Segmentation fault on aarch64 Linux

[floriangosse]

Steps to reproduce

I'm using Docker on a Linux host.

• Run docker run --platform linux/arm64 --entrypoint /bin/sh --rm -it -w /usr/app node:16-alpine
• Inside the container
  • npm install lmdb@2.5.3
  • node -e "require('lmdb').open('test')"

[kriszyp]

So far I haven't had any luck in reproducing this. I just tested this on an AWS graviton2 (with the node:16-alpine container), and not only does lmdb-js load and open dbs, but the whole unit test suite passes without issue. Any ideas of anything more specific that I might need to reproduce this?

[floriangosse]

I've used a t2.micro for testing.

This is how i've set it up:

# Update installed packages
sudo yum -y update

# Install amazon-linux-extras and enable some extras
sudo yum install -y amazon-linux-extras
sudo amazon-linux-extras enable docker

# Install docker and assign ec2-user to docker group
sudo yum install -y docker
sudo usermod -a -G docker ec2-user

# Start docker and make it auto-start
sudo service docker start
sudo chkconfig docker on

# Setup qemu for multiarch emulation
sudo docker run --rm --privileged multiarch/qemu-user-static --reset -p yes

# Install docker buildx
sudo curl --location "https://github.com/docker/buildx/releases/download/v0.9.1/buildx-v0.9.1.linux-amd64" --output "/usr/libexec/docker/cli-plugins/docker-buildx"
sudo chmod +x /usr/libexec/docker/cli-plugins/docker-buildx

After that I logged out and logged in again to be able to work with docker because of the group change and ran the comments in the original post.

[kriszyp]

Thank you for the detailed report, this was super helpful! I was able to easily identify the issue with this info. It turns out that this was due to a failure to set the robust setting on mutexes for platform that usually supports robust mutexes (must be some idiosyncrasy on the emulated platform), and some improper handling of that.

[floriangosse]

Thanks for you fast response. Do you already know in which version the fix will be made available?

[kriszyp]

Published the fix for this in v2.5.5. Let me know if you still have any problems with this.

[floriangosse]

Looks good. It didn't throw any error. :)

Thanks a lot for the quick bug fix.