worker_threads emergency aborted (std::invalid_argument exception from std::stoull)

[a-shvedov]

Version

node-v19.3.0

Platform

5.15.78-2.el7.3.x86_64 (RED OS release MUROM 7.3.2 - Centos based system)

Subsystem

Worker threads

What steps will reproduce the bug?

thread_minimize.js:

const { Worker } = require('worker_threads');
const path = require('path');
new Worker(path.join(__dirname), {
// new Worker(path.join(__dirname, 'test.js'), {
execArgv: [ '--cpu-prof-interval', process.TEST ] });

How often does it reproduce? Is there a required condition?

Standart build compiled with: gcc (version 8.3.1 2019112)
Configure params: --prefix="/target/dir"
Builded binary info: node: ELF 64-bit LSB executable, x86-64, version 1 (GNU/Linux), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=1bf428404ef62dd8b5c3f952c02e1d1e8da610a9, for GNU/Linux 3.2.0

What is the expected behavior?

No response

What do you see instead?

Reproduce an error with the following options: ASAN_OPTIONS=verbosity=1:detect_leaks=1:symbolize=1 LD_PRELOAD=/usr/lib64/libasan.so.5.0.0 /home/user/target/node-v19.0.1/node-v19.0.1/builded/bin/node ./thread_minimize.js

Asan log:

MemToShadow(shadow): 0x00008fff7000 0x000091ff6dff 0x004091ff6e00 0x02008fff6fff
redzone=16
max_redzone=2048
quarantine_size_mb=256M
thread_local_quarantine_size_kb=1024K
malloc_context_size=30
SHADOW_SCALE: 3
SHADOW_GRANULARITY: 8
SHADOW_OFFSET: 0x7fff8000
==1664269==Installed the sigaction for signal 11
==1664269==Installed the sigaction for signal 7
==1664269==Installed the sigaction for signal 8
==1664269==T0: stack [0x7fff1476b000,0x7fff14f6b000) size 0x800000; local=0x7fff14f68584
==1664269==AddressSanitizer Init done
==1664269==T1: stack [0x7faa62000000,0x7faa627fef00) size 0x7fef00; local=0x7faa627fee04
==1664269==T2: stack [0x7faa617ff000,0x7faa61ffdf00) size 0x7fef00; local=0x7faa61ffde04
==1664269==T3: stack [0x7faa60ffe000,0x7faa617fcf00) size 0x7fef00; local=0x7faa617fce04
==1664269==T4: stack [0x7faa607fd000,0x7faa60ffbf00) size 0x7fef00; local=0x7faa60ffbe04
==1664269==T5: stack [0x7faa5fffc000,0x7faa607faf00) size 0x7fef00; local=0x7faa607fae04
==1664269==T6: stack [0x7faa63026000,0x7faa6302cf00) size 0x6f00; local=0x7faa6302ce04
terminate called after throwing an instance of 'std::invalid_argument'
  what():  stoull
Aborted

Additional information

No response

[bnoordhuis]

Looks to have been introduced in 436f4de from 2018. The easy fix is to replace stoull(value) with strtoull(value.c_str(), nullptr, 10). Want to send a pull request?

Open question: how to handle invalid inputs? The kInteger case uses atoll() and simply ignores errors. atoll() is basically strtoll() with base=10.

[bnoordhuis]

Forgot to mention, I can easily reproduce like this:

$ node --cpu-prof-interval boom -p 42
libc++abi.dylib: terminating with uncaught exception of type std::invalid_argument: stoull: no conversion
Abort trap: 6

So yes, confirmed bug.

[bnoordhuis]

#46290

[a-shvedov]

#46290

Thanks for the quick solution to this problem!