https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125998
--- Comment #15 from Mikael Morin <mikael at gcc dot gnu.org> --- With that patch, the fix for PR97592 can probably be reverted or reconsidered. It's a question of choosing between gfc_conv_subref_array_arg which creates an unconditional temporary inline thus optimizable, and gfc_conv_array_parameter which calls internal_pack so temporary only if necessary but no inline code so no benefit from code optimizations. Not sure it's worth the trouble; probably not.
