https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124107
--- Comment #4 from Liam Powell <liam at liampwll dot com> --- Created attachment 63694 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=63694&action=edit testcase I have rewritten the testcase to avoid Unchecked_Access completely and attached it. I will also attach a patch as I hope that this is now sufficient to demonstrate that this is a GNAT bug. The patch adds non-trivial finalization to Root_Subpool, however it already contains a component with non-trivial finalization so this should not have any major impact. The restriction of generalized finalization to root types only means that a user can not provide their own finalization which would break this.
