[Bug tree-optimization/107952] tree-object-size: inconsistent size for flexible arrays nested in structs

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107952

--- Comment #18 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Qing Zhao from comment #15)
> > On Jan 25, 2023, at 11:12 AM, siddhesh at gcc dot gnu.org <gcc-bugzilla@gcc.gnu.org> wrote:
> >> 
> >>> The first is handled by the function just fine,
> >> 
> >> No, even the first case is not recognized by the current
> >> “array_ref_flexible_size_p”, it’s not been identified as a flexible array
> >> right now.
> >> Shall we include this case into “array_ref_flexible_size_p”?  (It’s a GCC
> >> extension).
> > 
> > In the first case, array_ref_flexible_size_p recognizes S2.flex.data as having
> > flexible size.  The tests in my patch[1] for this bug checks for this.
> Oh, yes. That’s right.
> > 
> > However, array_ref_flexible_size_p does not recognize S2.flex as having
> > flexible size.  It might make sense to support that, i.e. any struct or union
> > with the last element as a flex array should be recognized as having flexible
> > size.
> 
> Since S2.flex is not an “array_ref”, it’s correct for
> array_ref_fleixble_size_p to return false for it, I think. 

Yes, that's correct.

The routine is supposed to answer if an actual _access_ is to a flexible
size part.  The whole aggregate is not part of a flexible size part here.

tree-object-size usually wants to know if a pointer points to an object
with flexible size which is really something different and might not
share much with the logic in array_ref_fleixble_size_p