x86 npm on Windows randomly crashes with -1073741819 exit code

[FeodorFitsner]

• Version: 14.8.0 x86 (but looks like the entire 14.x line is affected)
• Platform: Windows Server 2019 and Windows 10
• Subsystem: npm

What steps will reproduce the bug?

npm from x86 version of Node.js 14.x sporadically crashes with -1073741819 exit code.

npm from x64 version of Node 14.x works normally.

To reproduce the issue run the following Windows batch script which runs npm -v 500 times in a loop:

@ECHO OFF
SETLOCAL
SET EL=0

IF %ERRORLEVEL% NEQ 0 GOTO ERROR

set iterations=500
:loop
set /a iterations -= 1
CALL npm -v
IF %ERRORLEVEL% NEQ 0 GOTO ERROR
if %iterations% GTR 0 GOTO loop


GOTO DONE

:ERROR
ECHO ~~~~~~~~~~~~~~~~~~~~~~ ERROR %~f0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ECHO ERRORLEVEL^: %ERRORLEVEL%
SET EL=%ERRORLEVEL%

:DONE
ECHO ~~~~~~~~~~~~~~~~~~~~~~ DONE %~f0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

EXIT /b %EL%

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

It could crash on the first or second iteration; sometimes it manages to do around 5 iterations.

What is the expected behavior?

All iterations should be successful:

c:\>test-node.cmd
6.14.7
6.14.7
6.14.7
6.14.7
6.14.7
...

What do you see instead?

c:\>test-node
6.14.7
~~~~~~~~~~~~~~~~~~~~~~ ERROR c:\test-node.cmd ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ERRORLEVEL: -1073741819
~~~~~~~~~~~~~~~~~~~~~~ DONE c:\test-node.cmd ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Additional information

Initially, we thought it was AppVeyor-specific issue, however we were able to reproduce the issue on Hyper-V and GCP virtual machines with Windows Server 2019 as well as bare-metal PC with Windows 10.

[Almenon]

Example: https://ci.appveyor.com/project/Almenon/arepl-backend/builds/34120966

[bnoordhuis]

-1073741819 is 0xC0000005, the status code for a general access violation. Unfortunately, that can be caused by just about anything.

Does it print a stack trace or anything else that could help troubleshoot this?

[FeodorFitsner]

Nope, it just silently crashes. There was an attempt to troubleshoot that: appveyor/ci#3407 (comment)

[bnoordhuis]

Thanks, that's helpful insofar that it tells me what happens, although not why. :-)

What: the garbage collector tries to move an object but its map pointer (its "hidden class" descriptor) is invalid

Why: unclear. I don't recall receiving similar bug reports.

cc @nodejs/platform-windows since this only seems to happen with the windows x86 build.

[jasnell]

Ah, these are "fun". It might not have anything to do with the garbage collector and may be something with the environment clean up logic. The fact that it's so intermittent means it could be initialized memory somewhere. Later on today or tomorrow I'll see if I can reproduce the issue.

[addaleax]

and may be something with the environment clean up logic

The stack trace makes that seem very unlikely, the crash is happening while JS is running.

[tniessen]

Node.js 14 is about to reach end-of-life. Is this known to be an issue in the 16, 18, 19, or 20 release lines?

[shnooshnoo]

tested with node 14,16,18,20 (both x86 and x64) and was not able to reproduce this (i'm on win11). If anyone can help me reproduce i'd be keen to look into it and try to fix, otherwise looks like it is no longer relevant.

[tniessen]

Since there has been no activity here in years and any evidently problematic configuration has reached end-of-life status, I'll close this issue.