This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH, ARM] Fix Dwarf register numbering for VFPv3/Neon registers
- From: Richard Earnshaw <rearnsha at arm dot com>
- To: Julian Brown <julian at codesourcery dot com>
- Cc: gcc-patches at gcc dot gnu dot org, paul at codesourcery dot com
- Date: Mon, 23 Nov 2009 10:11:23 +0000
- Subject: Re: [PATCH, ARM] Fix Dwarf register numbering for VFPv3/Neon registers
- References: <20091029163511.4a37c00b@rex.config> <1256897566.396.8.camel@e200601-lin.cambridge.arm.com> <20091120193516.206a4826@rex.config>
On Fri, 2009-11-20 at 19:35 +0000, Julian Brown wrote:
> On Fri, 30 Oct 2009 10:12:46 +0000
> Richard Earnshaw <rearnsha@arm.com> wrote:
>
> > On Thu, 2009-10-29 at 16:35 +0000, Julian Brown wrote:
> > > + The recommended encodings for S0-S31 is a DW_OP_bit_piece of
> > > the
> > > + corresponding D register. However gdb6.6 does not support
> > > this, so
> > > + we use the legacy encodings. We also use these encodings for
> > > D0-D15
> > > + for compatibility with older debuggers. */
> >
> > This comment implies that the problem was probably fixed in gdb-6.7
> > which is now over two years (and three releases ago). Is the comment
> > wrong, or do we now support this. Given the status of the VFP support
> > in GDB, I would really expect that someone wanting to debug VFP would
> > need a more recent version of the debugger than 6.6 anyway, so I
> > wouldn't see that as a reason for keeping the old numbering.
>
> As far as I can tell, current GDB doesn't have any support for
> DW_op_bit_piece either (so writing a specific version number in that
> comment was perhaps a bit optimistic). Using the deprecated numberings
> should work for now, but runs the risk of this issue never being fixed
> properly. Fixing GDB and GCC to use the proper encoding would mean that
> folks couldn't debug GCC-compiled VFPv3 code until they upgraded their
> GDB to one which contained the hypothetical DW_op_bit_piece support
> too.
>
> So, what's the best way forward?
>
> Julian
Remove the version information and change the comment to an XXX Fixme
type.
R.