[PATCH] D70172: [CUDA][HIP] Fix assertion due to dtor check on windows

Wed Dec 4 15:26:39 PST 2019

rnk added a comment.

In D70172#1746451 <https://reviews.llvm.org/D70172#1746451>, @yaxunl wrote:

> We are not using Itanium ABI when we do host compilation of CUDA/HIP on windows. During the host compilation on windows only MS C++ ABI is used.
>
> This issue is not due to mixing MS ABI with Itanium ABI.
>  ...

I think I might have understood all that.

Really, the problem is that, in C++, there are many kinds of special members created by the compiler that are not modeled in the AST. Deleting destructors are a good example. If we consistently used GlobalDecl throughout Sema, then we would be able to separate marking the deleting destructor referenced from marking the base destructor referenced, and this code would be easier to understand.

However, given the way things stand, your new approach seems like a reasonable way of detecting the case of referencing the deleting dtor here. So from my perspective, this is fine. @rjmccall, assuming that Richard doesn't have time to give any input, do you still think this needs his review?

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70172/new/

https://reviews.llvm.org/D70172