Re: MI: fix base members in references

[Jim Blandy <jimb at codesourcery dot com>]

Daniel Jacobowitz <drow@false.org> writes:
> On Wed, Dec 06, 2006 at 10:21:06AM +1300, Nick Roberts wrote:
>> The comment in varobj.c refers to "Baseclass" and I presume the original author
>> included the case
>> 
>>   TYPE_CODE (value_type (parent->value)) == TYPE_CODE_REF
>> 
>> for a specific reason.
>
> Probably, but as Vlad's explanation is correct, either the original
> author was wrong or the behavior of value_ind has changed.  Something
> in this code should change, but see below.
>
> On Tue, Dec 05, 2006 at 01:33:12PM -0800, Jim Blandy wrote:
>> (By the way --- it's handy to include a link to the post with the
>> patch, or ideally the patch itself.)
>
> Is a threading mailer really so much to ask? :-)

*blush*  (ding) Gnus is amazingly slow at finding parent messages.

>> How should this behave if parent->value is a reference to a pointer?
>> Shouldn't it follow the ref, and then behave the same as when it's a
>> pointer?  If so, then the fix would be something like this instead
>> (not that I understand this code):
>
> If this is always the same values affected by Vlad's other patch which
> calls coerce_ref when setting the value, then maybe we should just be
> asserting there is no reference here after that patch.
>
>> --- 2426,2439 ----
>>   	  /* Baseclass */
>>   	  if (parent->value != NULL)
>>   	    {
>> ! 	      struct value *temp = coerce_ref (parent->value);
>>   
>> ! 	      if (TYPE_CODE (value_type (temp)) == TYPE_CODE_PTR
>> ! 		  || TYPE_CODE (value_type (temp)) == TYPE_CODE_REF)
>>   		{
>> ! 		  if (!gdb_value_ind (temp, &temp))
>>   		    return NULL;
>
> In any case, you can drop the TYPE_CODE_REF check if you do it this
> way.

Oh, yes --- I meant to drop the TYPE_CODE_REF case.