Importing tensorflow causes a crash on exit

[rmarru]

🐛 Bug Reports

🌍 Environment

• Your operating system and version: Ubuntu 18.04.4 LTS (Bionic Beaver)
• Your python version: Python 3.7.8
• How did you install python (e.g. apt or pyenv)? Did you use a virtualenv?: Installed via apt. No virtualenv.
• Your Rust version (rustc --version): 1.45.0
• Your PyO3 version: 0.11.1
• Have you tried using latest PyO3 master (replace version = "0.x.y" with git = "https://github.com/PyO3/pyo3")?: Yes

💥 Reproducing

Tensorflow version: 2.1.0, installed via pip: pip3 install tensorflow==2.1.0

Minimal reproduction:

use pyo3::prelude::*;

fn main() {
    let gil = Python::acquire_gil();
    let py = gil.python();

    let _mod = PyModule::import(py, "tensorflow").unwrap();
}

On exit:

$ cargo run
    Finished dev [unoptimized + debuginfo] target(s) in 0.02s
          Running `target/debug/pyo3`
corrupted double-linked list
[1]    4387 abort      cargo run

or

$ cargo run
    Finished dev [unoptimized + debuginfo] target(s) in 0.02s
          Running `target/debug/pyo3`
corrupted size vs. prev_size
[1]    5299 abort      cargo run

Running import tensorflow directly in python (and exiting the interpreter) exits normally.

Looks like a similar issue to Importing mxnet causes a segfault on exit #1044.

If I comment out libc::atexit(finalize) as @kngwyu suggested in comment, the error does not happen and the process exits successfully.

[davidhewitt]

#1355 has now merged - we decided to disable finalizing by default because as we see it can lead to weird crashes like this.