datadog 2.15.0 → 2.16.0

data/lib/datadog/di/boot.rb

@@ -0,0 +1,34 @@
+# frozen_string_literal: true
+
+require_relative 'logger'
+require_relative 'base'
+require_relative 'error'
+require_relative 'code_tracker'
+require_relative 'component'
+require_relative 'instrumenter'
+require_relative 'probe'
+require_relative 'probe_builder'
+require_relative 'probe_manager'
+require_relative 'probe_notification_builder'
+require_relative 'probe_notifier_worker'
+require_relative 'redactor'
+require_relative 'serializer'
+require_relative 'transport/http'
+require_relative 'utils'
+
+if %w[1 true yes].include?(ENV['DD_DYNAMIC_INSTRUMENTATION_ENABLED']) # steep:ignore
+  # For initial release of Dynamic Instrumentation, activate code tracking
+  # only if DI is explicitly requested in the environment.
+  # Code tracking is required for line probes to work; see the comments
+  # above for the implementation of the method.
+  #
+  # If DI is enabled programmatically, the application can (and must,
+  # for line probes to work) activate tracking in an initializer.
+  # We seem to have Datadog.logger here already
+  Datadog.logger.debug("di: activating code tracking")
+  Datadog::DI.activate_tracking
+end
+
+require_relative 'contrib'
+
+Datadog::DI::Contrib.load_now_or_later

data/lib/datadog/di/remote.rb

@@ -18,6 +18,8 @@ module Datadog
         PRODUCT = 'LIVE_DEBUGGING'
 
         def products
+          # TODO: do not send our product on unsupported runtimes
+          # (Ruby 2.5 / JRuby)
           [PRODUCT]
         end
 

data/lib/datadog/di.rb

@@ -1,24 +1,8 @@
 # frozen_string_literal: true
 
-require_relative 'di/logger'
-require_relative 'di/base'
-require_relative 'di/error'
-require_relative 'di/code_tracker'
-require_relative 'di/component'
 require_relative 'di/configuration'
 require_relative 'di/extensions'
-require_relative 'di/instrumenter'
-require_relative 'di/probe'
-require_relative 'di/probe_builder'
-require_relative 'di/probe_manager'
-require_relative 'di/probe_notification_builder'
-require_relative 'di/probe_notifier_worker'
-require_relative 'di/redactor'
 require_relative 'di/remote'
-require_relative 'di/serializer'
-#require_relative 'di/transport'
-require_relative 'di/transport/http'
-require_relative 'di/utils'
 
 module Datadog
   # Namespace for Datadog dynamic instrumentation.
@@ -52,19 +36,8 @@ module Datadog
   end
 end
 
-if %w(1 true).include?(ENV['DD_DYNAMIC_INSTRUMENTATION_ENABLED']) # steep:ignore
-  # For initial release of Dynamic Instrumentation, activate code tracking
-  # only if DI is explicitly requested in the environment.
-  # Code tracking is required for line probes to work; see the comments
-  # above for the implementation of the method.
-  #
-  # If DI is enabled programmatically, the application can (and must,
-  # for line probes to work) activate tracking in an initializer.
-  # We seem to have Datadog.logger here already
-  Datadog.logger.debug("di: activating code tracking")
-  Datadog::DI.activate_tracking
-end
-
-require_relative 'di/contrib'
-
-Datadog::DI::Contrib.load_now_or_later
+# Line probes will not work on Ruby < 2.6 because of lack of :script_compiled
+# trace point. Activate DI automatically on supported Ruby versions but
+# always load its settings so that, for example, turning DI off when
+# we are on Ruby 2.5 does not produce exceptions.
+require_relative 'di/boot' if RUBY_VERSION >= '2.6' && RUBY_ENGINE != 'jruby'

data/lib/datadog/error_tracking/collector.rb

@@ -0,0 +1,87 @@
+# frozen_string_literal: true
+
+require_relative 'ext'
+
+module Datadog
+  module ErrorTracking
+    # The Collector is in charge, for a SpanOperation of storing the span events
+    # created when an error is handled. Each SpanOperation has a Collector as soon
+    # as a span event is created and the Collector has the same life time as the SpanOp.
+    #
+    # If an error is handled then rethrown, the SpanEvent corresponding to the error
+    # will be deleted. That is why we do not add directly the SpanEvent to the SpanOp.
+    #
+    # @api private
+    class Collector
+      SPAN_EVENTS_LIMIT = 100
+      LOCK = Mutex.new
+      # Proc called when the span_operation :after_stop event is published
+      def self.after_stop
+        @after_stop ||= proc do |span_op, error|
+          # if this proc is called, we are sure that span_op has a collector
+          collector = span_op.get_collector_or_initialize
+          # if an error exited the scope of the span, we delete the corresponding SpanEvent.
+          collector.on_error(span_op, error) if error
+
+          span_events = collector.span_events
+          span_op.span_events.concat(span_events)
+        end
+      end
+
+      def initialize
+        @span_event_per_error = {}
+      end
+
+      def add_span_event(span_op, span_event, error)
+        # When this is the first time we add a span event for a span,
+        # we suscribe to the :after_stop event
+        if @span_event_per_error.empty?
+          events = span_op.send(:events)
+          events.after_stop.subscribe(&self.class.after_stop)
+
+          # This tag is used by the Error Tracking product to report
+          # the error in Error Tracking
+          span_op.set_tag(Ext::SPAN_EVENTS_HAS_EXCEPTION, true)
+        end
+        # Set a limit to the number of span event we can store per SpanOp
+        # If an error has been handled several times in the same span we can still
+        # modify the event (even if the capacity is reached) in order to report
+        # the information of the last rescue
+        if @span_event_per_error.key?(error) || @span_event_per_error.length < SPAN_EVENTS_LIMIT
+          @span_event_per_error[error] = span_event
+        end
+      end
+
+      if RUBY_VERSION >= Ext::RUBY_VERSION_WITH_RESCUE_EVENT
+        # Starting from ruby3.3, as we are listening to :rescue event,
+        # we just want to remove the span event if the error was
+        # previously handled
+        def on_error(_span_op, error)
+          @span_event_per_error.delete(error)
+        end
+      else
+        # Up to ruby3.2, we are listening to :raise event. We need to ensure
+        # that an error exiting the scope of a span is not handled in a parent span.
+        # This function will propagate the span event to the parent span. If the
+        # error is not handled in the parent span, it will be deleted by design.
+        def on_error(span_op, error)
+          return unless @span_event_per_error.key?(error)
+
+          unless span_op.root?
+            parent = span_op.send(:parent)
+            LOCK.synchronize do
+              parent_collector = parent.get_collector_or_initialize { Collector.new }
+              parent_collector.add_span_event(parent, @span_event_per_error[error], error)
+            end
+          end
+
+          @span_event_per_error.delete(error)
+        end
+      end
+
+      def span_events
+        @span_event_per_error.values
+      end
+    end
+  end
+end

data/lib/datadog/error_tracking/component.rb

@@ -0,0 +1,167 @@
+# frozen_string_literal: true
+
+require 'set'
+require_relative 'collector'
+require_relative 'filters'
+
+module Datadog
+  module ErrorTracking
+    # Component for Error Tracking.
+    #
+    # Only one instance of the Component should ever be active.
+    #
+    # The component instance records every handled exceptions from the configured scopes
+    # (user, third_party packages, specified files or everything).
+    class Component
+      LOCK = Mutex.new
+
+      class << self
+        def build(settings, tracer, logger)
+          return if !settings.respond_to?(:error_tracking) || (settings.error_tracking.handled_errors.nil? &&
+            settings.error_tracking.handled_errors_include.empty?)
+
+          return unless environment_supported?(logger)
+
+          new(
+            tracer: tracer,
+            handled_errors: settings.error_tracking.handled_errors,
+            handled_errors_include: settings.error_tracking.handled_errors_include,
+          ).tap(&:start)
+        end
+
+        def environment_supported?(logger)
+          if RUBY_ENGINE != 'ruby'
+            logger.warn("error tracking: cannot enable error tracking: MRI is required, but running on #{RUBY_ENGINE}")
+            false
+          elsif RUBY_VERSION < '2.6'
+            logger.warn(
+              "error tracking: cannot enable error tracking: Ruby 2.6+ is required, but running
+              on #{RUBY_VERSION}"
+            )
+            false
+          else
+            true
+          end
+        end
+      end
+
+      def initialize(tracer:, handled_errors:, handled_errors_include:)
+        @tracer = tracer
+
+        # Hash containing the paths to the instrumented files
+        @instrumented_files = Set.new unless handled_errors_include.empty?
+        # Array containing file paths, file names and gems names to instrument.
+        # This is coming from the DD_ERROR_TRACKING_HANDLED_ERRORS_INCLUDE env variable
+        @handled_errors_include = handled_errors_include
+
+        # Filter function is used to filter out the exception
+        # we do not want to report. For instance exception from gems.
+        @filter_function = Filters.generate_filter(handled_errors, @instrumented_files)
+
+        # :rescue event was added in Ruby 3.3
+        #
+        # Before Ruby3.3 the TracePoint listen for :raise events.
+        # If an error is not handled, we will delete the according
+        # span event in the collector.
+        event = (RUBY_VERSION >= '3.3') ? :rescue : :raise
+
+        # This TracePoint is in charge of capturing the handled exceptions
+        # and of adding the corresponding span events to the collector
+        @handled_exc_tracker = create_exc_tracker_trace_point(event)
+
+        if @instrumented_files
+          # The only thing we know about the handled errors is the path of the file
+          # in which the error was rescued. Therefore, we need to retrieve the path
+          # of the files the user want to instrument. This TracePoint is used for that
+          # purpose
+          @include_path_getter = create_script_compiled_trace_point
+        end
+      end
+
+      def create_exc_tracker_trace_point(event)
+        TracePoint.new(event) do |tp|
+          active_span = @tracer.active_span
+          if active_span
+            raised_exception = tp.raised_exception
+            # Note that in 3.2, this will give the path of where the error was raised
+            # which may cause the handled_error_include env variable to malfunction.
+            rescue_file_path = tp.path
+            if @filter_function.call(rescue_file_path)
+              span_event = generate_span_event(raised_exception)
+              LOCK.synchronize do
+                collector = active_span.get_collector_or_initialize { Collector.new }
+                collector.add_span_event(active_span, span_event, raised_exception)
+              end
+            end
+          end
+        end
+      end
+
+      def create_script_compiled_trace_point
+        TracePoint.new(:script_compiled) do |tp|
+          next if tp.eval_script
+
+          path = tp.instruction_sequence.path
+          next if path.nil?
+
+          @handled_errors_include.each do |file_to_instr|
+            # The user can provide either
+            # - absolute_path starting with '/'. In that case the path of the file
+            #   should begin with file_to_instr
+            # - a relative_path starting with './'. In that case, we extend the path
+            #   and it is the same as above
+            # - otherwise we just check if the name provided is in the path and is
+            #   either the name of a folder or of a ruby file.
+            regex =
+              if file_to_instr.start_with?('/')
+                %r{\A#{Regexp.escape(file_to_instr)}(?:/|\.rb\z|\z)}
+              elsif file_to_instr.start_with?('./')
+                abs_path = File.expand_path(file_to_instr)
+                %r{\A#{Regexp.escape(abs_path)}(?:/|\.rb\z|\z)}
+              else
+                %r{/#{Regexp.escape(file_to_instr)}(?:/|\.rb\z|\z)}
+              end
+
+            add_instrumented_file(path) if path.match?(regex)
+          end
+        end
+      end
+
+      # Starts the TracePoints.
+      #
+      # Enables the script_compiled TracePoint if handled_errors_include is not empty.
+      def start
+        @handled_exc_tracker.enable
+        @include_path_getter&.enable
+      end
+
+      # Shuts down error tracker.
+      #
+      # Disables the TracePoints.
+      def shutdown!
+        @handled_exc_tracker.disable
+        @include_path_getter&.disable
+      end
+
+      private
+
+      # Generates a span event from the exception info.
+      #
+      # The event follows the otel semantics.
+      # https://opentelemetry.io/docs/specs/otel/trace/exceptions/
+      def generate_span_event(exception)
+        formatted_exception = Datadog::Core::Error.build_from(exception)
+        attributes = {
+          'exception.type' => formatted_exception.type,
+          'exception.message' => formatted_exception.message,
+          'exception.stacktrace' => formatted_exception.backtrace
+        }
+        Datadog::Tracing::SpanEvent.new('exception', attributes: attributes)
+      end
+
+      def add_instrumented_file(file_path)
+        @instrumented_files&.add(file_path)
+      end
+    end
+  end
+end

data/lib/datadog/error_tracking/configuration/settings.rb

@@ -0,0 +1,63 @@
+# frozen_string_literal: true
+
+require_relative '../ext'
+
+module Datadog
+  module ErrorTracking
+    module Configuration
+      # Settings
+      module Settings
+        def self.extended(base)
+          base = base.singleton_class unless base.is_a?(Class)
+          add_settings!(base)
+        end
+
+        def self.add_settings!(base)
+          base.class_eval do
+            # Error Tracking specific configurations.
+            # @public_api
+            settings :error_tracking do
+              # Enable automatic reporting of handled errors and defines the scope
+              # for which to report errors: user code, gems, or both. Possible
+              # values are: all | user | third_party.
+              #
+              # @default 'DD_ERROR_TRACKING_HANDLED_ERRORS' environment variable, otherwise `nil`
+              # @return [String, nil]
+              option :handled_errors do |o|
+                o.type :string, nilable: true
+                o.default Ext::DEFAULT_HANDLED_ERRORS
+                o.env Ext::ENV_HANDLED_ERRORS
+                o.setter do |value|
+                  next value if Ext::VALID_HANDLED_ERRORS.include?(value)
+
+                  unless value.nil?
+                    Datadog.logger.warn(
+                      "Invalid handled errors scope: #{value}. " \
+                      "Supported values are: #{Ext::VALID_HANDLED_ERRORS.join(" | ")}. " \
+                      'Deactivating the feature.'
+                    )
+                  end
+
+                  Ext::DEFAULT_HANDLED_ERRORS
+                end
+              end
+
+              # Enable automatic reporting of handled errors and specify what files should be
+              # instrumented. The value is a list of comma separated paths, filenames or gem names.
+              # The paths can be absolute, starting with '/' or relative to directory in which the program
+              # is launched, starting with './'.
+              #
+              # @default 'DD_ERROR_TRACKING_HANDLED_ERRORS_MODULES' environment variable, otherwise `nil`
+              # @return [String, nil]
+              option :handled_errors_include do |o|
+                o.type :array
+                o.default []
+                o.env Ext::ENV_HANDLED_ERRORS_INCLUDE
+              end
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/datadog/error_tracking/configuration.rb

@@ -0,0 +1,11 @@
+# frozen_string_literal: true
+
+module Datadog
+  module ErrorTracking
+    # Configuration for ErrorTracking
+    module Configuration
+    end
+  end
+end
+
+require_relative 'configuration/settings'

data/lib/datadog/error_tracking/ext.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+module Datadog
+  module ErrorTracking
+    # Error Tracking constants
+    module Ext
+      ENV_HANDLED_ERRORS = 'DD_ERROR_TRACKING_HANDLED_ERRORS'
+      ENV_HANDLED_ERRORS_INCLUDE = 'DD_ERROR_TRACKING_HANDLED_ERRORS_INCLUDE'
+      HANDLED_ERRORS_ALL = 'all'
+      HANDLED_ERRORS_USER = 'user'
+      HANDLED_ERRORS_THIRD_PARTY = 'third_party'
+      DEFAULT_HANDLED_ERRORS = nil
+      VALID_HANDLED_ERRORS = [HANDLED_ERRORS_ALL, HANDLED_ERRORS_USER, HANDLED_ERRORS_THIRD_PARTY].freeze
+      SPAN_EVENTS_HAS_EXCEPTION = '_dd.span_events.has_exception'
+      RUBY_VERSION_WITH_RESCUE_EVENT = '3.3'
+    end
+  end
+end

data/lib/datadog/error_tracking/extensions.rb

@@ -0,0 +1,16 @@
+# frozen_string_literal: true
+
+require_relative 'configuration'
+require_relative '../core/configuration'
+
+module Datadog
+  module ErrorTracking
+    # Extends Datadog tracing with ErrorTracking features
+    module Extensions
+      # Inject Error Tracking into global objects.
+      def self.activate!
+        Core::Configuration::Settings.extend(Configuration::Settings)
+      end
+    end
+  end
+end

data/lib/datadog/error_tracking/filters.rb

@@ -0,0 +1,77 @@
+# frozen_string_literal: true
+
+module Datadog
+  module ErrorTracking
+    # Based on configuration, the TracePoint listening to :rescue or :raise
+    # may report more handled errors than we want to report. Therefore we need
+    # a function to filter the events. As the filter function both depends
+    # on configuration and is called numerous time, we generate it during
+    # during the initialization of the feature to have the best performance
+    # possible.
+    #
+    # @api private
+    module Filters
+      module_function
+
+      def get_gem_name(file_path)
+        regex = %r{gems/([^/]+)-\d}
+        regex_match = regex.match(file_path)
+        return unless regex_match
+
+        gem_name = regex_match[1]
+
+        begin
+          Gem::Specification.find_by_name(gem_name) # steep:ignore
+        rescue Gem::MissingSpecError
+          nil
+        end
+      end
+
+      def user_code?(file_path)
+        !get_gem_name(file_path)
+      end
+
+      def datadog_code?(file_path)
+        file_path.include?('lib/datadog/')
+      end
+
+      def third_party_code?(file_path)
+        gem_name = get_gem_name(file_path)
+        gem_name && gem_name != "datadog"
+      end
+
+      def file_included?(file_path, instrumented_files)
+        instrumented_files.include?(file_path)
+      end
+
+      # Generate the proc used in the TracePoint
+      def generate_filter(to_instrument_scope, handled_errors_include = nil)
+        case to_instrument_scope
+        # If DD_ERROR_TRACKING_HANDLED_ERRORS is set
+        when 'all'
+          proc { |file_path| !datadog_code?(file_path) }
+        when 'user'
+          # If DD_ERROR_TRACKING_HANDLED_ERRORS_INCLUDE is set
+          if handled_errors_include
+            proc { |file_path|
+              user_code?(file_path) || file_included?(file_path, handled_errors_include)
+            }
+          else
+            proc { |file_path| user_code?(file_path) }
+          end
+        when 'third_party'
+          if handled_errors_include
+            proc { |file_path|
+              third_party_code?(file_path) || file_included?(file_path, handled_errors_include)
+            }
+          else
+            proc { |file_path| third_party_code?(file_path) }
+          end
+        else
+          # If only DD_ERROR_TRACKING_HANDLED_ERRORS_INCLUDE is set
+          proc { |file_path| file_included?(file_path, handled_errors_include) }
+        end
+      end
+    end
+  end
+end

data/lib/datadog/error_tracking.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+require_relative 'error_tracking/collector'
+require_relative 'error_tracking/component'
+require_relative 'error_tracking/configuration'
+require_relative 'error_tracking/ext'
+require_relative 'error_tracking/extensions'
+require_relative 'error_tracking/filters'
+
+module Datadog
+  # Namespace for Datadog ErrorTracking.
+  #
+  # @api private
+  module ErrorTracking
+    # Expose ErrorTracking to global shared objects
+    Extensions.activate!
+  end
+end

data/lib/datadog/kit/identity.rb

@@ -70,7 +70,7 @@ module Datadog
             if Datadog::AppSec.active_context
               active_span.set_tag('_dd.appsec.user.collection_mode', 'sdk')
 
-              user = ::Datadog::AppSec::Instrumentation::Gateway::User.new(id, others[:login])
+              user = ::Datadog::AppSec::Instrumentation::Gateway::User.new(id, others[:login], session_id)
               ::Datadog::AppSec::Instrumentation.gateway.push('identity.set_user', user)
             end
           end

data/lib/datadog/profiling/exporter.rb

@@ -49,7 +49,7 @@ module Datadog
         @internal_metadata = internal_metadata
         # NOTE: At the time of this comment collected info does not change over time so we'll hardcode
         #       it on startup to prevent serializing the same info on every flush.
-        @info_json = JSON.fast_generate(info_collector.info).freeze
+        @info_json = JSON.generate(info_collector.info).freeze
       end
 
       def flush

data/lib/datadog/tracing/analytics.rb

@@ -16,7 +16,7 @@ module Datadog
         def set_measured(span_op, value = true)
           return if span_op.nil?
 
-          value = value == true || value == 1 ? 1 : 0 # rubocop:disable Style/MultipleComparison
+          value = value == true || value == 1 ? 1 : 0
           span_op.set_metric(Metadata::Ext::Analytics::TAG_MEASURED, value)
         end
       end

data/lib/datadog/tracing/contrib/karafka/distributed/propagation.rb

@@ -31,6 +31,8 @@ module Datadog
                     Tracing::Distributed::Datadog.new(fetcher: Tracing::Distributed::Fetcher),
                   Tracing::Configuration::Ext::Distributed::PROPAGATION_STYLE_TRACE_CONTEXT =>
                     Tracing::Distributed::TraceContext.new(fetcher: Tracing::Distributed::Fetcher),
+                  Tracing::Configuration::Ext::Distributed::PROPAGATION_STYLE_BAGGAGE =>
+                    Tracing::Distributed::Baggage.new(fetcher: Tracing::Distributed::Fetcher),
                   Tracing::Configuration::Ext::Distributed::PROPAGATION_STYLE_NONE => Tracing::Distributed::None.new
                 },
                 propagation_style_inject: propagation_style_inject,

data/lib/datadog/tracing/contrib/karafka/monitor.rb

@@ -15,7 +15,7 @@ module Datadog
             worker.processed
           ].freeze
 
-          def instrument(event_id, payload = EMPTY_HASH, &block)
+          def instrument(event_id, payload = {}, &block)
             return super unless TRACEABLE_EVENTS.include?(event_id)
 
             Datadog::Tracing.trace(Ext::SPAN_WORKER_PROCESS) do |span|

data/lib/datadog/tracing/contrib/mongodb/configuration/settings.rb

@@ -48,6 +48,14 @@ module Datadog
               o.type :string, nilable: true
               o.env Ext::ENV_PEER_SERVICE
             end
+
+            # Serializes the command to JSON format, which is the desired format for the agent and Datadog UI.
+            # Setting this to false is deprecated and does not have any advantages.
+            option :json_command do |o|
+              o.type :bool
+              o.env Ext::ENV_JSON_COMMAND
+              o.default false
+            end
           end
         end
       end

data/lib/datadog/tracing/contrib/mongodb/ext.rb

@@ -13,6 +13,7 @@ module Datadog
           # @!visibility private
           ENV_ANALYTICS_ENABLED = 'DD_TRACE_MONGO_ANALYTICS_ENABLED'
           ENV_ANALYTICS_SAMPLE_RATE = 'DD_TRACE_MONGO_ANALYTICS_SAMPLE_RATE'
+          ENV_JSON_COMMAND = 'DD_TRACE_MONGO_JSON_COMMAND'
           DEFAULT_PEER_SERVICE_NAME = 'mongodb'
           SPAN_COMMAND = 'mongo.cmd'
           SPAN_TYPE_COMMAND = 'mongodb'

data/lib/datadog/tracing/contrib/mongodb/subscribers.rb

@@ -1,5 +1,7 @@
 # frozen_string_literal: true
 
+require 'json'
+
 require_relative '../analytics'
 require_relative 'ext'
 require_relative '../ext'
@@ -30,7 +32,7 @@ module Datadog
 
             # build a quantized Query using the Parser module
             query = MongoDB.query_builder(event.command_name, event.database_name, event.command)
-            serialized_query = query.to_s
+            serialized_query = serialize_query(query)
 
             if datadog_configuration[:peer_service]
               span.set_tag(
@@ -109,6 +111,21 @@ module Datadog
 
           private
 
+          def serialize_query(query)
+            if datadog_configuration[:json_command]
+              query.to_json
+            else
+              # Incorrect Hash#to_s serialization. The Mongo command should only be encoded as JSON.
+              # This code path should be removed, and is only kept to avoid a breaking change.
+              Datadog::Core.log_deprecation(key: :mongo_json_command) do
+                'MongoDB integration: `json_command: false` causes invalid command serialization. '\
+                'Use `json_command: true` or `DD_TRACE_MONGO_JSON_COMMAND=1` instead.'
+              end
+
+              query.to_s
+            end
+          end
+
           def get_span(event)
             Thread.current[:datadog_mongo_span] \
               && Thread.current[:datadog_mongo_span][event.request_id]