Re: [rfc] [03/18] Cell multi-arch: Save/restore current_gdbarch

["Ulrich Weigand" <uweigand at de dot ibm dot com>]

Hello Suzuki-san,

> I wonder if it is the right thread to reply, but I would like to
> mention about current_gdbarch, so I'll put my reply here.  

Thanks for looking at those patches!
 
> extract/store_*signed_integer expect that current_gdbarch points to
> the same one as what of the value to be extracted or stored.  But it
> seems that your patch set doesn't assure of them, for example, in
> regcache_cooked_read_*signed and regcache_cooked_write_*signed.  
> 
> It would not be a problem while Cell/B.E. is the only target which has
> multi-architecture for GDB, since the endian of both PPU(PPC64) and
> SPU are big.  Also current_gdbarch should be removed in the near
> future, however, we should take care of them now IMHO.  

That's right, I had ignored the gdbarch endian routines for now since
-as you say- this doesn't really matter on the Cell/B.E.

I fully agree that this needs to be fixed for a complete solution.
I'll look into whether we can fix this now, even while we still
have current_gdbarch ...

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com