I'm pretty sure that the find hotkey will also not follo= w any connections where rat lines are not shown.

I belive this was at least partly intentional (although the = reasoning is somewhat unclear to me), but is also fairly linked to our impl= ementation of the connectivity search.

Always up to date rat lines could solve the problem, but I b= elive some utilise the feature where you can erase all rats, and re add spe= cific ones from the netlist dialog box.

The idea to change search start point based on node clicked = sounds good to me. If clicking on the net in general (is this possible?), p= erhaps invoke only the purple highlighting?

Shorted connections are the aspect which needs careful consi= deration whatever we do... The way things behave in that case really needs = to be clear and predictable.

Igor had a nice min graph cutting based approach for identif= ying the probable location of a short in PCB-rnd iirc. That kind of thing m= ight be a nice pre ratline adding step, that would avoid pcb offering ratli= nes which further entangle and short out two (or more?) nets.

Our existing logic for highlighting the bad connection doesn= 't work well, and this was the reason Igor started working on the min c= ut solution.

Anyone who's ever needed to track down a gnd to power sh= ort on a complex board will attest to this being tricky if you don't sp= ot the problem right away!

On 15 Feb 2016 20:22, "Britton Kerin (britton DOT kerin AT gmail DOT com) [via geda-user AT delorie DOT com]" <= geda-user AT delorie DOT com> wrot= e:

On Sat, Feb 13, 2= 016 at 5:15 PM, DJ Delorie <dj AT delorie= .com> wrote:
>
>> That does sound useful.=C2=A0 I guess you do it with first Find bu= tton the
>> Ctrl-F?=C2=A0 The behavior I get is:
>
> I "select" one net, then "find" the other net.=C2= =A0 The select code uses
> the find flag internally so order matters.
>
>> Am I correct in thinking that things only appear under Key Binding= s if
>> they don't have a binding described elsewhere in the menu hier= archy?
>
> All key bindings are caused by menu entries, so all key bindings have<= br> > to be in there somewhere.=C2=A0 Any key binding that doesn't have = a "home"
> elsewhere in the menu gets put in the key bindings list.
>
>> I'd like the more obscure elements in the menus/hotkeys to be = less
>> forbidding to learn hence tooltip effort.=C2=A0 What do you think = of
>> putting everything in the hotkeys menu, regardless of whether it >> appears somewhere else as well?=C2=A0 Any other suggestion on how = best to
>> proceed?
>
> The menu system isn't designed for conflicting hotkeys.=C2=A0 I do= n't know
> what would happen if you did that.

If I cut a net in half by taking out a wire, then use Find in the
Netlist dialog, it doesn't show all the pins/pads connected to that
net.=C2=A0 This seems like a bug.=C2=A0 If rat lines are included Find foll= ows
them and gets it right, but shouldn't it get them all anyway, rather than picking an arbitrary start point?=C2=A0 Perhaps more usefully, if a particular node of a net is selected in the nodes pane of the netlist
dialog, then the search could start from there.=C2=A0 This last shouldn'= ;t
be too hard to implement and I'll do it if there aren't objections.=

Otherwise at the implementation level it's mostly good as it is.=C2=A0 = In
theory it might be nice to use private flags so for example Find (the
button) didn't nuke the "connected" set (Ctrl-F) and similarl= y with
selected and find, but in practice allowing that would open a can of
worms since there's no visual way to represent things that are in both<= br> sets.=C2=A0 With the order sensitivity at least it's obvious that they<= br> interact.

Its the interface that's problematic.=C2=A0 F stands for find, but actu= ally
means something different.=C2=A0 The Find button could get a name change but then it would be inconsistent with the underlying flag name
(exposed via Ctrl-R and the file format).=C2=A0 Changing the F binding is obviously a no-no.
F also handles rat lines by marking them and everything on the other
side of them as found (but not connected).

I'm just going to document all this, and maybe change the menu entry names for F and/or Ctrl-F somehow to make it clear that one clears
found and connected flags and the other doesn't.=C2=A0 To allow for
cross-refs that don't fall out of sync, hotkeys are going to get their<= br> own header and the rules that do e.g. gpcb-menu.res.in ->
gpcb-menu.res change slightly.

Britton