[Bug tree-optimization/108166] [12/13 Regression] Wrong code with -O2 since r12-8078-ga42aa68bf1ad745a

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108166

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 54142
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54142&action=edit
gcc13-pr108166.patch

Untested fix.  I wouldn't change replace_phi_edge_with_variable, because even
without the duplicate_ssa_name_range_info (say if new_tree already has
SSA_NAME_RANGE_INFO) we'd be lying to the compiler.
The right thing would be to union the global range of the phi result with the
oarg INTEGER_CST, not sure how hard would that be.