Hi Daniel,

I've attached a new patch, which includes improvements you suggested.

It has been tested. No regression has been detected.

-- Maxim

Daniel Jacobowitz wrote:
> On Fri, Nov 10, 2006 at 05:21:07PM -0800, Maxim Grigoriev wrote:
>   
>> I understood your concern is that the register number within the next 
>> frame is not set ( *realnump ). On Xtensa, it's going to be unchanged 
>> from the previous frame's value.
>>
>> Question: is replacing of
>>
>> frame_register_unwind (next_frame, regnum,
>> optimizedp, lvalp, addrp, realnump, valuep);
>>
>> with
>>
>> *optimizedp = 0;
>> *lvalp = lval_register;
>> *addrp = 0;
>> *realnump = regnum;
>> if (valuep)
>> frame_unwind_register (next_frame, (*realnump), valuep);
>>
>> the fix you are looking for ? I tested it. It works fine.
>>     
>
> Yes, I think that's right.
>
>