On Nov 10, 2010, at 2:41 PM, Tim Edwards wrote:

>> objinstbbox() is called from selection.c:genselectelement() with a float 
>> range.
>> It expects an int range. Do we lose data if extend parameter is int? Or 
>> should
>> range be an int as well?
> 
> You'll note that there is a RANGE_NARROW and RANGE_WIDE, which I put
> into the selection scheme noting that certain elements should be
> selected based on different criteria depending on whether they were
> being selected by themselves, or selected as part of the process of
> selecting an object instance (this avoids the problem of having
> something like a page border object, and having it get selected
> all the time because it covers the whole page---large areas of
> whitespace inside objects aren't selectable).  I found that the
> wire selection seemed to be about right (to my taste, anyway) at
> a range value of RANGE_NARROW = 11.5.  I didn't like it at 11, and
> I didn't like it at 12.  But that only applies to wire selection.
> Object instance bounding box calculations aren't so sensitive to
> range value changes, so converting to int is a reasonable approach
> (which is the current approach, albeit missing a type conversion
> specifier).  Range could probably be an integer if it were doubled
> before passing to pathselect(), and halved again inside the subroutine,
> but that seems sort of unnecessary unless you have a specific reason
> in mind for doing so.

OK, I just wanted to understand why it was done so. I won't be touching it.
-- Kuba
_______________________________________________
Xcircuit-dev mailing list
[email protected]
http://www.opencircuitdesign.com/mailman/listinfo/xcircuit-dev

Reply via email to