llvm-premerge-checks/scripts/phabtalk/phabtalk.py

#!/usr/bin/env python3
# Copyright 2019 Google LLC
#
# Licensed under the the Apache License v2.0 with LLVM Exceptions (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://llvm.org/LICENSE.txt
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Upload build results to Phabricator.

As I did not like the Jenkins plugin, we're using this script to upload the 
build status, a summary and the test reults to Phabricator."""

import argparse
import os
import re
import socket
import time
import urllib
import uuid
from typing import Optional, List, Dict

import pathspec
from lxml import etree
from phabricator import Phabricator


class PhabTalk:
    """Talk to Phabricator to upload build results.
       See https://secure.phabricator.com/conduit/method/harbormaster.sendmessage/
    """

    def __init__(self, token: Optional[str], host: Optional[str], dryrun: bool):
        self._phab = None  # type: Optional[Phabricator]
        if not dryrun:
            self._phab = Phabricator(token=token, host=host)
            self._phab.update_interfaces()

    @property
    def dryrun(self):
        return self._phab is None

    def get_revision_id(self, diff: str) -> Optional[str]:
        """Get the revision ID for a diff from Phabricator."""
        if self.dryrun:
            return None

        result = self._phab.differential.querydiffs(ids=[diff])
        return 'D' + result[diff]['revisionID']

    def comment_on_diff(self, diff_id: str, text: str):
        """Add a comment to a differential based on the diff_id"""
        print('Sending comment to diff {}:'.format(diff_id))
        print(text)
        revision_id = self.get_revision_id(diff_id)
        if revision_id is not None:
            self._comment_on_revision(revision_id, text)

    def _comment_on_revision(self, revision: str, text: str):
        """Add comment on a differential based on the revision id."""

        transactions = [{
            'type': 'comment',
            'value': text
        }]

        if self.dryrun:
            print('differential.revision.edit =================')
            print('Transactions: {}'.format(transactions))
            return

        # API details at
        # https://secure.phabricator.com/conduit/method/differential.revision.edit/
        self._phab.differential.revision.edit(objectIdentifier=revision,
                                              transactions=transactions)

    def update_build_status(self, diff_id: str, phid: str, working: bool, success: bool, lint: {}, unit: []):
        """Submit collected report to Phabricator.
        """

        result_type = 'pass'
        if working:
            result_type = 'working'
        elif not success:
            result_type = 'fail'

        # Group lint messages by file and line.
        lint_messages = []
        for v in lint.values():
            path = ''
            line = 0
            descriptions = []
            for e in v:
                path = e['path']
                line = e['line']
                descriptions.append('{}: {}'.format(e['name'], e['description']))
            lint_message = {
                'name': 'Pre-merge checks',
                'severity': 'warning',
                'code': 'llvm-premerge-checks',
                'path': path,
                'line': line,
                'description': '\n'.join(descriptions),
            }
            lint_messages.append(lint_message)

        if self.dryrun:
            print('harbormaster.sendmessage =================')
            print('type: {}'.format(result_type))
            print('unit: {}'.format(unit))
            print('lint: {}'.format(lint_messages))
            return

        _try_call(lambda: self._phab.harbormaster.sendmessage(
            buildTargetPHID=phid,
            type=result_type,
            unit=unit,
            lint=lint_messages))

    def add_artifact(self, phid: str, file: str, name: str, results_url: str):
        artifactKey=str(uuid.uuid4())
        artifactType='uri'
        artifactData={'uri': '{}/{}'.format(results_url, file),
                        'ui.external': True,
                        'name': name}
        if self.dryrun:
            print('harbormaster.createartifact =================')
            print('artifactKey: {}'.format(artifactKey))
            print('artifactType: {}'.format(artifactType))
            print('artifactData: {}'.format(artifactData))
            return

        _try_call(lambda: self._phab.harbormaster.createartifact(
            buildTargetPHID=phid,
            artifactKey=artifactKey,
            artifactType=artifactType,
            artifactData=artifactData))


def _parse_patch(patch) -> List[Dict[str, str]]:
    """Extract the changed lines from `patch` file.
    The return value is a list of dictionaries {filename, line, diff}.
    Diff must be generated with -U0 (no context lines).
    """
    entries = []
    lines = []
    filename = None
    line_number = 0
    for line in patch:
        match = re.search(r'^(\+\+\+|---) [^/]+/(.*)', line)
        if match:
            if len(lines) > 0:
                entries.append({
                    'filename': filename,
                    'diff': ''.join(lines),
                    'line': line_number,
                })
                lines = []
            filename = match.group(2).rstrip('\r\n')
            continue
        match = re.search(r'^@@ -(\d+)(,(\d+))? \+(\d+)(,(\d+))?', line)
        if match:
            if len(lines) > 0:
                entries.append({
                    'filename': filename,
                    'diff': ''.join(lines),
                    'line': line_number,
                })
                lines = []
            line_number = int(match.group(1))
            continue
        if line.startswith('+') or line.startswith('-'):
            lines.append(line)
    if len(lines) > 0:
        entries.append({
            'filename': filename,
            'diff': ''.join(lines),
            'line': line_number,
        })
    return entries


class BuildReport:

    def __init__(self, args):
        # self.args = args
        self.ph_id = args.ph_id  # type: str
        self.diff_id = args.diff_id  # type: str
        self.test_result_file = args.test_result_file  # type: str
        self.conduit_token = args.conduit_token  # type: str
        self.dryrun = args.dryrun  # type: bool
        self.buildresult = args.buildresult  # type: str
        self.clang_format_patch = args.clang_format_patch  # type: str
        self.clang_tidy_result = args.clang_tidy_result  # type: str
        self.clang_tidy_ignore = args.clang_tidy_ignore  # type: str
        self.results_dir = args.results_dir  # type: str
        self.results_url = args.results_url  # type: str
        self.workspace = args.workspace  # type: str
        self.failure_messages = args.failures  # type: str

        self.api = PhabTalk(args.conduit_token, args.host, args.dryrun)

        self.comments = []
        self.success = True
        self.working = False
        self.unit = []  # type: List
        self.lint = {}
        self.test_stats = {
            'pass': 0,
            'fail': 0,
            'skip': 0
        }  # type: Dict[str, int]

    def add_lint(self, m):
        key = '{}:{}'.format(m['path'], m['line'])
        if key not in self.lint:
            self.lint[key] = []
        self.lint[key].append(m)

    def final_report(self):
        if self.buildresult is not None:
            print('Jenkins result: {}'.format(self.buildresult))
            if self.buildresult.lower() == 'success':
                pass
            elif self.buildresult.lower() == 'null':
                self.working = True
        else:
            self.success = False

        try:
            self.add_test_results()
        except etree.XMLSyntaxError:
            # Sometimes we get an incomplete XML file.
            # In this case: 
            #   - fail the build (the safe thing to do)
            #   - continue so the user gets some feedback.
            print('Error parsing {}. Invalid XML syntax!'.format(self.test_result_file))
            self.success = False

        self.add_clang_tidy()
        self.add_clang_format()
        self.api.update_build_status(self.diff_id, self.ph_id, self.working, self.success, self.lint, self.unit)

        self.add_links_to_artifacts()

        title = 'Issue with build for {} ({})'.format(self.api.get_revision_id(self.diff_id), self.diff_id)
        self.comments.append(
            'Pre-merge checks is in beta <a href="https://github.com/google/llvm-premerge-checks/issues/new?assignees'
            '=&labels=bug&template=bug_report.md&title={}">report issue</a>.<br/>'
            'Please <a href="https://reviews.llvm.org/project/update/78/join/">join beta</a> or '
            '<a href="https://github.com/google/llvm-premerge-checks/issues/new?assignees=&labels=enhancement&template'
            '=&title=enable%20checks%20for%20{{PATH}}">enable it for your project</a>'.format(
                urllib.parse.quote(title)))
        with open(os.path.join(self.results_dir, 'summary.html'), 'w') as f:
            f.write('<html><head><style>'
                    'body { font-family: monospace; margin: 16px; }\n'
                    '.failure {color:red;}\n'
                    '.success {color:green;}\n'
                    '</style></head><body>')
            if self.failure_messages and len(self.failure_messages) > 0:
                for s in self.failure_messages.split('\n'):
                    f.write('<p class="failure">{}</p>'.format(s))
            f.write('<p>' + '</p><p>'.join(self.comments) + '</p>')
            f.write('</body></html>')
            self.api.add_artifact(self.ph_id, 'summary.html', 'summary', self.results_url)

    def add_clang_format(self):
        """Populates results from diff produced by clang format."""
        if self.clang_format_patch is None:
            return
        present = os.path.exists(
            os.path.join(self.results_dir, self.clang_format_patch))
        if not present:
            print('clang-format result {} is not found'.format(self.clang_format_patch))
            self.comments.append(section_title('clang-format', False, False))
            return
        p = os.path.join(self.results_dir, self.clang_format_patch)
        if os.stat(p).st_size != 0:
            self.api.add_artifact(self.ph_id, self.clang_format_patch, "clang-format", self.results_url)
        diffs = _parse_patch(open(p, 'r'))
        success = len(diffs) == 0
        for d in diffs:
            self.add_lint({
                'name': 'clang-format',
                'severity': 'autofix',
                'code': 'clang-format',
                'path': d['filename'],
                'line': d['line'],
                'char': 1,
                'description': 'please reformat the code\n```\n' + d['diff'] + '\n```',
            })
        comment = section_title('clang-format', success, present)
        if not success:
            comment += 'Please format your changes with clang-format by running `git-clang-format HEAD^` or applying ' \
                       'this <a href="{}">patch</a>.'.format(self.clang_format_patch)
        self.comments.append(comment)
        self.success = success and self.success

    def add_clang_tidy(self):
        # Typical message looks like
        # [..]/clang/include/clang/AST/DeclCXX.h:3058:20: error: no member named 'LifetimeExtendedTemporary' in 'clang::Decl' [clang-diagnostic-error]
        if self.clang_tidy_result is None:
            return
        pattern = '^{}/([^:]*):(\\d+):(\\d+): (.*): (.*)'.format(self.workspace)
        errors_count = 0
        warn_count = 0
        inline_comments = 0
        present = os.path.exists(
            os.path.join(self.results_dir, self.clang_tidy_result))
        if not present:
            print('clang-tidy result {} is not found'.format(self.clang_tidy_result))
            self.comments.append(section_title('clang-tidy', False, False))
            return
        present = (self.clang_tidy_ignore is not None) and os.path.exists(self.clang_tidy_ignore)
        if not present:
            print('clang-tidy ignore file {} is not found'.format(self.clang_tidy_ignore))
            self.comments.append(section_title('clang-tidy', False, False))
            return
        p = os.path.join(self.results_dir, self.clang_tidy_result)
        if os.stat(p).st_size > 4:
            self.api.add_artifact(self.ph_id, self.clang_tidy_result, "clang-tidy", self.results_url)
        ignore = pathspec.PathSpec.from_lines(pathspec.patterns.GitWildMatchPattern,
                                              open(self.clang_tidy_ignore, 'r').readlines())
        for line in open(p, 'r'):
            match = re.search(pattern, line)
            if match:
                file_name = match.group(1)
                line_pos = match.group(2)
                char_pos = match.group(3)
                severity = match.group(4)
                text = match.group(5)
                text += '\n[[{} | not useful]] '.format(
                    'https://github.com/google/llvm-premerge-checks/blob/master/docs/clang_tidy.md#warning-is-not'
                    '-useful')
                if severity in ['warning', 'error']:
                    if severity == 'warning':
                        warn_count += 1
                    if severity == 'error':
                        errors_count += 1
                    if ignore.match_file(file_name):
                        print('{} is ignored by pattern and no comment will be added'.format(file_name))
                    else:
                        inline_comments += 1
                        self.add_lint({
                            'name': 'clang-tidy',
                            'severity': 'warning',
                            'code': 'clang-tidy',
                            'path': file_name,
                            'line': int(line_pos),
                            'char': int(char_pos),
                            'description': '{}: {}'.format(severity, text),
                        })
        success = errors_count + warn_count == 0
        comment = section_title('clang-tidy', success, present)
        if not success:
            comment += 'clang-tidy found <a href="{}">{} errors and {} warnings</a>. ' \
                       '{} of them are added as review comments <a href="{}">why?</a>.'.format(
                self.clang_tidy_result, errors_count, warn_count, inline_comments,
                'https://github.com/google/llvm-premerge-checks/blob/master/docs/clang_tidy.md#review-comments')

        self.comments.append(comment)
        self.success = success and self.success

    def add_test_results(self):
        """Populates results from build test results XML.

         Only reporting failed tests as the full test suite is too large to upload.
         """

        success = True
        present = (self.test_result_file is not None) and os.path.exists(
            os.path.join(self.results_dir, self.test_result_file))
        if not present:
            print('Warning: Could not find test results file: {}'.format(self.test_result_file))
            self.comments.append(section_title('Unit tests', False, present))
            return

        root_node = etree.parse(os.path.join(self.results_dir, self.test_result_file))
        for test_case in root_node.xpath('//testcase'):
            test_result = _test_case_status(test_case)
            self.test_stats[test_result] += 1

            if test_result == 'fail':
                success = False
                failure = test_case.find('failure')
                test_result = {
                    'name': test_case.attrib['name'],
                    'namespace': test_case.attrib['classname'],
                    'result': test_result,
                    'duration': float(test_case.attrib['time']),
                    'details': failure.text
                }
                self.unit.append(test_result)

        comment = section_title('Unit tests', success, True)
        comment += '{} tests passed, {} failed and {} were skipped.<br/>'.format(
            self.test_stats['pass'],
            self.test_stats['fail'],
            self.test_stats['skip'],
        )
        if not success:
            comment += 'Failures:<br/>'
        for test_case in self.unit:
            if test_case['result'] == 'fail':
                comment += '{}/{}<br/>'.format(test_case['namespace'], test_case['name'])
        self.comments.append(comment)
        self.success = success and self.success

    def add_links_to_artifacts(self):
        """Comment on a diff, read text from file."""
        file_links = []
        for f in os.listdir(self.results_dir):
            if f == 'summary.html':
                continue
            p = os.path.join(self.results_dir, f)
            if not os.path.isfile(p):
                continue
            if os.stat(p).st_size == 0:
                continue
            file_links.append('<a href="{0}">{0}</a>'.format(f))
        if len(file_links) > 0:
            self.comments.append('<a href="./">Build artifacts</a>:<br/>' + '<br/>'.join(file_links))


def _test_case_status(test_case) -> str:
    """Get the status of a test case based on an etree node."""
    if test_case.find('failure') is not None:
        return 'fail'
    if test_case.find('skipped') is not None:
        return 'skip'
    return 'pass'


def section_title(title: str, ok: bool, present: bool) -> str:
    result = 'unknown'
    c = ''
    if present:
        c = 'success' if ok else 'failure'
        result = 'pass' if ok else 'fail'
    return '{} <span class="{}">{}</span>. '.format(title, c, result)


def _try_call(call):
    """Tries to call function several times retrying on socked.timeout."""
    c = 0
    while True:
        try:
            call()
        except socket.timeout as e:
            c += 1
            if c > 5:
                print('Connection to Pharicator failed, giving up: {}'.format(e))
                raise
            print('Connection to Pharicator failed, retrying: {}'.format(e))
            time.sleep(c * 10)
        break


def main():
    parser = argparse.ArgumentParser(
        description='Write build status back to Phabricator.')
    parser.add_argument('ph_id', type=str)
    parser.add_argument('diff_id', type=str)
    parser.add_argument('--test-result-file', type=str, dest='test_result_file', default='test-results.xml')
    parser.add_argument('--conduit-token', type=str, dest='conduit_token', required=True)
    parser.add_argument('--host', type=str, dest='host', default="https://reviews.llvm.org/api/",
                        help="full URL to API with trailing slash, e.g. https://reviews.llvm.org/api/")
    parser.add_argument('--dryrun', action='store_true',
                        help="output results to the console, do not report back to the server")
    parser.add_argument('--buildresult', type=str, default=None, choices=['SUCCESS', 'UNSTABLE', 'FAILURE', 'null'])
    parser.add_argument('--clang-format-patch', type=str, default=None,
                        dest='clang_format_patch',
                        help="path to diff produced by git-clang-format, relative to results-dir")
    parser.add_argument('--clang-tidy-result', type=str, default=None,
                        dest='clang_tidy_result',
                        help="path to diff produced by git-clang-tidy, relative to results-dir")
    parser.add_argument('--clang-tidy-ignore', type=str, default=None,
                        dest='clang_tidy_ignore',
                        help="path to file with patters to exclude commenting on for clang-tidy findings")
    parser.add_argument('--results-dir', type=str, default=None, required=True,
                        dest='results_dir',
                        help="directory of all build artifacts")
    parser.add_argument('--results-url', type=str, default=None,
                        dest='results_url',
                        help="public URL to access results directory")
    parser.add_argument('--workspace', type=str, required=True, help="path to workspace")
    parser.add_argument('--failures', type=str, default=None, help="optional failure messages separated by newline")
    args = parser.parse_args()

    reporter = BuildReport(args)
    reporter.final_report()


if __name__ == '__main__':
    main()