Re: ping: Re: PATCH : allow to set length of hw watchpoints (e.g. for Valgrind gdbserver)

[Yao Qi <yao@codesourcery.com>]

On 05/28/2011 01:58 AM, Pedro Alves wrote:
> On Friday 27 May 2011 18:53:19, Tom Tromey wrote:
>>>>>>> "Yao" == Yao Qi <yao@codesourcery.com> writes:
>>
>> Yao> Maybe, we need a new target `remote-valgrind' here, and move your
>> Yao> stuff there.
>>
>> We discussed this a bit (last year?), but Pedro was against adding a new
>> target.  I don't recall why; I would like to know though.
> 
> It was in a different context (some target where endianess matters
> depending on whether you're reading code or something else), but the
> reasons are the same.  There's no need for one, and it adds to
> user confusion, and IDE complication.  If the remote target needs
> to behave differently against some remote stub, that calls 
> for the remote end giving gdb enough information for gdb to 
> adjust itself automatically.
> 

Yes, I agree.

> I can't say I understand why was that being proposed in this case?
> What is the patch breaking?
> 

One new command "set remote hardware-watchpoint-length-limit"
 is added in the patch, which is only useful to gdbserver+valgrind.
When gdb is talking with normal gdbserver, it may be wrong to set
hardware-watchpoint-length-limit in gdb side.  Users should be careful
when using this command.

The ideal solution, IMO, is remote side gives GDB the value of
hardware-watchpoint-length-limit, however, I don't know it is easy or
hard to do such thing.

-- 
Yao (齐尧)