This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch: darwin specific
- From: Matt Austern <austern at apple dot com>
- To: Dale Johannesen <dalej at apple dot com>
- Cc: Zack Weinberg <zack at codesourcery dot com>, gcc-patches at gcc dot gnu dot org
- Date: Mon, 25 Nov 2002 14:56:45 -0800
- Subject: Re: Patch: darwin specific
On Monday, November 25, 2002, at 02:47 PM, Dale Johannesen wrote:
The basic problem is that the DECL nodes are not passed down far
enough.
The hooks available (ASM_xxx mostly) take strings or, in really
high-level
cases, RTL. As a result, new names like !_t__foo are coined to hold
information that really belongs in the DECL node (for example, the t
indicates a function declared, but not yet defined, in this file; if a
definition is seen a new name is constructed). In darwin's case there
are frequently multiple such names derived from more or less the same
source construct, and it's necessary to get from one to the others.
(Formerly this was met by the linked list nodes, which had two
children,
different forms of the same name. Now the list nodes are no longer
allocated, a win.) This interface (which I personally strongly
dislike)
predates anybody now working on it; I seem to recall it being
recommended
in the docs somewhere, but can't find it now.
Would it be useful to make changes where we pass the DECL
nodes down farther? Mark Mitchell and I have had some
preliminary discussions about that, and it would be useful for
some of the things that I'm doing.
That would be a long-term change, of course. My feeling is
that we should put your change in for now, and explore more
complete solutions later.
--Matt