This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: PATCH (take 2) - QNX bfd cleanup
- From: "Graeme Peterson" <gp at qnx dot com>
- To: amodra at bigpond dot net dot au (Alan Modra)
- Cc: binutils at sources dot redhat dot com
- Date: Fri, 1 Nov 2002 11:10:07 -0500 (EST)
- Subject: Re: PATCH (take 2) - QNX bfd cleanup
As Kris mentioned in his posting, we are not removing QNX support.
I don't think I can, as the code now belongs to the FSF, not me or
QNX (that copyright assignment thing ;-) ).
I had already addressed this in a response to another e-mail, which
you may have missed:
http://sources.redhat.com/ml/binutils/2002-10/msg00674.html
We are cleaning it up by removing the bfd_*qnx_vec entries, and
instead using the base vectors that we had originally extended.
There are a variety of reasons, including the ambiguous name issue,
but not limited to that. The code that required a set of qnx bfd's
(which then also may have required an ELFOSABI_QNX), has other
problems, and after researching things, we are not convinced that
it is the right way to solve the problem we want solved.
>From the diff, bfd/config.bfd:
*** 371,377 ****
targ_selvecs=i386coff_vec
;;
i[3456]86-*-nto-qnx*)
! targ_defvec=bfd_elf32_i386qnx_vec
targ_selvecs=i386coff_vec
;;
i[3456]86-*-chorus*)
--- 371,377 ----
targ_selvecs=i386coff_vec
;;
i[3456]86-*-nto-qnx*)
! targ_defvec=bfd_elf32_i386_vec
targ_selvecs=i386coff_vec
;;
i[3456]86-*-chorus*)
And similarly for arm, sh and ppc. Every binutils that QNX has ever
released used the base vectors. I want the GNU source to go back to
that so we can re-investigate things and re-implement a better solution.
I am fairly sure I have achieved this in the 2nd patch. Can you review
it and let me know? If I have in fact removed any QNX support, this is
an error, although I am pretty sure I have not.
Thanks and let me know.
GP
>
> Hi Graeme,
> Sorry for the delay in replying. A close lightning strike here
> took out my cable modem and ethernet. :-(
>
> On Thu, Oct 31, 2002 at 05:05:17PM -0500, Graeme Peterson wrote:
> [about removing QNX support from binutils]
>
> Can you tell me why your management people decided that QNX support
> ought to be removed from binutils? Is it just simply that you
> don't have time for support? It seems like some people outside
> your organization would like to keep QNX support in binutils, so
> perhaps there's reason to leave it in. We can solve the ambiguous
> bfd vector easily by removing the QNX vectors from the target list
> or use something like Kaz's patch.
>
> --
> Alan Modra
> IBM OzLabs - Linux Technology Centre
>