Support return job result as exitcode in hosted runner.

[TingluoHuang]

https://github.com/github/hosted-compute/issues/3923

[TingluoHuang]

We didn't use the bool other than set it to true before.
So i am expose the job result via it.

[Copilot]

Pull request overview

Adds support for hosted runners to return the completed job’s TaskResult as the listener process exit code, enabling improved telemetry/alerting on job failures.

Changes:

• Change IJobDispatcher.RunOnceJobCompleted from TaskCompletionSource<bool> to TaskCompletionSource<TaskResult>.
• Update Runner run-once loop to optionally translate the completed job TaskResult into the process return code when ACTIONS_RUNNER_RETURN_JOB_RESULT_FOR_HOSTED is enabled.
• Update L0 tests to reflect the new TaskResult-based completion signal.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/Runner.Listener/Runner.cs	Adds hosted-runner env-var toggle and returns exit code based on TaskResult for run-once execution.
src/Runner.Listener/JobDispatcher.cs	Propagates run-once job completion as a TaskResult instead of a boolean.
src/Test/L0/Listener/RunnerL0.cs	Updates mocks to complete run-once with TaskResult.Succeeded.
src/Test/L0/Listener/JobDispatcherL0.cs	Updates assertions to validate TaskResult completion.

Comments suppressed due to low confidence (2)

src/Runner.Listener/JobDispatcher.cs:352

• RunOnceAsync initializes jobResult to Succeeded and sets _runOnceJobCompleted in finally. If RunAsync(...) throws, the completion source will still be set to Succeeded, which can cause hosted runners to exit with a success code even though dispatch failed. Consider catching exceptions around RunAsync and setting an appropriate failure result (or setting an exception on the TCS) before signaling completion.

        private async Task RunOnceAsync(Pipelines.AgentJobRequestMessage message, string orchestrationId, WorkerDispatcher previousJobDispatch, CancellationToken jobRequestCancellationToken, CancellationToken workerCancelTimeoutKillToken)
        {
            var jobResult = TaskResult.Succeeded;
            try
            {
                jobResult = await RunAsync(message, orchestrationId, previousJobDispatch, jobRequestCancellationToken, workerCancelTimeoutKillToken);
            }
            finally
            {
                Trace.Info("Fire signal for one time used runner.");
                _runOnceJobCompleted.TrySetResult(jobResult);
            }

src/Runner.Listener/Runner.cs:592

• This PR adds new behavior to return the run-once job TaskResult as the listener process exit code, but there doesn’t appear to be an L0 test that sets returnRunOnceJobResult/ACTIONS_RUNNER_RETURN_JOB_RESULT_FOR_HOSTED and asserts the returned code matches TaskResultUtil.TranslateToReturnCode(...) (including a failure case). Adding a focused unit test would prevent regressions in the exit-code mapping.

                                    if (returnRunOnceJobResult)
                                    {
                                        try
                                        {
                                            var jobResult = await jobDispatcher.RunOnceJobCompleted.Task;
                                            return TaskResultUtil.TranslateToReturnCode(jobResult);
                                        }