Re: [PATCH v2 2/3] Adjust stat width calculations to take --graph output into account

>>>> +
>>>> +             /*
>>>> +              * If the remaining unreserved space will not accomodate
>>>> the
>>>> +              * filenames, adjust name_width to use all available
>>>> remaining space.
>>>> +              * Otherwise, assign any extra space to graph_width.
>>>> +              */
>>>> +             if (name_width>    width - reserved_character_count -
>>>> graph_width) {
>>>> +                     name_width = width - reserved_character_count -
>>>> graph_width;
>>>> +             } else {
>>>> +                     graph_width = width - reserved_character_count -
>>>> name_width;
>>>> +             }
>>>> +
>>>> +             /*
>>>> +              * If stat-graph-width was specified, limit graph_width to
>>>> its value.
>>>> +              */
>>>>               if (options->stat_graph_width&&
>>>> -                 graph_width>    options->stat_graph_width)
>>>> +                             graph_width>    options->stat_graph_width)
>>>> {
>>>>                       graph_width = options->stat_graph_width;
>>>> -             if (name_width>    width - number_width - 6 - graph_width)
>>>> -                     name_width = width - number_width - 6 -
>>>> graph_width;
>>>> -             else
>>>> -                     graph_width = width - number_width - 6 -
>>>> name_width;
>>>> +             }
>>>
>>> Here, the order of the two tests
>>> (1) if (options->stat_graph_width&&  graph_width>
>>>  options->stat_graph_width)
>>>
>>> (2) if (name_width>  width - number_width - 6 - graph_width)
>>> is reversed. This is not OK, because this means that
>>> options->stat_graph_width will be used unconditionally, while
>>> before it was subject to limiting by total width.
>>
>>
>> If options->stat_graph_width is specified, it should always limit the
>> value of graph_width, correct? Since (1) is the last test, it can only
>> decrease the value of graph_width, which would already be limited by
>> the total width.
>
> Right, but the way the tests are ordered now, we could end up decreasing
> name_width first (after (2)) and then graph_width (after (1)), actually
> using less than full width.