[Bug middle-end/78888] New: toupper(x) can be assumed not to be in the range 'a' - 'z'

["msebor at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78888

            Bug ID: 78888
           Summary: toupper(x) can be assumed not to be in the range 'a' -
                    'z'
           Product: gcc
           Version: 7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: msebor at gcc dot gnu.org
  Target Milestone: ---

The return value of the toupper function is guaranteed not to be in the range
'a' - 'z'.  Similarly, the tolower return value is guaranteed not to be in the
range 'A' - 'Z'.  For non-EBCDIC character sets, it would be useful to set the
range on the return value reflecting this constrained range.  (The range is
actually the intersection of the range [0, UCHAR_MAX] and the anti-range ~['a',
'z'], plus the value EOF).

In addition, it would be useful to issue a warning if the return value is
compared against a constant from its anti-range since such a comparison in
always false (this might naturally fall out of the optimization).

The following test case shows that GCC does not take advantage of this
optimization (it doesn't remove the call to f()) or issue the warning.

$ cat d.c && gcc -O2 -S -Wall -Wextra -Wpedantic
-fdump-tree-optimized=/dev/stdout d.c 
void f (void);

void g (int x)
{
  if (__builtin_toupper ((unsigned char)x) == 'a')
    f ();
}

;; Function g (g, funcdef_no=0, decl_uid=1797, cgraph_uid=0, symbol_order=0)

Removing basic block 5
g (int x)
{
  int _1;
  int _6;

  <bb 2> [100.00%]:
  _6 = x_3(D) & 255;
  _1 = __builtin_toupper (_6);
  if (_1 == 97)
    goto <bb 3>; [22.95%]
  else
    goto <bb 4>; [77.05%]

  <bb 3> [22.95%]:
  f (); [tail call]

  <bb 4> [100.00%]:
  return;

}