[llvm] [llvm] Improve llvm.objectsize computation by computing GEP, alloca a… (PR #117849)

[Harald van Dijk via llvm-commits]

hvdijk wrote:

> @hvdijk : gentle ping :-)

Apologies for the late reply, I have been on holiday. In #115504 I commented that I am concerned that that resulted in incorrect results in some cases, which you were intending to fix in a subsequent PR. I cannot tell whether the current implementation is or is not meant to address those cases. If it is, great! If not, then I cannot offer a meaningful review. For reviewing PRs, I want to make sure that even if they are not perfect yet, they at least do not make things any worse. But this being the rather complicated bit of logic that it is, if there are known incorrect results in some cases, I am not capable of grasping the logic well enough to make sure that a PR is an improvement.

https://github.com/llvm/llvm-project/pull/117849