On Wed, Aug 28, 2013 at 03:30:46PM +0200, Marek Polacek wrote:
> >From a quick look, it doesn't seem like we do. (I was searching for
> something about pointer_map in ggc* and gen* files.)
If we can't make it GC aware easily, I'm afraid we need to revert that change to
pointer_map. Now, the question is if the if_marked stuff can be easily done
for the previous implementation with C++ish hash table, or if we should just
use something similar to what tree.c uses for value_expr_for_decl and
similar (of course, with minor adjustments, because we want to map from
TYPE_UID rather than DECL_UID of the key to tree). Or with pointer_map we'd
need to push both the keys (types) and what they map to into (decls) into
some GC vector in addition to the pointer_map.
Jakub
