SHAPE_LINE_CHAIN::PathLength stuck

[Nick Østergaard]

I was trying to route the top layer on a six layer board. I had some trouble getting a pin routed out. Suddenly pcbnew seemed to get stuck, and I then tried to attach to it with gdb and it seemed to be stuck in:

SHAPE_LINE_CHAIN::PathLength(VECTOR2<int> const&)

I have attached a backtrace.

I did wait a couple of minutes or so before attaching gdb.

Application: kicad
Version: (6.0.0-rc1-dev-1168-gacfa13c03), release build
Libraries:
    wxWidgets 3.0.4
    libcurl/7.62.0 OpenSSL/1.1.1 zlib/1.2.11 libidn2/2.0.5 libpsl/0.20.2 (+libidn2/2.0.4) libssh2/1.8.0 nghttp2/1.34.0
Platform: Linux 4.18.16-arch1-1-ARCH x86_64, 64 bit, Little endian, wxGTK
Build Info:
    wxWidgets: 3.0.4 (wchar_t,wx containers,compatible with 2.8) GTK+ 3.24
    Boost: 1.68.0
    OpenCASCADE Community Edition: 6.9.1
    Curl: 7.62.0
    Compiler: GCC 8.2.1 with C++ ABI 1013

Build settings:
    USE_WX_GRAPHICS_CONTEXT=OFF
    USE_WX_OVERLAY=OFF
    KICAD_SCRIPTING=OFF
    KICAD_SCRIPTING_MODULES=OFF
    KICAD_SCRIPTING_PYTHON3=OFF
    KICAD_SCRIPTING_WXPYTHON=OFF
    KICAD_SCRIPTING_WXPYTHON_PHOENIX=OFF
    KICAD_SCRIPTING_ACTION_MENU=OFF
    BUILD_GITHUB_PLUGIN=ON
    KICAD_USE_OCE=ON
    KICAD_USE_OCC=OFF
    KICAD_SPICE=ON

[Nick Østergaard (nickoe)]

I don't have a PCB where I can reliably reproduce this.

[Wayne Stambaugh (stambaughw)]

@Tom, did you get a chance to take a look at this? Is it possible that some condition puts the SHAPE_LINE_CHAIN code into an infinite loop?

[Tomasz Wlostowski (twlostow)]

Hi Wayne,

There's no way for PathLength() to get stuck, but Nick's stack trace might also mean the infinite loop occurs somewhere in the walkaround code and just repeatedly calls PathLength().

Tom

[Wayne Stambaugh (stambaughw)]

@Tom, from a user perspective I not the difference is significant. If the walkaround code can get stuck in a loop and there is no way to exit out of the loop, then I see this as an issue because the process will have to be kill possibly causing loss of changes.

[Tomasz Wlostowski (twlostow)]

@Nick, do you have any PCB where this bug ever occurred? Maybe I'll try fuzzing the router a bit with random mouse events to see if it fails after a while...

[Nick Østergaard (nickoe)]

@Tom, I think this is the issue I mentioned at FOSDEM, I have the pcb file locally, and I will try to see if it still happens for master and report back.

[Tomasz Wlostowski (twlostow)]

@Nick, I really appreciate that. If you manage to freeze Kicad, could you send me a full coredump instead of just the stack trace? I suspect the freeze happens somewhere in the shove algorithm (PathLength() always exits in finite time).

[Nick Østergaard (nickoe)]

@Tomasz, sorry for the slow response. But I can't seem to make it get stuck here at the moment.

Version: (5.1.0-rc1-90-g2dd653a32), release build

But I will attach the file that I tried to reproduce it in. But this file show the issue with the diff pairs not always connecting. Maybe I should create a new bug for that, please advise.

[Tomasz Wlostowski (twlostow)]

@Nick,

Do you remember which part/component/track were you routing when the freeze occured?

Thx,
Tom

PS. Register the diff pair bug in a separate report, it's easier to follow.

[Nick Østergaard (nickoe)]

I was trying to route those diff pairs, and because they were "hard" to connect and it was not something I could reproduce consistently. I don't remember the pns settings. But I guess it was in walkaroud mode. Maybe it requires some more obstacles for the pns to get stuck. I guess we should close this bug for now.

[Tomasz Wlostowski (twlostow)]

@Nick ... and open a new one with an ugly diff pair routing bug causing difficulties drawing the pairs ;-) Thanks for a nice test case Nick!

T.

[KiCad Janitor (kicad-janitor)]

KiCad bug tracker has moved to Gitlab. This report is now available here: https://gitlab.com/kicad/code/kicad/-/issues/1789