This is the mail archive of the
mailing list for the GCC project.
Re: Move DECL_RESULT into function_decl
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Jason Merrill <jason at redhat dot com>
- Cc: Jan Hubicka <hubicka at ucw dot cz>, gcc-patches at gcc dot gnu dot org, rguenther at suse dot de
- Date: Wed, 16 Jul 2014 13:01:20 +0200
- Subject: Re: Move DECL_RESULT into function_decl
- Authentication-results: sourceware.org; auth=none
- References: <20140715141942 dot GD22085 at kam dot mff dot cuni dot cz> <53C5466E dot 20508 at redhat dot com> <20140715152922 dot GD11265 at atrey dot karlin dot mff dot cuni dot cz> <53C5741F dot 3020503 at redhat dot com>
> On 07/15/2014 11:29 AM, Jan Hubicka wrote:
> >Yep, you are right. I was looking into this, too. I think dump_ada_template
> >should be moved into C++ FE somehow because it deals with C++ trees.
> >What would be most appropriate way to do so?
> I suppose we could pass a pointer to it into dump_ada_specs like we
> do for cpp_check.
That sounds good. I will prepare patch and then prepare updated version of the RESULT_DECL
patch - I reused DECL_ATTIBUTES of INSTANCE_METHOD and CLASS_METHOD. It seems to mostly work
except for -Wuninitialized where we access these by accident. It is not hard to avoid this,
but I think it may be better to turn those into custom decl structures - I plan to get rid of
the ATTIBUTES pointer incrementally too by on-side hash that allows faster attribute
lookups. I also plan to move DECL_SIZE/DECL_SIZE_UNIT into separate datastructure as discussed
with Richard on IRC yesterday.
Basically we can have size descriptors containing size, size unit and mode that are off-tree
with pointers from both types and decls. This way we will avoid duplicating this info all
I sadly do not know how to turn Obj-C tree nodes into custom layout because
there are no prior cases in Obj-C FE and there is obj-c++/c/objc/c++ code
sharing that makes this a bit twisty. Any help would be welcome.