Re: MI: fix base members in references

[Vladimir Prus <ghost at cs dot msu dot su>]

On Sunday 10 December 2006 00:45, you wrote:
>  > > > I would have preferred if instead of adding if, the code was modified to
>  > > > look at
>  > > 
>  > > >         value_type (var->value)
>  > > 
>  > > > as opposed to
>  > > 
>  > > >         var->type
>  > > 
>  > > I'm not sure that I follow your point.  The patch just gets the target
>  > > type, after dereferencing, in the case of a pointer.
>  > > 
>  > > > The latter is the type of the varobj expression as it is in source
>  > > > program.  The former is the value we're actually showing. It makes sense
>  > > > to use value_type (var->value) for all presentation purposes.
>  > > 
>  > > The former appears to be a type also (not a value).
>  > 
>  > Slight typo: the former is the type of the value we're actually showing. So,
>  > you don't need to take original type and try to arrived to the type that
>  > should be shown to the user, you just use value_type (var->value), and don't
>  > need any further processing. One less thing that can be broken in future.
> 
> I'm still not sure that I follow.  My patch doesn't look at var->type directly.

Well, get_type does.

> Are you saying get_type should use value_type (var->value) instead of
> var->type?  

No. I'm saying that instead of using get_type and then additionally doing something
with the type, we should directly use value_type (var->value). If we want to know how
many children a value has, and you know the type of the value, it does not make
sense to take some other type, do some transformation on it, and the hope you'll
get the same type. You can just use the type that's right.

- Volodya