lib: fix AbortSignal.any() with timeout signals

[gurgunday]

Fixes #57736

This PR fixes an issue where AbortSignal.any([AbortSignal.timeout(ms)]) would sometimes fail. The problem occurred because timeout signals inside a composite signal (created by AbortSignal.any()) were being garbage collected before they could fire.

• Modified AbortSignal.any() to identify timeout signals and add them to the gcPersistentSignals set
• Added cleanup for source signals when a composite signal aborts

I don't know how to make a simpler test for this honestly. Open to suggestions.

I used the following test to check for memory leaks. I assume that it works because when I initially made a mistake trying to fix this, it did cause a memory leak and it was reflected in this test:

After:

gurgunday@DELL-G15:~/work/oss/node$ ./out/Release/node --expose-gc leak-test.js
Starting memory leak test for AbortSignal.any()
Initial memory usage: 3.62 MB
Batch 1/100, Memory: 3.74 MB
Batch 11/100, Memory: 3.83 MB
Batch 21/100, Memory: 3.84 MB
Batch 31/100, Memory: 3.84 MB
Batch 41/100, Memory: 3.86 MB
Batch 51/100, Memory: 3.87 MB
Batch 61/100, Memory: 3.86 MB
Batch 71/100, Memory: 3.90 MB
Batch 81/100, Memory: 3.92 MB
Batch 91/100, Memory: 3.92 MB
Batch 100/100, Memory: 3.92 MB

Test completed:
- Initial memory: 3.62 MB
- Final memory: 3.93 MB
- Difference: 0.30 MB

Before:

gurgunday@DELL-G15:~/work/oss/node$ node --expose-gc leak-test.js
Starting memory leak test for AbortSignal.any()
Initial memory usage: 3.65 MB
Batch 1/100, Memory: 3.76 MB
Batch 11/100, Memory: 3.78 MB
Batch 21/100, Memory: 3.82 MB
Batch 31/100, Memory: 3.87 MB
Batch 41/100, Memory: 3.87 MB
Batch 51/100, Memory: 3.88 MB
Batch 61/100, Memory: 3.88 MB
Batch 71/100, Memory: 3.93 MB
Batch 81/100, Memory: 3.94 MB
Batch 91/100, Memory: 3.94 MB
Batch 100/100, Memory: 3.95 MB

Test completed:
- Initial memory: 3.65 MB
- Final memory: 3.96 MB
- Difference: 0.30 MB

'use strict';

// This test creates many composite signals with timeouts 
// and checks if memory usage grows unbounded

// We use --expose-gc to manually trigger garbage collection
// Run with: node --expose-gc leak-test.js

const ITERATIONS = 10000;
const BATCH_SIZE = 100;
const TIMEOUT = 100; // Small timeout so it resolves quickly

// Record initial memory
const getMemoryUsage = () => process.memoryUsage().heapUsed / 1024 / 1024;
let initialMemory = 0;
let finalMemory = 0;
let signals = [];

async function runTest() {
  console.log('Starting memory leak test for AbortSignal.any()');
  
  // Force garbage collection before we start
  if (global.gc) {
    global.gc();
    await new Promise(resolve => setTimeout(resolve, 100));
    global.gc();
  } else {
    console.warn('Start node with --expose-gc to get more accurate results');
  }
  
  // Record initial memory after GC
  initialMemory = getMemoryUsage();
  console.log(`Initial memory usage: ${initialMemory.toFixed(2)} MB`);
  
  // Run multiple batches and force GC between them
  for (let batch = 0; batch < ITERATIONS / BATCH_SIZE; batch++) {
    signals = [];
    for (let i = 0; i < BATCH_SIZE; i++) {
      const timeoutSignal = AbortSignal.timeout(TIMEOUT);
      const compositeSignal = AbortSignal.any([timeoutSignal]);
      
      // Hold a reference to the composite signal
      signals.push(compositeSignal);
    }
    
    await new Promise(resolve => setTimeout(resolve, TIMEOUT * 2));
    
    // Clear references and force GC
    signals = [];
    if (global.gc) {
      global.gc();
      await new Promise(resolve => setTimeout(resolve, 100));
      global.gc();
    }
    
    // Log memory usage
    if (batch % 10 === 0 || batch === (ITERATIONS / BATCH_SIZE) - 1) {
      const currentMemory = getMemoryUsage();
      console.log(`Batch ${batch + 1}/${ITERATIONS / BATCH_SIZE}, Memory: ${currentMemory.toFixed(2)} MB`);
    }
  }
  
  finalMemory = getMemoryUsage();
  
  console.log(`\nTest completed:`);
  console.log(`- Initial memory: ${initialMemory.toFixed(2)} MB`);
  console.log(`- Final memory: ${finalMemory.toFixed(2)} MB`);
  console.log(`- Difference: ${(finalMemory - initialMemory).toFixed(2)} MB`);
}

runTest().catch(console.error);

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.24%. Comparing base (964e41c) to head (77f6539).
Report is 31 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff            @@
##             main   #57867    +/-   ##
========================================
  Coverage   90.23%   90.24%            
========================================
  Files         630      630            
  Lines      185518   185726   +208     
  Branches    36369    36410    +41     
========================================
+ Hits       167401   167604   +203     
+ Misses      11005    10999     -6     
- Partials     7112     7123    +11     

Files with missing lines	Coverage Δ
lib/internal/abort_controller.js	98.26% <100.00%> (+0.12%)	⬆️

... and 36 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[geeksilva97]

LGTM

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/66266/

[atlowChemi]

Do we really need this?
Can't we validate the reason is a timeout signal?

[gurgunday]

Updated, does it look better?

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/66317/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/66323/

[nodejs-github-bot]

Landed in 9d6626a