reverse-engineering-anything/src/application/javascript/JavaScriptPreloadDiscovery.ts

import { posix } from "node:path";

import { AnalysisError } from "../../domain/analysisErrorBase.js";
import { analysisErrorWithCleanupFailure } from "../../domain/analysisErrorCleanup.js";
import {
  AnalysisCancelledError,
  AnalysisResourceConstraintError,
} from "../../domain/analysisErrorCore.js";
import { translateArtifactFailure } from "../../artifacts/ArtifactOperationFailure.js";
import type { JavaScriptArtifactFileSet } from "../../domain/javascript/javascriptArtifactFiles.js";
import type { JavaScriptSourceAnalysisLimits } from "../../domain/javascript/javascriptSourceAnalysis.js";
import type { ArtifactResourceScope } from "../../artifacts/ArtifactResourceScope.js";
import { createJavaScriptSourceAnalysis } from "../../artifacts/javascript/JavaScriptSourceAnalysis.js";
import { createJavaScriptPackageTypeLookup } from "./JavaScriptArtifactAnalysis.js";
import { JavaScriptJsonDocuments } from "./JavaScriptJsonDocuments.js";
import { resolveArtifactPathByContext } from "./JavaScriptArtifactPathResolution.js";

/** Declared preload identities, with the discovery gaps that may hide others. */
export interface JavaScriptPreloadDiscovery {
  readonly paths: ReadonlySet<string>;
  readonly limitations: readonly string[];
}

/** Resolve declared preload identities before choosing any .js file's loader.
 * Discovery uses the same bounded worker and path owner as normal analysis.
 * It never infers a loader from renderer API text in the selected file.
 */
export const discoverJavaScriptPreloadPaths = async (
  fileSet: JavaScriptArtifactFileSet,
  limits: JavaScriptSourceAnalysisLimits,
  scope: ArtifactResourceScope,
  signal?: AbortSignal,
): Promise<JavaScriptPreloadDiscovery> => {
  const paths = new Set<string>();
  const limitations: string[] = [];
  const jsonDocuments = new JavaScriptJsonDocuments();
  const packageTypeFor = createJavaScriptPackageTypeLookup(
    fileSet,
    jsonDocuments,
  );
  if (
    !fileSet.files.some(
      (file) =>
        file.kind === "javascript" &&
        file.text.included &&
        posix.extname(file.path) === ".js" &&
        packageTypeFor(file) === "module",
    )
  )
    return { paths, limitations };
  const sources = fileSet.files.filter(
    (file) => file.kind === "javascript" && file.text.included,
  );
  const files = new Map(fileSet.files.map((file) => [file.path, file]));
  const worker = await createJavaScriptSourceAnalysis(limits, sources.length, {
    scope,
  });
  let outcome: { kind: "completed" } | { kind: "failed"; cause: unknown } = {
    kind: "completed",
  };
  try {
    for (const [index, file] of sources.entries()) {
      const result = await worker.analyze(
        file,
        { nodes: 0, unknowns: 0 },
        signal,
        undefined,
        packageTypeFor(file),
        "preload-discovery",
      );
      if (!result.ok) {
        const { error } = result.error;
        if (signal?.aborted === true || error instanceof AnalysisCancelledError)
          throw error;
        // Normal analysis meets the same source and owns its failure and the
        // partial Evidence. A per-source limit only loses this source's
        // declarations; any other failure would stop that pass here too. The
        // gap stays reported even if normal analysis then reads the source.
        const skipped =
          error instanceof AnalysisResourceConstraintError &&
          error.reportedLimits?.parent_result_budget_exhausted !== true;
        const later = sources.length - index - 1;
        limitations.push(
          skipped || later === 0
            ? `Preload discovery could not read ${file.path} (${error.message}); a .js preload declared only there keeps its package scope.`
            : `Preload discovery stopped at ${file.path} (${error.message}); a .js preload declared only there or in ${String(later)} later sources keeps its package scope.`,
        );
        if (skipped) continue;
        break;
      }
      for (const window of result.value.javascript.electron.browser_windows) {
        if (
          window.preload_path === null ||
          window.loader_authority === "candidate"
        )
          continue;
        const resolution = resolveArtifactPathByContext({
          declaredPath: window.preload_path,
          sourcePath: file.path,
          context: window.preload_resolution_context,
          files,
          jsonDocuments,
          unanalyzedFiles: fileSet.unanalyzed_files,
          runtimePathKind: "exact-file",
        });
        if (
          resolution.resolution_status !== "resolved" ||
          resolution.resolved_path === null
        )
          continue;
        const target = files.get(resolution.resolved_path);
        if (target?.container_sha256 === file.container_sha256)
          paths.add(`${target.container_sha256}:${target.path}`);
      }
    }
  } catch (cause: unknown) {
    outcome = { kind: "failed", cause };
    throw cause;
  } finally {
    // The scoped port parks a healthy process; the following checkout resets
    // its per-analysis accounting, and scope cleanup still owns failed closes.
    const cleanup = await scope.release({
      kind: "javascript-source-analysis",
      analysis: worker,
      resource: "JavaScript preload discovery",
    });
    if (cleanup.kind === "failed") {
      if (outcome.kind === "failed" && outcome.cause instanceof AnalysisError)
        throw analysisErrorWithCleanupFailure(
          outcome.cause,
          translateArtifactFailure(
            "analyze_javascript_application",
            scope.cleanupFailure({ kind: "completed" }, [cleanup]),
          ),
          "analyze_javascript_application",
        );
      throw scope.cleanupFailure(outcome, [cleanup]);
    }
  }
  return { paths, limitations };
};