[Bug tree-optimization/104528] [12 Regression] ICE: Segmentation fault (in gimple_bb)

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104528

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
At the start of IVOPTs we have

(gdb) p debug_tree (cfun->x_current_loops->larray->m_vecdata[4]->nb_iterations)
 <minus_expr 0x7ffff6348f28
    type <integer_type 0x7ffff670ce70 public unsigned SI
        size <integer_cst 0x7ffff6517e40 constant 32>
        unit-size <integer_cst 0x7ffff6517e58 constant 4>
        align:32 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff670ce70 precision:32 min <integer_cst 0x7ffff6321078 0> max <integer_cst
0x7ffff6321060 4294967295>>

    arg:0 <ssa_name 0x7ffff631e240 type <error_mark 0x7ffff6517be8>
        nothrow
        def_stmt 
        version:36 in-free-list>
    arg:1 <nop_expr 0x7ffff6349400 type <integer_type 0x7ffff670ce70>

        arg:0 <ssa_name 0x7ffff63345a0 type <integer_type 0x7ffff652f5e8
integer(kind=4)>
            visited
            def_stmt pretmp_90 = ii;
            version:90>>>
$12 = void

so that mentions a released SSA name.  It's released during DSE which
fails to reset niter info since we now prune dead SSA names there.