Daniel Jacobowitz wrote:

>> +     rather value_contents, will take care of this.
>> +     It might throw, but unlike var-update for -var-assign
>> +     there's just one variable we're working it, so we don't
>> +     need to catch the exception here.  */
> 
> Wait, what?  gdb_value_assign will never throw.  value_contents
> might, but gdb_value_assign will catch it.

Yep, confused gdb_value_assign with value_assign.

Updated patch attached. I also attach the delta to the previous patch.

-  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.