Re: [RFA/RFC] mips tracepoint: fix Bug 12013

[Hui Zhu <teawater@gmail.com>]

On Thu, Dec 23, 2010 at 10:57, Hui Zhu <teawater@gmail.com> wrote:
> On Wed, Dec 22, 2010 at 23:36, Kevin Buettner <kevinb@redhat.com> wrote:
>> On Wed, 22 Dec 2010 15:12:22 +0800
>> Hui Zhu <teawater@gmail.com> wrote:
>>
>>> > You might consider implementing a new gdbarch method which provides a
>>> > mapping from pseudo register numbers to raw register numbers. __The
>>> > trace machinery could use such a mapping to find the corresponding raw
>>> > register(s) when presented with a pseudo register. __I can think of
>>> > several potential pitfalls with this approach, but I think the idea is
>>> > worth exploring.
>>>
>>> Thanks Kevin.  I will do it.
>>
>> Please look at Pedro's reply.  He has outlined a better approach.
>>
>>> And I make a patch to add some comments from your mail to mips_register_name.
>>> Wish it can help other people.
>>>
>>> Please help me review it.
>>
>> Okay, see below...
>>
>>> 2010-12-22  Hui Zhu  <teawater@gmail.com>
>>>
>>>       * mips-tedp.c (mips_register_name): Add comments.
>>>
>>> ---
>>>  mips-tdep.c |    7 ++++++-
>>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> --- a/mips-tdep.c
>>> +++ b/mips-tdep.c
>>> @@ -454,7 +454,12 @@ mips_register_name (struct gdbarch *gdba
>>>    enum mips_abi abi = mips_abi (gdbarch);
>>>
>>>    /* Map [gdbarch_num_regs .. 2*gdbarch_num_regs) onto the raw registers,
>>> -     but then don't make the raw register names visible.  */
>>> +     but then don't make the raw register names visible.
>>> +     Because It is possible to debug a 64-bit device using a 32-bit programming
>>> +     model.  In such instances, the raw registers are configured to be
>>> +     64-bits wide, while the pseudo registers are configured to be 32-bits
>>> +     wide.  The registers that the user sees - the pseudo registers - match
>>> +     the user's expectations given the programming model being used.  */
>>
>> Could you revise the comment to read as follows?
>>
>>    /* Map [gdbarch_num_regs .. 2*gdbarch_num_regs) onto the raw registers,
>>       but do not make the raw register names visible.  This (upper)
>>       range of user visible register numbers are the
>>       pseudo-registers.
>>
>>       This approach was adopted accomodate the following scenario:
>>       It is possible to debug a 64-bit device using a 32-bit
>>       programming model.  In such instances, the raw registers are
>>       configured to be 64-bits wide, while the pseudo registers are
>>       configured to be 32-bits wide.  The reigsters that the user
>>       sees - the pseudo registers - match the users expectations
>>       given the programming model being used.  */
>>
>> Please allow several days for others to tweak my suggested wording.  If
>> there are no further comments on the above wording, feel free to commit
>> it.

Checked in.

Joel,  do you think this patch can check in to 7.2.1.

Thanks,
Hui

>>
>> Thanks,
>>
>> Kevin
>>
>
> OK.   Thanks Kevin.
>
> Best,
> Hui
>
>
> 2010-12-23  Kevin Buettner  <kevinb@redhat.com>
>            Hui Zhu  <teawater@gmail.com>
>
>        * mips-tedp.c (mips_register_name): Add comments.
>
> ---
>  mips-tdep.c |   11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
>
> --- a/mips-tdep.c
> +++ b/mips-tdep.c
> @@ -454,7 +454,16 @@ mips_register_name (struct gdbarch *gdba
>   enum mips_abi abi = mips_abi (gdbarch);
>
>   /* Map [gdbarch_num_regs .. 2*gdbarch_num_regs) onto the raw registers,
> -     but then don't make the raw register names visible.  */
> +     but then don't make the raw register names visible.  This (upper)
> +     range of user visible register numbers are the pseudo-registers.
> +
> +     This approach was adopted accommodate the following scenario:
> +     It is possible to debug a 64-bit device using a 32-bit
> +     programming model.  In such instances, the raw registers are
> +     configured to be 64-bits wide, while the pseudo registers are
> +     configured to be 32-bits wide.  The registers that the user
> +     sees - the pseudo registers - match the users expectations
> +     given the programming model being used.  */
>   int rawnum = regno % gdbarch_num_regs (gdbarch);
>   if (regno < gdbarch_num_regs (gdbarch))
>     return "";
>