Running release tests causes SIGILL about ntest HOT 3 OPEN

@phodina thanks for reporting and sorry for the inconvenience. I am not familiar with guix. Is this some kind of replacement of package manager? Like cargo? Is the failing test you describe somehow reproducible without using guix?

from ntest.

Hi,
sorry for the long delay.

GNU Guix is a package manager similar to Nix.

The "main" idea is to break the Filesystem Hierarchy Standard and store packages in /gnu/store/ or /nix/store.

This is done using prefixes and some modifications. The advantage is that we don't have a global state but each package is independent and we just symlink the whole system together. So we can have multiple version of the package and have generation of the OS therefore we can go always back if we break the system. Also the OS config is simple text config file.

This is very brief and high level overview.

Guix/Nix do follow the package build/test process. So in this case they also use cargo with some additional steps (patching the crates).

• https://guix.gnu.org/manual/en/html_node/Rust-Crates.html
• https://guix.gnu.org/en/blog/2018/bootstrapping-rust/

Is the failing test you describe somehow reproducible without using guix?

Unfortunately I don't have "classical" distro on my laptop. But, I could spin docker container e.g. Alpine or Arch and run the test there to check if it's Guix specific or distro agnostic.

I'll report once this is done, ok?

from ntest.

@phodina thanks for the explanation. I am not directly calling any OS specific functions but since the error ocures in timeout_inf_loop I am pretty sure that this line of code causes trouble:

thread::spawn(move || {
    match sender.send(code()) {
        Ok(()) => {} // All good
        Err(_) => {} // Released, don't panic
    }
});

What version of rust are you using?

from ntest.