@nx/nx-linux-arm64-gnu crashes with jemalloc "Unsupported system page size" on 16K page aarch64 systems (Asahi linux)

[theres]

Current Behavior

The native binding @nx/nx-linux-arm64-gnu statically links jemalloc compiled with 4K page size assumption. On aarch64 Linux systems with 16K page size (e.g., Asahi Linux on Apple Silicon), any Nx command immediately aborts:

<jemalloc>: Unsupported system page size
<jemalloc>: arena_new: allocation failed
<jemalloc>: background thread creation failed
<jemalloc>: Unsupported system page size
<jemalloc>: os_pages_map: mmap error
<jemalloc>: arena_new: malloc_extent_alloc failed
<jemalloc>: Unsupported system page size
<jemalloc>: os_overcommit_proc: /proc/sys/vm/overcommit_memory: Cannot allocate memory
<jemalloc>: Unsupported system page size
memory allocation of 576 bytes failed
Aborted (core dumped)

Even nx --version crashes. The WASI fallback (NAPI_RS_FORCE_WASI=1) partially works but fails with TypeError: WorkspaceContext is not a constructor, so it's not a viable workaround.

Expected Behavior

NX works on asahi linux.

GitHub Repo

No response

Steps to Reproduce

Have no other ideas how to test except run any nx command on such system.

Nx Report

❯ pnpm nx report
<jemalloc>: Unsupported system page size
memory allocation of 576 bytes failed
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Failure Logs

Package Manager Version

pnpm

Operating System

• macOS
• Linux
• Windows
• Other (Please specify)

Additional Information

Any aarch64 Linux with page size > 4K — primarily Asahi Linux (Apple Silicon), but also some ARM server configurations (e.g., Ampere Altra with 64K pages).

jemalloc is compiled with --with-lg-page=12 (4K pages). On 16K page systems it refuses to operate. This is a known jemalloc limitation: jemalloc/jemalloc#467.

[ItsBenCodes]

This same error is also happening when using Docker Sandboxes on a Mac. I'm guessing it is for the same reason because it's running with linux/arm64 instead of darwin/arm64.

[theres]

Workaround for Asahi:

1. Install muvm
   a) Fedora should have it in the repo
   b) For Ubuntu, you can use this PPA:

sudo add-apt-repository ppa:kaazoo/ubuntu-asahi-gaming
sudo apt update
sudo apt install muvm

2. Now you can use it to run commands. In my case it’s via pnpm, but you need to pass PATH and HOME variables:

muvm -it -e HOME -e USER -e SHELL -- /usr/bin/bash -lic 'cd /ABSOLUTE/PATH/TO/PROJECT && pnpm nx report'

3. You can create a helper function like this:

mpnpm() {
  muvm -it -e HOME -e USER -e SHELL -- /usr/bin/bash -lic "cd \"$(pwd)\" && pnpm $*"
}

From now on, instead of pnpm (yarn/whatever you use), you can run commands like:

mpnpm nx report

[leosvelperez]

Thanks for reporting this!

Could you please test with this version 22.7.0-pr.35356.e1b62ce and let me know if things work?

[theres]

It seems to work! 🚀

[github-actions]

This issue has been closed for more than 30 days. If this issue is still occuring, please open a new issue with more recent context.