https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115897
Patrick Palka <ppalka at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Target Milestone|--- |14.2 Assignee|unassigned at gcc dot gnu.org |ppalka at gcc dot gnu.org Known to fail| |14.1.0, 15.0 Known to work| |13.3.0 Status|UNCONFIRMED |ASSIGNED Last reconfirmed| |2024-07-12 Ever confirmed|0 |1 Summary|[14/15 Regression] |[14/15 Regression] |vector_si |vector_size attribute on | |alias template has no | |effect when used in a | |dependent variable | |template-id --- Comment #3 from Patrick Palka <ppalka at gcc dot gnu.org> --- ... which means the comment #2 testcase using a class instead of variable template-id has been failing for much longer, and r14-2170 just caused this preexisting issue to affect variable template-ids too. The problem seems to be that we're stripping the alias template-id _Vp<_Tp> to _Tp too soon, before we have an argument for _Tp, because we wrongly think the two types are equivalent, but they're not due to the attribute. complex_alias_template_id probably should return true for alias attributes that have an attribute. A workaround is to give _Vp a dummy defaulted template parameter, e.g. template <typename _Tp, typename _Up = _Tp> using _Vp [[gnu::vector_size(16)]] = int;