[Bug c/85800] A miscompilation bug with unsigned char

[juneyoung.lee at sf dot snu.ac.kr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85800

--- Comment #2 from Juneyoung Lee <juneyoung.lee at sf dot snu.ac.kr> ---
If address is not adjacent - you can reorder the definition of A and B and try
again.
```
  int A[4], B[4];
```

If changing

    if (c1 == c2)
      // Always true if p and q have same integer representation.
      arr3[i] = arr1[i];
    else
      arr3[i] = arr2[i];

to

    if (c1 == c2)
      // Always true if p and q have same integer representation.
      arr3[i] = arr2[i];
    else
      arr3[i] = arr1[i];

removed miscompilation, I think this implies that the former one is folded into
`arr[3] = arr2[i]` (which is incorrect, because arr2 is equivalent to `q`) but
the latter one is folded into `arr[3] = arr1[i]`.

This explains the generated assembly as well: Compiling store_10_to_p with -O3
generates

```
store_10_to_p(int*, int*):
  mov DWORD PTR [rdi], 1
  mov DWORD PTR [rsi], 10
  ret
```

meaning that it is actually trying to store 10 to *q, which is UB if the
function is inlined. (B[4] is out-of-bounds)