SIGSEGV when launching custom entrypoint/cmd

I’m getting this error when specifying a custom entrypoint and cmd:

2022-04-13T19:11:46Z   [info]Main child exited with signal (with signal 'SIGSEGV', core dumped? false)
2022-04-13T19:11:46Z   [info]Starting clean up.

My fly.toml looks like this:

[build]
  image = "ghcr.io/my/image:latest"

[experimental]
  entrypoint = ["/my/entrypoint"]
  cmd = ["node", "/path/to/app"]

Any ideas what could be causing the SIGSEGV?

What does the “Preparing to run” log line show?

See below:

2022-04-13T19:11:44Z   [info]Pulling container image
2022-04-13T19:11:44Z   [info]Unpacking image
2022-04-13T19:11:44Z   [info]Preparing kernel init
2022-04-13T19:11:45Z   [info]Configuring firecracker
2022-04-13T19:11:45Z   [info]Starting virtual machine
2022-04-13T19:11:45Z   [info]Starting init (commit: 6f9865f)...
2022-04-13T19:11:45Z   [info]Preparing to run: `/my/entrypoint node /path/to/app` as root
2022-04-13T19:11:45Z   [info]2022/04/13 19:11:45 listening on [fdaa:0:5abd:a7b:66:f8de:70ca:2]:22 (DNS: [fdaa::3]:53)

Ok that seems to be right behaviour.

Is it possible your program is segfaulting? Can you run it in docker with the same entrypoint and command as a test?

I thought about that as well, but I can’t get it to segfault myself. It works fine with entrypoint and command in a docker-compose for example, and runs great on Google Cloud Run.