> > +/* gdbstub.c */
> > +void ppc_gdb_init(CPUState *cs, PowerPCCPUClass *ppc);
> > +gchar *ppc_gdb_arch_name(CPUState *cs);
>
> These should probably go into internal.h and not cpu.h.
> These do not need to be exposed outside of target/ppc/.

Makes sense, I can do that. Is such a small change worth a v3, or do I fix this as I send the disable-tcg patch series?

> > +#include "exec/helper-proto.h"
>
> Not ideal, but ok temporarily.


yeah, this is only here until we figure out how to deal with the vscr stuff.

> > +        gdb_get_reg32(buf, helper_mfvscr(env));
> ...
> > +        helper_mtvscr(env, ldl_p(mem_buf));
>
> These should be modeled on e.g. store_fpscr, where there's a non-"helper"
> function to be called, which is then called by the "helper" function.
>
> Obviously, splitting that out should be a separate patch.

We already expected to fix this in the disable-tcg patch series, but this reference does help (:
However, checking now, store_fpscr is defined in fpu_helper.c, which is mostly TCG stuff. Any
idea where we could move the store_* functions to have them compile in the !tcg build?

My best guess is to do it in cpu.c, as the other files compiled with the same options are cpu-models.c and
gdbstub.c.

Bruno Piazera Larsen

Instituto de Pesquisas ELDORADO

Departamento Computação Embarcada

Analista de Software Trainee

Aviso Legal - Disclaimer