Buffer leaks about wgpu

[teoxoy]

Are the StagingBuffers not getting destroyed? Or both the Buffers and the StagingBuffers?

[teoxoy]

The issue might also be that we don't unmap the staging buffer once we are done with it

wgpu/wgpu-core/src/resource.rs

Lines 673 to 679 in 9796766

	if !staging_buffer.is_coherent {
	unsafe {
	device
	.raw()
	.flush_mapped_ranges(raw_staging_buffer, iter::once(0..self.size));
	}
	}

We could call .flush() on it like in all the other places

wgpu/wgpu-core/src/device/queue.rs

Lines 342 to 353 in 9796766

	unsafe fn flush(&self, device: &A::Device) -> Result<(), DeviceError> {
	if !self.is_coherent {
	unsafe {
	device.flush_mapped_ranges(
	self.raw.lock().as_ref().unwrap(),
	iter::once(0..self.size),
	)
	};
	}
	unsafe { device.unmap_buffer(self.raw.lock().as_ref().unwrap())? };
	Ok(())
	}

[xiaopengli89]

Are the StagingBuffers not getting destroyed? Or both the Buffers and the StagingBuffers?

It all seems to have leaked.

The issue might also be that we don't unmap the staging buffer once we are done with it

If we drop the resource before free suballocation, the leak is gone, which maybe cause by internal mechanism of gpu-allocator.