newrelic_rpm 9.3.0 → 9.4.0

data/test/agent_helper.rb

@@ -0,0 +1,1027 @@
+# This file is distributed under New Relic's license terms.
+# See https://github.com/newrelic/newrelic-ruby-agent/blob/main/LICENSE for complete details.
+# frozen_string_literal: true
+
+# These helpers should not have any gem dependencies except on newrelic_rpm
+# itself, and should be usable from within any multiverse suite.
+
+require 'json'
+require 'net/http'
+begin
+  require 'net/http/status'
+rescue LoadError
+  # NOP -- Net::HTTP::STATUS_CODES was introduced in Ruby 2.5
+end
+
+class ArrayLogDevice
+  def initialize(array = [])
+    @array = array
+  end
+  attr_reader :array
+
+  def write(message)
+    @array << message
+  end
+
+  def close; end
+end
+
+def fake_guid(length = 16)
+  NewRelic::Agent::GuidGenerator.generate_guid(length)
+end
+
+def assert_between(floor, ceiling, value, message = "expected #{floor} <= #{value} <= #{ceiling}")
+  assert((floor <= value && value <= ceiling), message)
+end
+
+def assert_in_delta(expected, actual, delta)
+  assert_between((expected - delta), (expected + delta), actual)
+end
+
+def harvest_error_traces!
+  NewRelic::Agent.instance.error_collector.error_trace_aggregator.harvest!
+end
+
+def reset_error_traces!
+  NewRelic::Agent.instance.error_collector.error_trace_aggregator.reset!
+end
+
+def assert_has_traced_error(error_class)
+  errors = harvest_error_traces!
+
+  refute_nil errors.find { |e| e.exception_class_name == error_class.name }, \
+    "Didn't find error of class #{error_class}"
+end
+
+def last_traced_error
+  harvest_error_traces!.last
+end
+
+def harvest_transaction_events!
+  NewRelic::Agent.instance.transaction_event_aggregator.harvest!
+end
+
+def last_transaction_event
+  harvest_transaction_events!.last.last
+end
+
+def harvest_span_events!
+  NewRelic::Agent.instance.span_event_aggregator.harvest!
+end
+
+def last_span_event
+  harvest_span_events!.last.last
+end
+
+def harvest_error_events!
+  NewRelic::Agent.instance.error_collector.error_event_aggregator.harvest!
+end
+
+def last_error_event
+  harvest_error_events!.last.last
+end
+
+unless defined? assert_includes
+  def assert_includes(collection, member, msg = nil)
+    msg = "Expected #{collection.inspect} to include #{member.inspect}"
+
+    assert_includes collection, member, msg
+  end
+end
+
+unless defined? assert_not_includes
+  def assert_not_includes(collection, member, msg = nil)
+    msg = "Expected #{collection.inspect} not to include #{member.inspect}"
+
+    refute_includes collection, member, msg
+  end
+end
+
+unless defined? assert_empty
+  def assert_empty(collection, msg = nil)
+    assert_empty collection, msg
+  end
+end
+
+def assert_equal_unordered(left, right)
+  assert_equal(left.length, right.length, "Lengths don't match. #{left.length} != #{right.length}")
+  left.each { |element| assert_includes(right, element) }
+end
+
+def assert_log_contains(log, message)
+  lines = log.array
+
+  assert (lines.any? { |line| line.match(message) }),
+    "Could not find message. Log contained: #{lines.join("\n")}"
+end
+
+def assert_audit_log_contains(audit_log_contents, needle)
+  # Original request bodies dumped to the log have symbol keys, but once
+  # they go through a dump/load, they're strings again, so we strip
+  # double-quotes and colons from the log, and the strings we searching for.
+  regex = /[:"]/
+  needle = needle.gsub(regex, '')
+  haystack = audit_log_contents.gsub(regex, '')
+
+  assert_includes(haystack, needle, "Expected log to contain '#{needle}'")
+end
+
+# Because we don't generate a strictly machine-readable representation of
+# request bodies for the audit log, the transformation into strings is
+# effectively one-way. This, combined with the fact that Hash traversal order
+# is arbitrary in Ruby 1.8.x means that it's difficult to directly assert that
+# some object graph made it into the audit log (due to different possible
+# orderings of the key/value pairs in Hashes that were embedded in the request
+# body). So, this method traverses an object graph and only makes assertions
+# about the terminal (non-Array-or-Hash) nodes therein.
+def assert_audit_log_contains_object(audit_log_contents, o, format = :json)
+  case o
+  when Hash
+    o.each do |k, v|
+      assert_audit_log_contains_object(audit_log_contents, v, format)
+      assert_audit_log_contains_object(audit_log_contents, k, format)
+    end
+  when Array
+
+    o.each do |el|
+      assert_audit_log_contains_object(audit_log_contents, el, format)
+    end
+  when NilClass
+
+    assert_audit_log_contains(audit_log_contents, format == :json ? 'null' : 'nil')
+  else
+    assert_audit_log_contains(audit_log_contents, o.inspect)
+  end
+end
+
+def compare_metrics(expected, actual)
+  actual.delete_if { |a| a.include?('GC/Transaction/') }
+
+  assert_equal(expected.to_a.sort, actual.to_a.sort, "extra: #{(actual - expected).to_a.inspect}; missing: #{(expected - actual).to_a.inspect}")
+end
+
+def metric_spec_from_specish(specish)
+  spec = case specish
+  when String then NewRelic::MetricSpec.new(specish)
+  when Array then NewRelic::MetricSpec.new(*specish)
+  end
+  spec
+end
+
+def _normalize_metric_expectations(expectations)
+  case expectations
+  when Array
+    hash = {}
+    # Just assert that the metric is present, nothing about the attributes
+    expectations.each { |k| hash[k] = {} }
+    hash
+  when String
+    {expectations => {}}
+  else
+    expectations
+  end
+end
+
+def dump_stats(stats)
+  str = +"  Call count:           #{stats.call_count}\n"
+  str << "  Total call time:      #{stats.total_call_time}\n"
+  str << "  Total exclusive time: #{stats.total_exclusive_time}\n"
+  str << "  Min call time:        #{stats.min_call_time}\n"
+  str << "  Max call time:        #{stats.max_call_time}\n"
+  str << "  Sum of squares:       #{stats.sum_of_squares}\n"
+  str << "  Apdex S:              #{stats.apdex_s}\n"
+  str << "  Apdex T:              #{stats.apdex_t}\n"
+  str << "  Apdex F:              #{stats.apdex_f}\n"
+  str
+end
+
+def assert_stats_has_values(stats, expected_spec, expected_attrs)
+  expected_attrs.each do |attr, expected_value|
+    actual_value = stats.send(attr)
+
+    msg = "Expected #{attr} for #{expected_spec} to be #{'~' unless attr == :call_count}#{expected_value}, " \
+          "got #{actual_value}.\nActual stats:\n#{dump_stats(stats)}"
+
+    if attr == :call_count
+      assert_stats_has_values_with_call_count(expected_value, actual_value, msg)
+    else
+      assert_in_delta(expected_value, actual_value, 0.0001, msg)
+    end
+  end
+end
+
+def assert_stats_has_values_with_call_count(expected_value, actual_value, msg)
+  # >, <, >=, <= comparisons
+  if expected_value.to_s =~ /([<>]=?)\s*(\d+)/
+    operator = Regexp.last_match(1).to_sym
+    count = Regexp.last_match(2).to_i
+
+    assert_operator(actual_value, operator, count, msg)
+  # == comparison
+  else
+    assert_equal(expected_value, actual_value, msg)
+  end
+end
+
+def assert_metrics_recorded(expected)
+  expected = _normalize_metric_expectations(expected)
+  expected.each do |specish, expected_attrs|
+    expected_spec = metric_spec_from_specish(specish)
+    actual_stats = NewRelic::Agent.instance.stats_engine.to_h[expected_spec]
+    if !actual_stats
+      all_specs = NewRelic::Agent.instance.stats_engine.to_h.keys.sort
+      matches = all_specs.select { |spec| spec.name == expected_spec.name }
+      matches.map! { |m| "  #{m.inspect}" }
+
+      msg = "Did not find stats for spec #{expected_spec.inspect}."
+      msg += "\nDid find specs: [\n#{matches.join(",\n")}\n]" unless matches.empty?
+      msg += "\nAll specs in there were: #{format_metric_spec_list(all_specs)}"
+
+      assert(actual_stats, msg)
+    end
+
+    assert_stats_has_values(actual_stats, expected_spec, expected_attrs)
+  end
+end
+
+# Use this to assert that *only* the given set of metrics has been recorded.
+#
+# If you want to scope the search for unexpected metrics to a particular
+# namespace (e.g. metrics matching 'Controller/'), pass a Regex for the
+# :filter option. Only metrics matching the regex will be searched when looking
+# for unexpected metrics.
+#
+# If you want to *allow* unexpected metrics matching certain patterns, use
+# the :ignore_filter option. This will allow you to specify a Regex that
+# allowlists broad swathes of metric territory (e.g. 'Supportability/').
+#
+def assert_metrics_recorded_exclusive(expected, options = {})
+  expected = _normalize_metric_expectations(expected)
+
+  assert_metrics_recorded(expected)
+
+  recorded_metrics = NewRelic::Agent.instance.stats_engine.to_h.keys
+
+  if options[:filter]
+    recorded_metrics = recorded_metrics.select { |m| m.name.match(options[:filter]) }
+  end
+  if options[:ignore_filter]
+    recorded_metrics.reject! { |m| m.name.match(options[:ignore_filter]) }
+  end
+
+  expected_metrics = expected.keys.map { |s| metric_spec_from_specish(s) }
+
+  unexpected_metrics = recorded_metrics - expected_metrics
+  unexpected_metrics.reject! { |m| m.name.include?('GC/Transaction') }
+
+  assert_equal(0, unexpected_metrics.size, "Found unexpected metrics: #{format_metric_spec_list(unexpected_metrics)}")
+end
+
+def assert_newrelic_metadata_present(metadata)
+  assert metadata.key?('newrelic')
+  refute_nil metadata['newrelic']
+end
+
+def assert_distributed_tracing_payload_created_for_transaction(transaction)
+  assert transaction.distributed_tracer.instance_variable_get(:@distributed_trace_payload_created)
+end
+
+# The clear_metrics! method prevents metrics from "leaking" between tests by resetting
+# the @stats_hash instance variable in the current instance of NewRelic::Agent::StatsEngine.
+
+module NewRelic
+  module Agent
+    class StatsEngine
+      def reset_for_test!
+        @stats_hash = StatsHash.new
+      end
+    end
+  end
+end
+
+def clear_metrics!
+  NewRelic::Agent.instance.stats_engine.reset_for_test!
+end
+
+def assert_metrics_not_recorded(not_expected)
+  not_expected = _normalize_metric_expectations(not_expected)
+  found_but_not_expected = []
+  not_expected.each do |specish, _|
+    spec = metric_spec_from_specish(specish)
+    if NewRelic::Agent.instance.stats_engine.to_h[spec]
+      found_but_not_expected << spec
+    end
+  end
+
+  assert_empty(found_but_not_expected, "Found unexpected metrics: #{format_metric_spec_list(found_but_not_expected)}")
+end
+
+alias :refute_metrics_recorded :assert_metrics_not_recorded
+
+def assert_no_metrics_match(regex)
+  matching_metrics = []
+  NewRelic::Agent.instance.stats_engine.to_h.keys.map(&:to_s).each do |metric|
+    matching_metrics << metric if metric.match(regex)
+  end
+
+  assert_empty(
+    matching_metrics,
+    "Found unexpected metrics:\n" + matching_metrics.map { |m| "  '#{m}'" }.join("\n") + "\n\n"
+  )
+end
+
+alias :refute_metrics_match :assert_no_metrics_match
+
+def format_metric_spec_list(specs)
+  spec_strings = specs.map do |spec|
+    "#{spec.name} (#{spec.scope.empty? ? '<unscoped>' : spec.scope})"
+  end
+  "[\n  #{spec_strings.join(",\n  ")}\n]"
+end
+
+def assert_truthy(expected, msg = nil)
+  msg ||= "Expected #{expected.inspect} to be truthy"
+
+  refute !expected, msg
+end
+
+def assert_falsy(expected, msg = nil)
+  msg ||= "Expected #{expected.inspect} to be falsy"
+
+  refute expected, msg
+end
+
+unless defined? assert_false
+  def assert_false(expected)
+    refute expected
+  end
+end
+
+unless defined? refute
+  alias refute assert_false
+end
+
+# Mock up a transaction for testing purposes, optionally specifying a name and
+# transaction category. The given block will be executed within the context of the
+# dummy transaction.
+#
+# Examples:
+#
+# With default name ('dummy') and category (:other):
+#   in_transaction { ... }
+#
+# With an explicit transaction name and default category:
+#   in_transaction('foobar') { ... }
+#
+# With default name and explicit category:
+#   in_transaction(:category => :controller) { ... }
+#
+# With a transaction name plus category:
+#   in_transaction('foobar', :category => :controller) { ... }
+#
+def in_transaction(*args, &blk)
+  opts = args.last&.is_a?(Hash) ? args.pop : {}
+  category = (opts&.delete(:category)) || :other
+
+  # At least one test passes `:transaction_name => nil`, so handle it gently
+  name = opts.key?(:transaction_name) ? opts.delete(:transaction_name) : args.first || 'dummy'
+
+  state = NewRelic::Agent::Tracer.state
+  txn = nil
+
+  NewRelic::Agent::Tracer.in_transaction(name: name, category: category, options: opts) do
+    txn = state.current_transaction
+    yield(state.current_transaction)
+  end
+
+  txn
+end
+
+# Temporarily disables default transformer so tests with invalid inputs can be tried
+def with_disabled_defaults_transformer(key)
+  begin
+    transformer = NewRelic::Agent::Configuration::DEFAULTS[key][:transform]
+    NewRelic::Agent::Configuration::DEFAULTS[key][:transform] = nil
+    yield
+  ensure
+    NewRelic::Agent::Configuration::DEFAULTS[key][:transform] = transformer
+  end
+end
+
+# Convenience wrapper to stand up a transaction and provide a segment within
+# that transaction to work with.  The same arguments as provided to in_transaction
+# may be supplied.
+def with_segment(*args, &blk)
+  segment = nil
+  txn = in_transaction(*args) do |t|
+    segment = t.current_segment
+    yield(segment, t)
+  end
+  [segment, txn]
+end
+
+# building error attributes on segments are deferred until it's time
+# to publish/harvest them as spans, so for testing, we'll explicitly
+# build 'em as appropriate so we can test 'em
+def build_deferred_error_attributes(segment)
+  return unless segment.noticed_error
+  return if segment.noticed_error_attributes.frozen?
+
+  segment.noticed_error.build_error_attributes
+end
+
+def capture_segment_with_error
+  begin
+    segment_with_error = nil
+    with_segment do |segment|
+      segment_with_error = segment
+      raise 'oops!'
+    end
+  rescue Exception => exception
+    assert segment_with_error, 'expected to have a segment_with_error'
+    build_deferred_error_attributes(segment_with_error)
+    return segment_with_error, exception
+  end
+end
+
+def stub_transaction_guid(guid)
+  NewRelic::Agent::Transaction.tl_current.instance_variable_set(:@guid, guid)
+end
+
+# Convenience wrapper around in_transaction that sets the category so that it
+# looks like we are in a web transaction
+def in_web_transaction(name = 'dummy')
+  in_transaction(name, :category => :controller, :request => stub(:path => '/')) do |txn|
+    yield(txn)
+  end
+end
+
+def in_background_transaction(name = 'silly')
+  in_transaction(name, :category => :task) do |txn|
+    yield(txn)
+  end
+end
+
+def refute_contains_request_params(attributes)
+  attributes.keys.each do |key|
+    refute_match(/^request\.parameters\./, key.to_s)
+  end
+end
+
+def last_transaction_trace
+  return unless last_sample = NewRelic::Agent.agent.transaction_sampler.last_sample
+
+  NewRelic::Agent::Transaction::TraceBuilder.build_trace(last_sample)
+end
+
+def last_transaction_trace_request_params
+  agent_attributes = attributes_for(last_transaction_trace, :agent)
+  agent_attributes.inject({}) do |memo, (key, value)|
+    memo[key] = value if key.to_s.start_with?('request.parameters.')
+    memo
+  end
+end
+
+def find_sql_trace(metric_name)
+  NewRelic::Agent.agent.sql_sampler.sql_traces.values.detect do |trace|
+    trace.database_metric_name == metric_name
+  end
+end
+
+def last_sql_trace
+  NewRelic::Agent.agent.sql_sampler.sql_traces.values.last
+end
+
+def find_last_transaction_node(transaction_sample = nil)
+  if transaction_sample
+    root_node = transaction_sample.root_node
+  else
+    root_node = last_transaction_trace.root_node
+  end
+
+  last_node = nil
+  root_node.each_node { |s| last_node = s }
+
+  return last_node
+end
+
+def find_node_with_name(transaction_sample, name)
+  transaction_sample.root_node.each_node do |node|
+    if node.metric_name == name
+      return node
+    end
+  end
+
+  nil
+end
+
+def find_node_with_name_matching(transaction_sample, regex)
+  transaction_sample.root_node.each_node do |node|
+    if node.metric_name.match(regex)
+      return node
+    end
+  end
+
+  nil
+end
+
+def find_all_nodes_with_name_matching(transaction_sample, regexes)
+  regexes = [regexes].flatten
+  matching_nodes = []
+
+  transaction_sample.root_node.each_node do |node|
+    regexes.each do |regex|
+      if node.metric_name.match(regex)
+        matching_nodes << node
+      end
+    end
+  end
+
+  matching_nodes
+end
+
+def with_config(config_hash, at_start = true)
+  config = NewRelic::Agent::Configuration::DottedHash.new(config_hash, true)
+  NewRelic::Agent.config.add_config_for_testing(config, at_start)
+  NewRelic::Agent.instance.refresh_attribute_filter
+  begin
+    yield
+  ensure
+    NewRelic::Agent.config.remove_config(config)
+    NewRelic::Agent.instance.refresh_attribute_filter
+  end
+end
+
+def with_server_source(config_hash, at_start = true)
+  with_config(config_hash, at_start) do
+    NewRelic::Agent.config.notify_server_source_added
+    yield
+  end
+end
+
+def with_config_low_priority(config_hash)
+  with_config(config_hash, false) do
+    yield
+  end
+end
+
+def with_transaction_renaming_rules(rule_specs)
+  original_engine = NewRelic::Agent.agent.instance_variable_get(:@transaction_rules)
+  begin
+    new_engine = NewRelic::Agent::RulesEngine.create_transaction_rules('transaction_name_rules' => rule_specs)
+    NewRelic::Agent.agent.instance_variable_set(:@transaction_rules, new_engine)
+    yield
+  ensure
+    NewRelic::Agent.agent.instance_variable_set(:@transaction_rules, original_engine)
+  end
+end
+
+# Need to guard against double-installing this patch because in 1.8.x the same
+# file can be required multiple times under different non-canonicalized paths.
+unless Time.respond_to?(:__original_now)
+  Time.instance_eval do
+    class << self
+      attr_accessor :__frozen_now
+      alias_method :__original_now, :now
+
+      def now
+        __frozen_now || __original_now
+      end
+    end
+  end
+end
+
+def nr_freeze_time(now = Time.now)
+  Time.__frozen_now = now
+end
+
+def nr_unfreeze_time
+  Time.__frozen_now = nil
+end
+
+def advance_time(seconds)
+  Time.__frozen_now = Time.now + seconds
+end
+
+unless Process.respond_to?(:__original_clock_gettime)
+  Process.instance_eval do
+    class << self
+      attr_accessor :__frozen_clock_gettime
+      alias_method :__original_clock_gettime, :clock_gettime
+
+      def clock_gettime(clock_id, unit = :float_second)
+        __frozen_clock_gettime || __original_clock_gettime(clock_id, unit)
+      end
+    end
+  end
+end
+
+def advance_process_time(seconds, clock_id = Process::CLOCK_REALTIME)
+  Process.__frozen_clock_gettime = Process.clock_gettime(clock_id) + seconds
+end
+
+def nr_freeze_process_time(now = Process.clock_gettime(Process::CLOCK_REALTIME))
+  Process.__frozen_clock_gettime = now
+end
+
+def nr_unfreeze_process_time
+  Process.__frozen_clock_gettime = nil
+end
+
+def with_constant_defined(constant_symbol, implementation = Module.new)
+  const_path = constant_path(constant_symbol.to_s)
+
+  if const_path
+    # Constant is already defined, nothing to do
+    return yield
+  else
+    const_path = constant_path(constant_symbol.to_s, :allow_partial => true)
+    parent = const_path[-1]
+    constant_symbol = constant_symbol.to_s.split('::').last.to_sym
+  end
+
+  begin
+    parent.const_set(constant_symbol, implementation)
+    yield
+  ensure
+    parent.send(:remove_const, constant_symbol)
+  end
+end
+
+def constant_path(name, opts = {})
+  allow_partial = opts[:allow_partial]
+  path = [Object]
+  parts = name.gsub(/^::/, '').split('::')
+  parts.each do |part|
+    if !path.last.constants.include?(part.to_sym)
+      return allow_partial ? path : nil
+    end
+
+    path << path.last.const_get(part)
+  end
+  path
+end
+
+def get_parent(constant_name)
+  parent_name = constant_name.gsub(/::[^:]*$/, '')
+  const_path = constant_path(parent_name)
+  const_path ? const_path[-1] : nil
+end
+
+def undefine_constant(constant_symbol)
+  const_str = constant_symbol.to_s
+  parent = get_parent(const_str)
+  const_name = const_str.gsub(/.*::/, '')
+  return yield unless parent&.constants&.include?(const_name.to_sym)
+
+  removed_constant = parent.send(:remove_const, const_name)
+  yield
+ensure
+  parent.const_set(const_name, removed_constant) if removed_constant
+end
+
+def with_debug_logging
+  orig_logger = NewRelic::Agent.logger
+  $stderr.puts '', '---', ''
+  NewRelic::Agent.logger =
+    NewRelic::Agent::AgentLogger.new('', Logger.new($stderr))
+
+  with_config(:log_level => 'debug') do
+    yield
+  end
+ensure
+  NewRelic::Agent.logger = orig_logger
+end
+
+def create_agent_command(args = {})
+  NewRelic::Agent::Commands::AgentCommand.new([-1, {'name' => 'command_name', 'arguments' => args}])
+end
+
+def wait_for_backtrace_service_poll(opts = {})
+  defaults = {
+    :timeout => 10.0,
+    :service => NewRelic::Agent.agent.instance_variable_get(:@agent_command_router).backtrace_service,
+    :iterations => 1
+  }
+  opts = defaults.merge(opts)
+  deadline = Process.clock_gettime(Process::CLOCK_REALTIME) + opts[:timeout]
+
+  service = opts[:service]
+  worker_loop = service.worker_loop
+  worker_loop.setup(0, service.method(:poll))
+
+  until worker_loop.iterations > opts[:iterations]
+    sleep(0.01)
+    if Process.clock_gettime(Process::CLOCK_REALTIME) > deadline
+      raise "Timed out waiting #{opts[:timeout]} s for backtrace service poll\n" +
+        "Worker loop ran for #{opts[:service].worker_loop.iterations} iterations\n\n" +
+        Thread.list.map { |t|
+          "#{t.to_s}: newrelic_label: #{t[:newrelic_label].inspect}\n\n" +
+            (t.backtrace || []).join("\n\t")
+        }.join("\n\n")
+    end
+  end
+end
+
+def with_array_logger(level = :info)
+  orig_logger = NewRelic::Agent.logger
+  config = {:log_level => level}
+  logdev = ArrayLogDevice.new
+  override_logger = Logger.new(logdev)
+
+  with_config(config) do
+    NewRelic::Agent.logger = NewRelic::Agent::AgentLogger.new('', override_logger)
+    yield
+  end
+
+  return logdev
+ensure
+  NewRelic::Agent.logger = orig_logger
+end
+
+# The EnvUpdater was introduced due to random fails in JRuby environment
+# whereby attempting to set ENV[key] = some_value randomly failed.
+# It is conjectured that this is thread related, but may also be
+# a core bug in the JVM implementation of Ruby.  Root cause was not
+# discovered, but it was found that a combination of retrying and using
+# mutex lock around the update operation was the only consistently working
+# solution as the error continued to surface without the mutex and
+# retry alone wasn't enough, either.
+#
+# JRUBY: oraclejdk8 + jruby-9.2.6.0
+#
+# NOTE: Singleton pattern to ensure one mutex lock for all threads
+class EnvUpdater
+  MAX_RETRIES = 5
+
+  def initialize
+    @mutex = Mutex.new
+  end
+
+  # Will attempt the given block up to MAX_RETRIES before
+  # surfacing the exception down the chain.
+  def with_retry(retry_limit = MAX_RETRIES)
+    retries ||= 0
+    sleep(retries)
+    yield
+  rescue
+    (retries += 1) < retry_limit ? retry : raise
+  end
+
+  # Locks and updates the ENV
+  def safe_update(env)
+    with_retry do
+      @mutex.synchronize do
+        env.each { |key, val| ENV[key] = val.to_s }
+      end
+    end
+  end
+
+  # Locks and restores the ENV
+  def safe_restore(old_env)
+    with_retry do
+      @mutex.synchronize do
+        old_env.each { |key, val| val ? ENV[key] = val : ENV.delete(key) }
+      end
+    end
+  end
+
+  # Singleton pattern implemented via @@instance
+  def self.instance
+    @@instance ||= EnvUpdater.new
+  end
+
+  def self.safe_update(env)
+    instance.safe_update(env)
+  end
+
+  def self.safe_restore(old_env)
+    instance.safe_restore(old_env)
+  end
+
+  # Effectively saves current ENV settings for given env's key/values,
+  # runs given block, then restores ENV to original state before returning.
+  def self.inject(env, &block)
+    old_env = {}
+    env.each { |key, val| old_env[key] = ENV[key] }
+    begin
+      safe_update(env)
+      yield
+    ensure
+      safe_restore(old_env)
+    end
+  end
+
+  # must call instance here to ensure only one @mutex for all threads.
+  instance
+end
+
+# Changes ENV settings to given and runs given block and restores ENV
+# to original values before returning.
+def with_environment(env, &block)
+  EnvUpdater.inject(env) { yield }
+end
+
+def with_argv(argv)
+  old_argv = ARGV.dup
+  ARGV.clear
+  ARGV.concat(argv)
+
+  begin
+    yield
+  ensure
+    ARGV.clear
+    ARGV.concat(old_argv)
+  end
+end
+
+def with_ignore_error_filter(filter, &blk)
+  original_filter = NewRelic::Agent.ignore_error_filter
+  NewRelic::Agent.ignore_error_filter(&filter)
+
+  yield
+ensure
+  NewRelic::Agent::ErrorCollector.ignore_error_filter = original_filter
+end
+
+def json_dump_and_encode(object)
+  Base64.encode64(JSON.dump(object))
+end
+
+def get_last_analytics_event
+  NewRelic::Agent.agent.transaction_event_aggregator.harvest![1].last
+end
+
+def swap_instance_method(target, method_name, new_method_implementation, &blk)
+  old_method_implementation = target.instance_method(method_name)
+  target.send(:define_method, method_name, new_method_implementation)
+  yield
+rescue NameError => e
+  puts "Your target does not have the instance method #{method_name}"
+  puts e.inspect
+ensure
+  target.send(:define_method, method_name, old_method_implementation)
+end
+
+def cross_agent_tests_dir
+  File.expand_path(File.join(File.dirname(__FILE__), 'fixtures', 'cross_agent_tests'))
+end
+
+def load_cross_agent_test(name)
+  test_file_path = File.join(cross_agent_tests_dir, "#{name}.json")
+  data = File.read(test_file_path)
+  data.gsub!('callCount', 'call_count')
+  data = JSON.load(data)
+  data.each { |testcase| testcase['testname'].tr!(' ', '_') if String === testcase['testname'] }
+  data
+end
+
+def each_cross_agent_test(options)
+  options = {:dir => nil, :pattern => '*'}.update(options)
+  path = File.join([cross_agent_tests_dir, options[:dir], options[:pattern]].compact)
+  Dir.glob(path).each { |file| yield(file) }
+end
+
+def assert_event_attributes(event, test_name, expected_attributes, non_expected_attributes)
+  incorrect_attributes = []
+
+  event_attrs = event[0]
+
+  expected_attributes.each do |name, expected_value|
+    actual_value = event_attrs[name]
+    incorrect_attributes << name unless actual_value == expected_value
+  end
+
+  msg = +"Found missing or incorrect attribute values in #{test_name}:\n"
+
+  incorrect_attributes.each do |name|
+    msg << "  #{name}: expected = #{expected_attributes[name].inspect}, actual = #{event_attrs[name].inspect}\n"
+  end
+  msg << "\n"
+
+  msg << "All event values:\n"
+  event_attrs.each do |name, actual_value|
+    msg << "  #{name}: #{actual_value.inspect}\n"
+  end
+
+  assert_empty(incorrect_attributes, msg)
+
+  non_expected_attributes.each do |name|
+    refute event_attrs[name], "Found value '#{event_attrs[name]}' for attribute '#{name}', but expected nothing in #{test_name}"
+  end
+end
+
+def attributes_for(sample, type)
+  sample.attributes.instance_variable_get("@#{type}_attributes")
+end
+
+def uncache_trusted_account_key
+  NewRelic::Agent::Transaction::TraceContext::AccountHelpers.instance_variable_set(:@trace_state_entry_key, nil)
+end
+
+def reset_buffers_and_caches
+  NewRelic::Agent.drop_buffered_data
+  uncache_trusted_account_key
+end
+
+def message_for_status_code(code)
+  # Net::HTTP::STATUS_CODES was introduced in Ruby 2.5
+  if defined?(Net::HTTP::STATUS_CODES)
+    return Net::HTTP::STATUS_CODES[code]
+  end
+
+  case code
+  when 200 then 'OK'
+  when 404 then 'Not Found'
+  when 403 then 'Forbidden'
+  else 'Unknown'
+  end
+end
+
+# wraps the given headers in a Net::HTTPResponse which has accompanying
+# http status code associated with it.
+# a "status_code" may be passed in the headers to alter the HTTP Status Code
+# that is wrapped in the response.
+def mock_http_response(headers, wrap_it = true)
+  status_code = (headers.delete('status_code') || 200).to_i
+  net_http_resp = Net::HTTPResponse.new(1.0, status_code, message_for_status_code(status_code))
+  headers.each do |key, value|
+    net_http_resp.add_field(key.to_s, value)
+  end
+  return net_http_resp unless wrap_it
+
+  NewRelic::Agent::HTTPClients::NetHTTPResponse.new(net_http_resp)
+end
+
+# +expected+ can be a string or regular expression
+def assert_match_or_equal(expected, value)
+  if expected.is_a?(Regexp)
+    assert_match expected, value
+  else
+    assert_equal expected, value
+  end
+end
+
+# selects the last segment with a noticed_error and checks
+# the expectations against it.
+def assert_segment_noticed_error(txn, segment_name, error_classes, error_message)
+  error_segment = txn.segments.reverse.detect { |s| s.noticed_error }
+
+  assert error_segment, 'Expected at least one segment with a noticed_error'
+
+  assert_match_or_equal segment_name, error_segment.name
+
+  noticed_error = error_segment.noticed_error
+
+  assert_match_or_equal error_classes, noticed_error.exception_class_name
+  assert_match_or_equal error_message, noticed_error.message
+end
+
+def assert_transaction_noticed_error(txn, error_classes)
+  refute_empty txn.exceptions, 'Expected transaction to notice the error'
+  assert_match_or_equal error_classes, txn.exceptions.keys.first.class.name
+end
+
+def refute_transaction_noticed_error(txn, error_class)
+  error_segment = txn.segments.reverse.detect { |s| s.noticed_error }
+
+  assert error_segment, 'Expected at least one segment with a noticed_error'
+  assert_empty txn.exceptions, 'Expected transaction to NOT notice any segment errors'
+end
+
+def refute_raises(*exp)
+  msg = "#{exp.pop}.\n" if String === exp.last
+
+  begin
+    yield
+  rescue MiniTest::Skip => e
+    puts "SKIP REPORTS: #{e.inspect}"
+    return e if exp.include?(MiniTest::Skip)
+
+    raise e
+  rescue Exception => e
+    puts "EXCEPTION RAISED: #{e.inspect}\n#{e.backtrace}"
+    exp = exp.first if exp.size == 1
+
+    flunk(msg || "unexpected exception raised: #{e}")
+  end
+end
+
+def assert_implements(instance, method, *args)
+  fail_message = "expected #{instance.class}##{method} method to be implemented"
+  refute_raises NotImplementedError, fail_message do
+    instance.send(method, *args)
+  end
+end
+
+def defer_testing_to_min_supported_rails(test_file, min_rails_version, supports_jruby = true)
+  if defined?(Rails) &&
+      defined?(Rails::VERSION::STRING) &&
+      (Rails::VERSION::STRING.to_f >= min_rails_version) &&
+      (supports_jruby || !NewRelic::LanguageSupport.jruby?)
+
+    yield
+  else
+    puts "Skipping tests in #{File.basename(test_file)} because Rails >= #{min_rails_version} is unavailable" if ENV['VERBOSE_TEST_OUTPUT']
+  end
+end