Nets have gone haywaire

Thanks for that info, I might have made a bug report…

Amazing!! @Sparky_Labs found a serious bug, but quite the corner case and everybody scrambled around to see what was going on, 15 Hours ago the bug got fixed, really cool job guy!!! This is why I love KiCAD and the community!

Take that Eagle, Altium, OrCAD ! :stuck_out_tongue:

https://forum.kicad.info/t/kicad-growth-numbers/19138/27

2 Likes

This is why I abandoned circuit studio having paid £1000 for the licence and £100 for a further subscription. I decided that KiCad was about as good and that contrary to CS KiCad IS being developed. CS has been abandoned effectively, but they are still selling licences.

1 Like

I have to remind that we still can’t be sure if the way how JP could reproduce a similar situation is the same way Sparky_Labs did it. But at least one bug was fixed, and if the problem comes again, we can go through the same process :slight_smile:

1 Like

Lets hope v6 will get rid of the “connection” entity in the file format and determine where to draw junction dots based on the connectivity information. This would be the full fix.

That is the most radical move which I have proposed. As far as I can see, the only caveat would be that there might be situations when the user wants to force a junction where one can’t be inferred implicitly, for example with non-horizontal/vertical wires which don’t hit every grid (or even coordinate) point. Even that could be taken care of if junctions would be found by “near enough” principle instead of “same coordinates”. Or a forced junction could be a property of a wire end and use “near enough” principle.

I can’t see any reason to allow free-floating junction dots. But they aren’t a very big problem. The biggest problem is an implicit junction which isn’t marked with a dot. The second one is a dot on top of two wires which actually doesn’t form a junction. Those two situation just MUST not happen IMO. Everthing else is secondary from a user’s point of view. But how that would be achieved is another question. Getting rid of dedicated junction items in implementation level would be the cleanest but also most far-reaching and radical solution, touching everything from UI to file format.

I know. That is why I have written:

But it is not 100% true. Bus stubs are not only graphics - there is some connection logic behind them. If net label is not at the wire end then you have a square at wire end showing it is not connected. When you put bus stub there then square disappears - in that sense bus graphics is not only graphic.
I think it would be possible that if bus stub touches two wire segments then dot is placed there. Of course bus stubs would still not be a wires so if they touches each other at bus that makes no connection.
But I have written about that dot only by association with what is discussed here and what I noticed last week. I don’t think it is in any way important.

You write:
Wire touches bus - no dot.
Bus touches bus - no dot.

But my case is:
Two wires touches bus stub.
From what you write I’m not sure if that case was also discussed. From my point of view it has a very little importance. I have written it only because I noticed the discussion about dots and remind what I noticed last week.

That’s worth considering.

  • One wire end and a bus end: no dot.
    • Similar to two wires ends: no dot.
  • Two wire ends and a bus end: no dot.
    • Similar to three wire ends: YES dot.
  • Three wire ends and a bus end: YES dot.
    • similar to four wire ends: YES dot.

So there’s some inconcistency in the “two wire ends and a bus end” case because the dot comes from wire ends and the bus end doesn’t count.

This one is more serious:
image

“a” doesn’t form a continous net, the bus entry is separate. “b” is one net, there are 3 endpoints in the same coordinates.

EDIT: well, isn’t it strange… I highlighted the nets, and in “b” the bus entry was highlighted, too. Now I can’t get it highlighted anymore.

EDIT2: Oops, 5.1 and 5.99 behave differently. Only 5.99 does what I described. No wonder because the bus system has changed.

I found another way to create a no-dot 4-way junction, reported here: https://bugs.launchpad.net/kicad/+bug/1848714.

Another problem with implicit junctions without dots: https://bugs.launchpad.net/kicad/+bug/1848720.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.