Add macros for anonymous union support

[joselopeqti]

Introduce __CL_HAS_ANON_UNION__ and __CL_ANON_UNION__ for anonymous union support.
Anonymous unions are supported in C++11 and later versions, but not anonymous structs.
This change explicitly enables anonymous unions for C++11 or later, otherwise, it will fall back to the previous anonymous structs logic.

[CLAassistant]

All committers have signed the CLA.

[Kerilk]

Thanks for the contribution, I think this is a good idea.
It would require to update the loader code as well:
https://github.com/KhronosGroup/OpenCL-ICD-Loader/blob/02134b05bdff750217bf0c4c11a9b13b63957b04/loader/icd.h#L185-L189

#if __CL_HAS_ANON_STRUCT__
#define ICD_ANON_UNION_INIT_MEMBER(a) {a}
#else
#define ICD_ANON_UNION_INIT_MEMBER(a) a
#endif

[Kerilk]

@karolherbst Would this renaming change cause a problem with the way you use the headers for rusticl?

[karolherbst]

@karolherbst Would this renaming change cause a problem with the way you use the headers for rusticl?

No, the only thing that matters is how the structs look like after preprocessing and this change doesn't change that as the bindings are generated in C mode, not C++.

[joselopeqti]

Thanks for the contribution, I think this is a good idea. It would require to update the loader code as well: https://github.com/KhronosGroup/OpenCL-ICD-Loader/blob/02134b05bdff750217bf0c4c11a9b13b63957b04/loader/icd.h#L185-L189

#if __CL_HAS_ANON_STRUCT__
#define ICD_ANON_UNION_INIT_MEMBER(a) {a}
#else
#define ICD_ANON_UNION_INIT_MEMBER(a) a
#endif

I have just pushed a PR for it KhronosGroup/OpenCL-ICD-Loader#264, I guess the CI will pass once this change is merged.

[Kerilk]

Looking good, I think we should accept this.

[bashbaug]

Merging as discussed in the October 14th teleconference.