// Welcome to the triangle example!
//
// This is the only example that is entirely detailed. All the other examples avoid code
// duplication by using helper functions.
//
// This example assumes that you are already more or less familiar with graphics programming and
// that you want to learn Vulkan. This means that for example it won't go into details about what a
// vertex or a shader is.

use std::{error::Error, sync::Arc};
use vulkano::{
    buffer::{Buffer, BufferContents, BufferCreateInfo, BufferUsage},
    command_buffer::{
        allocator::StandardCommandBufferAllocator, CommandBufferBeginInfo, CommandBufferLevel,
        CommandBufferUsage, RecordingCommandBuffer, RenderPassBeginInfo, SubpassBeginInfo,
        SubpassContents,
    },
    device::{
        physical::PhysicalDeviceType, Device, DeviceCreateInfo, DeviceExtensions, QueueCreateInfo,
        QueueFlags,
    },
    image::{view::ImageView, Image, ImageUsage},
    instance::{Instance, InstanceCreateFlags, InstanceCreateInfo},
    memory::allocator::{AllocationCreateInfo, MemoryTypeFilter, StandardMemoryAllocator},
    pipeline::{
        graphics::{
            color_blend::{ColorBlendAttachmentState, ColorBlendState},
            input_assembly::InputAssemblyState,
            multisample::MultisampleState,
            rasterization::RasterizationState,
            vertex_input::{Vertex, VertexDefinition},
            viewport::{Viewport, ViewportState},
            GraphicsPipelineCreateInfo,
        },
        layout::PipelineDescriptorSetLayoutCreateInfo,
        DynamicState, GraphicsPipeline, PipelineLayout, PipelineShaderStageCreateInfo,
    },
    render_pass::{Framebuffer, FramebufferCreateInfo, RenderPass, Subpass},
    swapchain::{
        acquire_next_image, Surface, Swapchain, SwapchainCreateInfo, SwapchainPresentInfo,
    },
    sync::{self, GpuFuture},
    Validated, VulkanError, VulkanLibrary,
};
use winit::{
    event::{Event, WindowEvent},
    event_loop::{ControlFlow, EventLoop},
    window::WindowBuilder,
};

fn main() -> Result<(), impl Error> {
    let event_loop = EventLoop::new().unwrap();

    let library = VulkanLibrary::new().unwrap();

    // The first step of any Vulkan program is to create an instance.
    //
    // When we create an instance, we have to pass a list of extensions that we want to enable.
    //
    // All the window-drawing functionalities are part of non-core extensions that we need to
    // enable manually. To do so, we ask `Surface` for the list of extensions required to draw to
    // a window.
    let required_extensions = Surface::required_extensions(&event_loop).unwrap();

    // Now creating the instance.
    let instance = Instance::new(
        library,
        InstanceCreateInfo {
            // Enable enumerating devices that use non-conformant Vulkan implementations.
            // (e.g. MoltenVK)
            flags: InstanceCreateFlags::ENUMERATE_PORTABILITY,
            enabled_extensions: required_extensions,
            ..Default::default()
        },
    )
    .unwrap();

    // The objective of this example is to draw a triangle on a window. To do so, we first need to
    // create the window. We use the `WindowBuilder` from the `winit` crate to do that here.
    //
    // Before we can render to a window, we must first create a `vulkano::swapchain::Surface`
    // object from it, which represents the drawable surface of a window. For that we must wrap the
    // `winit::window::Window` in an `Arc`.
    let window = Arc::new(WindowBuilder::new().build(&event_loop).unwrap());
    let surface = Surface::from_window(instance.clone(), window.clone()).unwrap();

    // Choose device extensions that we're going to use. In order to present images to a surface,
    // we need a `Swapchain`, which is provided by the `khr_swapchain` extension.
    let device_extensions = DeviceExtensions {
        khr_swapchain: true,
        ..DeviceExtensions::empty()
    };

    // We then choose which physical device to use. First, we enumerate all the available physical
    // devices, then apply filters to narrow them down to those that can support our needs.
    let (physical_device, queue_family_index) = instance
        .enumerate_physical_devices()
        .unwrap()
        .filter(|p| {
            // Some devices may not support the extensions or features that your application, or
            // report properties and limits that are not sufficient for your application. These
            // should be filtered out here.
            p.supported_extensions().contains(&device_extensions)
        })
        .filter_map(|p| {
            // For each physical device, we try to find a suitable queue family that will execute
            // our draw commands.
            //
            // Devices can provide multiple queues to run commands in parallel (for example a draw
            // queue and a compute queue), similar to CPU threads. This is something you have to
            // have to manage manually in Vulkan. Queues of the same type belong to the same queue
            // family.
            //
            // Here, we look for a single queue family that is suitable for our purposes. In a
            // real-world application, you may want to use a separate dedicated transfer queue to
            // handle data transfers in parallel with graphics operations. You may also need a
            // separate queue for compute operations, if your application uses those.
            p.queue_family_properties()
                .iter()
                .enumerate()
                .position(|(i, q)| {
                    // We select a queue family that supports graphics operations. When drawing to
                    // a window surface, as we do in this example, we also need to check that
                    // queues in this queue family are capable of presenting images to the surface.
                    q.queue_flags.intersects(QueueFlags::GRAPHICS)
                        && p.surface_support(i as u32, &surface).unwrap_or(false)
                })
                // The code here searches for the first queue family that is suitable. If none is
                // found, `None` is returned to `filter_map`, which disqualifies this physical
                // device.
                .map(|i| (p, i as u32))
        })
        // All the physical devices that pass the filters above are suitable for the application.
        // However, not every device is equal, some are preferred over others. Now, we assign each
        // physical device a score, and pick the device with the lowest ("best") score.
        //
        // In this example, we simply select the best-scoring device to use in the application.
        // In a real-world setting, you may want to use the best-scoring device only as a "default"
        // or "recommended" device, and let the user choose the device themself.
        .min_by_key(|(p, _)| {
            // We assign a lower score to device types that are likely to be faster/better.
            match p.properties().device_type {
                PhysicalDeviceType::DiscreteGpu => 0,
                PhysicalDeviceType::IntegratedGpu => 1,
                PhysicalDeviceType::VirtualGpu => 2,
                PhysicalDeviceType::Cpu => 3,
                PhysicalDeviceType::Other => 4,
                _ => 5,
            }
        })
        .expect("no suitable physical device found");

    // Some little debug infos.
    println!(
        "Using device: {} (type: {:?})",
        physical_device.properties().device_name,
        physical_device.properties().device_type,
    );

    // Now initializing the device. This is probably the most important object of Vulkan.
    //
    // An iterator of created queues is returned by the function alongside the device.
    let (device, mut queues) = Device::new(
        // Which physical device to connect to.
        physical_device,
        DeviceCreateInfo {
            // A list of optional features and extensions that our program needs to work correctly.
            // Some parts of the Vulkan specs are optional and must be enabled manually at device
            // creation. In this example the only thing we are going to need is the `khr_swapchain`
            // extension that allows us to draw to a window.
            enabled_extensions: device_extensions,

            // The list of queues that we are going to use. Here we only use one queue, from the
            // previously chosen queue family.
            queue_create_infos: vec![QueueCreateInfo {
                queue_family_index,
                ..Default::default()
            }],

            ..Default::default()
        },
    )
    .unwrap();

    // Since we can request multiple queues, the `queues` variable is in fact an iterator. We only
    // use one queue in this example, so we just retrieve the first and only element of the
    // iterator.
    let queue = queues.next().unwrap();

    // Before we can draw on the surface, we have to create what is called a swapchain. Creating a
    // swapchain allocates the color buffers that will contain the image that will ultimately be
    // visible on the screen. These images are returned alongside the swapchain.
    let (mut swapchain, images) = {
        // Querying the capabilities of the surface. When we create the swapchain we can only pass
        // values that are allowed by the capabilities.
        let surface_capabilities = device
            .physical_device()
            .surface_capabilities(&surface, Default::default())
            .unwrap();

        // Choosing the internal format that the images will have.
        let image_format = device
            .physical_device()
            .surface_formats(&surface, Default::default())
            .unwrap()[0]
            .0;

        // Please take a look at the docs for the meaning of the parameters we didn't mention.
        Swapchain::new(
            device.clone(),
            surface,
            SwapchainCreateInfo {
                // Some drivers report an `min_image_count` of 1, but fullscreen mode requires at
                // least 2. Therefore we must ensure the count is at least 2, otherwise the program
                // would crash when entering fullscreen mode on those drivers.
                min_image_count: surface_capabilities.min_image_count.max(2),

                image_format,

                // The size of the window, only used to initially setup the swapchain.
                //
                // NOTE:
                // On some drivers the swapchain extent is specified by
                // `surface_capabilities.current_extent` and the swapchain size must use this
                // extent. This extent is always the same as the window size.
                //
                // However, other drivers don't specify a value, i.e.
                // `surface_capabilities.current_extent` is `None`. These drivers will allow
                // anything, but the only sensible value is the window size.
                //
                // Both of these cases need the swapchain to use the window size, so we just
                // use that.
                image_extent: window.inner_size().into(),

                image_usage: ImageUsage::COLOR_ATTACHMENT,

                // The alpha mode indicates how the alpha value of the final image will behave. For
                // example, you can choose whether the window will be opaque or transparent.
                composite_alpha: surface_capabilities
                    .supported_composite_alpha
                    .into_iter()
                    .next()
                    .unwrap(),

                ..Default::default()
            },
        )
        .unwrap()
    };

    let memory_allocator = Arc::new(StandardMemoryAllocator::new_default(device.clone()));

    // We now create a buffer that will store the shape of our triangle. We use `#[repr(C)]` here
    // to force rustc to use a defined layout for our data, as the default representation has *no
    // guarantees*.
    #[derive(BufferContents, Vertex)]
    #[repr(C)]
    struct Vertex {
        #[format(R32G32_SFLOAT)]
        position: [f32; 2],
    }

    let vertices = [
        Vertex {
            position: [-0.5, -0.25],
        },
        Vertex {
            position: [0.0, 0.5],
        },
        Vertex {
            position: [0.25, -0.1],
        },
    ];
    let vertex_buffer = Buffer::from_iter(
        memory_allocator,
        BufferCreateInfo {
            usage: BufferUsage::VERTEX_BUFFER,
            ..Default::default()
        },
        AllocationCreateInfo {
            memory_type_filter: MemoryTypeFilter::PREFER_DEVICE
                | MemoryTypeFilter::HOST_SEQUENTIAL_WRITE,
            ..Default::default()
        },
        vertices,
    )
    .unwrap();

    // The next step is to create the shaders.
    //
    // The raw shader creation API provided by the vulkano library is unsafe for various reasons,
    // so The `shader!` macro provides a way to generate a Rust module from GLSL source - in the
    // example below, the source is provided as a string input directly to the shader, but a path
    // to a source file can be provided as well. Note that the user must specify the type of shader
    // (e.g. "vertex", "fragment", etc.) using the `ty` option of the macro.
    //
    // The items generated by the `shader!` macro include a `load` function which loads the shader
    // using an input logical device. The module also includes type definitions for layout
    // structures defined in the shader source, for example uniforms and push constants.
    //
    // A more detailed overview of what the `shader!` macro generates can be found in the
    // vulkano-shaders crate docs. You can view them at https://docs.rs/vulkano-shaders/
    mod vs {
        vulkano_shaders::shader! {
            ty: "vertex",
            src: r"
                #version 450

                layout(location = 0) in vec2 position;

                void main() {
                    gl_Position = vec4(position, 0.0, 1.0);
                }
            ",
        }
    }

    mod fs {
        vulkano_shaders::shader! {
            ty: "fragment",
            src: r"
                #version 450

                layout(location = 0) out vec4 f_color;

                void main() {
                    f_color = vec4(1.0, 0.0, 0.0, 1.0);
                }
            ",
        }
    }

    // At this point, OpenGL initialization would be finished. However in Vulkan it is not. OpenGL
    // implicitly does a lot of computation whenever you draw. In Vulkan, you have to do all this
    // manually.

    // The next step is to create a *render pass*, which is an object that describes where the
    // output of the graphics pipeline will go. It describes the layout of the images where the
    // colors, depth and/or stencil information will be written.
    let render_pass = vulkano::single_pass_renderpass!(
        device.clone(),
        attachments: {
            // `color` is a custom name we give to the first and only attachment.
            color: {
                // `format: <ty>` indicates the type of the format of the image. This has to be one
                // of the types of the `vulkano::format` module (or alternatively one of your
                // structs that implements the `FormatDesc` trait). Here we use the same format as
                // the swapchain.
                format: swapchain.image_format(),
                // `samples: 1` means that we ask the GPU to use one sample to determine the value
                // of each pixel in the color attachment. We could use a larger value
                // (multisampling) for antialiasing. An example of this can be found in
                // msaa-renderpass.rs.
                samples: 1,
                // `load_op: Clear` means that we ask the GPU to clear the content of this
                // attachment at the start of the drawing.
                load_op: Clear,
                // `store_op: Store` means that we ask the GPU to store the output of the draw in
                // the actual image. We could also ask it to discard the result.
                store_op: Store,
            },
        },
        pass: {
            // We use the attachment named `color` as the one and only color attachment.
            color: [color],
            // No depth-stencil attachment is indicated with empty brackets.
            depth_stencil: {},
        },
    )
    .unwrap();

    // Before we draw, we have to create what is called a **pipeline**. A pipeline describes how
    // a GPU operation is to be performed. It is similar to an OpenGL program, but it also contains
    // many settings for customization, all baked into a single object. For drawing, we create
    // a **graphics** pipeline, but there are also other types of pipeline.
    let pipeline = {
        // First, we load the shaders that the pipeline will use:
        // the vertex shader and the fragment shader.
        //
        // A Vulkan shader can in theory contain multiple entry points, so we have to specify which
        // one.
        let vs = vs::load(device.clone())
            .unwrap()
            .entry_point("main")
            .unwrap();
        let fs = fs::load(device.clone())
            .unwrap()
            .entry_point("main")
            .unwrap();

        // Automatically generate a vertex input state from the vertex shader's input interface,
        // that takes a single vertex buffer containing `Vertex` structs.
        let vertex_input_state = Vertex::per_vertex()
            .definition(&vs.info().input_interface)
            .unwrap();

        // Make a list of the shader stages that the pipeline will have.
        let stages = [
            PipelineShaderStageCreateInfo::new(vs),
            PipelineShaderStageCreateInfo::new(fs),
        ];

        // We must now create a **pipeline layout** object, which describes the locations and types
        // of descriptor sets and push constants used by the shaders in the pipeline.
        //
        // Multiple pipelines can share a common layout object, which is more efficient.
        // The shaders in a pipeline must use a subset of the resources described in its pipeline
        // layout, but the pipeline layout is allowed to contain resources that are not present in
        // the shaders; they can be used by shaders in other pipelines that share the same
        // layout. Thus, it is a good idea to design shaders so that many pipelines have
        // common resource locations, which allows them to share pipeline layouts.
        let layout = PipelineLayout::new(
            device.clone(),
            // Since w only have one pipeline in this example, and thus one pipeline layout,
            // we automeatically generate the creation info for it from the resources used in the
            // shaders. In a real application, you would specify this information manually so that
            // you can re-use one layout in multiple pipelines.
            PipelineDescriptorSetLayoutCreateInfo::from_stages(&stages)
                .into_pipeline_layout_create_info(device.clone())
                .unwrap(),
        )
        .unwrap();

        // We have to indicate which subpass of which render pass this pipeline is going to be used
        // in. The pipeline will only be usable from this particular subpass.
        let subpass = Subpass::from(render_pass.clone(), 0).unwrap();

        // Finally, create the pipeline.
        GraphicsPipeline::new(
            device.clone(),
            None,
            GraphicsPipelineCreateInfo {
                stages: stages.into_iter().collect(),
                // How vertex data is read from the vertex buffers into the vertex shader.
                vertex_input_state: Some(vertex_input_state),
                // How vertices are arranged into primitive shapes.
                // The default primitive shape is a triangle.
                input_assembly_state: Some(InputAssemblyState::default()),
                // How primitives are transformed and clipped to fit the framebuffer.
                // We use a resizable viewport, set to draw over the entire window.
                viewport_state: Some(ViewportState::default()),
                // How polygons are culled and converted into a raster of pixels.
                // The default value does not perform any culling.
                rasterization_state: Some(RasterizationState::default()),
                // How multiple fragment shader samples are converted to a single pixel value.
                // The default value does not perform any multisampling.
                multisample_state: Some(MultisampleState::default()),
                // How pixel values are combined with the values already present in the framebuffer.
                // The default value overwrites the old value with the new one, without any
                // blending.
                color_blend_state: Some(ColorBlendState::with_attachment_states(
                    subpass.num_color_attachments(),
                    ColorBlendAttachmentState::default(),
                )),
                // Dynamic states allows us to specify parts of the pipeline settings when
                // recording the command buffer, before we perform drawing.
                // Here, we specify that the viewport should be dynamic.
                dynamic_state: [DynamicState::Viewport].into_iter().collect(),
                subpass: Some(subpass.into()),
                ..GraphicsPipelineCreateInfo::layout(layout)
            },
        )
        .unwrap()
    };

    // Dynamic viewports allow us to recreate just the viewport when the window is resized.
    // Otherwise we would have to recreate the whole pipeline.
    let mut viewport = Viewport {
        offset: [0.0, 0.0],
        extent: [0.0, 0.0],
        depth_range: 0.0..=1.0,
    };

    // The render pass we created above only describes the layout of our framebuffers. Before we
    // can draw we also need to create the actual framebuffers.
    //
    // Since we need to draw to multiple images, we are going to create a different framebuffer for
    // each image.
    let mut framebuffers = window_size_dependent_setup(&images, render_pass.clone(), &mut viewport);

    // Before we can start creating and recording command buffers, we need a way of allocating
    // them. Vulkano provides a command buffer allocator, which manages raw Vulkan command pools
    // underneath and provides a safe interface for them.
    let command_buffer_allocator = Arc::new(StandardCommandBufferAllocator::new(
        device.clone(),
        Default::default(),
    ));

    // Initialization is finally finished!

    // In some situations, the swapchain will become invalid by itself. This includes for example
    // when the window is resized (as the images of the swapchain will no longer match the
    // window's) or, on Android, when the application went to the background and goes back to the
    // foreground.
    //
    // In this situation, acquiring a swapchain image or presenting it will return an error.
    // Rendering to an image of that swapchain will not produce any error, but may or may not work.
    // To continue rendering, we need to recreate the swapchain by creating a new swapchain. Here,
    // we remember that we need to do this for the next loop iteration.
    let mut recreate_swapchain = false;

    // In the loop below we are going to submit commands to the GPU. Submitting a command produces
    // an object that implements the `GpuFuture` trait, which holds the resources for as long as
    // they are in use by the GPU.
    //
    // Destroying the `GpuFuture` blocks until the GPU is finished executing it. In order to avoid
    // that, we store the submission of the previous frame here.
    let mut previous_frame_end = Some(sync::now(device.clone()).boxed());

    event_loop.run(move |event, elwt| {
        elwt.set_control_flow(ControlFlow::Poll);

        match event {
            Event::WindowEvent {
                event: WindowEvent::CloseRequested,
                ..
            } => {
                elwt.exit();
            }
            Event::WindowEvent {
                event: WindowEvent::Resized(_),
                ..
            } => {
                recreate_swapchain = true;
            }
            Event::WindowEvent {
                event: WindowEvent::RedrawRequested,
                ..
            } => {
                // Do not draw the frame when the screen size is zero. On Windows, this can
                // occur when minimizing the application.
                let image_extent: [u32; 2] = window.inner_size().into();

                if image_extent.contains(&0) {
                    return;
                }

                // It is important to call this function from time to time, otherwise resources
                // will keep accumulating and you will eventually reach an out of memory error.
                // Calling this function polls various fences in order to determine what the GPU
                // has already processed, and frees the resources that are no longer needed.
                previous_frame_end.as_mut().unwrap().cleanup_finished();

                // Whenever the window resizes we need to recreate everything dependent on the
                // window size. In this example that includes the swapchain, the framebuffers and
                // the dynamic state viewport.
                if recreate_swapchain {
                    // Use the new dimensions of the window.

                    let (new_swapchain, new_images) = swapchain
                        .recreate(SwapchainCreateInfo {
                            image_extent,
                            ..swapchain.create_info()
                        })
                        .expect("failed to recreate swapchain");

                    swapchain = new_swapchain;

                    // Because framebuffers contains a reference to the old swapchain, we need to
                    // recreate framebuffers as well.
                    framebuffers = window_size_dependent_setup(
                        &new_images,
                        render_pass.clone(),
                        &mut viewport,
                    );

                    recreate_swapchain = false;
                }

                // Before we can draw on the output, we have to *acquire* an image from the
                // swapchain. If no image is available (which happens if you submit draw commands
                // too quickly), then the function will block. This operation returns the index of
                // the image that we are allowed to draw upon.
                //
                // This function can block if no image is available. The parameter is an optional
                // timeout after which the function call will return an error.
                let (image_index, suboptimal, acquire_future) =
                    match acquire_next_image(swapchain.clone(), None).map_err(Validated::unwrap) {
                        Ok(r) => r,
                        Err(VulkanError::OutOfDate) => {
                            recreate_swapchain = true;
                            return;
                        }
                        Err(e) => panic!("failed to acquire next image: {e}"),
                    };

                // `acquire_next_image` can be successful, but suboptimal. This means that the
                // swapchain image will still work, but it may not display correctly. With some
                // drivers this can be when the window resizes, but it may not cause the swapchain
                // to become out of date.
                if suboptimal {
                    recreate_swapchain = true;
                }

                // In order to draw, we have to record a *command buffer*. The command buffer object
                // holds the list of commands that are going to be executed.
                //
                // Recording a command buffer is an expensive operation (usually a few hundred
                // microseconds), but it is known to be a hot path in the driver and is expected to
                // be optimized.
                //
                // Note that we have to pass a queue family when we create the command buffer. The
                // command buffer will only be executable on that given queue family.
                let mut builder = RecordingCommandBuffer::new(
                    command_buffer_allocator.clone(),
                    queue.queue_family_index(),
                    CommandBufferLevel::Primary,
                    CommandBufferBeginInfo {
                        usage: CommandBufferUsage::OneTimeSubmit,
                        ..Default::default()
                    },
                )
                .unwrap();

                builder
                    // Before we can draw, we have to *enter a render pass*.
                    .begin_render_pass(
                        RenderPassBeginInfo {
                            // A list of values to clear the attachments with. This list contains
                            // one item for each attachment in the render pass. In this case, there
                            // is only one attachment, and we clear it with a blue color.
                            //
                            // Only attachments that have `AttachmentLoadOp::Clear` are provided
                            // with clear values, any others should use `None` as the clear value.
                            clear_values: vec![Some([0.0, 0.0, 1.0, 1.0].into())],

                            ..RenderPassBeginInfo::framebuffer(
                                framebuffers[image_index as usize].clone(),
                            )
                        },
                        SubpassBeginInfo {
                            // The contents of the first (and only) subpass.
                            // This can be either `Inline` or `SecondaryCommandBuffers`.
                            // The latter is a bit more advanced and is not covered here.
                            contents: SubpassContents::Inline,
                            ..Default::default()
                        },
                    )
                    .unwrap()
                    // We are now inside the first subpass of the render pass.
                    //
                    // TODO: Document state setting and how it affects subsequent draw commands.
                    .set_viewport(0, [viewport.clone()].into_iter().collect())
                    .unwrap()
                    .bind_pipeline_graphics(pipeline.clone())
                    .unwrap()
                    .bind_vertex_buffers(0, vertex_buffer.clone())
                    .unwrap();

                unsafe {
                    builder
                        // We add a draw command.
                        .draw(vertex_buffer.len() as u32, 1, 0, 0)
                        .unwrap();
                }

                builder
                    // We leave the render pass. Note that if we had multiple subpasses we could
                    // have called `next_subpass` to jump to the next subpass.
                    .end_render_pass(Default::default())
                    .unwrap();

                // Finish recording the command buffer by calling `end`.
                let command_buffer = builder.end().unwrap();

                let future = previous_frame_end
                    .take()
                    .unwrap()
                    .join(acquire_future)
                    .then_execute(queue.clone(), command_buffer)
                    .unwrap()
                    // The color output is now expected to contain our triangle. But in order to
                    // show it on the screen, we have to *present* the image by calling
                    // `then_swapchain_present`.
                    //
                    // This function does not actually present the image immediately. Instead it
                    // submits a present command at the end of the queue. This means that it will
                    // only be presented once the GPU has finished executing the command buffer
                    // that draws the triangle.
                    .then_swapchain_present(
                        queue.clone(),
                        SwapchainPresentInfo::swapchain_image_index(swapchain.clone(), image_index),
                    )
                    .then_signal_fence_and_flush();

                match future.map_err(Validated::unwrap) {
                    Ok(future) => {
                        previous_frame_end = Some(future.boxed());
                    }
                    Err(VulkanError::OutOfDate) => {
                        recreate_swapchain = true;
                        previous_frame_end = Some(sync::now(device.clone()).boxed());
                    }
                    Err(e) => {
                        panic!("failed to flush future: {e}");
                        // previous_frame_end = Some(sync::now(device.clone()).boxed());
                    }
                }
            }
            Event::AboutToWait => window.request_redraw(),
            _ => (),
        }
    })
}

/// This function is called once during initialization, then again whenever the window is resized.
fn window_size_dependent_setup(
    images: &[Arc<Image>],
    render_pass: Arc<RenderPass>,
    viewport: &mut Viewport,
) -> Vec<Arc<Framebuffer>> {
    let extent = images[0].extent();
    viewport.extent = [extent[0] as f32, extent[1] as f32];

    images
        .iter()
        .map(|image| {
            let view = ImageView::new_default(image.clone()).unwrap();
            Framebuffer::new(
                render_pass.clone(),
                FramebufferCreateInfo {
                    attachments: vec![view],
                    ..Default::default()
                },
            )
            .unwrap()
        })
        .collect::<Vec<_>>()
}