Re: [patch/rfc/rfa:sh] gdbarch_register_read() -> gdbarch_cooked_register_read()

[Andrew Cagney <ac131313@ges.redhat.com>]

>  >  > One NB.  The parameter set is technically redundant.  The assertion:
>  >  > 
>  >  > 	gdbarch == regcache_gdbarch (regcache)
>  >  > 
>  >  > always holds so the gdbarch parameter isn't strictly needed.  I figured 
>  >  > that it is more convenient to include the gdbarch as a parameter and 
>  >  > save everyone the hassle of including the above line in their *-tdep code.
>  >  > 
>  >  > The SH stuff should be reviewed.  I think its on the right track since 
>  >  > sh-tdep.c no longer refers to current_regcache!!!
>  >  > 
> 
> 
> So, wait a minute, is the sh the only port that overwrites the generic
> register reads and writes to handle pseudo registers?

Gnew speak: s/pseudo/cooked/.  Pseudo registers are implemented 
differently :-)

In GDB (i.e. the official GDB sources)? Yes.

The only architecture using the gdbarch register read/write (i.e. the 
new cooked interface) is the SH.

The only platform using the old fetch/store pseudo register interface is 
the m68hc11.  The conversion looks easy.

The MIPS is the platform that desperatly needs to start using cooked 
registers.  The Arm is waiting, ready to pounce, when all the changes 
are in.

enjoy,
Andrew