This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch: darwin specific
Zack Weinberg wrote:
This interferes with my plans to make IDENTIFIERs not be trees. Can
you find an alternate implementation please?
I too think it would be great to make IDENTIFIERs not be trees.
But I wonder how difficult that would be. For example, TREE_TYPE
could be changed to point to a non-tree struct identifier - except
that sometimes it's a TYPE_DECL. And I suspect there are other
problems. That is why I've toyed with the dea of a hybrid:
A tree_base base class that contains the tree_code and the flags,
but not the TREE_TYPE or TREE_CHAIN fields. Then tree_common
would extend tree_base, but tree_identifier would extend tree_base
directly.
But if you have more radical plans, I'd love to hear about them.
Along similar lines, one might define a simple_decl type that
contains only a subset of the tree_decl fields, and could be used
for CONST_DECL, VAR_DECL, and PARAM_DECL, saving considerable space.
--
--Per Bothner
per@bothner.com http://www.bothner.com/per/