On Mon Nov 20 13:05:32 2023 +0000, Francisco Casas wrote:
Without thinking it through fully, that sounds like copy-prop is even
more broken. Fundamentally we shouldn't be propagating a store that happens *after* the load we're reaching through. Yep, at some point I realized that and the current version of the MR handles the problem. Check my other comment in the [MR](https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/482#note_52713) or the commit message on 3/3.
Well, it still begs the question of why 1/3 is there then, especially with the commit message that it has.