mirror of
https://github.com/Mikaela/Limnoria.git
synced 2024-12-11 20:59:42 +01:00
303 lines
12 KiB
Python
303 lines
12 KiB
Python
#!/usr/bin/env python
|
|
|
|
###
|
|
# Copyright (c) 2003, Daniel Berlin
|
|
# Based on code from kibot
|
|
# All rights reserved.
|
|
#
|
|
# Redistribution and use in source and binary forms, with or without
|
|
# modification, are permitted provided that the following conditions are met:
|
|
#
|
|
# * Redistributions of source code must retain the above copyright notice,
|
|
# this list of conditions, and the following disclaimer.
|
|
# * Redistributions in binary form must reproduce the above copyright notice,
|
|
# this list of conditions, and the following disclaimer in the
|
|
# documentation and/or other materials provided with the distribution.
|
|
# * Neither the name of the author of this software nor the name of
|
|
# contributors to this software may be used to endorse or promote products
|
|
# derived from this software without specific prior written consent.
|
|
#
|
|
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
|
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
# POSSIBILITY OF SUCH DAMAGE.
|
|
###
|
|
|
|
"""
|
|
Bugzilla bug retriever
|
|
"""
|
|
|
|
import os
|
|
import re
|
|
import string
|
|
import urllib2
|
|
import xml.dom.minidom as minidom
|
|
from itertools import imap, ifilter
|
|
from htmlentitydefs import entitydefs as entities
|
|
|
|
import conf
|
|
import utils
|
|
import plugins
|
|
import ircutils
|
|
import privmsgs
|
|
import callbacks
|
|
import structures
|
|
|
|
dbfilename = os.path.join(conf.dataDir, 'Bugzilla.db')
|
|
def makeDb(filename):
|
|
if os.path.exists(filename):
|
|
d = structures.PersistentDictionary(filename)
|
|
else:
|
|
d = structures.PersistentDictionary(filename)
|
|
d['gcc'] = ['http://gcc.gnu.org/bugzilla', 'GCC']
|
|
d['rh'] = ['http://bugzilla.redhat.com/bugzilla', 'Red Hat']
|
|
d['gnome'] = ['http://bugzilla.gnome.org/bugzilla', 'Gnome']
|
|
d['mozilla'] = ['http://bugzilla.mozilla.org', 'Mozilla']
|
|
d['ximian'] = ['http://bugzilla.ximian.com/bugzilla', 'Ximian Gnome']
|
|
d.flush()
|
|
return d
|
|
|
|
class BugzillaError(Exception):
|
|
"""A bugzilla error"""
|
|
pass
|
|
|
|
def configure(onStart, afterConnect, advanced):
|
|
from questions import expect, anything, yn
|
|
onStart.append('load Bugzilla')
|
|
if advanced:
|
|
print 'The Bugzilla plugin has the functionality to watch for URLs'
|
|
print 'that match a specific pattern (we call this a snarfer). When'
|
|
print 'supybot sees such a URL, he will parse the web page for'
|
|
print 'information and reply with the results.\n'
|
|
if yn('Do you want the Bugzilla snarfer enabled by default?') == 'n':
|
|
onStart.append('Bugzilla config bug-snarfer off')
|
|
|
|
replyNoBugzilla = 'I don\'t have a bugzilla %r'
|
|
|
|
class Bugzilla(callbacks.PrivmsgCommandAndRegexp, plugins.Configurable):
|
|
"""Show a link to a bug report with a brief description"""
|
|
threaded = True
|
|
regexps = ['bzSnarfer']
|
|
configurables = plugins.ConfigurableDictionary(
|
|
[('bug-snarfer', plugins.ConfigurableBoolType, True,
|
|
"""Determines whether the bug snarfer will be enabled, such that any
|
|
Bugzilla URLs seen in the channel will have their information reported
|
|
into the channel.""")]
|
|
)
|
|
def __init__(self):
|
|
plugins.Configurable.__init__(self)
|
|
callbacks.PrivmsgCommandAndRegexp.__init__(self)
|
|
self.entre = re.compile('&(\S*?);')
|
|
# Schema: {name, [url, description]}
|
|
self.db = makeDb(dbfilename)
|
|
self.shorthand = utils.abbrev(self.db.keys())
|
|
|
|
def die(self):
|
|
plugins.Configurable.die(self)
|
|
self.db.close()
|
|
del self.db
|
|
|
|
def add(self, irc, msg, args):
|
|
"""<name> <url> <description>
|
|
|
|
Add a bugzilla <url> to the list of defined bugzillae. <name>
|
|
is the name that will be used to reference the zilla in all
|
|
commands. Unambiguous abbreviations of <name> will be accepted also.
|
|
<description> is the common name for the bugzilla and will
|
|
be listed with the bugzilla query.
|
|
"""
|
|
(name, url, description) = privmsgs.getArgs(args, required=3)
|
|
cursor = self.db.cursor()
|
|
self.db[name] = [url, description]
|
|
self.shorthand = utils.abbrev(self.db.keys())
|
|
irc.reply(msg, conf.replySuccess)
|
|
|
|
def remove(self, irc, msg, args):
|
|
"""<abbreviation>
|
|
|
|
Remove the bugzilla associated with <abbreviation> from the list of
|
|
defined bugzillae.
|
|
"""
|
|
name = privmsgs.getArgs(args)
|
|
try:
|
|
name = self.shorthand[name]
|
|
del self.db[name]
|
|
self.shorthand = utils.abbrev(self.db.keys())
|
|
irc.reply(msg, conf.replySuccess)
|
|
except KeyError:
|
|
irc.error(msg, replyNoBugzilla % name)
|
|
|
|
def list(self, irc, msg, args):
|
|
"""[<abbreviation>]
|
|
|
|
List defined bugzillae. If <abbreviation> is specified, list the
|
|
information for that bugzilla.
|
|
"""
|
|
name = privmsgs.getArgs(args, required=0, optional=1)
|
|
if name:
|
|
try:
|
|
name = self.shorthand[name]
|
|
(url, description) = self.db[name]
|
|
irc.reply(msg, '%s: %s, %s' % (name, description, url))
|
|
except KeyError:
|
|
irc.error(msg, replyNoBugzilla % name)
|
|
else:
|
|
if self.db:
|
|
L = self.db.keys()
|
|
L.sort()
|
|
irc.reply(msg, utils.commaAndify(L))
|
|
else:
|
|
irc.reply(msg, 'I have no defined bugzillae.')
|
|
|
|
def bzSnarfer(self, irc, msg, match):
|
|
r"(http://\S+)/show_bug.cgi\?id=([0-9]+)"
|
|
if not self.configurables.get('bug-snarfer', channel=msg.args[0]):
|
|
return
|
|
queryurl = '%s/xml.cgi?id=%s' % (match.group(1), match.group(2))
|
|
try:
|
|
summary = self._get_short_bug_summary(queryurl, 'Snarfed '\
|
|
'Bugzilla URL', match.group(2))
|
|
except BugzillaError, e:
|
|
irc.reply(msg, str(e))
|
|
return
|
|
except IOError, e:
|
|
msgtouser = '%s. Try yourself: %s' % (e, queryurl)
|
|
irc.reply(msg, msgtouser)
|
|
return
|
|
report = {}
|
|
report['id'] = match.group(2)
|
|
report['url'] = str('%s/show_bug.cgi?id=%s' % (match.group(1),
|
|
match.group(2)))
|
|
report['title'] = str(summary['title'])
|
|
report['summary'] = str(self._mk_summary_string(summary))
|
|
report['product'] = str(summary['product'])
|
|
s = '%(product)s bug #%(id)s: %(title)s %(summary)s' % report
|
|
irc.reply(msg, s, prefixName=False)
|
|
bzSnarfer = privmsgs.urlSnarfer(bzSnarfer)
|
|
|
|
def bug(self, irc, msg, args):
|
|
"""<abbreviation> <number>
|
|
|
|
Look up bug <number> in the bugzilla associated with <abbreviation>.
|
|
"""
|
|
(name, number) = privmsgs.getArgs(args, required=2)
|
|
try:
|
|
name = self.shorthand[name]
|
|
(url, description) = self.db[name]
|
|
except KeyError:
|
|
irc.error(msg, replyNoBugzilla % name)
|
|
return
|
|
queryurl = '%s/xml.cgi?id=%s' % (url, number)
|
|
try:
|
|
summary = self._get_short_bug_summary(queryurl,description,number)
|
|
except BugzillaError, e:
|
|
irc.error(msg, str(e))
|
|
return
|
|
except IOError, e:
|
|
s = '%s. Try yourself: %s' % (e, queryurl)
|
|
irc.error(msg, s)
|
|
report = {}
|
|
report['zilla'] = description
|
|
report['id'] = number
|
|
report['url'] = '%s/show_bug.cgi?id=%s' % (url, number)
|
|
report['title'] = str(summary['title'])
|
|
report['summary'] = self._mk_summary_string(summary)
|
|
s = '%(zilla)s bug #%(id)s: %(title)s %(summary)s %(url)s' % report
|
|
irc.reply(msg, s)
|
|
|
|
def _mk_summary_string(self, summary):
|
|
L = []
|
|
if 'product' in summary:
|
|
L.append(ircutils.bold('Product: ') + summary['product'])
|
|
if 'component' in summary:
|
|
L.append(ircutils.bold('Component: ') + summary['component'])
|
|
if 'severity' in summary:
|
|
L.append(ircutils.bold('Severity: ') + summary['severity'])
|
|
if 'assigned to' in summary:
|
|
L.append(ircutils.bold('Assigned to: ') + summary['assigned to'])
|
|
if 'status' in summary:
|
|
L.append(ircutils.bold('Status: ') + summary['status'])
|
|
if 'resolution' in summary:
|
|
L.append(ircutils.bold('Resolution: ') + summary['resolution'])
|
|
return ', '.join(imap(str, L))
|
|
|
|
def _get_short_bug_summary(self, url, desc, number):
|
|
bugxml = self._getbugxml(url, desc)
|
|
try:
|
|
zilladom = minidom.parseString(bugxml)
|
|
except Exception, e:
|
|
s = 'Could not parse XML returned by %s bugzilla: %s' % (desc, e)
|
|
raise BugzillaError, s
|
|
bug_n = zilladom.getElementsByTagName('bug')[0]
|
|
if bug_n.hasAttribute('error'):
|
|
errtxt = bug_n.getAttribute('error')
|
|
s = 'Error getting %s bug #%s: %s' % (desc, number, errtxt)
|
|
raise BugzillaError, s
|
|
summary = {}
|
|
try:
|
|
node = bug_n.getElementsByTagName('short_desc')[0]
|
|
summary['title'] = self._getnodetxt(node)
|
|
node = bug_n.getElementsByTagName('bug_status')[0]
|
|
summary['status'] = self._getnodetxt(node)
|
|
try:
|
|
node = bug_n.getElementsByTagName('resolution')[0]
|
|
summary['resolution'] = self._getnodetxt(node)
|
|
except:
|
|
pass
|
|
node = bug_n.getElementsByTagName('assigned_to')[0]
|
|
summary['assigned to'] = self._getnodetxt(node)
|
|
node = bug_n.getElementsByTagName('product')[0]
|
|
summary['product'] = self._getnodetxt(node)
|
|
node = bug_n.getElementsByTagName('component')[0]
|
|
summary['component'] = self._getnodetxt(node)
|
|
node = bug_n.getElementsByTagName('bug_severity')[0]
|
|
summary['severity'] = self._getnodetxt(node)
|
|
except Exception, e:
|
|
s = 'Could not parse XML returned by %s bugzilla: %s' % (desc, e)
|
|
raise BugzillaError, s
|
|
return summary
|
|
|
|
def _getbugxml(self, url, desc):
|
|
try:
|
|
fh = urllib2.urlopen(url)
|
|
except urllib2.UrlError, e:
|
|
raise IOError, 'Connection to %s bugzilla failed' % desc
|
|
bugxml = fh.read()
|
|
fh.close()
|
|
if not bugxml:
|
|
raise IOError, 'Error getting bug content from %s' % desc
|
|
return bugxml
|
|
|
|
def _getnodetxt(self, node):
|
|
L = []
|
|
for childnode in node.childNodes:
|
|
if childnode.nodeType == childnode.TEXT_NODE:
|
|
L.append(childnode.data)
|
|
val = ''.join(L)
|
|
if node.hasAttribute('encoding'):
|
|
encoding = node.getAttribute('encoding')
|
|
if encoding == 'base64':
|
|
try:
|
|
val = val.decode('base64')
|
|
except:
|
|
val = 'Cannot convert bug data from base64.'
|
|
while self.entre.search(val):
|
|
entity = self.entre.search(val).group(1)
|
|
if entity in entities:
|
|
val = self.entre.sub(entities[entity], val)
|
|
else:
|
|
val = self.entre.sub('?', val)
|
|
return val
|
|
|
|
|
|
Class = Bugzilla
|
|
|
|
# vim:set shiftwidth=4 tabstop=8 expandtab textwidth=78:
|