signal: 11, SIGSEGV: invalid memory reference

[oli-obk]

So I'm not sure where to post this, and it's hard to reproduce since it's UB.

I added a test to this crate

Cargo.toml changes

[dev-dependencies]
piston_window = "0.40"
rand = "0.3"

tests/smoke.rs

extern crate gfx_texture;
extern crate piston_window;
extern crate rand;
extern crate image as im;

use piston_window::*;

#[test]
fn create() {
    let window: PistonWindow = WindowSettings::new("Hello Piston!", (640, 480))
        .build()
        .expect("Failed to build PistonWindow");

    let mut img = im::ImageBuffer::new(2, 2);
    for x in 0..2 {
        for y in 0..2 {
            img.put_pixel(x, y,
                im::Rgba([rand::random(), rand::random(), rand::random(), 255]));
        }
    }
    let mut factory = window.factory.borrow_mut();
    let factory = &mut *factory;
    Texture::from_image(factory, &img, &TextureSettings::new()).unwrap();
}

#[test]
fn update() {
    let window: PistonWindow = WindowSettings::new("Hello Piston!", (640, 480))
        .build()
        .expect("Failed to build PistonWindow");

    let mut img = im::ImageBuffer::new(2, 2);
    for x in 0..2 {
        for y in 0..2 {
            img.put_pixel(x, y,
                im::Rgba([rand::random(), rand::random(), rand::random(), 255]));
        }
    }
    let mut factory = window.factory.borrow_mut();
    let factory = &mut *factory;
    let mut tex = Texture::from_image(factory, &img, &TextureSettings::new()).unwrap();
    tex.update(&mut *window.encoder.borrow_mut(), &img).unwrap();
}

renaming the update test to segfault prevents the segfault and gives a normal panic. So this might not be reproducable on the target machine.

splitting the window.encoder.borrow_mut() like the factory into multiple bindings will also not segfault on my machine.

I'm using rustc 1.9.0-nightly (30a3849f2 2016-03-30)

[kvark]

Do you have a call stack?

[bvssvni]

The window might be running out of scope before the texture creation/update.

Try drop(window); at the end.

[kvark]

@oli-obk is it related to gfx-rs/gfx#928 ?

[oli-obk]

nope, completely unrelated, it was just found while trying to create an MCVE for that.