[Bug tree-optimization/98467] gcc optimizes tapping code away

[bernd.edlinger at hotmail dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98467

--- Comment #2 from Bernd Edlinger <bernd.edlinger at hotmail dot de> ---
I debugged a bit in when we decide this function is const.
That appears to be in gcc/ipa-fnsummary.c:

/* Return true if T is a pointer pointing to memory location that is local
   for the function (that means, dead after return) or read-only.  */

bool
points_to_local_or_readonly_memory_p (tree t)
{
  /* See if memory location is clearly invalid.  */
  if (integer_zerop (t))
    return flag_delete_null_pointer_checks;
  if (TREE_CODE (t) == SSA_NAME)
    return !ptr_deref_may_alias_global_p (t);
  if (TREE_CODE (t) == ADDR_EXPR)
    return refs_local_or_readonly_memory_p (TREE_OPERAND (t, 0));
  return false;
}



And indeed the "problem" can be fixed by using -fno-delete-null-pointer-checks.
>From the documentation in gcc/doc/invoke.texi I would never have guessed
what this option does here:

"@item -fdelete-null-pointer-checks
@opindex fdelete-null-pointer-checks
Assume that programs cannot safely dereference null pointers, and that
no code or data element resides at address zero.
This option enables simple constant
folding optimizations at all optimization levels.  In addition, other
optimization passes in GCC use this flag to control global dataflow
analyses that eliminate useless checks for null pointers; these assume
that a memory access to address zero always results in a trap, so
that if a pointer is checked after it has already been dereferenced,
it cannot be null."


It is only supposed to remove checks for null-pointer, not the other
way round...