Missed optimization: fold (select C, (x | a), x) | b -> x | select C, (a | b), b

[zxt5]

define i8 @src(i8 %arg0, i8 %arg1) {
  %v0 = icmp eq i8 %arg1, -1
  %v1 = or i8 %arg0, 4
  %v2 = select i1 %v0, i8 %v1, i8 %arg0
  %v3 = or i8 %v2, 1
  ret i8 %v3
}

define i8 @tgt(i8 %arg0, i8 %arg1) {
  %v0 = icmp eq i8 %arg1, -1
  %v1 = select i1 %v0, i8 5, i8 1
  %v2 = or i8 %arg0, %v1
  ret i8 %v2
}

Alive2: https://alive2.llvm.org/ce/z/Nq6b4_
Godbolt: https://godbolt.org/z/dnv46Kede

Pattern found in https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/linux/optimized/fib_trie.ll