This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [PATCH]: memory region attributes


"J.T. Conklin" wrote:
> 
> >>>>> "Fernando" == Fernando Nasser <fnasser@redhat.com> writes:
> Fernando> Can we just pass zero as the value for the new argument?
> 
> It's hard to say no, seeing how I did exactly that to patch up
> sol-thread.c.  But I don't think it's the best thing.
> 
> Ideally the vector functions are always being called through
> target_xfer_memory() (or the dcache functions, which also look up
> attributes), so in every case an attrib argument will be available.
> It turns out there are some cases where this is not true, but IMHO
> this indicates a problem within GDB's internals.

I don't remember that being a requirement.  Was that
documented anywhere?


> In the meantime, we have to handle the case where an attrib argument
> is not available.  I'd like the interface to be defined so the target
> vector is always able to dereference the attrib.  Otherwise each
> vector function has to "invent" semantics for the case where a null
> attrib pointer is passed.  For those target vectors that don't care
> what the attributes are (all of them as of now), it doesn't really
> matter.  But in time, this will change.
> 
> One possibility is to expose the default attribute.  At the present,
> it is wholly contained within memattr.c.  But we could expose it for
> now so that it could be passed when no other attribute can be found.
> 
>         --jtc
> 
> --
> J.T. Conklin
> RedBack Networks

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]