[Bug tree-optimization/86415] New: TREE_STRING_LENGTH incorrect for constant arrays

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86415

            Bug ID: 86415
           Summary: TREE_STRING_LENGTH incorrect for constant arrays
           Product: gcc
           Version: 9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: msebor at gcc dot gnu.org
  Target Milestone: ---

While testing a solution to bug 77357 I noticed that GCC fails to fold strlen()
calls with constant char array arguments initialized by constant strings with
an  offset that's past the terminating NUL character unless the elements past
the NUL are initialized explicitly as opposed to implicitly.  I tracked it down
to the c_getstr() function which normally returns a pointer to the string at
the offset.  The function uses TREE_STRING_LENGTH() to obtain the size of the
string (i.e., the result of the sizeof(str) expression, as it should) but the
macro returns the length of the string rather than its size.

$ cat d.c && gcc -O2 -S -Wall -Wextra -fdump-tree-optimized=/dev/stdout d.c
const char a[7] = "123\000\000\000";

int f (void)
{
  return __builtin_strlen (a + 5);   // folded
}

const char b[7] = "123";

int g (void)
{
  return __builtin_strlen (b + 5);   // not folded
}


;; Function f (f, funcdef_no=0, decl_uid=1899, cgraph_uid=1, symbol_order=1)

f ()
{
  <bb 2> [local count: 1073741825]:
  return 0;

}



;; Function g (g, funcdef_no=1, decl_uid=1903, cgraph_uid=2, symbol_order=3)

g ()
{
  long unsigned int _1;
  int _3;

  <bb 2> [local count: 1073741825]:
  _1 = __builtin_strlen (&MEM[(void *)&b + 5B]);
  _3 = (int) _1;
  return _3;

}