Re: [PATCH v2] Fix inconsistent breakpoint kinds between breakpoints and tracepoints in GDBServer.

[Antoine Tremblay <antoine dot tremblay at ericsson dot com>]

On 10/22/2015 12:16 PM, Antoine Tremblay wrote:
> On 10/22/2015 12:11 PM, Pedro Alves wrote:
> > On 10/22/2015 05:06 PM, Antoine Tremblay wrote:
> > > On 10/22/2015 12:04 PM, Pedro Alves wrote:
> >
> > > > (I think you'll need to move the function to target.c to
> > > > fix !Linux ports, but I'm OK with doing that as a separate step.)
> > >
> > > I don't think so since I fix !linux ports like so :
> > >
> > > /* Implementation of the target_ops method
> > > "breakpoint_kind_from_pc".  */
> > >
> > > static int
> > > win32_breakpoint_kind_from_pc (CORE_ADDR *pcptr)
> > > {
> > >     return the_low_target.breakpoint_len;
> > > }
> >
> > You don't need this one nor the equivalent in other ports if you
> > add this to target.h:
> >
> > #define target_breakpoint_kind_from_pc(PCPTR) \
> >    (the_target->breakpoint_kind_from_pc \
> >      ? (*the_target->breakpoint_kind_from_pc) (PCPTR) \
> >      : default_breakpoint_kind_from_pc ())
> >
> > (see the other similar macros there)
> >
> > You'll need to adjust callers to call target_breakpoint_kind_from_pc
> > instead, of course.
>
> Yes but then I would need a os_arch_sw_breakpoint_from_kind operation in
> all the arch that this os supports (and that I can't test), you think
> it's still better that way ?

Unless I have another macro that checks for the sw_breakpoint_from_kind 
and if abscent returns breakpoint_len... I could do that I guess..