[PATCH] c++: Private parent access check for using decls [PR19377]

[Jason Merrill]

On 2/24/21 4:17 PM, Anthony Sharp wrote:
>> "special"
> 
> 
> It wouldn't be my code if it didn't have sp3ling mstakes innit!
> Actually to be fair I already changed that spelling mistake a few days
> ago in my local code ;)
> 
> I was actually thinking about this last night as I was falling asleep
> (as you do) and I realised that the whole of my using decl lookup is
> redundant. I can simply do this (formatting probably messes up here):
> 
> /* 1.  If the "using" keyword is used to inherit DECL within the parent,
>       this may cause DECL to be private, so we should return the using
>       statement as the source of the problem.
> 
>       Scan the fields of PARENT_BINFO and see if there are any using decls.  If
>       there are, see if they inherit DECL.  If they do, that's where DECL must
>       have been declared private.  */
> 
>    for (tree parent_field = TYPE_FIELDS (BINFO_TYPE (parent_binfo));
>         parent_field;
>         parent_field = DECL_CHAIN (parent_field))
>      {
>        /* Not necessary, but also check TREE_PRIVATE for the sake of
>        eliminating obviously non-relevant using decls.  */
>        if (TREE_CODE (parent_field) == USING_DECL
>   && TREE_PRIVATE (parent_field))
> {
> /* If the using statement inherits DECL, it is the source of the
>   access failure, so return it.  */
>   if (cp_tree_equal (strip_using_decl (parent_field), decl))
>     return parent_field;
> }
>      }
> 
> I was wrong to say that the using decl does not store "where it came
> from/what it inherits" - that's exactly what strip_using_decl
> achieves. I think the problem was that when I did my initial testing
> in trying out ways to get the original decl, I didn't strip it, so the
> comparison failed, which led me to make the whole redundant lookup,
> blah blah blah.
> 
> I've run a quick test and it seems to work, even with the overloads.
> 
> Will test it some more and if all's good I will probably send a new
> patch some time this weekend.

Sounds good, though strip_using_decl (parent_field) may be overloaded if 
the using-decl brings in multiple functions with that name.

Jason