# -*- coding: utf-8 -*- from datetime import date import md5 from pprint import pformat import re from urllib2 import HTTPError import requests import osc.core import ReviewBot from osclib.comments import CommentAPI from suse import SUSEUpdate try: from xml.etree import cElementTree as ET except ImportError: from xml.etree import ElementTree as ET try: import simplejson as json except ImportError: import json QA_UNKNOWN = 0 QA_INPROGRESS = 1 QA_FAILED = 2 QA_PASSED = 3 pkgname_re = re.compile(r'(?P.+)-(?P[^-]+)-(?P[^-]+)\.(?P[^.]+)\.rpm') comment_marker_re = re.compile( r'') class OpenQABot(ReviewBot.ReviewBot): """ check ABI of library packages """ def __init__(self, *args, **kwargs): super(OpenQABot, self).__init__(*args, **kwargs) self.tgt_repo = {} self.project_settings = {} self.api_map = {} self.force = False self.openqa = None self.commentapi = CommentAPI(self.apiurl) self.update_test_builds = {} self.pending_target_repos = set() self.openqa_jobs = {} def gather_test_builds(self): for prj, u in self.tgt_repo[self.openqa.baseurl].items(): buildnr = 0 cjob = 0 for j in self.jobs_for_target(u): # avoid going backwards in job ID if cjob > int(j['id']): continue buildnr = j['settings']['BUILD'] cjob = int(j['id']) self.update_test_builds[prj] = buildnr jobs = self.jobs_for_target(u, build=buildnr) self.openqa_jobs[prj] = jobs if self.calculate_qa_status(jobs) == QA_INPROGRESS: self.pending_target_repos.add(prj) # reimplemention from baseclass def check_requests(self): if self.ibs: self.check_suse_incidents() else: self.check_opensuse_incidents() # first calculate the latest build number for current jobs self.gather_test_builds() started = [] # then check progress on running incidents for req in self.requests: jobs = self.request_get_openqa_jobs(req, incident=True, test_repo=True) ret = self.calculate_qa_status(jobs) if ret != QA_UNKNOWN: started.append(req) all_requests = self.requests self.requests = started self.logger.debug("check started requests") super(OpenQABot, self).check_requests() self.requests = all_requests skipped_one = False # now make sure the jobs are for current repo for prj, u in self.tgt_repo[self.openqa.baseurl].items(): if prj in self.pending_target_repos: skipped_one = True continue self.trigger_build_for_target(prj, u) # do not schedule new incidents unless we finished # last wave if skipped_one: return self.logger.debug("Check all requests") super(OpenQABot, self).check_requests() # check a set of repos for their primary checksums @staticmethod def calculate_repo_hash(repos): m = md5.new() # if you want to force it, increase this number m.update('b') for url in repos: url += '/repodata/repomd.xml' try: root = ET.parse(osc.core.http_GET(url)).getroot() except HTTPError: raise cs = root.find( './/{http://linux.duke.edu/metadata/repo}data[@type="primary"]/{http://linux.duke.edu/metadata/repo}checksum') m.update(cs.text) return m.hexdigest() def is_incident_in_testing(self, incident): # hard coded for now as we only run this code for SUSE Maintenance workflow project = 'SUSE:Maintenance:{}'.format(incident) xpath = "(state/@name='review') and (action/source/@project='{}' and action/@type='maintenance_release')".format(project) res = osc.core.search(self.apiurl, request=xpath)['request'] # return the one and only (or None) return res.find('request') def calculate_incidents(self, incidents): """ get incident numbers from SUSE:Maintenance:Test project returns dict with openQA var name : string with numbers """ self.logger.debug("calculate_incidents: {}".format(pformat(incidents))) l_incidents = [] for kind, prj in incidents.items(): packages = osc.core.meta_get_packagelist(self.apiurl, prj) incidents = [] # filter out incidents in staging for incident in packages: # remove patchinfo. prefix incident = incident.replace('_', '.').split('.')[1] req = self.is_incident_in_testing(incident) # without release request it's in staging if not req: continue # skip kgraft patches from aggregation req_ = osc.core.Request() req_.read(req) src_prjs = {a.src_project for a in req_.actions} if SUSEUpdate.kgraft_target(self.apiurl, src_prjs.pop()): self.logger.debug("calculate_incidents: Incident is kgraft - {} ".format(incident)) continue incidents.append(incident) l_incidents.append((kind + '_TEST_ISSUES', ','.join(incidents))) self.logger.debug("Calculate incidents:{}".format(pformat(l_incidents))) return l_incidents def jobs_for_target(self, data, build=None): settings = data['settings'] values = { 'distri': settings['DISTRI'], 'version': settings['VERSION'], 'arch': settings['ARCH'], 'flavor': settings['FLAVOR'], 'scope': 'relevant', 'latest': '1', } if build: values['build'] = build else: values['test'] = data['test'] self.logger.debug("Get jobs: {}".format(pformat(values))) return self.openqa.openqa_request('GET', 'jobs', values)['jobs'] # we don't know the current BUILD and querying all jobs is too expensive # so we need to check for one known TEST first # if that job doesn't contain the proper hash, we trigger a new one # and then we know the build def trigger_build_for_target(self, prj, data): today = date.today().strftime("%Y%m%d") try: repohash = self.calculate_repo_hash(data['repos']) except HTTPError as e: self.logger.debug("REPOHAS not calculated with response {}".format(e)) return buildnr = None jobs = self.jobs_for_target(data) for job in jobs: if job['settings'].get('REPOHASH', '') == repohash: # take the last in the row buildnr = job['settings']['BUILD'] self.update_test_builds[prj] = buildnr # ignore old build numbers, we want a fresh run every day # to find regressions in the tests and to get data about # randomly failing tests if buildnr and buildnr.startswith(today): return buildnr = 0 # not found, then check for the next free build nr for job in jobs: build = job['settings']['BUILD'] if build and build.startswith(today): try: nr = int(build.split('-')[1]) if nr > buildnr: buildnr = nr except ValueError: continue buildnr = "{!s}-{:d}".format(today, buildnr + 1) s = data['settings']: # now schedule it for real if 'incidents' in data.keys(): for x, y in self.calculate_incidents(data['incidents']): s[x] = y s['BUILD'] = buildnr s['REPOHASH'] = repohash self.logger.debug("Prepared: {}".format(pformat(s))) if not self.dryrun: try: self.logger.info("Openqa isos POST {}".format(pformat(s))) self.openqa.openqa_request('POST', 'isos', data=s, retries=1) except Exception as e: self.logger.error(e) self.update_test_builds[prj] = buildnr def request_get_openqa_jobs(self, req, incident=True, test_repo=False): ret = None types = {a.type for a in req.actions} if 'maintenance_release' in types: src_prjs = {a.src_project for a in req.actions} if len(src_prjs) != 1: raise Exception("can't handle maintenance_release from different incidents") build = src_prjs.pop() tgt_prjs = {a.tgt_project for a in req.actions} ret = [] if incident: ret += self.openqa_jobs[build] for prj in sorted(tgt_prjs): repo_settings = self.tgt_repo.get(self.openqa.baseurl, {}) if test_repo and prj in repo_settings: repo_jobs = self.openqa_jobs[prj] ret += repo_jobs return ret def calculate_qa_status(self, jobs=None): if not jobs: return QA_UNKNOWN j = {} has_failed = False in_progress = False for job in jobs: if job['clone_id']: continue name = job['name'] if name in j and int(job['id']) < int(j[name]['id']): continue j[name] = job if job['state'] not in ('cancelled', 'done'): in_progress = True else: if job['result'] != 'passed' and job['result'] != 'softfailed': has_failed = True if not j: return QA_UNKNOWN if in_progress: return QA_INPROGRESS if has_failed: return QA_FAILED return QA_PASSED def add_comment(self, msg, state, request_id=None, result=None): if not self.do_comments: return comment = "\n".format(state, ' result={!s}'.format(result) if result else '') comment += "\n" + msg info = self.find_obs_request_comment(request_id=request_id) comment_id = info.get('id', None) if state == info.get('state', 'missing'): lines_before = len(info['comment'].split('\n')) lines_after = len(comment.split('\n')) if lines_before == lines_after: self.logger.info("not worth the update, previous comment %s is state %s", comment_id, info['state']) return self.logger.info("adding comment to %s, state %s result %s", request_id, state, result) self.logger.info("message: %s", msg) if not self.dryrun: if comment_id: self.commentapi.delete(comment_id) self.commentapi.add_comment(request_id=request_id, comment=str(comment)) # escape markdown @staticmethod def emd(str): return str.replace('_', r'\_') @staticmethod def get_step_url(testurl, modulename): failurl = testurl + '/modules/{!s}/fails'.format(modulename) fails = requests.get(failurl).json() failed_step = fails.get('first_failed_step', 1) return "[{!s}]({!s}#step/{!s}/{:d})".format(OpenQABot.emd(modulename), testurl, modulename, failed_step) @staticmethod def job_test_name(job): return "{!s}@{!s}".format(OpenQABot.emd(job['settings']['TEST']), OpenQABot.emd(job['settings']['MACHINE'])) def summarize_one_openqa_job(self, job): testurl = osc.core.makeurl(self.openqa.baseurl, ['tests', str(job['id'])]) if not job['result'] in ['passed', 'failed', 'softfailed']: rstring = job['result'] if rstring == 'none': return None return '\n- [{!s}]({!s}) is {!s}'.format(self.job_test_name(job), testurl, rstring) modstrings = [] for module in job['modules']: if module['result'] != 'failed': continue modstrings.append(self.get_step_url(testurl, module['name'])) if modstrings: return '\n- [{!s}]({!s}) failed in {!s}'.format(self.job_test_name(job), testurl, ','.join(modstrings)) elif job['result'] == 'failed': # rare case: fail without module fails return '\n- [{!s}]({!s}) failed'.format(self.job_test_name(job), testurl) return '' def summarize_openqa_jobs(self, jobs): groups = {} for job in jobs: gl = "{!s}@{!s}".format(self.emd(job['group']), self.emd(job['settings']['FLAVOR'])) if gl not in groups: groupurl = osc.core.makeurl(self.openqa.baseurl, ['tests', 'overview'], {'version': job['settings']['VERSION'], 'groupid': job['group_id'], 'flavor': job['settings']['FLAVOR'], 'distri': job['settings']['DISTRI'], 'build': job['settings']['BUILD'], }) groups[gl] = {'title': "__Group [{!s}]({!s})__\n".format(gl, groupurl), 'passed': 0, 'unfinished': 0, 'failed': []} job_summary = self.summarize_one_openqa_job(job) if job_summary is None: groups[gl]['unfinished'] = groups[gl]['unfinished'] + 1 continue # None vs '' if not len(job_summary): groups[gl]['passed'] = groups[gl]['passed'] + 1 continue # if there is something to report, hold the request # TODO: what is this ? # qa_state = QA_FAILED # gmsg = groups[gl] groups[gl]['failed'].append(job_summary) msg = '' for group in sorted(groups.keys()): msg += "\n\n" + groups[group]['title'] infos = [] if groups[group]['passed']: infos.append("{:d} tests passed".format(groups[group]['passed'])) if len(groups[group]['failed']): infos.append("{:d} tests failed".format(len(groups[group]['failed']))) if groups[group]['unfinished']: infos.append("{:d} unfinished tests".format(groups[group]['unfinished'])) msg += "(" + ', '.join(infos) + ")\n" for fail in groups[group]['failed']: msg += fail return msg.rstrip('\n') def check_one_request(self, req): ret = None try: jobs = self.request_get_openqa_jobs(req) qa_state = self.calculate_qa_status(jobs) self.logger.debug("request %s state %s", req.reqid, qa_state) msg = None if self.force or qa_state == QA_UNKNOWN: ret = super(OpenQABot, self).check_one_request(req) jobs = self.request_get_openqa_jobs(req) if self.force: # make sure to delete previous comments if we're forcing info = self.find_obs_request_comment(request_id=req.reqid) if 'id' in info: self.logger.debug("deleting old comment %s", info['id']) if not self.dryrun: self.commentapi.delete(info['id']) if jobs: # no notification until the result is done osc.core.change_review_state(self.apiurl, req.reqid, newstate='new', by_group=self.review_group, by_user=self.review_user, message='now testing in openQA') else: msg = "no openQA tests defined" self.add_comment(msg, 'done', request_id=req.reqid, result='accepted') ret = True elif qa_state == QA_FAILED or qa_state == QA_PASSED: # don't take test repo results into the calculation of total # this is for humans to decide which incident broke the test repo jobs += self.request_get_openqa_jobs(req, incident=False, test_repo=True) if self.calculate_qa_status(jobs) == QA_INPROGRESS: self.logger.info( "incident tests for request %s are done, but need to wait for test repo", req.reqid) return if qa_state == QA_PASSED: msg = "openQA tests passed\n" result = 'accepted' ret = True else: msg = "openQA tests problematic\n" result = 'declined' ret = False msg += self.summarize_openqa_jobs(jobs) self.add_comment(msg, 'done', result=result, request_id=req.reqid) elif qa_state == QA_INPROGRESS: self.logger.info("request %s still in progress", req.reqid) else: raise Exception("unknown QA state %d", qa_state) except Exception: import traceback self.logger.error("unhandled exception in openQA Bot") self.logger.error(traceback.format_exc()) ret = None return ret def find_obs_request_comment(self, request_id=None, project_name=None): """Return previous comments (should be one).""" if self.do_comments: comments = self.commentapi.get_comments(request_id=request_id, project_name=project_name) for c in comments.values(): m = comment_marker_re.match(c['comment']) if m: return { 'id': c['id'], 'state': m.group('state'), 'result': m.group('result'), 'comment': c['comment'], 'revision': m.group('revision')} return {} def check_product_arch(self, job, product_prefix, pmap, arch): need = False settings = {'VERSION': pmap['version']} settings['ARCH'] = arch if arch else 'x86_64' settings['DISTRI'] = 'sle' if 'distri' not in pmap else pmap['distri'] issues = pmap.get('issues', {}) issues['OS_TEST_ISSUES'] = issues.get('OS_TEST_ISSUES', product_prefix) required_issue = pmap.get('required_issue', False) for key, prefix in issues.items(): self.logger.debug("KP {} {}".format(key, prefix) + str(job)) channel = prefix if arch: channel += arch if channel in job['channels']: settings[key] = str(job['id']) need = True if required_issue: if required_issue not in settings: need = False if not need: return [] product_key = product_prefix if arch: product_key += arch update = self.project_settings[product_key] update.apiurl = self.apiurl update.logger = self.logger posts = [] for j in update.settings( update.maintenance_project + ':' + str(job['id']), product_key): if not job.get('openqa_build'): job['openqa_build'] = update.get_max_revision(job) if not job.get('openqa_build'): return [] j['BUILD'] += '.' + str(job['openqa_build']) j.update(settings) # kGraft jobs can have different version if 'real_version' in j: j['VERSION'] = j['real_version'] del j['real_version'] posts.append(j) return posts def check_product(self, job, product_prefix): pmap = self.api_map[product_prefix] posts = [] if 'archs' in pmap: for arch in pmap['archs']: posts += self.check_product_arch(job, product_prefix, pmap, arch) else: posts += self.check_product_arch(job, product_prefix, pmap, None) self.logger.debug("Pmap: {} Posts: {}".format(pmap, posts)) return posts def incident_openqa_jobs(self, s): return self.openqa.openqa_request( 'GET', 'jobs', { 'distri': s['DISTRI'], 'version': s['VERSION'], 'arch': s['ARCH'], 'flavor': s['FLAVOR'], 'build': s['BUILD'], 'scope': 'relevant', 'latest': '1' })['jobs'] # for SUSE we use mesh for openSUSE we limit the jobs to open release requests def check_opensuse_incidents(self): for req in self.requests: types = set([a.type for a in req.actions]) if not 'maintenance_release' in types: next src_prjs = set([a.src_project for a in req.actions]) if len(src_prjs) != 1: raise Exception("can't handle maintenance_release from different incidents") build = src_prjs.pop() tgt_prjs = set([a.tgt_project for a in req.actions]) ret = [] for prj in tgt_prjs: # ignore e.g. Backports if not prj in self.project_settings: next incident_id = build.split(':')[-1] self.test_job({'project': build, 'id': incident_id, 'channels': [prj]}) print("TGT", self.tgt_repo[self.openqa.baseurl][prj]['settings']) #['OS_TEST_ISSUES']) def check_suse_incidents(self): for inc in requests.get('https://maintenance.suse.de/api/incident/active/').json(): self.logger.info("Incident number: {}".format(inc)) mesh_job = requests.get('https://maintenance.suse.de/api/incident/' + inc).json() if mesh_job['meta']['state'] in ['final', 'gone']: continue # required in mesh_job: project, id, channels self.test_job(mesh_job['base']) def test_job(self, mesh_job): self.logger.debug("Called test_job with: {}".format(mesh_job)) incident_project = str(mesh_job['project']) try: comment_info = self.find_obs_request_comment(project_name=incident_project) except HTTPError as e: self.logger.debug("Couldn't load comments - {}".format(e)) return comment_id = comment_info.get('id', None) comment_build = str(comment_info.get('revision', '')) openqa_posts = [] for prod in self.api_map.keys(): self.logger.debug("{} -- product in apimap".format(prod)) openqa_posts += self.check_product(mesh_job, prod) openqa_jobs = [] for s in openqa_posts: jobs = self.incident_openqa_jobs(s) # take the project comment as marker for not posting jobs if not len(jobs) and comment_build != str(mesh_job['openqa_build']): if self.dryrun: self.logger.info('WOULD POST:{}'.format(pformat(json.dumps(s, sort_keys=True)))) else: self.logger.info("Posted: {}".format(pformat(json.dumps(s, sort_keys=True)))) self.openqa.openqa_request('POST', 'isos', data=s, retries=1) openqa_jobs += self.incident_openqa_jobs(s) else: self.logger.info("{} got {}".format(pformat(s), len(jobs))) openqa_jobs += jobs self.openqa_jobs[incident_project] = openqa_jobs if len(openqa_jobs) == 0: self.logger.debug("No openqa jobs defined") return # print openqa_jobs msg = self.summarize_openqa_jobs(openqa_jobs) state = 'seen' result = 'none' qa_status = self.calculate_qa_status(openqa_jobs) if qa_status == QA_PASSED: result = 'accepted' state = 'done' if qa_status == QA_FAILED: result = 'declined' state = 'done' comment = "\n".format( state, result, mesh_job.get('openqa_build')) comment += msg if comment_id and state != 'done': self.logger.info("%s is already commented, wait until done", incident_project) return if comment_info.get('comment', '').rstrip('\n') == comment.rstrip('\n'): self.logger.info("%s comment did not change", incident_project) return self.logger.info("adding comment to %s, state %s", incident_project, state) if not self.dryrun: if comment_id: self.logger.debug("delete comment: {}".format(comment_id)) self.commentapi.delete(comment_id) self.commentapi.add_comment(project_name=str(incident_project), comment=str(comment))