process: argv1 property to preserve original argv[1]

doc/api/process.md

@@ -958,6 +958,32 @@ $ bash -c 'exec -a customArgv0 ./node'
 'customArgv0'
 ```
 
+## `process.argv1`
+
+<!-- YAML
+added: REPLACEME
+-->
+
+* {string | undefined}
+
+The `process.argv1` property stores a read-only copy of the original value of
+`argv[1]` passed when Node.js starts.
+
+```js
+// test.js
+console.log('argv1  ', process.argv1);
+console.log('argv[1]', process.argv[1]);
+```
+
+```shell
+$ node ./test.js
+argv1   ./test.js
+argv[1] /Users/geoffrey/Sites/node/test.js
+```
+
+`process.argv1` is `undefined` for cases where Node.js is run without a main
+entry point, such as `node --eval`.
+
 ## `process.channel`
 
 <!-- YAML

lib/internal/main/run_main_module.js

@@ -7,17 +7,18 @@ const {
   markBootstrapComplete,
 } = require('internal/process/pre_execution');
 
-prepareMainThreadExecution(true);
+const mainEntry = prepareMainThreadExecution(true);
 
 markBootstrapComplete();
 
 // Necessary to reset RegExp statics before user code runs.
 RegExpPrototypeExec(/^/, '');
 
-// Note: this loads the module through the ESM loader if the module is
-// determined to be an ES module. This hangs from the CJS module loader
-// because we currently allow monkey-patching of the module loaders
-// in the preloaded scripts through require('module').
-// runMain here might be monkey-patched by users in --require.
-// XXX: the monkey-patchability here should probably be deprecated.
-require('internal/modules/cjs/loader').Module.runMain(process.argv[1]);
+/**
+ * To support legacy monkey-patching of `Module.runMain`, we call `runMain` here to have the CommonJS loader begin
+ * the execution of the main entry point, even if the ESM loader immediately takes over because the main entry is an
+ * ES module or one of the other opt-in conditions (such as the use of `--import`) are met. Users can monkey-patch
+ * before the main entry point is loaded by doing so via scripts loaded through `--require`. This monkey-patchability
+ * is undesirable and should be deprecated.
+ */
+require('internal/modules/cjs/loader').Module.runMain(mainEntry);

lib/internal/modules/run_main.js

@@ -5,7 +5,6 @@ const {
 } = primordials;
 
 const { getOptionValue } = require('internal/options');
-const path = require('path');
 
 /**
  * Get the absolute path to the main entry point.
@@ -16,7 +15,8 @@ function resolveMainPath(main) {
   // future major.
   // Module._findPath is monkey-patchable here.
   const { Module, toRealPath } = require('internal/modules/cjs/loader');
-  let mainPath = Module._findPath(path.resolve(main), null, true);
+  const { resolve } = require('path');
+  let mainPath = Module._findPath(resolve(main), null, true);
   if (!mainPath) { return; }
 
   const preserveSymlinksMain = getOptionValue('--preserve-symlinks-main');
@@ -29,7 +29,7 @@ function resolveMainPath(main) {
 
 /**
  * Determine whether the main entry point should be loaded through the ESM Loader.
- * @param {string} mainPath Absolute path to the main entry point
+ * @param {string} mainPath - Absolute path to the main entry point
  */
 function shouldUseESMLoader(mainPath) {
   /**
@@ -53,22 +53,22 @@ function shouldUseESMLoader(mainPath) {
 
 /**
  * Run the main entry point through the ESM Loader.
- * @param {string} mainPath Absolute path to the main entry point
+ * @param {string} mainPath - Absolute path for the main entry point
  */
 function runMainESM(mainPath) {
-  const { loadESM } = require('internal/process/esm_loader');
+  const { isAbsolute } = require('path');
   const { pathToFileURL } = require('internal/url');
+  const mainHref = isAbsolute(mainPath) ? pathToFileURL(mainPath).href : mainPath;
 
+  const { loadESM } = require('internal/process/esm_loader');
   handleMainPromise(loadESM((esmLoader) => {
-    const main = path.isAbsolute(mainPath) ?
-      pathToFileURL(mainPath).href : mainPath;
-    return esmLoader.import(main, undefined, { __proto__: null });
+    return esmLoader.import(mainHref, undefined, { __proto__: null });
   }));
 }
 
 /**
  * Handle process exit events around the main entry point promise.
- * @param {Promise} promise Main entry point promise
+ * @param {Promise} promise - Main entry point promise
  */
 async function handleMainPromise(promise) {
   const {
@@ -86,7 +86,8 @@ async function handleMainPromise(promise) {
  * Parse the CLI main entry point string and run it.
  * For backwards compatibility, we have to run a bunch of monkey-patchable code that belongs to the CJS loader (exposed
  * by `require('module')`) even when the entry point is ESM.
- * @param {string} main CLI main entry point string
+ * Because of backwards compatibility, this function is exposed publicly via `import { runMain } from 'node:module'`.
+ * @param {string} main - Resolved absolute path for the main entry point, if found
  */
 function executeUserEntryPoint(main = process.argv[1]) {
   const resolvedMain = resolveMainPath(main);

lib/internal/process/pre_execution.js

@@ -52,29 +52,19 @@ const {
 } = require('internal/v8/startup_snapshot');
 
 function prepareMainThreadExecution(expandArgv1 = false, initializeModules = true) {
-  prepareExecution({
-    expandArgv1,
-    initializeModules,
-    isMainThread: true,
-  });
+  return prepareExecution(expandArgv1, initializeModules, true);
 }
 
 function prepareWorkerThreadExecution() {
-  prepareExecution({
-    expandArgv1: false,
-    initializeModules: false,  // Will need to initialize it after policy setup
-    isMainThread: false,
-  });
+  prepareExecution(undefined, false, false); // Will need to initialize modules after policy setup
 }
 
-function prepareExecution(options) {
-  const { expandArgv1, initializeModules, isMainThread } = options;
-
+function prepareExecution(expandArgv1, initializeModules, isMainThread) {
   refreshRuntimeOptions();
   reconnectZeroFillToggle();
 
-  // Patch the process object with legacy properties and normalizations
-  patchProcessObject(expandArgv1);
+  // Patch the process object and get the resolved main entry point.
+  const mainEntry = patchProcessObject(expandArgv1);
   setupTraceCategoryState();
   setupInspectorHooks();
   setupWarningHandler();
@@ -131,6 +121,8 @@ function prepareExecution(options) {
   if (initializeModules) {
     setupUserModules();
   }
+
+  return mainEntry;
 }
 
 function setupSymbolDisposePolyfill() {
@@ -176,12 +168,21 @@ function refreshRuntimeOptions() {
   refreshOptions();
 }
 
+/**
+ * Patch the process object with legacy properties and normalizations.
+ * Replace `process.argv[0]` with `process.execPath`, preserving the original `argv[0]` value as `process.argv0`.
+ * Replace `process.argv[1]` with the resolved file path of the entry point, if requested and a replacement can be
+ * found; preserving the original value as `process.argv1`.
+ * @param {boolean} expandArgv1 - Whether to replace `process.argv[1]` with the resolved file path of the main entry
+ * point.
+ */
 function patchProcessObject(expandArgv1) {
   const binding = internalBinding('process_methods');
   binding.patchProcessObject(process);
 
   require('internal/process/per_thread').refreshHrtimeBuffer();
 
+  // Since we replace process.argv[0] below, preserve the original value in case the user needs it.
   ObjectDefineProperty(process, 'argv0', {
     __proto__: null,
     enumerable: true,
@@ -190,16 +191,27 @@ function patchProcessObject(expandArgv1) {
     value: process.argv[0],
   });
 
+  // Since we usually replace process.argv[1] below, preserve the original value in case the user needs it.
+  ObjectDefineProperty(process, 'argv1', {
+    __proto__: null,
+    enumerable: true,
+    // Only set it to true during snapshot building.
+    configurable: isBuildingSnapshot(),
+    value: process.argv[1],
+  });
+
   process.exitCode = undefined;
   process._exiting = false;
   process.argv[0] = process.execPath;
 
-  if (expandArgv1 && process.argv[1] &&
-      !StringPrototypeStartsWith(process.argv[1], '-')) {
-    // Expand process.argv[1] into a full path.
-    const path = require('path');
+  // If a requested, update process.argv[1] to replace whatever the user provided with the resolved absolute URL or file
+  // path of the entry point.
+  /** @type {ReturnType<determineMainEntry>} */
+  let mainEntry;
+  if (expandArgv1) {
+    mainEntry = determineMainEntry(process.argv[1]);
     try {
-      process.argv[1] = path.resolve(process.argv[1]);
+      process.argv[1] = mainEntry;
     } catch {
       // Continue regardless of error.
     }
@@ -226,6 +238,21 @@ function patchProcessObject(expandArgv1) {
   addReadOnlyProcessAlias('traceDeprecation', '--trace-deprecation');
   addReadOnlyProcessAlias('_breakFirstLine', '--inspect-brk', false);
   addReadOnlyProcessAlias('_breakNodeFirstLine', '--inspect-brk-node', false);
+
+  return mainEntry;
+}
+
+/**
+ * For non-STDIN input, we need to resolve what was specified on the command line into a path.
+ * @param {string} input - What was specified on the command line as the main entry point.
+ */
+function determineMainEntry(input = process.argv[1]) {
+  if (!input || StringPrototypeStartsWith(input, '-')) { // STDIN
+    return undefined;
+  }
+  // Expand the main entry point into a full path.
+  const { resolve } = require('path');
+  return resolve(input);
 }
 
 function addReadOnlyProcessAlias(name, option, enumerable = true) {

test/fixtures/process-argv1.js

@@ -0,0 +1,2 @@
+console.log(process.argv1);
+console.log(process.argv[1]);

test/parallel/test-process-argv1.mjs

@@ -0,0 +1,20 @@
+import { spawnPromisified } from '../common/index.mjs';
+import * as fixtures from '../common/fixtures.mjs';
+import { strictEqual } from 'node:assert';
+import { describe, it } from 'node:test';
+import { dirname } from 'node:path';
+
+describe('process.argv1', { concurrency: true }, () => {
+  it('should be the string for the entry point before the entry was resolved into an absolute path', async () => {
+    const fixtureAbsolutePath = fixtures.path('process-argv1.js');
+    const fixtureRelativePath = './process-argv1.js';
+    const { code, signal, stdout, stderr } = await spawnPromisified(process.execPath, [fixtureRelativePath], {
+      cwd: dirname(fixtureAbsolutePath),
+    });
+
+    strictEqual(stderr, '');
+    strictEqual(stdout, `${fixtureRelativePath}\n${fixtureAbsolutePath}\n`);
+    strictEqual(code, 0);
+    strictEqual(signal, null);
+  });
+});