bwendling wrote:

The first PR attempt was here: https://github.com/llvm/llvm-project/pull/78526

It was NACK'ed because it used the LLVM IR representation of the structure, 
which wasn't appropriate. To solve that issue, I chose to expand the 
`llvm.objectsize()` builtin to contain the size and offset of the sub-object, 
which is determined in the front-end.

Note that there are many other things wrong with our 
`__builtin_{dynamic_}object_size` implementations. For one, they perform loads 
of the pointer which isn't necessary and contrary to the idea that the builtins 
don't allow for side-effects. I'll be addressing those issues in future patches.

https://github.com/llvm/llvm-project/pull/83204
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to