template <typename T> struct outer { template <T...> struct inner {}; template <T... Ts> struct inner<T{}, Ts...> {}; }; This fails to compile, although the code is well-formed (in particular, [temp.class.spec]/(8.1) and (8.2) are not violated); Apparently, GCC fails to correctly distinguish nested levels of template parameters.
It still fails even like this: template <typename T> struct outer { static constexpr T t{}; template <T...> struct inner {}; template <T... Ts> struct inner<t, Ts...> {}; };
The master branch has been updated by Patrick Palka <ppalka@gcc.gnu.org>: https://gcc.gnu.org/g:9b94785dedb08b006419bec1a402614d9241317a commit r12-1093-g9b94785dedb08b006419bec1a402614d9241317a Author: Patrick Palka <ppalka@redhat.com> Date: Thu May 27 12:36:39 2021 -0400 c++: Relax rule for non-type arguments in partial specs [CWG1315] This implements the wording changes of CWG 1315, which permits non-type template arguments in a partial specialization to use template parameters more freely. Delightfully, it seems the only change needed is to remove a single check from process_partial_specialization (amidst a bunch of whitespace changes). But that change alone revealed a latent problem with for_each_template_parm: it ends up looking into some non-deduced contexts even when include_nondeduced_p is false. This causes us to silently accept some partial specializations within the testsuite that contain non-deducible non-type template parameters (and that were previously rejected due to the rule that CWG 1315 relaxed). For now this patch makes a minimal amount of changes to for_each_template_parm_r so that we continue to reject existing ill-formed partial specializations within the testsuite. I opened PR c++/100779 to track this issue. DR 1315 PR c++/67593 PR c++/96555 gcc/cp/ChangeLog: * pt.c (process_partial_specialization): Don't error on a non-simple non-type template argument that involves template parameters. (for_each_template_parm_r): Don't walk TRAIT_EXPR, PLUS_EXPR, MULT_EXPR, or SCOPE_REF when include_nondeduced_p is false. gcc/testsuite/ChangeLog: * g++.dg/template/partial16.C: New test. * g++.dg/template/partial17.C: New test. * g++.dg/template/partial18.C: New test. * g++.dg/template/partial19.C: New test. * g++.dg/cpp0x/pr68724.C: Adjust expected diagnostic for ill-formed partial specialization. * g++.dg/cpp0x/variadic38.C: Likewise. * g++.dg/cpp1z/pr81016.C: Likewise. * g++.dg/template/partial5.C: Likewise. * g++.old-deja/g++.pt/spec21.C: Likewise.
Fixed for GCC 12.
*** Bug 51155 has been marked as a duplicate of this bug. ***