Kai-Martin Knaak wrote: >Rubber band mode is (still) much too sloppy about the decision which >segment endpoints to move. > >Imagine a short segment between longer ones like I constantly get because >of the metric/imperial nuisance. Thes almost never rubber move in a >decent way. Suppose a track looks like this: >_____ >1 2\_____ > 3 >If I drag the top corners at segment 2 there are two sensible rubber >modes: >a) Segment 1 and 2 stretch. Segment 3 stays in place. >b) Segment 1 and 3 stretch. Segment 2 moves a matching amount, no stretch > >However, rubber band mode almost invariably chooses a third mode: >c) All three segments stretch. > >This results in > > /|\ > / | \ > / | \ > / | \ >/1 2| 3\ > >This is rarely a useful transform. > >---<(kaimartin)>--- > >
I agree completely with your suggested behavior and with the statement that what it does now is rarely useful. My approach usually involves deleting the whole mess (all three segments) and starting over. This problem has been there as long as I can remember (several years/versions). It is affected by things like: - the distance (relative to grid spacing) of the two vertices (the ends of segment 2 in this example) - the width of he lines themselves I believe the problem is in the logic or distance calculations which look for the nearest vertex to the selection point. I think you will encounter it most often when the vertex spacing is less than 1/2 line width. If you haven't seen this problem, your board isn't dense enough :). There was a bug report submitted on this issue well over a year ago. There was also a partial fix submitted to the developers' list at that time. As far as I ever saw, neither got any attention. In my opinion, this is the one of the most serious flaws that PCB has since it can cause so much re-work. Joe T _______________________________________________ geda-user mailing list geda-user@moria.seul.org http://www.seul.org/cgi-bin/mailman/listinfo/geda-user