Jason Merrill <ja...@redhat.com> a écrit: > On 12/03/2011 04:25 PM, Dodji Seketeli wrote: >> This might be a theoretical nit, but it looks like if expr is >> error_mark_node, we'd crash as well, because of the line below that comes >> right after the comment above: >> >> gcc_assert (!MAYBE_CLASS_TYPE_P (fromtype) || !MAYBE_CLASS_TYPE_P (totype) >> || !DERIVED_FROM_P (totype, fromtype)); > > That should be OK, since TREE_TYPE (error_mark_node) is > error_mark_node.
Oh, right.
--
Dodji
