cw-datadog 2.23.0.2

data/lib/datadog/core/configuration.rb

@@ -0,0 +1,328 @@
+# frozen_string_literal: true
+
+require_relative 'configuration/components'
+require_relative 'configuration/settings'
+require_relative 'telemetry/emitter'
+require_relative 'logger'
+require_relative 'pin'
+
+module Datadog
+  module Core
+    # Configuration provides a unique access point for configurations
+    module Configuration
+      # Used to ensure that @components initialization/reconfiguration is performed one-at-a-time, by a single thread.
+      #
+      # This is important because components can end up being accessed from multiple application threads (for instance on
+      # a threaded webserver), and we don't want their initialization to clash (for instance, starting two profilers...).
+      #
+      # Note that a Mutex **IS NOT** reentrant: the same thread cannot grab the same Mutex more than once.
+      # This means below we are careful not to nest calls to methods that would trigger initialization and grab the lock.
+      #
+      # Every method that directly or indirectly mutates @components should be holding the lock (through
+      # #safely_synchronize) while doing so.
+      COMPONENTS_WRITE_LOCK = Mutex.new
+      private_constant :COMPONENTS_WRITE_LOCK
+
+      # We use a separate lock when reading the @components, so that they continue to be accessible during reconfiguration.
+      # This was needed because we ran into several issues where we still needed to read the old
+      # components while the COMPONENTS_WRITE_LOCK was being held (see https://github.com/DataDog/dd-trace-rb/pull/1387
+      # and https://github.com/DataDog/dd-trace-rb/pull/1373#issuecomment-799593022 ).
+      #
+      # Technically on MRI we could get away without this lock, but on non-MRI Rubies, we may run into issues because
+      # we fall into the "UnsafeDCLFactory" case of https://shipilev.net/blog/2014/safe-public-construction/ .
+      # Specifically, on JRuby reads from the @components do NOT have volatile semantics, and on TruffleRuby they do
+      # BUT just as an implementation detail, see https://github.com/jruby/jruby/wiki/Concurrency-in-jruby#volatility and
+      # https://github.com/DataDog/dd-trace-rb/pull/1329#issuecomment-776750377 .
+      # Concurrency is hard.
+      COMPONENTS_READ_LOCK = Mutex.new
+      private_constant :COMPONENTS_READ_LOCK
+
+      attr_writer :configuration
+
+      # Current Datadog configuration.
+      #
+      # Access to non-global configuration will raise an error.
+      #
+      # To modify the configuration, use {.configure}.
+      #
+      # @return [Datadog::Core::Configuration::Settings]
+      # @!attribute [r] configuration
+
+      # @public_api
+      def configuration
+        @configuration ||= Settings.new
+      end
+
+      # Apply global configuration changes to `Datadog`. An example of a {.configure} call:
+      #
+      # ```
+      # Datadog.configure do |c|
+      #   c.service = 'my-service'
+      #   c.env = 'staging'
+      #   # c.diagnostics.debug = true # Enables debug output
+      # end
+      # ```
+      #
+      # See {Datadog::Core::Configuration::Settings} for all available options, defaults, and
+      # available environment variables for configuration.
+      #
+      # Only permits access to global configuration settings; others will raise an error.
+      # If you wish to configure a setting for a specific Datadog component (e.g. Tracing),
+      # use the corresponding `Datadog::COMPONENT.configure` method instead.
+      #
+      # Because many configuration changes require restarting internal components,
+      # invoking {.configure} is the only safe way to change `Datadog` configuration.
+      #
+      # Successive calls to {.configure} maintain the previous configuration values:
+      # configuration is additive between {.configure} calls.
+      #
+      # The yielded configuration `c` comes pre-populated from environment variables, if
+      # any are applicable.
+      #
+      # @yieldparam [Datadog::Core::Configuration::Settings] c the mutable configuration object
+      def configure
+        configuration = self.configuration
+        yield(configuration)
+
+        safely_synchronize do |write_components|
+          write_components.call(
+            if components?
+              replace_components!(configuration, @components)
+            else
+              build_components(configuration)
+            end
+          )
+        end
+
+        configuration
+      end
+
+      # Apply configuration changes only to a specific Ruby object.
+      #
+      # Certain integrations or Datadog features may use these
+      # settings to customize behavior for this object.
+      #
+      # An example of a {.configure_onto} call:
+      #
+      # ```
+      # client = Net::HTTP.new(host, port)
+      # Datadog.configure_onto(client, service_name: 'api-requests', split_by_domain: true)
+      # ```
+      #
+      # In this example, it will configure the `client` object with custom options
+      # `service_name: 'api-requests', split_by_domain: true`. The `Net::HTTP` integration
+      # will then use these customized options when the `client` is used, whereas other
+      # clients will use the `service_name: 'http-requests'` configuration provided to the
+      # `Datadog.configure` call block.
+      #
+      # {.configure_onto} is used to separate cases where spans generated by certain objects
+      # require exceptional options.
+      #
+      # The configuration keyword arguments provided should match well known options defined
+      # in the integration or feature that would use them.
+      #
+      # For example, for `Datadog.configure_onto(redis_client, **opts)`, `opts` can be
+      # any of the options in the Redis {Datadog::Tracing::Contrib::Redis::Configuration::Settings} class.
+      #
+      # @param [Object] target the object to receive configuration options
+      # @param [Hash] opts keyword arguments respective to the integration this object belongs to
+      # @public_api
+      def configure_onto(target, **opts)
+        Pin.set_on(target, **opts)
+      end
+
+      # Get configuration changes applied only to a specific Ruby object, via {.configure_onto}.
+      # An example of an object with specific configuration:
+      #
+      # ```
+      # client = Net::HTTP.new(host, port)
+      # Datadog.configure_onto(client, service_name: 'api-requests', split_by_domain: true)
+      # config = Datadog.configuration_for(client)
+      # config[:service_name] # => 'api-requests'
+      # config[:split_by_domain] # => true
+      # ```
+      #
+      # @param [Object] target the object to receive configuration options
+      # @param [Object] option an option to retrieve from the object configuration
+      # @public_api
+      def configuration_for(target, option = nil)
+        pin = Pin.get_from(target)
+        return pin unless option
+
+        pin[option] if pin
+      end
+
+      # Internal {Datadog::Statsd} metrics collection.
+      #
+      # The list of metrics collected can be found in {Datadog::Core::Diagnostics::Ext::Health::Metrics}.
+      # @public_api
+      def health_metrics
+        components.health_metrics
+      end
+
+      def logger
+        # avoid initializing components if they didn't already exist
+        current_components = components(allow_initialization: false)
+
+        if current_components
+          @temp_logger = nil
+          current_components.logger
+        else
+          logger_without_components
+        end
+      end
+
+      def config_init_logger
+        configuration? ? logger : logger_without_configuration
+      end
+
+      # Gracefully shuts down all components.
+      #
+      # Components will still respond to method calls as usual,
+      # but might not internally perform their work after shutdown.
+      #
+      # This avoids errors being raised across the host application
+      # during shutdown, while allowing for graceful decommission of resources.
+      #
+      # Components won't be automatically reinitialized after a shutdown.
+      def shutdown!
+        safely_synchronize do
+          @components.shutdown! if components?
+        end
+      end
+
+      protected
+
+      def components(allow_initialization: true)
+        current_components = COMPONENTS_READ_LOCK.synchronize { defined?(@components) && @components }
+        return current_components if current_components || !allow_initialization
+
+        safely_synchronize do |write_components|
+          if defined?(@components) && @components
+            @components
+          else
+            write_components.call(build_components(configuration))
+          end
+        end
+      end
+
+      private
+
+      # Gracefully shuts down Datadog components and disposes of component references,
+      # allowing execution to start anew.
+      #
+      # In contrast with +#shutdown!+, components will be automatically
+      # reinitialized after a reset.
+      #
+      # Used internally to ensure a clean environment between test runs.
+      def reset!
+        safely_synchronize do |write_components|
+          if components?
+            @components.shutdown!
+            @temp_logger = nil # Reset to ensure instance and log level are reset for next run
+          end
+
+          write_components.call(nil)
+          configuration.reset!
+        end
+      end
+
+      def safely_synchronize
+        # Writes to @components should only happen through this proc. Because this proc is only accessible to callers of
+        # safely_synchronize, this forces all writers to go through this method.
+        write_components = proc do |new_value|
+          COMPONENTS_READ_LOCK.synchronize { @components = new_value }
+        end
+
+        COMPONENTS_WRITE_LOCK.synchronize do
+          yield write_components
+        rescue ThreadError => e
+          logger_without_components.error(
+            "Detected deadlock during datadog initialization: #{e.class}: #{e}. " \
+            'Please report this at https://github.com/datadog/dd-trace-rb/blob/master/CONTRIBUTING.md#found-a-bug' \
+            "\n\tSource:\n\t#{Array(e.backtrace).join("\n\t")}"
+          )
+          nil
+        end
+      end
+
+      def configuration?
+        (defined?(@configuration) && @configuration) != nil
+      end
+
+      def components?
+        # This does not need to grab the COMPONENTS_READ_LOCK because it's not returning the components
+        (defined?(@components) && @components) != nil
+      end
+
+      def build_components(settings)
+        components = Components.new(settings)
+        components.startup!(settings)
+        components
+      end
+
+      def replace_components!(settings, old)
+        components = Components.new(settings)
+
+        # Carry over state from existing components to the new ones.
+        # Currently:
+        # 1. If we already started the remote component (which
+        #    happens after a request goes through installed Rack middleware),
+        #    we will start the new remote component as well.
+        # 2. If telemetry has been enabled and is enabled in the new
+        #    component tree, we send AppConfigurationChange event instead
+        #    of AppStarted event.
+        old_state = old.state
+
+        old.shutdown!(components)
+        components.startup!(settings, old_state: old_state)
+        components
+      end
+
+      def logger_without_components
+        # Use default logger without initializing components.
+        # This enables logging during initialization, otherwise we'd run into deadlocks.
+
+        @temp_logger ||= begin
+          logger = configuration.logger.instance || Core::Logger.new($stdout)
+          logger.level = configuration.diagnostics.debug ? ::Logger::DEBUG : configuration.logger.level
+          logger
+        end
+      end
+
+      def logger_without_configuration
+        # There's rare cases where we need to use logger during configuration initialization,
+        # such as reading stable config. In this case we cannot access configuration.
+
+        @temp_config_logger ||= begin
+          debug_env_value = DATADOG_ENV[Ext::Diagnostics::ENV_DEBUG_ENABLED]&.strip&.downcase
+          debug_value = debug_env_value == 'true' || debug_env_value == '1'
+
+          logger = Core::Logger.new($stdout)
+          # We cannot access config and the default level is INFO, so we need to set the level manually
+          logger.level = debug_value ? ::Logger::DEBUG : ::Logger::INFO
+          logger
+        end
+      end
+
+      # Called from our at_exit hook whenever there was a pending Interrupt exception (e.g. typically due to ctrl+c)
+      # to print a nice message whenever we're taking a bit longer than usual to finish the process.
+      def handle_interrupt_shutdown!
+        logger = Datadog.logger
+        shutdown_thread = Thread.new { shutdown! }
+        shutdown_thread.name = Datadog::Core::Configuration.name
+
+        print_message_treshold_seconds = 0.2
+
+        slow_shutdown = shutdown_thread.join(print_message_treshold_seconds).nil?
+
+        if slow_shutdown
+          logger.info 'Reporting remaining data... Press ctrl+c to exit immediately.'
+          shutdown_thread.join
+        end
+
+        nil
+      end
+    end
+  end
+end

data/lib/datadog/core/contrib/rails/utils.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module Datadog
+  module Core
+    module Contrib
+      module Rails
+        # common utilities for Rails
+        module Utils
+          def self.app_name
+            if ::Rails::VERSION::MAJOR >= 6
+              ::Rails.application.class.module_parent_name.underscore
+            else
+              ::Rails.application.class.parent_name.underscore
+            end
+          end
+
+          def self.railtie_supported?
+            !!defined?(::Rails::Railtie)
+          end
+        end
+      end
+    end
+  end
+end

data/lib/datadog/core/crashtracking/component.rb

@@ -0,0 +1,105 @@
+# frozen_string_literal: true
+
+require 'libdatadog'
+
+require_relative 'tag_builder'
+require_relative '../utils/only_once'
+require_relative '../utils/at_fork_monkey_patch'
+
+module Datadog
+  module Core
+    module Crashtracking
+      # Used to report Ruby VM crashes.
+      #
+      # NOTE: The crashtracker native state is a singleton;
+      # so even if you create multiple instances of `Crashtracking::Component` and start them,
+      # it only works as "last writer wins". Same for stop -- there's only one state, so calling stop
+      # on it will stop the crash tracker, regardless of which instance started it.
+      #
+      # Methods prefixed with _native_ are implemented in `crashtracker.c`
+      class Component
+        ONLY_ONCE = Core::Utils::OnlyOnce.new
+
+        def self.build(settings, agent_settings, logger:)
+          tags = TagBuilder.call(settings)
+          agent_base_url = agent_settings.url
+
+          ld_library_path = ::Libdatadog.ld_library_path
+          logger.warn('Missing ld_library_path; cannot enable crash tracking') unless ld_library_path
+
+          path_to_crashtracking_receiver_binary = ::Libdatadog.path_to_crashtracking_receiver_binary
+          unless path_to_crashtracking_receiver_binary
+            logger.warn('Missing path_to_crashtracking_receiver_binary; cannot enable crash tracking')
+          end
+
+          return unless agent_base_url
+          return unless ld_library_path
+          return unless path_to_crashtracking_receiver_binary
+
+          new(
+            tags: tags,
+            agent_base_url: agent_base_url,
+            ld_library_path: ld_library_path,
+            path_to_crashtracking_receiver_binary: path_to_crashtracking_receiver_binary,
+            logger: logger
+          ).tap(&:start)
+        end
+
+        def initialize(tags:, agent_base_url:, ld_library_path:, path_to_crashtracking_receiver_binary:, logger:)
+          @tags = tags
+          @agent_base_url = agent_base_url
+          @ld_library_path = ld_library_path
+          @path_to_crashtracking_receiver_binary = path_to_crashtracking_receiver_binary
+          @logger = logger
+        end
+
+        def start
+          Utils::AtForkMonkeyPatch.apply!
+
+          start_or_update_on_fork(action: :start, tags: tags)
+
+          ONLY_ONCE.run do
+            Utils::AtForkMonkeyPatch.at_fork(:child) do
+              # Must NOT reference `self` here, as only the first instance will
+              # be captured by the ONLY_ONCE and we want to pick the latest active one
+              # (which may have different tags or agent config)
+              Datadog.send(:components, allow_initialization: false)&.crashtracker&.update_on_fork
+            end
+          end
+        end
+
+        def update_on_fork(settings: Datadog.configuration)
+          # Here we pick up the latest settings, so that we pick up any tags that change after forking
+          # such as the pid or runtime-id
+          start_or_update_on_fork(action: :update_on_fork, tags: TagBuilder.call(settings))
+        end
+
+        def stop
+          self.class._native_stop
+          logger.debug('Crash tracking stopped successfully')
+        rescue => e
+          logger.error("Failed to stop crash tracking: #{e.message}")
+        end
+
+        private
+
+        attr_reader :tags, :agent_base_url, :ld_library_path, :path_to_crashtracking_receiver_binary, :logger
+
+        def start_or_update_on_fork(action:, tags:)
+          self.class._native_start_or_update_on_fork(
+            action: action,
+            agent_base_url: agent_base_url,
+            path_to_crashtracking_receiver_binary: path_to_crashtracking_receiver_binary,
+            ld_library_path: ld_library_path,
+            tags_as_array: tags.to_a,
+            # @ivoanjo: On my machine this needs to be > 5 seconds, and seems to work with 10; the extra 15 is extra margin
+            upload_timeout_seconds: 15,
+          )
+          logger.debug("Crash tracking action: #{action} successful")
+        rescue => e
+          logger.error("Failed to #{action} crash tracking: #{e.message}")
+        end
+      end
+    end
+  end
+end

data/lib/datadog/core/crashtracking/tag_builder.rb

@@ -0,0 +1,21 @@
+# frozen_string_literal: true
+
+require_relative '../tag_builder'
+require_relative '../utils'
+
+module Datadog
+  module Core
+    module Crashtracking
+      # This module builds a hash of tags
+      module TagBuilder
+        def self.call(settings)
+          hash = Core::TagBuilder.tags(settings).merge(
+            'is_crash' => 'true',
+          )
+
+          Utils.encode_tags(hash)
+        end
+      end
+    end
+  end
+end

data/lib/datadog/core/ddsketch.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+module Datadog
+  module Core
+    # Used to access ddsketch APIs.
+    # APIs in this class are implemented as native code.
+    class DDSketch
+      def self.supported?
+        Datadog::Core::LIBDATADOG_API_FAILURE.nil?
+      end
+
+      def initialize
+        unless self.class.supported?
+          raise(ArgumentError, "DDSketch is not supported: #{Datadog::Core::LIBDATADOG_API_FAILURE}")
+        end
+      end
+    end
+  end
+end

data/lib/datadog/core/deprecations.rb

@@ -0,0 +1,58 @@
+# frozen_string_literal: true
+
+module Datadog
+  module Core
+    # Contains behavior for handling deprecated functions in the codebase.
+    module Deprecations
+      # Records the occurrence of a deprecated operation in this library.
+      #
+      # Currently, these operations are logged to `Datadog.logger` at `warn` level.
+      #
+      # `disallowed_next_major` adds a message informing that the deprecated operation
+      # won't be allowed in the next major release.
+      #
+      # @yieldreturn [String] a String with the lazily evaluated deprecation message.
+      # @param [Boolean] disallowed_next_major whether this deprecation will be enforced in the next major release.
+      # @param [Object] key A unique key for the deprecation. Only the first message with the same key will be logged.
+      def log_deprecation(disallowed_next_major: true, key: nil, logger: Datadog.logger)
+        return unless log_deprecation?(key)
+
+        logger.warn do
+          message = yield
+          message += ' This will be enforced in the next major release.' if disallowed_next_major
+          message
+        end
+
+        # Track the deprecation being logged.
+        deprecation_logged!(key)
+
+        nil
+      end
+
+      private
+
+      # Determines whether a deprecation message should be logged.
+      #
+      # Internal use only.
+      def log_deprecation?(key)
+        return true if key.nil?
+
+        # Only allow a deprecation to be logged once.
+        !logged_deprecations.key?(key)
+      end
+
+      def deprecation_logged!(key)
+        return if key.nil?
+
+        logged_deprecations[key] += 1
+      end
+
+      # Tracks what deprecation warnings have already been logged
+      #
+      # Internal use only.
+      def logged_deprecations
+        @logged_deprecations ||= Hash.new(0)
+      end
+    end
+  end
+end