Cannot modify nested struct field with #[pyo3(set)]

[fschutt]

Environment

• Your operating system and version: Windows 8. 64-bit
• Your python version: Python 3.6.5
• How did you install python (e.g. apt or pyenv)? Did you use a virtualenv?: Windows Installer, no
• Your Rust version (rustc --version): stable rustc 1.52.1 (9bc8c42bb 2021-05-09)
• Your PyO3 version: 1ae3d87
• Have you tried using latest PyO3 main (replace version = "0.x.y" with git = "https://github.com/PyO3/pyo3")?: Yes

Reproducing

git clone https://github.com/fschutt/pyo3-bug
cargo build --release
cd target/release
cp ./azul.dll ./azul.pyd
# put the test.py file in /target/release
python ./test.py # see README

Rust code:

extern crate pyo3;

use pyo3::prelude::*;
use pyo3::Python;

#[pyclass]
#[repr(C)]
struct Wrapper {
    #[pyo3(get, set)]
    child: Child,
}

impl Clone for Wrapper {
    fn clone(&self) -> Self {
        Self {
            child: self.child.clone(),
        }
    }
}

#[pymethods]
impl Wrapper {
    #[new]
    fn new() -> Self {
        Self {
            child: Child { flag: true }
        }
    }
}

#[pyclass]
struct Child {
    #[pyo3(get, set)]
    flag: bool,
}

#[pymethods]
impl Child {
    #[new]
    fn new() -> Self {
        Self {
            flag: true
        }
    }
}

impl Clone for Child {
    fn clone(&self) -> Self {
        Self {
            flag: self.flag,
        }
    }
}

#[pymodule]
fn azul(py: Python, m: &PyModule) -> PyResult<()> {
    m.add_class::<Wrapper>()?;
    m.add_class::<Child>()?;

    Ok(())
}

Python code:

from azul import *

wrapper = Wrapper()
child = Child()

print(child.flag)               # ok: prints True (default)
print(wrapper.child.flag)       # ok: prints True (default)

# set both to False
child.flag = False
wrapper.child.flag = False

print(child.flag)               # ok: prints False (changed)
print(wrapper.child.flag)       # error: still prints True (not changed)

bug.zip

[davidhewitt]

Hi @fschutt thanks for the report.

This is unfortunately known behaviour at the moment, though I hope one day we can improve this. Please see the duplicate ticket #1358 (comment)