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.jsconstITERATIONS=10000;constBATCH_SIZE=100;constTIMEOUT=100;// Small timeout so it resolves quickly// Record initial memoryconstgetMemoryUsage=()=>process.memoryUsage().heapUsed/1024/1024;letinitialMemory=0;letfinalMemory=0;letsignals=[];asyncfunctionrunTest(){console.log('Starting memory leak test for AbortSignal.any()');// Force garbage collection before we startif(global.gc){global.gc();awaitnewPromise(resolve=>setTimeout(resolve,100));global.gc();}else{console.warn('Start node with --expose-gc to get more accurate results');}// Record initial memory after GCinitialMemory=getMemoryUsage();console.log(`Initial memory usage: ${initialMemory.toFixed(2)} MB`);// Run multiple batches and force GC between themfor(letbatch=0;batch<ITERATIONS/BATCH_SIZE;batch++){signals=[];for(leti=0;i<BATCH_SIZE;i++){consttimeoutSignal=AbortSignal.timeout(TIMEOUT);constcompositeSignal=AbortSignal.any([timeoutSignal]);// Hold a reference to the composite signalsignals.push(compositeSignal);}awaitnewPromise(resolve=>setTimeout(resolve,TIMEOUT*2));// Clear references and force GCsignals=[];if(global.gc){global.gc();awaitnewPromise(resolve=>setTimeout(resolve,100));global.gc();}// Log memory usageif(batch%10===0||batch===(ITERATIONS/BATCH_SIZE)-1){constcurrentMemory=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