[PATCH] OpenCL/SPIR - kernel_arg_base_type metadata node

[Pekka Jääskeläinen]

On 07/30/2014 12:22 PM, Fraser Cormack wrote:
> For instance, if someone did "typedef uint myunsignedinteger" then from
> what I can see, kernel_arg_type would report it as "myuinteger" which
> would probably not be conformant. What do you think?

Yes it seems a bit fragile, and while I couldn't come up with a breaking
case yesterday, there seems to be some, like the one you pointed out.

> I'll work on that test case for this, then I can perhaps later provide a
> patch which fixes the possible bug in the unsigned -> u logic.

Good.

-- 
--Pekka