On 12/29/2017 12:06 PM, David Malcolm wrote:
One issue I ran into was that fold_for_warn doesn't eliminate
location wrappers when processing_template_decl, leading to
failures of the template-based cases in
g++.dg/warn/Wmemset-transposed-args-1.C.
This is due to the early bailout when processing_template_decl
within cp_fold:
2078 if (processing_template_decl
2079 || (EXPR_P (x) && (!TREE_TYPE (x) || TREE_TYPE (x) ==
error_mark_node)))
2080 return x;
which dates back to the merger of the C++ delayed folding branch.
I've fixed that in this version of the patch by removing that
"processing_template_decl ||" condition from that cp_fold early
bailout.
Hmm, that makes me nervous. We might want to fold in templates when
called from fold_for_warn, but not for other occurrences. But I see
that we check processing_template_decl in cp_fully_fold and in the call
to cp_fold_function, so I guess this is fine.
+ case VIEW_CONVERT_EXPR:
+ case NON_LVALUE_EXPR:
case CAST_EXPR:
case REINTERPRET_CAST_EXPR:
case CONST_CAST_EXPR:
@@ -14937,6 +14940,15 @@ tsubst_copy (tree t, tree args, tsubst_flags_t
complain, tree in_decl)
case CONVERT_EXPR:
case NOP_EXPR:
{
+ if (location_wrapper_p (t))
+ {
+ /* Handle location wrappers by substituting the wrapped node
+ first, *then* reusing the resulting type. Doing the type
+ first ensures that we handle template parameters and
+ parameter pack expansions. */
+ tree op0 = tsubst_copy (TREE_OPERAND (t, 0), args, complain,
in_decl);
+ return build1 (code, TREE_TYPE (op0), op0);
+ }
I'd rather handle location wrappers separately, and abort if
VIEW_CONVERT_EXPR or NON_LVALUE_EXPR appear other than as wrappers.
@@ -24998,6 +25018,8 @@ build_non_dependent_expr (tree expr)
&& !expanding_concept ())
fold_non_dependent_expr (expr);
+ STRIP_ANY_LOCATION_WRAPPER (expr);
Why is this needed?
Jason
