[Bug tree-optimization/87859] [8/9 Regression] store-merging miscompilation of mesa

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87859

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I didn't have access to the preprocessed source and was lazy to prepare it
myself, so I just copied the corresponding structure definition without C++
ctors/dtors, figured out what I thought would be a bitset_t, and tried if I can
reproduce it with that.  As I could, I have cleaned that up and replaced the
fields so that they are named in a way to make it more readable what is going
on.
Now that I think about it, I think I could fix this differently by adding a
flag to the store structures that it has been merged already, but I'll postpone
that to next week, need to work on stage1 stuff this week.