Code using AccelerationStructureInstanceKHR does not compile

[ashpil]

vk.zig:4902:5: error: packed structs cannot contain fields of type 'vk.TransformMatrixKHR'
    transform: TransformMatrixKHR,
vk.zig:4902:5: note: only packed structs layout are allowed in packed types
vk.zig:4897:39: note: struct declared here
pub const TransformMatrixKHR = extern struct {
                               ~~~~~~~^~~~~~
referenced by:
    AccelerationStructureInstanceKHR: vk.zig:4901:53

The issue appears to be that AccelerationStructureInstanceKHR is defined to be a packed struct, but it contains TransformMatrixKHR -- an extern struct, and doing so is now illegal in Zig. Not sure what the solution is here -- presumably, the only thing that actually works is to ensure all structs in packed structs are packed as well, but not sure if we can actually rely on this behaving as expected, as that would require replacing extern with packed in some scenarios, which the C ABI may not like.

Though a cursory glance makes me think only TransformMatrixKHR and SRTDataNV would need to undergo the extern -> packed transition, so at least for now it should be fine.

Also, since packed structs cannot be in extern structs, there's a couple places where that would need to be amended as well -- just the simple replacement solution as above seems even sketchier here. There are more scenarios affected by this.

[ashpil]

Also, unrelated, but this not being noticed earlier makes me think this project should have one of those things some Zig projects have where it just references every struct to ensure it typechecks (though this specific issue seems to need an initialization attempt to typecheck, not just a reference).

[Snektron]

Ah yes, this type has given some trouble before. It used to work before the packed struct overhaul, I think.

Also, unrelated, but this not being noticed earlier makes me think this project should have one of those things some Zig projects have where it just references every struct to ensure it typechecks (though this specific issue seems to need an initialization attempt to typecheck, not just a reference).

Yes, thats a good idea

[Snektron]

Added this refAllDecls-like test also (even more elaborate than the STD one): https://github.com/Snektron/vulkan-zig/blob/master/test/ref_all_decls.zig

Of course it uncovered a bunch of issues, those are also all fixed now.

[ashpil]

Thanks! Looks like vulkan-zig received a lot of improvements today! 😄