Re: SH follow up, part 2 (was Re: [RFA] sh-tdep.c: Follow up patch to implement two different ABIs)

[Corinna Vinschen <vinschen@redhat.com>]

On Thu, Sep 25, 2003 at 05:54:43PM -0400, Elena Zannoni wrote:
> Corinna Vinschen writes:
>  > Currently there's no way to differ between these two ABIs automatically
>  > though that might change at some later point.  The user has to change
>  > the ABI on the command line by hand.  The compiler option was -mhitachi,
>  > gcc is just changing to -mrenesas to reflect the company name change.
> 
> Hmm but then the user can set the calling convention to the wrong one.
> Also the finish command, I am afraid, will be confused. There should
> be something detecting the calling convention from the executable. Is
> the info recorded in the file anywhere? I.e. would readelf/objdump
> show it?

No.  I started a discussion about this on gdb four weeks ago but at this
point there is no chance to find this out.  The automatism requires some
work in gcc and gdb which is not done yet and which is not target
dependent.  It's another step.  This patch only contains the first step
necessary to allow to debug two ABIs for sh.  It requires user
intervention.  If the changes for automating the ABI stuff are done, the
next target dependent step is to add the "auto" flag and to make it the
default.

> These descriptions should be comments in the file, the more verbiage
> the better.

The comments are in the code.

> I don't have anything in particular against these patches, but I'd like
> to have a better idea about how to solve this calling convention issue
> before the changes go in.

http://sources.redhat.com/ml/gdb/2003-08/msg00252.html

Corinna

-- 
Corinna Vinschen
Cygwin Developer
Red Hat, Inc.