llvm dev - Jan 2019 - Issue with "t -> signature is meaningless, use custom typechecking"

Hello,

I was implementing the llvm_any_type in my intrinsic
def int_csa_xxx :  Intrinsic<[llvm_any_ty], [llvm_i32_ty]>;

 as the following in its corresponding builtins in Builtins.def:
BUILTIN(__builtin_xxx, "v.", "nt")

the "t" was sufficient here to not perform any type checking.

The type checking was handled in CGBuiltin.cpp.

This was working until recently.

Do you know which change might have caused this?

Thanks,
Dounia
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181230/ce58163d/attachment.html>

On 12/30/2018 6:35 PM, Dounia Khaldi via llvm-dev wrote:
> Hello,
>
> I was implementing the llvm_any_type in my intrinsic
> def int_csa_xxx :  Intrinsic<[llvm_any_ty], [llvm_i32_ty]>;
>
>  as the following in its corresponding builtins in Builtins.def:
> BUILTIN(__builtin_xxx, "v.", "nt")
>
> the "t" was sufficient here to not perform any type checking.
>
> The type checking was handled in CGBuiltin.cpp.
>
> This was working until recently.
>
> Do you know which change might have caused this?
Adding cfe-dev; generally, you'll get better responses there to 
questions about clang.

You forgot to include a description of what, exactly, is broken.

In general, you should be doing custom type-checking in 
SemaChecking.cpp; some important operations can't be performed correctly 
in CGBuiltin, like lvalue-to-rvalue conversions.

-Eli

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux
Foundation Collaborative Project