This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [rfa] Eliminate supply_gregset etc. from -tdep files
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: drow at false dot org (Daniel Jacobowitz)
- Cc: gdb-patches at sourceware dot org
- Date: Sun, 29 Apr 2007 21:52:18 +0200 (CEST)
- Subject: Re: [rfa] Eliminate supply_gregset etc. from -tdep files
Daniel Jacobowitz wrote:
> On Sun, Apr 29, 2007 at 06:18:24PM +0200, Ulrich Weigand wrote:
> > there are two -tdep files (m68k-tdep.c and mips-linux-tdep.c) that contain
> > implementations of the supply_greget family of functions. This is incorrect;
> > those function should only ever be provided by native support files.
>
> "Incorrect" may be a bit strong. core-regset.c needs them to link; at
> one point, that was the right way to implement core file support. I
> don't remember how the history worked out for mips, since it doesn't
> look like the version in the FSF repository ever used core-regset.o,
> but I think that was the original motivation.
Right, but I thought core-regset.o was also supposed to used only as a
native support file, never in an .mt file. Was this ever done in the
past?
Anyway, the reason why I'd like to get of the use in -tdep files is that
it would prevent building GDB with multiple architecture targets built-in
at the same time, if multiple tdep files define different copies of
routines with the same name ...
(B.t.w. it doesn't look too hard to get completely rid of core-regset.o
at this time, but it would need someone to test on i386-gnu, alpha-freebsd,
sparc-solaris ... I can take care of ia64-linux.)
> Anyway, now there are no targets that use core-regset.o in .mt files,
> and no good reason to do so. So now it's incorrect :-) This patch is
> OK.
Thanks, committed.
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
Ulrich.Weigand@de.ibm.com