SEGFAULT on Node Fermium

[SalathielGenese]

• Version: v14.1.0 for me

• Platform: Linux salathiel 5.4.0-39-generic The binary and long term compatibility with node #43-Ubuntu SMP Fri Jun 19 10:28:31 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux

• Subsystem: (Ubuntu for me)

What steps will reproduce the bug?

I tried to build https://github.com/ever-co/gauzy - yarn nx build api --configuration=production and it landed me here nrwl/nx#2956 where other developers noticed that downgrading Node fixed the issue.

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

To the exception of Node version, I have not noticed any other condition.

What is the expected behavior?

A successful build.

What do you see instead?

In Docker build

error command failed with signal "SIGSEGV"

On host, it is silent and the exit status - echo $? - returned is 139, which happens to be NodeJS exit status for SEGFAULT.

Additional information

It works after I downgraded to node:erbium-alpine. Since the issue is not happening during development, I have not [yet] downgraded my environment.

[addaleax]

@SalathielGenese Any chance you could share the core dump from this? That might be helpful.

[addaleax]

Thanks for the report! I’ve reproduced this. It’s happening because the BackingStore for a Buffer created inside a Worker (possibly from a native addon) does not keep the ArrayBuffer::Allocator itself alive, leading to a use-after-free bug. This is an oversight on my part.

However, the good news is that #33291 fixes this, and is scheduled to be released today, as part of #34093. So all that we really need to do here is write a regression test. (The other good news is that this would typically only happen on process exit, so while the error is annoying, it doesn’t affect the functionality of the program.)