kicad-developers team mailing list archive

[Tomasz Wlostowski <tomasz.wlostowski@xxxxxxx>]

On 18/09/18 12:17, Jeff Young wrote:
> When we check for collisions we do it first with the
> worst-case-clearance value (m_maxClearance).  I imagine this is so that
> we can search the RTree uniformly.
> 
Hi Jeff,

Yes, that's ^^^ the reason.

> But don’t we then have to check each collision for actual clearance
> violation (ie: with the actual, not worst-case clearance values)?  I
> can’t seem to find where this happens while routing.  (I’m looking
> at DIFF_PAIR_PLACER::attemptWalk().)

Have a look at PNS::NODE::DEFAULT_OBSTACLE_VISITOR::operator(). It does
exactly what you've mentioned.

Cheers,
Tom