> Daniel Jacobowitz wrote:
>> I think this is the key bit - my_value_equal called value_fetch_lazy,
>> the new code calls gdb_value_fetch_lazy.
>> 
>> Vlad, I noticed that the old code used coerce_array and the new one
>> doesn't.  Is that a problem?
> 
> This should not be a problem on -var-update path, because we never try
> to compare values of array types, since for them type_changeable returns
> false.
> 
> However, it looks to be a problem on -var-assign path. Given:
> 
>    int b[] = {1,2,3};
>    int *a  = b;
> 
> if we create varobj for 'a' and assign it the value of 'b', we should not
> mark this variable as changed.
> 
> I'll double-check this (and other coercions that coerce_array silently
> does) and add new test cases as needed.

This version of patch fixes the problem and adds new tests.

- Volodya

        Fetch values from memory in a single place,
        and only fetch the values that are really needed.
        * varobj.c (struct varobj): Clarify comment.
        (my_value_equal): Remove.
        (install_new_value): New function.
        (type_of_child): Remove.
        (varobj_create): Use install_new_value.
        (varobj_set_value): Use value_contents_equal, not
        my_value_equal.
        (varobj_update): Use install_new_value.
        (create_child): Likewise. Inline type_of_child here.
        (value_of_child): Don't fetch the value.
        (c_value_of_root): Likewise.
        (c_value_of_variable): Likewise.
        (type_changeable): Improve comments.

        gdb/testsuite/
        * gdb.mi/mi-var-cmd.exp: Check -var-update after
        assignement of arrays and function pointers.
        * gdb.mi/var-cmd.c: Add declaration necessary for above
        tests.