[HLSL] static in cbuffer legality? Might require a test fix.

[Keenuts]

Hello!

from clang/test/CodeGenHLSL/static_global_and_function_in_cb.hlsl

cbuffer {
    static float b = 3;
}

This test checks b is correctly emitted as a global variable. But is that even legal?
DXC kinda doesn't allow that for Vulkan: we warn, but compile and then fail validation.
MSDN doesn't mentions that: https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/dx-graphics-hlsl-constants

And since cbuffers are supposed to be a "view" on an external buffer, having a static defined inside doesn't seem right.

[damyanp]

@hekota, this might be interesting to you?

[hekota]

This is mirroring DXC behavior by emitting a global variable b as is it was declared outside of the cbuffer context. We should consider making this an error in Clang. Same goes for function definitions inside cbuffer - I don't see any reason why that needs to be allowed.

We are tracking that here: #118524

[damyanp]

We need to make a decision on what to do here before we can estimate and assign it to someone to work on.

[damyanp]

Preference is to disallow statics in cbuffers.

Next steps:

• get this in to a HLSL202x proposal (llvm-beanz)
• Use this issue to track implementing disallowing this in cbuffers (hekotas)