This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [trunk][patch committed] New default_def_flag in tree_base
- From: "Andrew Pinski" <pinskia at gmail dot com>
- To: "Oleg Ryjkov" <olegr at google dot com>
- Cc: gcc-patches at gcc dot gnu dot org, "Diego Novillo" <dnovillo at google dot com>
- Date: Fri, 11 Apr 2008 20:17:14 -0700
- Subject: Re: [trunk][patch committed] New default_def_flag in tree_base
- References: <14be89990804102229x36cdac2dv15644dd79dca5f89@mail.gmail.com>
On Thu, Apr 10, 2008 at 10:29 PM, Oleg Ryjkov <olegr@google.com> wrote:
> Hello.
>
> This patch introduces a new default_def_flag, to tree_base. This
> is needed because I've run into a flag collision, when volatile_flag
> is used for indicating a default definition. Thus some variables end
> up being marked as volatile, when in fact they should not be.
>
> Tested on i686-linux, no new regressions.
Index: tree.h
===================================================================
--- tree.h (revision 134191)
+++ tree.h (working copy)
@@ -389,6 +389,7 @@ struct tree_base GTY(())
unsigned deprecated_flag : 1;
unsigned invariant_flag : 1;
unsigned saturating_flag : 1;
+ unsigned default_def_flag : 1;
unsigned lang_flag_0 : 1;
unsigned lang_flag_1 : 1;
@@ -399,7 +400,7 @@ struct tree_base GTY(())
unsigned lang_flag_6 : 1;
unsigned visited : 1;
- unsigned spare : 23;
+ unsigned spare : 22;
These bit-fields were grouped by 8 bits each and now they are not, can
you fix that? Note you were not the first one who caused this either,
when saturating_flag was added it, the fields became messed up. I
think the lang_flag_* fields really should start on a 8bit boundary
too.
Thanks,
Andrew Pinski