[PATCH] D104505: [HIP] Defer operator overloading errors

[Yaxun Liu via Phabricator via cfe-commits]

yaxunl added a comment.

In D104505#2833746 <https://reviews.llvm.org/D104505#2833746>, @tra wrote:

> In D104505#2833271 <https://reviews.llvm.org/D104505#2833271>, @yaxunl wrote:
>
>> Such host/device overloading resolution induced issue is not limited to device functions calling host functions.
>
> It does not change the fact that the code in the test above is invalid, regardless of whether we compile it on the host or on the device.
>
>> It could also happen to host device functions calling host functions, which is less controversial for deferring.
>
> H/D functions are special, because their overloading is affected whether compilation is done on the host or on the device side and we often can not tell whether the diagnostic is appropriate until codegen.
>
> I still think that deferring diags for unambiguously invalid code is not a good idea. The fact that NVCC can only diagnose such errors during device-side compilation is not a good enough reason, IMO, to make clang ignore real errors, even if we'd still end up eventually failing later, during device-side compilation.
>
> @rsmith, @rjmccall  -- any thoughts?

We don't defer such diags by default. We only defer them under option -fgpu-defer-diags, which users have to specify explicitly.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D104505/new/

https://reviews.llvm.org/D104505