inferno_core 1.1.2 → 1.2.0

data/lib/inferno/apps/cli/execute_script.rb

@@ -0,0 +1,918 @@
+require 'English'
+require 'yaml'
+require_relative 'session/create_session'
+require_relative 'session/start_run'
+require_relative 'session/cancel_run'
+require_relative 'session/session_status'
+require_relative 'session/session_results'
+require_relative 'session/session_compare'
+require_relative 'session/session_details'
+require_relative 'session/session_data'
+
+module Inferno
+  module CLI
+    # Orchestrates multi-session Inferno test runs from a YAML configuration file.
+    #
+    # YAML format:
+    #
+    #   sessions:                                  # Controls the creation of Inferno sessions for the execution
+    #                                              #   a session for each indicated suite will be created and
+    #                                              #   a successful run will have the expected results for all sessions.
+    #     - suite: my_suite                        # internal ID, title, or short title
+    #       name: my_name                          # optional; used as key in multi-session
+    #       preset: my-preset                      # optional; internal ID or title
+    #       suite_options:                         # optional; option_key and option_value can be the
+    #         option_key: option_value             #           internal values or the displayed titles
+    #
+    #
+    #   comparison_config:                         # optional; Controls the comparison of actual results during a run
+    #                                              #    to the expected results for each created Inferno session.
+    #                                              #    When the configured expected results file for a session does
+    #                                              #    not exist, the run will be considered a failure and the expected
+    #                                              #    results file will be generated using the results from the run.
+    #                                              #    Developers are responsible for verifying that the results match
+    #                                              #    their expectations before committing those expected results.
+    #                                              #    When expected results are present, they must match for the run
+    #                                              #    to be successful. When the results for a session do not match
+    #                                              #    the expected results, an actual results JSON file and a CSV diff
+    #                                              #    are written to the directory of the expected results file for
+    #                                              #    use in evaluating the failure.
+    #     normalized_strings:                      # optional; global normalization rules applied to
+    #       - "http://my-server.example.com"       # both expected and actual before comparing.
+    #       - "http://other-value.example.com"     # plain string: replaced with <NORMALIZED>;
+    #                                              # URL-encoded form is also replaced automatically.
+    #       - "/code_challenge=[A-Za-z0-9+\\/=_-]{20,}/"  # regex string (wrapped in /…/): compiled
+    #                                              # to a Regexp and replaced with <NORMALIZED>.
+    #                                              # Supports flags: /pattern/i, /pattern/m, etc.
+    #                                              # URL-encoded form is NOT auto-replaced for regex.
+    #       - pattern: '/code_challenge=[A-Za-z0-9+\\/=_-]{20,}/'  # hash form: use when you need
+    #         replacement: '<CODE_CHALLENGE>'      # a named placeholder instead of <NORMALIZED>.
+    #       - patterns:                            # 'patterns' (plural) shares one replacement
+    #           - '/code_challenge=[A-Za-z0-9+\\/=_-]{20,}/'  # across multiple patterns.
+    #           - '/code_verifier=[A-Za-z0-9+\\/=_-]{20,}/'
+    #         replacement: '<PKCE_VALUE>'
+    #
+    #     # Single-session scripts: expected file config lives directly under comparison_config.
+    #     expected_results_file: expected.json     # optional; relative to yaml file; defaults to
+    #                                              #   <yaml basename>_expected.json
+    #     alternate_expected_files:                # optional; tried in order; first matching wins
+    #       - file: alt_expected.json              # required; relative to yaml file
+    #         when:                                # required; all conditions must match (AND logic)
+    #           - field: inputs.url                # required: can be inputs.<name>, configuration_messages,
+    #                                              #   or inferno_base_url
+    #             matches: ^http://                # other values are top-level session detail fields.
+    #                                              # Evaluated against GET api/test_sessions/{id}.
+    #
+    #     # Multi-session scripts: per-session config is nested under sessions.<name>.
+    #     sessions:
+    #       my_name:                               # matches session name key (sessions[*].name or suite)
+    #         expected_results_file: expected.json # optional; defaults to <yaml basename>_<name>_expected.json
+    #         alternate_expected_files:            # optional; same structure as single-session above
+    #           - file: alt_expected.json
+    #             when:
+    #               - field: inputs.url
+    #                 matches: ^http://
+    #
+    #   steps:                                     # Details the steps taken in the execution of the script
+    #     - status: created|done|waiting           # required; other status values cannot be matched on
+    #       last_completed: "1.01"                 # optional; required unless the status is created
+    #                                              #           can be a full ID, short ID (e.g. "1.01"), or 'suite'
+    #       session: my_name                       # optional; required when multiple sessions to indicate which session
+    #                                              #           can match this step
+    #       action: END_SCRIPT|NOOP|WAIT           # optional; built-in action (case-insensitive)
+    #                                              #           (mutually exclusive with command/start_run)
+    #       # OR
+    #       command: "bundle exec ..."             # optional; arbitrary shell command (requires
+    #                                              #           --allow-commands CLI flag)
+    #       # OR
+    #       start_run:
+    #         session: "my_name"                   # optional; session name or template token
+    #                                              #           (e.g. {session_id}); defaults to current session
+    #         runnable: "1.01"                     # required; can be a short id from the UI,
+    #                                              #           an internal long id, or 'suite'
+    #         inputs:                              # optional; the key must be the internal name
+    #           input_name: "value"                #           for the input
+    #           input_name: "@path/to/file.txt"    #           prefix value with @ to read from a file;
+    #                                              #           relative paths are resolved from the
+    #                                              #           directory containing this script file
+    #           input_name:                        #           YAML mappings and sequences are
+    #             key: value                       #           automatically JSON-serialized, useful
+    #           input_name:                        #           for auth_info and other JSON inputs
+    #             - item1
+    #       timeout: 300                           # optional; seconds to wait for next match (Default is 120)
+    #       next_poll_session: other_name          # optional; switch polling target after command
+    #       state_description: "..."               # optional; logged when step is matched
+    #       action_description: "..."              # optional; logged when step is matched
+    #
+    # Built-in action values:
+    #   END_SCRIPT  — terminate script successfully
+    #   NOOP        — no-op; keep polling with (optionally updated) timeout or session
+    #   WAIT        — keep polling without breaking out of the current poll loop
+    #                 (unlike noop, does not restart the loop with a new timeout or session)
+    #
+    # Security note:
+    #   Steps using command: execute arbitrary shell commands via the system() call. This is
+    #   intentional for use cases like browser automation (e.g. Selenium) that must interact
+    #   with Inferno's waiting state mid-test. Because of this risk, command: steps are blocked
+    #   by default and require the --allow-commands CLI flag to run. Scripts that contain no
+    #   command: steps are unaffected and do not need the flag.
+    #
+    # Template tokens in command strings and start_run input values:
+    #   {session_id}              — current session's Inferno session ID
+    #   {NAME.session_id}         — named session's ID
+    #   {result_message}          — current session's wait_result_message
+    #   {NAME.result_message}     — named session's wait_result_message
+    #   {wait_outputs.KEY}        — current session's wait output by name
+    #   {NAME.wait_outputs.KEY}   — named session's wait output by name
+    #   {inferno_base_url}        — the Inferno base URL (--inferno-base-url option)
+    class ExecuteScript
+      SHORT_ID_PATTERN = /\A[0-9][0-9.]*\z/
+      # Seconds subtracted from the initial last_log_time so the first
+      # active-status line is logged immediately rather than after one interval.
+      LOG_INTERVAL_SECONDS = 30
+
+      ScriptSession = Struct.new(
+        :key, :suite_id, :session_id, :short_id_map,
+        keyword_init: true
+      )
+
+      ExecutionStatus = Struct.new(
+        :done, :failed, :timed_out, :current_session, :current_timeout, :last_log_time,
+        :cross_session_status, :last_step_signatures
+      )
+
+      attr_accessor :yaml_file, :options, :execution_status
+
+      def initialize(yaml_file, options)
+        self.yaml_file = yaml_file
+        self.options = options
+        validate_yaml_file!
+        validate_commands_allowed!
+        self.execution_status = ExecutionStatus.new(
+          done: false,
+          failed: false,
+          timed_out: false,
+          current_session: sessions.first,
+          current_timeout: options[:default_poll_timeout],
+          cross_session_status: {},
+          last_step_signatures: {}
+        )
+      end
+
+      def run
+        exit(orchestrate)
+      end
+
+      private
+
+      def script_config
+        @script_config ||= YAML.safe_load_file(File.expand_path(yaml_file))
+      end
+
+      def sessions
+        @sessions ||= create_sessions
+      end
+
+      def session_name_to_id_map
+        @session_name_to_id_map ||= sessions.to_h { |session| [session.key, session.session_id] }
+      end
+
+      def steps
+        @steps ||= scripted_steps || []
+      end
+
+      def multi_session_script?
+        script_config['sessions'].present? &&
+          (script_config['sessions'].length > 1)
+      end
+
+      def yaml_directory
+        @yaml_directory ||= File.dirname(File.expand_path(yaml_file))
+      end
+
+      def yaml_basename
+        @yaml_basename ||= File.basename(yaml_file, '.yaml')
+      end
+
+      def validate_yaml_file!
+        unless File.exist?(yaml_file)
+          puts JSON.pretty_generate({ errors: "File not found: #{yaml_file}" })
+          exit(1)
+        end
+        return if yaml_file.end_with?('.yaml', '.yml')
+
+        puts JSON.pretty_generate(
+          { errors: "'#{yaml_file}' does not appear to be a YAML file (.yaml or .yml extension required)." }
+        )
+        exit(1)
+      end
+
+      def validate_commands_allowed!
+        return if options[:allow_commands]
+        return if Array(script_config['steps']).none? { |step| step['command'].present? }
+
+        puts JSON.pretty_generate(
+          { errors: "Script contains 'command' steps but --allow-commands was not set. " \
+                    'Re-run with --allow-commands to permit arbitrary shell command execution.' }
+        )
+        exit(1)
+      end
+
+      # ---------------------------------------------------------------------------
+      # Session creation
+      # ---------------------------------------------------------------------------
+
+      def create_sessions
+        script_config['sessions']&.map do |session_config|
+          create_session_from_config(session_config)
+        end || []
+      end
+
+      def create_session_from_config(session_config)
+        suite = session_config['suite']
+        warn "Creating '#{suite}' session..."
+        creator = Session::CreateSession.new(suite, session_create_options(session_config))
+        session_details = creator.create_session
+        key = session_config['name'] || suite
+        warn "Session created: #{session_details['id']}"
+        ScriptSession.new(
+          key: key,
+          suite_id: session_details['test_suite_id'],
+          session_id: session_details['id'],
+          short_id_map: extract_short_ids_from_session_details(session_details)
+        )
+      end
+
+      def session_create_options(session_config)
+        {
+          preset: session_config['preset'],
+          suite_options: session_config['suite_options'],
+          inferno_base_url: options[:inferno_base_url]
+        }
+      end
+
+      def extract_short_ids_from_session_details(session_details)
+        results = {}
+        results['suite'] = session_details['test_suite_id']
+        suite_details = session_details['test_suite']
+        suite_details['test_groups']&.each do |group|
+          extract_short_ids(group, results)
+        end
+        suite_details['tests']&.each do |test|
+          extract_short_ids(test, results)
+        end
+
+        results
+      end
+
+      # Recursive runnable walk collecting { short_id => full_id } pairs.
+      def extract_short_ids(runnable, results)
+        results[runnable['short_id']] = runnable['id'] if runnable['short_id'] && runnable['id']
+        runnable['test_groups']&.each do |group|
+          extract_short_ids(group, results)
+        end
+        runnable['tests']&.each do |test|
+          extract_short_ids(test, results)
+        end
+      end
+
+      # ---------------------------------------------------------------------------
+      # Extract Steps - resolve short ids in the last_completed entries
+      # ---------------------------------------------------------------------------
+
+      def scripted_steps
+        script_config['steps']&.map do |step|
+          last_completed = step['last_completed'].to_s
+          next step unless last_completed.match?(SHORT_ID_PATTERN) || last_completed == 'suite'
+
+          resolved_test_id = resolve_short_id(last_completed, step['session'])
+          warn "Resolved short ID \"#{last_completed}\" -> \"#{resolved_test_id}\""
+          step.merge('last_completed' => resolved_test_id)
+        end
+      end
+
+      def resolve_short_id(short_id, session_name)
+        if multi_session_script? && session_name.blank?
+          puts JSON.pretty_generate(
+            {
+              errors: "Short ID '#{short_id}' used in step 'last_completed' " \
+                      "without a 'session' in a multi-session script."
+            }
+          )
+          exit(3)
+        end
+
+        target_session = session_for_name(session_name)
+        unless target_session
+          puts JSON.pretty_generate(
+            { errors: "Session '#{session_name}' not found in script sessions" }
+          )
+          exit(3)
+        end
+
+        test_id = target_session.short_id_map&.dig(short_id)
+        unless test_id
+          puts JSON.pretty_generate(
+            { errors: "Short ID '#{short_id}' not found in session '#{target_session.key}'" }
+          )
+          exit(3)
+        end
+        test_id
+      end
+
+      def session_for_name(session_name)
+        if session_name.present?
+          sessions.find { |session| session.key == session_name }
+        else
+          sessions.first
+        end
+      end
+
+      # ---------------------------------------------------------------------------
+      # Orchestration loop
+      # ---------------------------------------------------------------------------
+
+      def orchestrate
+        loop do
+          matched_step = poll_for_next_step
+          execution_status.done = true if [nil, 'END_SCRIPT'].include?(matched_step[:command])
+          break if execution_status.done
+
+          take_step(matched_step)
+          break if execution_status.done
+        end
+
+        warn ''
+        warn "All runs complete #{execution_status.failed ? 'with errors' : 'successfully'}."
+
+        if results_match_expected?(sessions)
+          execution_status.failed ? 3 : 0
+        else
+          3
+        end
+      end
+
+      # perform the specified command to continue the script
+      def take_step(matched_step)
+        execution_status.current_timeout = next_step_timeout(matched_step)
+        if matched_step[:next_poll_session].present?
+          execution_status.current_session = session_for_name(matched_step[:next_poll_session])
+        end
+        return if matched_step[:command] == 'NOOP'
+
+        if matched_step[:command] == 'START_RUN'
+          target_session_id = start_run_target_session(matched_step[:start_run])
+          warn "Executing: #{start_run_description(matched_step[:start_run], session_id: target_session_id)}"
+          execute_start_run(matched_step[:start_run], target_session_id)
+        else
+          warn "Executing: #{matched_step[:command]}"
+          execution_status.failed = true unless execute_command(matched_step[:command])
+          execution_status.done = execution_status.failed
+        end
+      end
+
+      def next_step_timeout(matched_step)
+        matched_step[:timeout].present? ? matched_step[:timeout] : options[:default_poll_timeout]
+      end
+
+      def start_run_target_session(start_run_details)
+        if start_run_details['session'].present?
+          if session_name_to_id_map[start_run_details['session']].present?
+            session_name_to_id_map[start_run_details['session']]
+          else
+            start_run_details['session']
+          end
+        elsif multi_session_script?
+          puts JSON.pretty_generate({ errors: 'Start run steps must have `session` defined when multiple sessions.' })
+          exit(3)
+        else
+          execution_status.current_session.session_id
+        end
+      end
+
+      # templates already resolved
+      def execute_start_run(start_run_config, target_session_id)
+        start_run_options = {
+          runnable: start_run_config['runnable'],
+          inputs: start_run_config['inputs'],
+          inferno_base_url: options[:inferno_base_url]
+        }
+
+        runner = Session::StartRun.new(target_session_id, start_run_options)
+        runner.start_run
+      end
+
+      # ---------------------------------------------------------------------------
+      # Polling
+      # ---------------------------------------------------------------------------
+
+      def poll_for_next_step
+        session = execution_status.current_session
+        timeout = execution_status.current_timeout
+
+        warn ''
+        warn "Polling session: #{session.key} (#{session.session_id}) timeout=#{timeout}s"
+        deadline = Time.now + timeout
+        execution_status.last_log_time = Time.now - LOG_INTERVAL_SECONDS
+
+        loop do
+          status = fetch_session_status(session.session_id)
+          run_status = status['status']
+
+          case run_status
+          when 'running', 'queued', 'cancelling'
+            log_poll_if_needed(status, session.key)
+          when 'waiting', 'done', 'created'
+            execution_status.cross_session_status = {} # reset per poll cycle
+            execution_status.cross_session_status[session.key] = status
+            result = handle_actionable_status(status, session, timeout)
+            return result if result
+          end
+
+          if Time.now >= deadline
+            warn "Timeout after #{timeout}s: session=#{session.key} status=#{run_status}"
+            execution_status.failed = true
+            execution_status.timed_out = true
+            return { command: nil, timeout: timeout, next_poll_session: nil }
+          end
+
+          sleep options[:poll_interval]
+        end
+      end
+
+      # Returns a step hash to act on, or nil to keep polling.
+      def handle_actionable_status(status, session, timeout)
+        matched_step = match_step(status, session.key)
+
+        if matched_step
+          handle_matched_step(matched_step, status, session, timeout)
+        elsif status['status'] == 'waiting'
+          handle_unmatched_wait(status, session)
+        else
+          handle_unmatched_status(status, session, timeout)
+        end
+      end
+
+      def handle_matched_step(matched_step, status, session, timeout)
+        return nil if matched_step[:command] == 'WAIT'
+
+        verify_step(matched_step, status, session, timeout)
+      end
+
+      def handle_unmatched_wait(status, session)
+        last_completed = format_last_completed(last_completed_from_status(status), session.key)
+        warn "UNHANDLED WAIT - Canceling: session=#{session.key} last_completed=#{last_completed}"
+        execution_status.failed = true
+        attempt_cancel(session.session_id, status)
+        nil
+      end
+
+      def handle_unmatched_status(status, session, timeout)
+        run_status = status['status']
+        last_completed = format_last_completed(last_completed_from_status(status), session.key)
+        warn "UNMATCHED: session=#{session.key} status=#{run_status} last_completed=#{last_completed}"
+        execution_status.failed = true
+        { command: nil, timeout: timeout, next_poll_session: nil }
+      end
+
+      # Checks for a repeated steps that aren't NOOPs; returns nil to keep polling or the step to act on.
+      def verify_step(matched_step, status, session, timeout)
+        run_status = status['status']
+        last_completed = last_completed_from_status(status)
+        step_sig = [run_status, last_completed]
+
+        if step_sig == execution_status.last_step_signatures[session.key] && matched_step[:command] != 'NOOP'
+          execution_status.failed = true
+          if run_status == 'waiting'
+            warn "Loop detected - Canceling: session=#{session.key} last_completed=#{last_completed}"
+            attempt_cancel(session.session_id, status)
+            return nil
+          else
+            warn "Loop detected: session=#{session.key} status=#{run_status} " \
+                 "last_completed=#{last_completed}"
+            return { command: nil, timeout: timeout, next_poll_session: nil }
+          end
+        end
+
+        execution_status.last_step_signatures[session.key] = step_sig
+        matched_step
+      end
+
+      def log_poll_if_needed(status, session_key)
+        return unless Time.now - execution_status.last_log_time >= 30
+
+        last_completed = last_completed_from_status(status)
+        poll_status_last_test =
+          last_completed.present? ? " - last test: #{format_last_completed(last_completed, session_key)}" : ''
+        warn "  [#{session_key}] #{status['status']}#{poll_status_last_test}"
+        execution_status.last_log_time = Time.now
+      end
+
+      def fetch_session_status(session_id)
+        Session::SessionStatus.new(session_id, options).status_for_session
+      end
+
+      def attempt_cancel(session_id, status)
+        Session::CancelRun.new(session_id, options).cancel_run(status)
+      end
+
+      # ---------------------------------------------------------------------------
+      # Step matching
+      # ---------------------------------------------------------------------------
+
+      def match_step(status, session_key)
+        run_status = status['status']
+        last_completed = last_completed_from_status(status)
+
+        matched = find_matching_step(run_status, last_completed, session_key)
+        return nil unless matched
+
+        log_matched_step(matched, last_completed, session_key)
+        step_details = resolve_command(matched, status, session_key)
+        step_details[:timeout] = matched['timeout'].to_i if matched['timeout'].present?
+        step_details[:next_poll_session] = matched['next_poll_session'] if matched['next_poll_session'].present?
+
+        step_details
+      end
+
+      # when status['status'] is done, return the executed runnable that was completed
+      # otherwise, return last_test_executed
+      def last_completed_from_status(status)
+        if status['status'] == 'done'
+          (status['test_id'] || status['test_group_id'] || status['test_suite_id']).to_s
+        else
+          status['last_test_executed'].to_s
+        end
+      end
+
+      def log_matched_step(matched, last_completed, session_key)
+        warn 'Matched step:'
+        warn "  State: #{matched['state_description']}" if matched['state_description'].present?
+        warn "  status=#{matched['status']} last_completed=#{format_last_completed(last_completed, session_key)}"
+        warn "  Command: #{step_command_description(matched)}"
+        matched_step_optional_lines(matched)
+      end
+
+      def matched_step_optional_lines(matched)
+        warn "  Action: #{matched['action_description']}" if matched['action_description'].present?
+        warn "  Next poll session: #{matched['next_poll_session']}" if matched['next_poll_session'].present?
+        warn "  Timeout: #{matched['timeout']}" if matched['timeout'].present?
+      end
+
+      def format_last_completed(last_completed, session_key)
+        return '(none)' if last_completed.empty?
+
+        short_id = session_for_name(session_key)&.short_id_map&.key(last_completed)
+        short_id ? "#{last_completed} (#{short_id})" : last_completed
+      end
+
+      def find_matching_step(run_status, last_completed, session_key)
+        steps.find do |step|
+          step['status'] == run_status &&
+            step['last_completed'].to_s == last_completed &&
+            (step['session'].blank? || step['session'] == session_key)
+        end
+      end
+
+      # ---------------------------------------------------------------------------
+      # Command resolution
+      # ---------------------------------------------------------------------------
+
+      # Returns a human-readable description of the step's command for logging.
+      def step_command_description(step)
+        if step.key?('start_run')
+          start_run_description(step['start_run'] || {})
+        elsif step.key?('action')
+          "action: #{step['action']}"
+        else
+          step['command'].to_s
+        end
+      end
+
+      def start_run_description(start_run_details, session_id: nil)
+        session_token =
+          if session_id.present?
+            session_id
+          elsif start_run_details['session'].present?
+            "{#{start_run_details['session']}.session_id}"
+          else
+            '{session_id}'
+          end
+        runnable_token = start_run_details['runnable'].present? ? " '#{start_run_details['runnable']}'" : ''
+        parts = ["bundle exec inferno session start_run '#{session_token}'#{runnable_token}"]
+        if start_run_details['inputs'].present?
+          parts << "-i #{start_run_details['inputs'].map do |k, v|
+            "#{k}:#{v}"
+          end.join(' ')}"
+        end
+        parts.join(' ')
+      end
+
+      def resolve_command(step, status, session_key)
+        if step.key?('start_run')
+          start_run = apply_templates_to_start_run(step['start_run'] || {}, status, session_key)
+          { command: 'START_RUN', start_run: }
+        elsif step.key?('action')
+          { command: step['action'].upcase }
+        else
+          cmd = step['command'].to_s
+          { command: cmd.include?('{') ? apply_templates(cmd, status, session_key) : cmd }
+        end
+      end
+
+      # ---------------------------------------------------------------------------
+      # Template token substitution
+      # ---------------------------------------------------------------------------
+
+      # Substitutes all {token} placeholders in +str+.
+      # Cross-session tokens trigger an on-demand status fetch (cached in
+      # execution_status.cross_session_status, reset each poll cycle in poll_for_next_step).
+      #
+      # Returns the resolved string or exits 3 if a token cannot be resolved.
+      def apply_templates(str, status, session_key)
+        result = str.dup
+
+        # {inferno_base_url} — the Inferno base URL
+        result.gsub!('{inferno_base_url}') { options[:inferno_base_url] }
+
+        # {session_id} — current session
+        result.gsub!('{session_id}') { session_name_to_id_map[session_key] }
+
+        # {NAME.session_id} — named session
+        result.gsub!(/\{([^}.]+)\.session_id\}/) do
+          name = Regexp.last_match(1)
+          id   = session_name_to_id_map[name]
+          unless id
+            puts JSON.pretty_generate({ errors: "Unknown session name '#{name}' in token {#{name}.session_id}" })
+            exit(3)
+          end
+          id
+        end
+
+        # {result_message} — current session wait result message
+        result.gsub!('{result_message}') do
+          msg = status['wait_result_message']
+          unless msg
+            puts JSON.pretty_generate({ errors: 'Token {result_message} used but session is not in waiting state' })
+            exit(3)
+          end
+          msg
+        end
+
+        # {NAME.result_message} — another session's wait result message
+        result.gsub!(/\{([^}.]+)\.result_message\}/) do
+          name       = Regexp.last_match(1)
+          other_stat = cross_session_status(name)
+          msg        = other_stat['wait_result_message']
+          unless msg
+            puts JSON.pretty_generate(
+              { errors: "Token {#{name}.result_message} used but session '#{name}' is not in waiting state" }
+            )
+            exit(3)
+          end
+          msg
+        end
+
+        # {wait_outputs.KEY} — current session wait output
+        result.gsub!(/\{wait_outputs\.([^}]+)\}/) do
+          key    = Regexp.last_match(1)
+          output = find_wait_output(status['wait_outputs'], key)
+          unless output
+            puts JSON.pretty_generate({ errors: "Wait output '#{key}' not found in current session" })
+            exit(3)
+          end
+          output
+        end
+
+        # {NAME.wait_outputs.KEY} — another session's wait output
+        result.gsub!(/\{([^}.]+)\.wait_outputs\.([^}]+)\}/) do
+          name       = Regexp.last_match(1)
+          key        = Regexp.last_match(2)
+          other_stat = cross_session_status(name)
+          output     = find_wait_output(other_stat['wait_outputs'], key)
+          unless output
+            puts JSON.pretty_generate(
+              { errors: "Wait output '#{key}' not found in session '#{name}'" }
+            )
+            exit(3)
+          end
+          output
+        end
+
+        result
+      end
+
+      def find_wait_output(wait_outputs, key)
+        Array(wait_outputs).find { |o| o['name'] == key }&.dig('value')
+      end
+
+      # Returns the cached enriched status for a named session, fetching it on
+      # first access within a poll cycle.
+      def cross_session_status(name)
+        execution_status.cross_session_status[name] ||= begin
+          session_id = session_name_to_id_map[name]
+          unless session_id
+            puts JSON.pretty_generate({ errors: "Unknown session name '#{name}'" })
+            exit(3)
+          end
+          fetch_session_status(session_id)
+        end
+      end
+
+      def apply_templates_to_start_run(start_run, status, session_key)
+        if start_run['session'].present?
+          start_run['session'] = apply_templates(start_run['session'], status, session_key)
+        end
+        start_run['inputs']&.each_key do |input_name|
+          raw = start_run['inputs'][input_name]
+          raw = raw.to_json if raw.is_a?(Array) || raw.is_a?(Hash)
+          value = apply_templates(raw.to_s, status, session_key)
+          start_run['inputs'][input_name] = expand_file_input_path(value)
+        end
+
+        start_run
+      end
+
+      def expand_file_input_path(value)
+        return value unless value.start_with?('@')
+
+        path = value[1..]
+        expanded = Pathname.new(path).absolute? ? path : File.expand_path(path, File.dirname(yaml_file))
+
+        unless File.exist?(expanded)
+          puts JSON.pretty_generate({ errors: "File input not found: #{expanded}" })
+          exit(3)
+        end
+
+        "@#{expanded}"
+      end
+
+      # ---------------------------------------------------------------------------
+      # Command execution
+      # ---------------------------------------------------------------------------
+
+      def execute_command(cmd)
+        system(cmd)
+        $CHILD_STATUS.success?
+      end
+
+      # ---------------------------------------------------------------------------
+      # check results
+      # ---------------------------------------------------------------------------
+
+      def results_match_expected?(sessions)
+        sessions.map do |session|
+          warn "Checking results for #{session.key} session (#{session.session_id})"
+          warn "  View session at #{session_display_url(session)}"
+
+          if execution_status.timed_out
+            warn '  Session timed out - skipping comparison'
+            false
+          elsif any_error_results?(session)
+            warn '  Session contained execution errors - skipping comparison'
+            false
+          else
+            compare_session(session)
+          end
+        end.all?
+      end
+
+      def session_display_url(session)
+        base_url = (options[:inferno_base_url].presence || Inferno::Application['base_url']).to_s.delete_suffix('/')
+        "#{base_url}/#{session.suite_id}/#{session.session_id}"
+      end
+
+      def any_error_results?(session)
+        results = Session::SessionResults.new(session.session_id, options).results_for_session(session.session_id)
+        results.any? { |result| result['result'] == 'error' }
+      end
+
+      # ---------------------------------------------------------------------------
+      # Compare / save results
+      # ---------------------------------------------------------------------------
+
+      def resolve_expected_file(key)
+        session_cfg = comparison_session_config(key)
+        if session_cfg['expected_results_file'].present?
+          File.expand_path(session_cfg['expected_results_file'], yaml_directory)
+        elsif multi_session_script?
+          File.join(yaml_directory, "#{yaml_basename}_#{key}_expected.json")
+        else
+          File.join(yaml_directory, "#{yaml_basename}_expected.json")
+        end
+      end
+
+      def comparison_session_config(key)
+        if multi_session_script?
+          (comparison_config['sessions'] || {})[key] || {}
+        else
+          comparison_config
+        end
+      end
+
+      def compare_session(session)
+        expected_file = resolve_expected_file_for_comparison(session)
+
+        if File.exist?(expected_file)
+          cmp = Session::SessionCompare.new(session.session_id, compare_options(expected_file))
+          matched = cmp.results_match?
+          warn "  Comparing against #{File.basename(expected_file)}"
+          warn "  Actual results matched expected results? #{matched}"
+          unless matched
+            cmp.save_actual_results_to_file
+            cmp.save_comparison_csv_to_file
+          end
+          matched
+        else
+          warn "  Expected results file not found; writing actual results to #{expected_file}"
+          results = Session::SessionResults.new(session.session_id, options).results_for_session(session.session_id)
+          File.write(expected_file, results.to_json)
+          false
+        end
+      end
+
+      def resolve_expected_file_for_comparison(session)
+        default_file = resolve_expected_file(session.key)
+        alternates = Array(comparison_session_config(session.key)['alternate_expected_files'])
+        return default_file if alternates.empty?
+
+        session_details = Session::SessionDetails.new(session.session_id, options).details_for_session
+        session_inputs = Session::SessionData.new(session.session_id, options).session_data
+
+        alternates.each do |alt|
+          conditions = Array(alt['when'])
+          next if conditions.empty?
+          unless conditions.all? { |cond| session_detail_condition_matches?(cond, session_details, session_inputs) }
+            next
+          end
+
+          file = File.expand_path(alt['file'], yaml_directory)
+          return file
+        end
+
+        default_file
+      end
+
+      def session_detail_condition_matches?(condition, session_details, session_inputs)
+        return false unless condition_has_field_and_pattern?(condition)
+
+        field = condition['field']
+        matches_pattern = condition['matches']
+        not_matches_pattern = condition['not_matches']
+
+        if field == 'configuration_messages'
+          configuration_messages_match?(session_details['test_suite'] || {}, matches_pattern, not_matches_pattern)
+        elsif field == 'inferno_base_url'
+          string_matches?(options[:inferno_base_url].to_s, matches_pattern, not_matches_pattern)
+        elsif field.start_with?('inputs.')
+          input_name = field.delete_prefix('inputs.')
+          test_suite_input_matches?(session_inputs || [], input_name, matches_pattern, not_matches_pattern)
+        else
+          false
+        end
+      end
+
+      def condition_has_field_and_pattern?(condition)
+        condition['field'].present? && (condition['matches'].present? || condition['not_matches'].present?)
+      end
+
+      def string_matches?(value, matches_pattern, not_matches_pattern)
+        return false if matches_pattern && !Regexp.new(matches_pattern).match?(value)
+        return false if not_matches_pattern && Regexp.new(not_matches_pattern).match?(value)
+
+        true
+      end
+
+      def configuration_messages_match?(test_suite, matches_pattern, not_matches_pattern)
+        messages = Array(test_suite['configuration_messages'])
+        if matches_pattern && messages.none? { |msg| Regexp.new(matches_pattern).match?(msg['message'].to_s) }
+          return false
+        end
+        if not_matches_pattern && messages.any? { |msg| Regexp.new(not_matches_pattern).match?(msg['message'].to_s) }
+          return false
+        end
+
+        true
+      end
+
+      def test_suite_input_matches?(suite_data, input_name, matches_pattern, not_matches_pattern)
+        value = suite_data.find { |input| input['name'] == input_name }&.dig('value')
+        string_matches?(value.to_s, matches_pattern, not_matches_pattern)
+      end
+
+      def comparison_config
+        @comparison_config ||= script_config['comparison_config'] || {}
+      end
+
+      def compare_options(expected_file)
+        {
+          expected_results_file: expected_file,
+          compare_messages: options[:compare_messages],
+          compare_result_message: options[:compare_result_message],
+          inferno_base_url: options[:inferno_base_url],
+          normalized_strings: Array(comparison_config['normalized_strings'])
+        }
+      end
+    end
+  end
+end