push_debug_group requires cfg!(debug_assertions) under Vulkan

[Aaron1011]

Description

The RenderPass::push_debug_group method is very useful for debugging a complex application. However, wgpu-core only enables the required InstanceFlags::DEBUG when cfg!(debug_assertions) is enabled:

https://github.com/gfx-rs/wgpu/blob/4ab9dbea8f7efbad518c3e39db07c30177a1a6af/wgpu-core/src/instance.rs#L77C20-L80

I couldn't find this documented anywhere. Ideally, this could be explicitly enabled when calling wgpu::Instance::new, without the need to turn on debug assertions for the entire program.

Repro steps
Run any of the examples that use encoder.push_debug_group (e.g. examples/cube) with and without debug assertions (for example, in both debug and release mode) under a program like RenderDoc.

Expected vs observed behavior
Debug groups are not shown in Renderdoc unless debug assertions are enabled.

Extra materials

Platform
wgpu 0.16.2
Linux
Vulkan

[cwfitzgerald]

Yes, this is something we need to disinstangle. Specifically we want to raise all of the instance flags up to the wgpu user as it's their decision, not ours.

[teoxoy]

This was resolved by #4230.