[Openmp-commits] [openmp] [OpenMP] [OMPT] A pointer to HostOpId should be passed in EMI callbacks. (PR #75574)

[via Openmp-commits]

dhruvachak wrote:

> > Thanks for taking care of this, Dhruva. Do we have a test that covers this?
> 
> Though we have tests for OMPT that exercise this code path, it is not clear how to test for it unless we have support for trace records, which we don't yet in upstream code. Before this patch, the tool is getting a pointer to the wrong location. Unless there is a consumer of the correct location, I am not sure how to have a test to detect this problem.

There is a test downstream (with support for trace records) if that helps. https://github.com/ROCm/aomp/pull/762

https://github.com/llvm/llvm-project/pull/75574