Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look
OK for trunk and eventually backports?
-- >8 --
Here the normal form of the two immediately-declared D<<placeholder>, V>
constraints is the same, so we rightfully share the normal form between
them. However we first normalize the constraint from the context of
auto deduction for W in which case the placeholder has level 2 and so
the set of in-scope template parameters has depth 2 (a dummy level is
added from normalize_placeholder_type_constraints).
This depth 2 means that find_template_parameters when it sees V within
the requires-expr will recurse into its TREE_TYPE, an auto of level 2,
and mark the atomic constraint as depending on the template parameter of
depth 2 index 0 (alongside the obvious depth 1 index 0 dependence),
which is clearly wrong. Later during satisfaction of B's constraints we
lack two levels of template arguments and crash when consulting the
satisfaction cache for the shared atomic constraint.
I think when find_template_parameters sees an NTTP, it doesn't need to
walk its TREE_TYPE because NTTP substitution is done oblviously with
respect to its type -- only the corresponding NTTP argument matters.
This is more clearly true for (unconstrained) auto NTTPs, but I think
applies to all NTTPs. Doing so fixes the testcase because we no longer
record any depth 2 dependence when walking V.
PR c++/121981
gcc/cp/ChangeLog:
* pt.cc (any_template_parm_r) <case TEMPLATE_TYPE_PARM>:
Don't walk TREE_TYPE.
gcc/testsuite/ChangeLog:
* g++.dg/cpp2a/concepts-placeholder15.C: New test.
---
gcc/cp/pt.cc | 4 +++-
.../g++.dg/cpp2a/concepts-placeholder15.C | 17 +++++++++++++++++
2 files changed, 20 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-placeholder15.C
diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index b7cb807d8daa..276e1a4c7bb7 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -11125,7 +11125,9 @@ any_template_parm_r (tree t, void *data)
break;
case TEMPLATE_PARM_INDEX:
- WALK_SUBTREE (TREE_TYPE (t));
+ /* No need to walk TREE_TYPE: substitution into an NTTP is done
+ directly with the corresponding template argument, and its
+ type only came into play earlier during coercion. */
break;
case TEMPLATE_DECL:
diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-placeholder15.C
b/gcc/testsuite/g++.dg/cpp2a/concepts-placeholder15.C
new file mode 100644
index 000000000000..e6571e9f18cd
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp2a/concepts-placeholder15.C
@@ -0,0 +1,17 @@
+// PR c++/121981
+// { dg-do compile { target c++20 } }
+
+template<auto V>
+concept C = requires { V; };
+
+template<class T, auto V>
+concept D = C<V>;
+
+template<auto V, D<V> auto W>
+struct A { };
+
+template<auto V, D<V> T>
+struct B { };
+
+A<0, 1> a;
+B<0, int> b;
--
2.51.0.268.ga483264b01
