[geda-user] PCB: polygon dicer bug?

[Gabriel Paubert]

Hello,

I have a problem with the PCB in the attached file. Can someone please opne it and tell
me whether some pads in the polygon also appear shorted when they should not.

Gabriel

[Stephen Ecob]

Yes, I see two shorts using PCB of March 2013.

Stephen

[Gabriel Paubert]

Post by Stephen Ecob
Yes, I see two shorts using PCB of March 2013.

Thanks for confirming. Actually, if I delete the traces and recreate them,
everything is fine. But if I save and reload, the problem comes back.

If I export the pcb to Gerber, the two shorts are in the photoplotter file,
which is really annoying and shows that this is not a spurious display artifact.

This is a a reduced test case from a larger PCB. I can further remove
J1 and still see the problem. At this point I think that the test case is
simple enough (1 polygon, 2 lines, 4 pads) for the people who wrote/know t
he dicer code to step in.

I use the latest git code, with even a few local changes to the git
exporter, but they should not affect this issue.

Gabriel

[Stefan Salewski]

Post by Gabriel Paubert
If I export the pcb to Gerber, the two shorts are in the photoplotter file,
which is really annoying and shows that this is not a spurious display artifact.

That is interesting and good to know!

In 2009, when I did my large, still unfinished DSO board, I have not
seen such shortcut problems, and my feeling was that PCB was working not
bad at that time. But I had done no gerber check at that time. At least
my board with the AVR 1287 in 2006/2007 war really OK.

[Stefan Salewski]

Post by Gabriel Paubert
Hello,
I have a problem with the PCB in the attached file. Can someone please opne it and tell
me whether some pads in the polygon also appear shorted when they should not.
Gabriel

I can not really help you -- when I open that file with PCB version
20110918 for routing style signal line width is 0.0025 mm, so I guess
you are using some latest git version of pcb, and my release shipped
with gentoo can not really handle that file format. And pressing 'O' key
shows DRC errors, so I guess you have connected pads which should not be
connected following net list.

Do you care only about visual appearance?
Indeed there seems to be a thin connection from pad to surrounding
copper polygon. The funny part is: If I delete the trace, and press 'U'
key to undo delete operation, then all is OK. And seems to stay OK.

Unfortunately my PCB version has much more serious problems:
After some operations, GTK GUI behaviour becomes very sluggish.
I noted that multiple times when I imported my rubber band traces and
thought that it was only a problem of that import. But it occured now
with your file also. It may be related with my current PCB release, or
maybe with the fact that I am using Gnome 3, which uses GL drawing, and
PCB uses GL also.

[Levente Kovacs]

This bug is still exists in git HEAD.

If I modify the thickness of the trace, the bug is not triggered. I pressed s
and S. Maybe the problem is only triggered when the clearance and the thickness
are equal?

Please find the attached file.

HTH,
Levente

[Kai-Martin Knaak]

Post by Levente Kovacs
This bug is still exists in git HEAD.
If I modify the thickness of the trace, the bug is not triggered. I
pressed s and S. Maybe the problem is only triggered when the
clearance and the thickness are equal?
Please find the attached file.

I am confused.
Should the attached layout show the bug?
What is the bug about?

Is there a bug report on launchpad?

---<)kaimartin(>---

[Levente Kovacs]

On Fri, 03 Oct 2014 22:26:06 +0200

Post by Kai-Martin Knaak
I am confused.
Should the attached layout show the bug?
What is the bug about?

Sorry, I wasn't clear. If the thinkness is altered, the bug is NOT triggered.