tools/pubsub2rake.rb

#!/usr/bin/env ruby

$LOAD_PATH.unshift '/srv/whimsy/lib'

require 'net/http'
require 'json'
require 'whimsy/asf/config'
require 'whimsy/asf/svn'

def stamp(*s)
  '%s: %s' % [Time.now.gmtime.to_s, s.join(' ')]
end

# need to fetch all topics to ensure mixed commits are seen
PUBSUB_URL = 'https://pubsub.apache.org:2070/svn'

class PubSub

  require 'fileutils'
  ALIVE = File.join('/tmp', "#{File.basename(__FILE__)}.alive") # TESTING ONLY

  @restartable = false
  @updated = false
  def self.listen(url, creds, options={})
    debug = options[:debug]
    mtime = File.mtime(__FILE__)
    FileUtils.touch(ALIVE) # Temporary debug - ensure exists
    done = false
    ps_thread = Thread.new do
      begin
        uri = URI.parse(url)
        Net::HTTP.start(uri.host, uri.port,
          open_timeout: 20, read_timeout: 20, ssl_timeout: 20,
          use_ssl: uri.scheme == 'https') do |http|
          request = Net::HTTP::Get.new uri.request_uri
          request.basic_auth(*creds) if creds
          http.request request do |response|
            response.each_header do |h, v|
              puts stamp [h, v].inspect if h.start_with? 'x-' or h == 'server'
            end
            body = ''
            response.read_body do |chunk|
              # Long time no see?
              lasttime = File.mtime(ALIVE)
              diff = (Time.now - lasttime).to_i
              if diff > 60
                puts stamp 'HUNG?', diff, lasttime
              end
              FileUtils.touch(ALIVE) # Temporary debug
              body += chunk
              # All chunks are terminated with \n. Since 2070 can split events into 64kb sub-chunks
              # we wait till we have gotten a newline, before trying to parse the JSON.
              if chunk.end_with? "\n"
                event = JSON.parse(body.chomp)
                body = ''
                if event['stillalive'] # pingback
                  @restartable = true
                  puts stamp event if debug
                else
                  yield event
                end
              else
                puts stamp 'Partial chunk' if debug
              end
              unless mtime == File.mtime(__FILE__)
                puts stamp 'File updated' if debug
                @updated = true
                done = true
              end
              break if done
            end # reading chunks
            puts stamp 'Done reading chunks' if debug
            break if done
          end # read response
          puts stamp 'Done reading response' if debug
          break if done
        end # net start
        puts stamp 'Done with start' if debug
      rescue Errno::ECONNREFUSED => e
        @restartable = true
        $stderr.puts stamp e.inspect
        sleep 3
      rescue StandardError => e
        $stderr.puts stamp e.inspect
        $stderr.puts stamp e.backtrace
      end
      puts stamp 'Done with thread' if debug
    end # thread
    puts stamp "Pubsub thread started #{url} ..."
    ps_thread.join
    puts stamp 'Pubsub thread finished %s...' % (@updated ? '(updated) ' : '')
    if @restartable
      $stderr.puts stamp 'restarting'

      # relaunch script after a one second delay
      sleep 1
      exec RbConfig.ruby, __FILE__, *ARGV
    end
  end
end

if $0 == __FILE__
  $stdout.sync = true

  $hits = 0 # items matched
  $misses = 0 # items not matched

  options = {}
  args = ARGV.dup # preserve ARGV for relaunch
  options[:debug] = args.delete('--debug')
  pubsub_URL = args[0]  || PUBSUB_URL
  pubsub_FILE = args[1] || File.join(Dir.home, '.pubsub')
  pubsub_CRED = File.read(pubsub_FILE).chomp.split(':') rescue nil

  WATCH = Hash.new{|h, k| h[k] = Array.new}
  # determine which paths we are interested in
  # depth: 'skip' == ignore completely
  # files: only need to update if path matches one of the files
  # depth: 'files' only need to update for top level files

  # The first segment of the url is the repo name, e.g. asf or infra
  # The second segment is the subdir within the repo.
  # The combination of the two are used for the pubsub path, for example:
  # asf/infrastructure/site/trunk/content/foundation/records/minutes
  # relates to pubsub_path: svn/asf/infrastructure

  def process(event)
    path = event['pubsub_path']
    if WATCH.include? path # WATCH auto-vivifies so cannot use [] here
      $hits += 1
      log = event['commit']['log'].sub(/\n.*/m, '') # keep only first line
      id = event['commit']['id']
      puts ''
      puts stamp id, path, log
      matches = Hash.new{|h, k| h[k] = Array.new} # key alias, value = array of matching files
      watching = WATCH[path]
      watching.each do |svn_prefix, svn_alias, files|
        changed = event['commit']['changed']
        changed.each_key do |ck|
          if ck.start_with? svn_prefix # file matches target path
            if files && files.size > 0 # but does it match exactly?
              files.each do |file|
                if ck == File.join(svn_prefix, file)
                  matches[svn_alias] << ck
                  break # no point checking other files
                end
              end
            else
              matches[svn_alias] << ck
            end
          end
        end
      end
      matches.each do |k, _v|
        puts stamp "Updating #{k} #{$hits}/#{$misses}"
        cmd = ['rake', "svn:update[#{k}]"]
        unless system(*cmd, {chdir: '/srv/whimsy'})
          puts stamp "Error #{$?} processing #{cmd}"
        end
        $stdout.flush # Ensure we see the output
      end
    else
      $misses += 1
      if File.exist? '/srv/svn/pubsub2rake.trace'
        log = event['commit']['log'].sub(/\n.*/m, '') # keep only first line
        id = event['commit']['id']
        puts ''
        puts stamp id, path, 'DBG', log
      end
    end # possible match
  end

  ASF::SVN.repo_entries(true).each do |name, desc|
    next if desc['depth'] == 'skip' # not needed

    # Drop the dist.a.o prefix
    url = desc['url'].sub(%r{https?://.+?/repos/}, '')

    one, two, three = url.split('/', 3)
    path_prefix = %w{asf dist}.include?(one) ? ['/svn'] : ['/private', 'svn']
    pubsub_key = [path_prefix, one, two, 'commit'].join('/')
    svn_relpath = [two, three].join('/')
    WATCH[pubsub_key] << [svn_relpath, name, desc['files']]
    # N.B. A commit that includes more than one top-level directory
    # does not include either directory in the pubsub path.
    # e.g. dev->release dist renames have the path /svn/dist/commit
    # Allow for this by adding the parent path as well
    # This is only likely to be needed for dist, but there may
    # be other commits that mix directories.
    pubsub_key = [path_prefix, one, 'commit'].join('/')
    WATCH[pubsub_key] << [svn_relpath, name, desc['files']]
    # The whimsy user does not have full access to private commits.
    # As a work-round, commits that touch both documents and foundation are given the topic documents
    # This means that foundation commits may also be found under documents
    if two == 'foundation'
      pubsub_key = [path_prefix, one, 'documents', 'commit'].join('/')
      WATCH[pubsub_key] << [svn_relpath, name, desc['files']]
    end
  end

  if pubsub_URL == 'WATCH' # dump keys for use in constructing URL
    WATCH.sort.each do |k, v|
      puts k
      v.sort.each do |e|
        print '- '
        p e
      end
    end
    exit
  end

  if File.exist? pubsub_URL
    puts '** Unit testing **'
    File.open(pubsub_URL).each_line do |line|
      event = nil
      begin
        event = JSON.parse(line.chomp)
      rescue StandardError => e
        p e
        puts line
      end
      process(event) unless event.nil? || event['stillalive']
    end
  else
    puts stamp(pubsub_URL)
    PubSub.listen(pubsub_URL, pubsub_CRED, options) do |event|
      process(event)
    end
  end
end