On 12/13/2016 12:49 PM, Nathan Sidwell wrote:
+/* Template information for an alias template type. */
+#define TYPE_ALIAS_TEMPLATE_INFO(NODE) \
+ (DECL_LANG_SPECIFIC (TYPE_NAME (NODE)) \
+ ? DECL_TEMPLATE_INFO (TYPE_NAME (NODE)) \
+ : NULL_TREE)
+
+/* If NODE is a specialization of an alias template, this accessor
+ returns the template info for the alias template. Otherwise behave
+ as TYPE_TEMPLATE_INFO. */
+#define TYPE_TEMPLATE_INFO_MAYBE_ALIAS(NODE) \
+ (TYPE_ALIAS_P (NODE) && DECL_LANG_SPECIFIC (TYPE_NAME (NODE))
\
+ ? TYPE_ALIAS_TEMPLATE_INFO (NODE) \
+ : TYPE_TEMPLATE_INFO (NODE))
Looks like this is checking DECL_LANG_SPECIFIC twice again; we should be
able to drop the check from TYPE_TEMPLATE_INFO_MAYBE_ALIAS.
/* An alias template name is never deduced. */
if (TYPE_ALIAS_P (arg))
arg = strip_typedefs (arg);
- tree argvec = INNERMOST_TEMPLATE_ARGS (TYPE_TI_ARGS (arg));
+
+ tree tinfo = TYPE_TEMPLATE_INFO_MAYBE_ALIAS (arg);
You don't need _MAYBE_ALIAS here. And without it, we should be able to
drop the strip_typedefs.
OK with those changes.
Jason
