[Bug middle-end/95619] New: apparently in-bounds subscript in -Warray-bounds message

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95619

            Bug ID: 95619
           Summary: apparently in-bounds subscript in -Warray-bounds
                    message
           Product: gcc
           Version: 10.0
            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: ---

When the type of the access is obscured by a typedef (or macro, or, in C++, by
a template parameter), -Warray-bounds warnings like the one below must be
confusing.  The warnings don't make it clear how "array subscript 2" can be 
"outside array bounds of ‘X[3]’ {aka ‘short int[3]’}".

Either the warning should also mention the type of the access, maybe in the
note, or perhaps better, the byte offset into the array (when the the types
aren't the same).

$ cat z.c && gcc -O2 -S -Wall -Wextra -fdump-tree-optimized=/dev/stdout z.c
typedef short X;
typedef int Y;

void f (void*);

void g (void)
{
  X a[3];
  Y *p = (Y*)a;
  p[2] = 1;
  f (p);
}
z.c: In function ‘g’:
z.c:10:4: warning: array subscript 2 is outside array bounds of ‘X[3]’ {aka
‘short int[3]’} [-Warray-bounds]
   10 |   p[2] = 1;
      |   ~^~~
z.c:8:5: note: while referencing ‘a’
    8 |   X a[3];
      |     ^

;; Function g (g, funcdef_no=0, decl_uid=1935, cgraph_uid=1, symbol_order=0)

g ()
{
  X a[3];

  <bb 2> [local count: 1073741824]:
  MEM[(Y *)&a + 8B] = 1;
  f (&a);
  a ={v} {CLOBBER};
  return;

}