C++ PATCH for c++/69223 (ICE with deduced return type)
Jason Merrill
jason@redhat.com
Fri Jul 22 03:06:00 GMT 2016
Here when we deduced the return type for a lambda to an uninstantiated
class template specialization, we built the new RESULT_DECL, and
called layout_decl, before we called complete_type_or_else, so the
RESULT_DECL wasn't properly laid out.
Tested x86_64-pc-linux-gnu, applying to trunk, 6, 5.
-------------- next part --------------
commit f0113b081c53fa70be614f6d1f63953a3919b85f
Author: Jason Merrill <jason@redhat.com>
Date: Thu Jul 21 18:20:17 2016 -0400
PR c++/69223 - ICE with deduced template return type.
* semantics.c (apply_deduced_return_type): Call
complete_type_or_else before building the new RESULT_DECL.
diff --git a/gcc/cp/semantics.c b/gcc/cp/semantics.c
index 19daeff..63063b8 100644
--- a/gcc/cp/semantics.c
+++ b/gcc/cp/semantics.c
@@ -9279,6 +9279,10 @@ apply_deduced_return_type (tree fco, tree return_type)
if (TREE_TYPE (result) == return_type)
return;
+ if (!processing_template_decl && !VOID_TYPE_P (return_type)
+ && !complete_type_or_else (return_type, NULL_TREE))
+ return;
+
/* We already have a DECL_RESULT from start_preparsed_function.
Now we need to redo the work it and allocate_struct_function
did to reflect the new type. */
@@ -9294,8 +9298,6 @@ apply_deduced_return_type (tree fco, tree return_type)
if (!processing_template_decl)
{
- if (!VOID_TYPE_P (TREE_TYPE (result)))
- complete_type_or_else (TREE_TYPE (result), NULL_TREE);
bool aggr = aggregate_value_p (result, fco);
#ifdef PCC_STATIC_STRUCT_RETURN
cfun->returns_pcc_struct = aggr;
diff --git a/gcc/testsuite/g++.dg/cpp0x/lambda/lambda-deduce3.C b/gcc/testsuite/g++.dg/cpp0x/lambda/lambda-deduce3.C
new file mode 100644
index 0000000..68ac29c
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp0x/lambda/lambda-deduce3.C
@@ -0,0 +1,15 @@
+// PR c++/69223
+// { dg-do compile { target c++11 } }
+
+template <class T> struct A
+{
+ T x[20];
+};
+
+int main()
+{
+ auto l = [](const A<int>& i){ return i; };
+ A<int> a;
+
+ l(a);
+}
More information about the Gcc-patches
mailing list