Islands remain in polygons

[Gabriel Paubert]

In the attached file (simplified from the original), there are 3 islands remaining for the polygon on the bottom layer.
The largest island has vias that actually connected it before simplifying the PCB so it was not originally disconnected,
I nevertheless believe that this is a bug.

[Bert Timmerman (bert-timmerman)]

Hi Gabriel,

I can't load the attached file because of the line (#11), although pcb whines about line #8 in the log window:

<quote>
Styles["Signal,0.0000,0.0000,0.0000,0.0000:Power,0.0000,0.0000,0.0000,0.0000:Fat,0.0000,0.0000,0.0000,0.0000:Skinny,0.0000,0.0000,0.0000,0.1500mm"]
</quote>

Looks like "weird" values to me.

Kind regards,

Bert Timmerman.

[Gabriel Paubert (paubert)]

 Hi Bert,

On Mon, Aug 06, 2012 at 08:45:34PM -0000, Bert Timmerman wrote:
> Hi Gabriel,
>
> I can't load the attached file because of the line (#11), although pcb
> whines about line #8 in the log window:
>
> <quote>
> Styles["Signal,0.0000,0.0000,0.0000,0.0000:Power,0.0000,0.0000,0.0000,0.0000:Fat,0.0000,0.0000,0.0000,0.0000:Skinny,0.0000,0.0000,0.0000,0.1500mm"]
> </quote>

Strange, none of my PCB versions complains about it.

However I believe that we can close this bug, I had not understood
what the effect of setting "full" polygons was.

Now there are other bugs in this area:
- the GtkGL version of PCB does not show the islands in "full" polygons
- it is impossible to change the "fullpoly" flag from the GUI
- I believe that the morphpolygon command should remove the fullpoly flag
  from the parts it generates.

 Kind regards,
 Gabriel Paubert

[Peter Clifton (pcjc2)]

I'm planning a major overhaul of "full" polygon behaviour at some point, including possible connectivity based island removal. This code currently lives in my "pours" branch, however it is a little out of date w.r.t. git HEAD, as it has got stuck behind some blocking API redesign in my OpenGL branches.

Once those API changes are defined and pushed upstream, I'll rebase the "pours" branch properly and revisit it. The rebase is too complex (and manually intensive) to keep doing right now whilst the APIs are in flux.

The major "bug" with full polygons in git HEAD at the moment is that the smaller pieces aren't scanned correctly for connectivity.