PNS is crashing about every 5 PNS operations

[Dick Hollenbeck]

Most recent PNS code is segfaulting more than I have ever seen it.

I would guess it can be duplicated by moving tracks and vias on a busy board. Mine is 10 layers, 6 of which are busy routing layers.

[Jon Neal (reportingsjr)]

Can you provide a specific build number that you're having issues with and what OS you're on?

Also any chance you can provide a board as a test case?

Thanks!

[Dick Hollenbeck (dickelbeck)]

I am on linux, and I generally use the latest source code, typically within a couple of hours or days of HEAD. I did most recent PNS code, and I don't speak git hash codes because they are totally out of order, in more ways than one.

No I cannot supply a board.

After building pcbnew as a Debug release and running it standalone under a debugger, I have yet to see it crash. Summarizing the changes in my usage environment:

1) Compiled differently, Debug vs. Release
2) Running standalone.
3) Under debugger.

1) strikes me as most likely, although there is still a very remote possibility that I have reduced my pushing and shoving since this Debug built was put to use. From a pure probability point of view, I am currently leanging towards: code generation issue.

[Maciej Suminski (orsonmmz)]

Dick,

I suspect the crashes you have experienced were fixed in 65821328. Can you confirm your board contains locked tracks with vias attached?

[Dick Hollenbeck (dickelbeck)]

Orson,

Yes it does, but I have that fix in and it still crashes. So the program still has one or more bugs re: PNS.

In the latest crash there was involvement with the undo buffer. I was running with no debugger, single_top. I went to route a new segment, rounding off a corner, and after entering routing mode, there was ghosting from tracks that were not really there, but that were in the undo buffer. These appeared just as I entered routing mode, and they were concurrently visible with the *current* segments. After moving the mouse just a little, then segfault.

[Launchpad Janitor (janitor)]

[Expired for KiCad because there has been no activity for 60 days.]