Substituting into the fold-expression was producing an
EXPR_PACK_EXPANSION, but it would be better to keep it as a
fold-expression. tsubst_unary_left_fold et al try to do this, but
weren't recognizing the case where tsubst_pack_expansion returns a
TREE_VEC containing a single pack expansion. In that case, let's just
return the pack expansion without the TREE_VEC, as we do earlier in
tsubst_pack_expansion for similar situations.
Tested x86_64-pc-linux-gnu, applying to trunk and 6.
commit 994ccc9de1117b6388859663cb9817e20ed66c50
Author: Jason Merrill <ja...@redhat.com>
Date: Wed Aug 3 19:19:32 2016 -0400
PR c++/72415 - member template with fold-expression constraint
* pt.c (tsubst_pack_expansion): Pull a single pack expansion out
of the TREE_VEC.
diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
index bf729ea..60c87e0 100644
--- a/gcc/cp/pt.c
+++ b/gcc/cp/pt.c
@@ -11160,6 +11160,12 @@ tsubst_pack_expansion (tree t, tree args,
tsubst_flags_t complain,
local_specializations = saved_local_specializations;
}
+ /* If the dependent pack arguments were such that we end up with only a
+ single pack expansion again, there's no need to keep it in a TREE_VEC. */
+ if (len == 1 && TREE_CODE (result) == TREE_VEC
+ && PACK_EXPANSION_P (TREE_VEC_ELT (result, 0)))
+ return TREE_VEC_ELT (result, 0);
+
return result;
}
diff --git a/gcc/testsuite/g++.dg/concepts/memfun2.C
b/gcc/testsuite/g++.dg/concepts/memfun2.C
new file mode 100644
index 0000000..c186a18
--- /dev/null
+++ b/gcc/testsuite/g++.dg/concepts/memfun2.C
@@ -0,0 +1,21 @@
+// PR c++/72415
+// { dg-options "-std=c++1z -fconcepts" }
+
+template<int... Indices>
+struct indices {};
+
+template<typename Dummy>
+struct foo_type {
+ template<int... Indices>
+ static void impl(indices<Indices...>)
+ requires (... && (Indices, true));
+
+ static auto caller()
+ { return impl(indices<0, 1, 2> {}); }
+};
+
+int main()
+{
+ // internal compiler error: in satisfy_predicate_constraint, at
cp/constraint.cc:2013
+ foo_type<void>::caller();
+}
