https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71446
Jakub Jelinek <jakub at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #45856|0 |1 is obsolete| | --- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Created attachment 45857 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=45857&action=edit gcc9-pr71446-wip.patch So, that would be following. Except that it breaks desig{4,6}.C testcases, but those try to implement what [over.ics.list]/2 says. So, it is unclear to me what exactly is build_aggr_conv actually supposed to do. If the desig11.C testcase is valid, then we can't just ignore the designators as we were doing before, but on the other side if we must ignore the order, then would we need to lookup up each field?