Re: [rfc][2/3] gdbserver bi-arch support: core s390x part

[Daniel Jacobowitz <drow at false dot org>]

You're welcome to prod me when I drop out in the middle of a
conversation like this.  Really.  I'm sorry about the delay.

On Wed, Jan 30, 2008 at 12:33:33AM +0100, Ulrich Weigand wrote:
> Yes.  Note that there are two different issues: whether we need
> to clear the high bit depends on the architecture of gdbserver,
> but the size of the register depends on the architecture of
> the inferior.  To simplify the latter issue, maybe it would be
> nice if generic code had a "collect_register_as_addr" helper
> that would check the register's size and convert its contents
> to CORE_ADDR as appropriate?

Then we'd have room to squabble about whether that function sign
extended or not.  Let's hold off on this one.

> 	* configure.srv [s390x-*-linux*]: Set srv_regobj to include both
> 	reg-s390.o and reg-s390x.o.
> 
> 	* linux-low.c (new_inferior): New global variable.
> 	(linux_create_inferior, linux_attach): Set it.
> 	(linux_wait_for_process): Call the_low_target.arch_setup after the
> 	target has stopped for the first time.
> 	(initialize_low): Do not call the_low_target.arch_setup.
> 
> 	* linux-s390-low.c (s390_get_pc): Support bi-arch operation.
> 	(s390_set_pc): Likewise.
> 	(s390_arch_setup): New function.
> 	(the_low_target): Use s390_arch_setup as arch_setup routine.
> 
> 	* regcache.c (realloc_register_cache): New function.
> 	(set_register_cache): Call it for each existing regcache.

This looks great to me.

-- 
Daniel Jacobowitz
CodeSourcery