On 9/30/20 3:57 AM, Jakub Jelinek wrote:
On Tue, Sep 29, 2020 at 03:40:40PM -0600, Martin Sebor via Gcc-patches wrote:
I will commit this patch later this week unless I hear concerns
or suggestions for changes.
That is not how the patch review process works.
The review process hasn't been working well for me, but thankfully,
the commit policy lets me make these types of "obvious" fixes on
my own, without waiting for approval. But if I could get simple
changes reviewed in a few days instead of having to ping them for
weeks there would be no reason for me to take advantage of this
latitude (and for us to rehash this topic yet again).
+ arat = tree_cons (get_identifier ("array"), flag, NULL_TREE);
Better
arat = build_tree_list (get_identifier ("array"), flag);
then, tree_cons is when you have a meaningful TREE_CHAIN you want to supply
too.
Okay. I checked to make sure they both do the same thing and
create a tree with the size and committed the updated patch in
r11-3570.
Martin
}
- TYPE_ATOMIC (artype) = TYPE_ATOMIC (type);
- TYPE_READONLY (artype) = TYPE_READONLY (type);
- TYPE_RESTRICT (artype) = TYPE_RESTRICT (type);
- TYPE_VOLATILE (artype) = TYPE_VOLATILE (type);
- type = artype;
+ const int quals = TYPE_QUALS (type);
+ type = build_array_type (eltype, index_type);
+ type = build_type_attribute_qual_variant (type, arat, quals);
}
/* Format the type using the current pretty printer. The generic tree
@@ -2309,10 +2304,6 @@ attr_access::array_as_string (tree type) const
typstr = pp_formatted_text (pp);
delete pp;
- if (this->str)
- /* Remove the attribute that wasn't installed by decl_attributes. */
- TYPE_ATTRIBUTES (type) = NULL_TREE;
-
return typstr;
}
Otherwise LGTM.
Jakub
