> +/* GTY((user)) hooks for symtab_node_base (and its subclasses). > + We could use virtual functions for this, but given the presence of the > + "type" field and the trivial size of the class hierarchy, switches are > + perhaps simpler and faster. */
Generally I am not really happy about the hand marking - why can't GTY just handle it by itself? Do we have some eisting exmaple of this? GTY was in a way of getting proper class hiearchy for quite a while and this is probably less ugly than C-syntax-classes we have now. But it would be nice to have some longer term plan what to do here. Obviously this is going to make any changes to GGC implementation close to imposible since all the implementation details are exposed now. > +/* Workarounds for deficiencies in gengtype's handling of GTY((user)). > + > + gengtype walks the graph of the *types* in the code, and produces > + functions in gtype-desc.[ch] for all types that are reachable from a > + GTY root. > + > + However, it ignores the contents of GTY((user)) types when walking > + this graph. > + > + Hence if you have a subgraph of types that are only reachable > + via fields in GTY((user)) types, gengtype won't generate helper code > + for those types. > + > + Ideally there would be some way to mark a GTY((user)) type to say > + which types it references, to avoid having to mark these types as > + GTY((user)). > + > + For now, work around this issue by providing explicit roots of the > + missing types, of dummy variables. */ > + > +/* Force gengtype into providing GTY hooks for the type of this field: > + htab_t GTY((param_is (struct cgraph_edge))) call_site_hash; > + within cgraph_node. */ > +static GTY(()) htab_t GTY((param_is (struct cgraph_edge))) > dummy_call_site_hash; I think call site hash can safely go off GGC memory. All statements are linked from CFG. > + > +/* Similarly for this field of symtab_node_base: > + struct ipa_ref_list ref_list; > + */ > +static GTY((deletable)) struct ipa_ref_list *dummy_ipa_ref_list; > + > +/* Similarly for this field of cgraph_node: > + struct cgraph_clone_info clone; > +*/ > +static GTY((deletable)) struct cgraph_clone_info *dummy_cgraph_clone_info; > + > +/* Non-existent pointer, to trick gengtype into (correctly) realizing that > + ipa_replace_map is indeed used in the code, and thus should have > ggc_alloc_* > + routines. */ > +static GTY((deletable)) struct ipa_replace_map > *dummy_ipa_replace_map_pointer; > + > +/* Non-existent pointer, to trick gengtype into (correctly) realizing that > + varpool_node is indeed used in the code, and thus should have ggc_alloc_* > + routines. */ > +static GTY((deletable)) varpool_node *dummy_varpool_node_ptr; > + I would really like some GGC/middle-end or global maintainer to comment on these changes. The rest of changes seems fine to me with the comments from first email ;) Honza
