[Bug ipa/101502] New: Inconsistent behavior in maybe_record_node()

[Bug ipa/101502] New: Inconsistent behavior in maybe_record_node()

[fxue at os dot amperecomputing.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101502

            Bug ID: 101502
           Summary: Inconsistent behavior in maybe_record_node()
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ipa
          Assignee: unassigned at gcc dot gnu.org
          Reporter: fxue at os dot amperecomputing.com
                CC: marxin at gcc dot gnu.org
  Target Milestone: ---

For code snippet in maybe_record_node(), behavior might be inconsistent when
SANITIZE_UNREACHABLE is turned on. Suppose we are adding two functions
(method_fn/__cxa_pure_virtual), content of nodes is order-sensitive.

  o. method_fn, __cxa_pure_virtual ->  nodes: [ method_fn ]
  o. __cxa_pure_virtual, method_fn ->  nodes: [ __cxa_pure_virtual, method_fn ]

BTW: assertion on target_node->real_symbol_p() is redundant, since the
condition is contained.

  else if (target_node != NULL
           && (TREE_PUBLIC (target)
               || DECL_EXTERNAL (target)
               || target_node->definition)
           && target_node->real_symbol_p ())
    {
      gcc_assert (!target_node->inlined_to);
      gcc_assert (target_node->real_symbol_p ());
      /* When sanitizing, do not assume that __cxa_pure_virtual is not called
         by valid program.  */
      if (flag_sanitize & SANITIZE_UNREACHABLE)
        ;
      /* Only add pure virtual if it is the only possible target.  This way
         we will preserve the diagnostics about pure virtual called in many
         cases without disabling optimization in other.  */
      else if (pure_virtual)
        {
          if (nodes.length ())
            return;
        }
      /* If we found a real target, take away cxa_pure_virtual.  */
      else if (!pure_virtual && nodes.length () == 1
               && is_cxa_pure_virtual_p (nodes[0]->decl))
        nodes.pop ();
      if (pure_virtual && nodes.length ())
        return;
      if (!inserted->add (target))
        {
          cached_polymorphic_call_targets->add (target_node);
          nodes.safe_push (target_node);
        }
    }

[Bug ipa/101502] Inconsistent behavior in maybe_record_node()

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101502

--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
@Honza: Can you please take a look?
@Feng: Do you have a test-case for it, please?

[Bug ipa/101502] Inconsistent behavior in maybe_record_node()

[fxue at os dot amperecomputing.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101502

--- Comment #2 from Feng Xue <fxue at os dot amperecomputing.com> ---
(In reply to Martin Liška from comment #1)
> @Honza: Can you please take a look?
> @Feng: Do you have a test-case for it, please?

No. I just got this from code logic, but not 100% sure.