[Bug tree-optimization/117965] phiprop fails in some cases

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117965

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> Confirmed. Hmm, phiprop2 can handle it though; that is
> -fdisable-tree-phiprop1 and see that phiprop2 handles it.
> 
> I have not looked into why yet.

That is definitely eh is related. pass_cleanup_eh removes the finally block
which just had clobbers and that is how it works in phiprop2.