNodes[2].nodeValue) + err_desc = p_nodes[1].childNodes[0].nodeValue.encode('utf-8') + err_desc = err_desc.replace('"', '\\"').replace("\'", "\\'") + # do some checking + if not err_hex.startswith('0x'): continue + if not self._is_error_code_name(err_name): + self._errors_skipped.append("%s - %s - %d" % (err_name, err_hex, int(err_hex, 16))) + continue + # create entry + err_name = self._make_werr_name(err_name) + err_def = {'err_hex': err_hex, + 'err_name': err_name, + 'err_desc': err_desc, + 'code': int(err_hex, 16)} + errors[err_def['code']] = err_def + + # print skipped errors + if self.opt.print_skipped and len(self._errors_skipped): + print "\nErrors skipped during HTML parsing:" + pprint.pprint(self._errors_skipped) + print "\n" + + return errors + + def _load_url(self, url): + html_str = "" + try: + fp = urllib.urlopen(url) + for line in fp: + html_str += line.strip() + fp.close() + except IOError, e: + print "error loading url: " + e.strerror + pass + + # currently ERROR codes are rendered as table + # locate table chunk with ERROR_SUCCESS + html = [x for x in html_str.split('<table ') if "ERROR_SUCCESS" in x] + html = '<table ' + html[0] + pos = html.find('</table>') + if pos == -1: + return ''; + html = html[:pos] + '</table>' + + # html clean up + html = re.sub(r'<a[^>]*>(.*?)</a>', r'\1', html) + + return html + + +class WerrorGenerator(object): + """ + provides methods to generate parts of werror.h and doserr.c files + """ + + FNAME_WERRORS = 'w32errors.lst' + FNAME_WERROR_DEFS = 'werror_defs.h' + FNAME_DOSERR_DEFS = 'doserr_defs.c' + FNAME_DOSERR_DESC = 'doserr_desc.c' + + def __init__(self, opt): + self.opt = opt + self._out_dir = opt.out_dir + pass + + def _open_out_file(self, fname): + fname = os.path.join(self._out_dir, fname) + return open(fname, "w") + + def _gen_werrors_list(self, errors): + """uses 'errors' dictionary to display list of Win32 Errors""" + + fp = self._open_out_file(self.FNAME_WERRORS) + for err_code in sorted(errors.keys()): + err_name = errors[err_code]['err_name'] + fp.write(err_name) + fp.write("\n") + fp.close() + + def _gen_werror_defs(self, errors): + """uses 'errors' dictionary to generate werror.h file""" + + fp = self._open_out_file(self.FNAME_WERROR_DEFS) + for err_code in sorted(errors.keys()): + err_name = errors[err_code]['err_name'] + err_hex = errors[err_code]['err_hex'] + fp.write('#define %s\tW_ERROR(%s)' % (err_name, err_hex)) + fp.write("\n") + fp.close() + + def _gen_doserr_defs(self, errors): + """uses 'errors' dictionary to generate defines in doserr.c file""" + + fp = self._open_out_file(self.FNAME_DOSERR_DEFS) + for err_code in sorted(errors.keys()): + err_name = errors[err_code]['err_name'] + fp.write('\t{ "%s", %s },' % (err_name, err_name)) + fp.write("\n") + fp.close() + + def _gen_doserr_descriptions(self, errors): + """uses 'errors' dictionary to generate descriptions in doserr.c file""" + + fp = self._open_out_file(self.FNAME_DOSERR_DESC) + for err_code in sorted(errors.keys()): + err_name = errors[err_code]['err_name'] + fp.write('\t{ %s, "%s" },' % (err_name, errors[err_code]['err_desc'])) + fp.write("\n") + fp.close() + + def _lookup_error_by_name(self, err_name, defined_errors): + for err in defined_errors.itervalues(): + if err['err_name'] == err_name: + return err + return None + + def _filter_errors(self, errors, defined_errors): + """ + returns tuple (new_erros, diff_code_errors, diff_name_errors) + new_errors - dictionary of errors not in defined_errors + diff_code_errors - list of errors found in defined_errors + but with different value + diff_name_errors - list of errors found with same code in + defined_errors, but with different name + Most critical is diff_code_errors list to be empty! + """ + new_errors = {} + diff_code_errors = [] + diff_name_errors = [] + for err_def in errors.itervalues(): + add_error = True + # try get defined error by code + if defined_errors.has_key(err_def['code']): + old_err = defined_errors[err_def['code']] + if err_def['err_name'] != old_err['err_name']: + warning = {'msg': 'New and Old errors has different error names', + 'err_new': err_def, + 'err_old': old_err} + diff_name_errors.append(warning) + + # sanity check for errors with same name but different values + old_err = self._lookup_error_by_name(err_def['err_name'], defined_errors) + if old_err: + if err_def['code'] != old_err['code']: + warning = {'msg': 'New and Old error defs has different error value', + 'err_new': err_def, + 'err_old': old_err} + diff_code_errors.append(warning) + # exclude error already defined with same name + add_error = False + # do add the error in new_errors if everything is fine + if add_error: + new_errors[err_def['code']] = err_def + pass + return (new_errors, diff_code_errors, diff_name_errors) + + def generate(self, errors): + # load already defined error codes + werr_parser = WerrorParser(self.opt) + (defined_errors, + no_value_errors) = werr_parser.load_err_codes(self.opt.werror_file) + if not defined_errors: + print "\nUnable to load existing errors file: %s" % self.opt.werror_file + sys.exit(1) + if self.opt.verbose and len(no_value_errors): + print "\nWarning: there are errors defines using macro value:" + pprint.pprint(no_value_errors) + print "" + # filter generated error codes + (new_errors, + diff_code_errors, + diff_name_errors) = self._filter_errors(errors, defined_errors) + if diff_code_errors: + print("\nFound %d errors with same names but different error values! Aborting." + % len(diff_code_errors)) + pprint.pprint(diff_code_errors) + sys.exit(2) + + if diff_name_errors: + print("\nFound %d errors with same values but different names (should be normal)" + % len(diff_name_errors)) + pprint.pprint(diff_name_errors) + + # finally generate output files + self._gen_werror_defs(new_errors) + self._gen_doserr_defs(new_errors) + self._gen_werrors_list(errors) + self._gen_doserr_descriptions(errors) + pass + +class WerrorParser(object): + """ + Parses errors defined in werror.h file + """ + + def __init__(self, opt): + self.opt = opt + pass + + def _parse_werror_line(self, line): + m = re.match('#define[ \t]*(.*?)[ \t]*W_ERROR\((.*?)\)', line) + if not m or (len(m.groups()) != 2): + return None + if len(m.group(1)) == 0: + return None + if str(m.group(2)).startswith('0x'): + err_code = int(m.group(2), 16) + elif m.group(2).isdigit(): + err_code = int(m.group(2)) + else: + self.err_no_values.append(line) + return None + return {'err_name': str(m.group(1)), + 'err_hex': "0x%08X" % err_code, + 'code': err_code} + pass + + def load_err_codes(self, fname): + """ + Returns tuple of: + dictionary of "hex_err_code" => {code, name} + "hex_err_code" is string + "code" is int value for the error + list of errors that was ignored for some reason + """ + # reset internal variables + self.err_no_values = [] + err_codes = {} + fp = open(fname) + for line in fp.readlines(): + err_def = self._parse_werror_line(line) + if err_def: + err_codes[err_def['code']] = err_def + fp.close(); + return (err_codes, self.err_no_values) + + + +def _generate_files(opt): + parser = WerrorHtmlParser(opt) + errors = parser.parse_url(opt.url) + + out = WerrorGenerator(opt) + out.generate(errors) + pass + + +if __name__ == '__main__': + _cur_dir = os.path.abspath(os.path.dirname(__file__)) + opt_parser = OptionParser(usage="usage: %prog [options]", version="%prog 0.3") + opt_group = OptionGroup(opt_parser, "Main options") + opt_group.add_option("--url", dest="url", + default=_wspp_werror_url, + help="url for w32 error codes html - may be local file") + opt_group.add_option("--out", dest="out_dir", + default=_cur_dir, + help="output dir for generated files") + opt_group.add_option("--werror", dest="werror_file", + default=os.path.join(_cur_dir, 'werror.h'), + help="path to werror.h file") + opt_group.add_option("--print_skipped", + action="store_true", dest="print_skipped", default=False, + help="print errors skipped during HTML parsing") + opt_group.add_option("-q", "--quiet", + action="store_false", dest="verbose", default=True, + help="don't print warnings to stdout") + + opt_parser.add_option_group(opt_group) + + (options, args) = opt_parser.parse_args() + + # add some options to be used internally + options.err_defs_file = os.path.join(options.out_dir, WerrorGenerator.FNAME_WERROR_DEFS) + options.dos_defs_file = os.path.join(options.out_dir, WerrorGenerator.FNAME_DOSERR_DEFS) + options.dos_desc_file = os.path.join(options.out_dir, WerrorGenerator.FNAME_DOSERR_DESC) + + # check options + _generate_files(options) |