src/status-report.ts

import * as os from "os";

import * as core from "@actions/core";

import {
  getWorkflowEventName,
  getOptionalInput,
  getWorkflowRunID,
  getWorkflowRunAttempt,
  getActionVersion,
  getRequiredInput,
  isSelfHostedRunner,
} from "./actions-util";
import { getAnalysisKey, getApiClient } from "./api-client";
import { type Config } from "./config-utils";
import { DocUrl } from "./doc-url";
import { EnvVar } from "./environment";
import { getRef } from "./git-utils";
import { Logger } from "./logging";
import { getRepositoryNwo } from "./repository";
import {
  ConfigurationError,
  isHTTPError,
  getRequiredEnvParam,
  getCachedCodeQlVersion,
  isInTestMode,
  GITHUB_DOTCOM_URL,
  DiskUsage,
  assertNever,
  BuildMode,
  getErrorMessage,
} from "./util";

export enum ActionName {
  Autobuild = "autobuild",
  Analyze = "finish",
  Init = "init",
  InitPost = "init-post",
  ResolveEnvironment = "resolve-environment",
  UploadSarif = "upload-sarif",
}

/**
 * @returns a boolean indicating whether the analysis is considered to be first party.
 *
 * This is based on whether the init action has been used, which is only used for first party analysis.
 * When a SARIF file has been generated by other means and submitted using the upload action, this is
 * considered to be a third party analysis and is treated differently when calculating SLOs. To ensure
 * misconfigured workflows are not treated as third party, only the upload-sarif action can return false.
 */
export function isFirstPartyAnalysis(actionName: ActionName): boolean {
  if (actionName !== ActionName.UploadSarif) {
    return true;
  }
  return process.env[EnvVar.INIT_ACTION_HAS_RUN] === "true";
}

/**
 * @returns true if the analysis is considered to be third party.
 */
export function isThirdPartyAnalysis(actionName: ActionName): boolean {
  return !isFirstPartyAnalysis(actionName);
}

export type ActionStatus =
  | "aborted" // Only used in the init Action, if init failed before initializing the tracer due to something other than a configuration error.
  | "failure"
  | "starting"
  | "success"
  | "user-error";

/** Overall status of the entire job. String values match the Hydro schema. */
export enum JobStatus {
  UnknownStatus = "JOB_STATUS_UNKNOWN",
  SuccessStatus = "JOB_STATUS_SUCCESS",
  FailureStatus = "JOB_STATUS_FAILURE",
  ConfigErrorStatus = "JOB_STATUS_CONFIGURATION_ERROR",
}

export interface StatusReportBase {
  /** Name of the action being executed. */
  action_name: ActionName;
  /** Version of the action being executed, as a commit oid. */
  action_oid: string;
  /** Version of the action being executed, as a ref. */
  action_ref?: string;
  /** Time this action started. */
  action_started_at: string;
  /** Action version (x.y.z from package.json). */
  action_version: string;
  /** The name of the Actions event that triggered the workflow. */
  actions_event_name?: string;
  /** Analysis key, normally composed from the workflow path and job name. */
  analysis_key: string;
  /** Build mode, if specified. */
  build_mode?: BuildMode;
  /** Cause of the failure (or undefined if status is not failure). */
  cause?: string;
  /** CodeQL CLI version (x.y.z from the CLI). */
  codeql_version?: string;
  /** Commit oid that the workflow was triggered on. */
  commit_oid: string;
  /** Time this action completed, or undefined if not yet completed. */
  completed_at?: string;
  /** Stack trace of the failure (or undefined if status is not failure). */
  exception?: string;
  /** Whether this is a first-party (CodeQL) run of the action. */
  first_party_analysis: boolean;
  /** Job name from the workflow. */
  job_name: string;
  /**
   * UUID representing the job run that this status report belongs to. We
   * generate our own UUID here because Actions currently does not expose a
   * unique job run identifier. This UUID will allow us to more easily match
   * reports from different steps in the same workflow job.
   *
   * If and when Actions does expose a unique job ID, we plan to populate a
   * separate int field, `job_run_id`, with the Actions-generated identifier,
   * as it will allow us to more easily join our telemetry data with Actions
   * telemetry tables.
   */
  job_run_uuid: string;
  /**
   * Comma-separated list of languages that analysis was run for.
   *
   * This may be from the workflow file or may be calculated from the contents of the repository.
   */
  languages?: string;
  /** Value of the matrix for this instantiation of the job. */
  matrix_vars?: string;
  /**
   * Information about the enablement of the ML-powered JS query pack.
   *
   * @see {@link util.getMlPoweredJsQueriesStatus}
   */
  ml_powered_javascript_queries?: string;
  /** Ref that the workflow was triggered on. */
  ref: string;
  /** Action runner hardware architecture (context runner.arch). */
  runner_arch?: string;
  /** Available disk space on the runner, in bytes. */
  runner_available_disk_space_bytes?: number;
  /**
   * Version of the runner image, for workflows running on GitHub-hosted runners. Absent otherwise.
   */
  runner_image_version?: string;
  /** Action runner operating system (context runner.os). */
  runner_os: string;
  /** Action runner operating system release (x.y.z from os.release()). */
  runner_os_release?: string;
  /** Total disk space on the runner, in bytes. */
  runner_total_disk_space_bytes?: number;
  /** Time the first action started. Normally the init action. */
  started_at: string;
  /** State this action is currently in. */
  status: ActionStatus;
  /** Whether this run is part of a steady-state, and not new, default setup run. */
  steady_state_default_setup: boolean;
  /**
   * Testing environment: Set if non-production environment.
   * The server accepts one of the following values:
   *  `["", "qa-rc", "qa-rc-1", "qa-rc-2", "qa-experiment-1", "qa-experiment-2", "qa-experiment-3"]`.
   */
  testing_environment: string;
  /** Workflow name. Converted to analysis_name further down the pipeline.. */
  workflow_name: string;
  /** Attempt number of the run containing the action run. */
  workflow_run_attempt: number;
  /** ID of the workflow run containing the action run. */
  workflow_run_id: number;
}

export interface DatabaseCreationTimings {
  scanned_language_extraction_duration_ms?: number;
  trap_import_duration_ms?: number;
}

export function getActionsStatus(
  error?: unknown,
  otherFailureCause?: string,
): ActionStatus {
  if (error || otherFailureCause) {
    return error instanceof ConfigurationError ? "user-error" : "failure";
  } else {
    return "success";
  }
}

export function getJobStatusDisplayName(status: JobStatus): string {
  switch (status) {
    case JobStatus.SuccessStatus:
      return "success";
    case JobStatus.FailureStatus:
      return "failure";
    case JobStatus.ConfigErrorStatus:
      return "configuration error";
    case JobStatus.UnknownStatus:
      return "unknown";
    default:
      assertNever(status);
  }
}

/**
 * Sets the overall job status environment variable to configuration error
 * or failure, unless it's already been set to one of these values in a
 * previous step.
 */
function setJobStatusIfUnsuccessful(actionStatus: ActionStatus) {
  if (actionStatus === "user-error") {
    core.exportVariable(
      EnvVar.JOB_STATUS,
      process.env[EnvVar.JOB_STATUS] ?? JobStatus.ConfigErrorStatus,
    );
  } else if (actionStatus === "failure" || actionStatus === "aborted") {
    core.exportVariable(
      EnvVar.JOB_STATUS,
      process.env[EnvVar.JOB_STATUS] ?? JobStatus.FailureStatus,
    );
  }
}

// Any status report may include an array of EventReports associated with it.
export interface EventReport {
  /** Time this event ended. */
  completed_at: string;
  /** An enumerable description of the event. */
  event: string;
  /** eg: `success`, `failure`, `timeout`, etc. */
  exit_status?: string;
  /** If the event is language-specific. */
  language?: string;
  /**
   * A generic JSON blob of data related to this event.
   * Use Object.assign() to append additional fields to the object.
   */
  properties?: object;
  /** Time this event started. */
  started_at: string;
}

/**
 * Compose a StatusReport.
 *
 * @param actionName The name of the action, e.g. 'init', 'finish', 'upload-sarif'
 * @param status The status. Must be 'success', 'failure', or 'starting'
 * @param startedAt The time this action started executing.
 * @param cause  Cause of failure (only supply if status is 'failure')
 * @param exception Exception (only supply if status is 'failure')
 * @returns undefined if an exception was thrown.
 */
export async function createStatusReportBase(
  actionName: ActionName,
  status: ActionStatus,
  actionStartedAt: Date,
  config: Config | undefined,
  diskInfo: DiskUsage | undefined,
  logger: Logger,
  cause?: string,
  exception?: string,
): Promise<StatusReportBase | undefined> {
  try {
    const commitOid =
      getOptionalInput("sha") || process.env["GITHUB_SHA"] || "";
    const ref = await getRef();
    const jobRunUUID = process.env[EnvVar.JOB_RUN_UUID] || "";
    const workflowRunID = getWorkflowRunID();
    const workflowRunAttempt = getWorkflowRunAttempt();
    const workflowName = process.env["GITHUB_WORKFLOW"] || "";
    const jobName = process.env["GITHUB_JOB"] || "";
    const analysis_key = await getAnalysisKey();
    let workflowStartedAt = process.env[EnvVar.WORKFLOW_STARTED_AT];
    if (workflowStartedAt === undefined) {
      workflowStartedAt = actionStartedAt.toISOString();
      core.exportVariable(EnvVar.WORKFLOW_STARTED_AT, workflowStartedAt);
    }
    const runnerOs = getRequiredEnvParam("RUNNER_OS");
    const codeQlCliVersion = getCachedCodeQlVersion();
    const actionRef = process.env["GITHUB_ACTION_REF"] || "";
    const testingEnvironment = process.env[EnvVar.TESTING_ENVIRONMENT] || "";
    // re-export the testing environment variable so that it is available to subsequent steps,
    // even if it was only set for this step
    if (testingEnvironment !== "") {
      core.exportVariable(EnvVar.TESTING_ENVIRONMENT, testingEnvironment);
    }
    const isSteadyStateDefaultSetupRun =
      process.env["CODE_SCANNING_IS_STEADY_STATE_DEFAULT_SETUP"] === "true";

    const statusReport: StatusReportBase = {
      action_name: actionName,
      action_oid: "unknown", // TODO decide if it's possible to fill this in
      action_ref: actionRef,
      action_started_at: actionStartedAt.toISOString(),
      action_version: getActionVersion(),
      analysis_key,
      build_mode: config?.buildMode,
      commit_oid: commitOid,
      first_party_analysis: isFirstPartyAnalysis(actionName),
      job_name: jobName,
      job_run_uuid: jobRunUUID,
      ref,
      runner_os: runnerOs,
      started_at: workflowStartedAt,
      status,
      steady_state_default_setup: isSteadyStateDefaultSetupRun,
      testing_environment: testingEnvironment,
      workflow_name: workflowName,
      workflow_run_attempt: workflowRunAttempt,
      workflow_run_id: workflowRunID,
    };

    try {
      statusReport.actions_event_name = getWorkflowEventName();
    } catch (e) {
      logger.warning(`Could not determine the workflow event name: ${e}.`);
    }

    if (config) {
      statusReport.languages = config.languages.join(",");
    }

    if (diskInfo) {
      statusReport.runner_available_disk_space_bytes =
        diskInfo.numAvailableBytes;
      statusReport.runner_total_disk_space_bytes = diskInfo.numTotalBytes;
    }

    // Add optional parameters
    if (cause) {
      statusReport.cause = cause;
    }
    if (exception) {
      statusReport.exception = exception;
    }
    if (
      status === "success" ||
      status === "failure" ||
      status === "aborted" ||
      status === "user-error"
    ) {
      statusReport.completed_at = new Date().toISOString();
    }
    const matrix = getRequiredInput("matrix");
    if (matrix) {
      statusReport.matrix_vars = matrix;
    }
    if ("RUNNER_ARCH" in process.env) {
      // RUNNER_ARCH is available only in GHES 3.4 and later
      // Values other than X86, X64, ARM, or ARM64 are discarded server side
      statusReport.runner_arch = process.env["RUNNER_ARCH"];
    }
    if (!(runnerOs === "Linux" && isSelfHostedRunner())) {
      // We do not report the release number for Linux self-hosted runners
      // because the custom build suffix may be private customer information.
      statusReport.runner_os_release = os.release();
    }
    if (codeQlCliVersion !== undefined) {
      statusReport.codeql_version = codeQlCliVersion.version;
    }
    const imageVersion = process.env["ImageVersion"];
    if (imageVersion) {
      statusReport.runner_image_version = imageVersion;
    }

    return statusReport;
  } catch (e) {
    logger.warning(
      `Caught an exception while gathering information for telemetry: ${e}. Will skip sending status report.`,
    );
    return undefined;
  }
}

const OUT_OF_DATE_MSG =
  "CodeQL Action is out-of-date. Please upgrade to the latest version of codeql-action.";
const INCOMPATIBLE_MSG =
  "CodeQL Action version is incompatible with the code scanning endpoint. Please update to a compatible version of codeql-action.";

/**
 * Send a status report to the code_scanning/analysis/status endpoint.
 *
 * Optionally checks the response from the API endpoint and sets the action
 * as failed if the status report failed. This is only expected to be used
 * when sending a 'starting' report.
 *
 * The `/code-scanning/analysis/status` endpoint is internal and it is not critical that it succeeds:
 * https://github.com/github/codeql/issues/15462#issuecomment-1919186317
 *
 * Failures while calling this endpoint are logged as warings.
 */
export async function sendStatusReport<S extends StatusReportBase>(
  statusReport: S,
): Promise<void> {
  setJobStatusIfUnsuccessful(statusReport.status);

  const statusReportJSON = JSON.stringify(statusReport);
  core.debug(`Sending status report: ${statusReportJSON}`);
  // If in test mode we don't want to upload the results
  if (isInTestMode()) {
    core.debug("In test mode. Status reports are not uploaded.");
    return;
  }

  const nwo = getRepositoryNwo();
  const client = getApiClient();

  try {
    await client.request(
      "PUT /repos/:owner/:repo/code-scanning/analysis/status",
      {
        owner: nwo.owner,
        repo: nwo.repo,
        data: statusReportJSON,
      },
    );
  } catch (e) {
    if (isHTTPError(e)) {
      switch (e.status) {
        case 403:
          if (
            getWorkflowEventName() === "push" &&
            process.env["GITHUB_ACTOR"] === "dependabot[bot]"
          ) {
            core.warning(
              'Workflows triggered by Dependabot on the "push" event run with read-only access. ' +
                "Uploading Code Scanning results requires write access. " +
                'To use Code Scanning with Dependabot, please ensure you are using the "pull_request" event for this workflow and avoid triggering on the "push" event for Dependabot branches. ' +
                `See ${DocUrl.SCANNING_ON_PUSH} for more information on how to configure these events.`,
            );
          } else {
            core.warning(e.message);
          }
          return;
        case 404:
          core.warning(e.message);
          return;
        case 422:
          // schema incompatibility when reporting status
          // this means that this action version is no longer compatible with the API
          // we still want to continue as it is likely the analysis endpoint will work
          if (getRequiredEnvParam("GITHUB_SERVER_URL") !== GITHUB_DOTCOM_URL) {
            core.debug(INCOMPATIBLE_MSG);
          } else {
            core.debug(OUT_OF_DATE_MSG);
          }
          return;
      }
    }

    // something else has gone wrong and the request/response will be logged by octokit
    // it's possible this is a transient error and we should continue scanning
    core.warning(
      `An unexpected error occurred when sending code scanning status report: ${getErrorMessage(
        e,
      )}`,
    );
  }
}