Help / FAQ

[barnaclebill22]

Thanks, that was good advice. It wasn't a memory issue (4GB RPI with 1GB used and no swapping going on).
But switching to the non-flatpak version (installed with apt-get) definitely improved performance a lot. However, the new version has crashed twice in 5 minutes and doesn't render charts properly.
Oddly, flatpak is how the OpenCPN docs say to install on RPI4: https://opencpn.org/wiki/dokuwiki/do...tallation:rpi4
I'm getting the feeling that it's not practical to retrofit OpenCPN to my existing RPI installation. I'm suspicious that it might be because I installed headless and added what I needed for VNC, because I'm seeing things like this:

*** BUG ***
In pixman_region32_init_rect: Invalid rectangle passed
Set a breakpoint on '_pixman_log_error' to debug

 SENC: nCPU: 4    m_max_jobs :3

(opencpn:10784): Gtk-CRITICAL **: 06:31:37.345: gtk_box_gadget_distribute: assertion 'size >= 0' failed in GtkScrollbar

I'll start with a fresh RPI image with OpenPlotter and just copy my SignalK settings over.

[cas206]

Originally Posted by barnaclebill22

Thanks, that was good advice. It wasn't a memory issue (4GB RPI with 1GB used and no swapping going on).
But switching to the non-flatpak version (installed with apt-get) definitely improved performance a lot. However, the new version has crashed twice in 5 minutes and doesn't render charts properly.
Oddly, flatpak is how the OpenCPN docs say to install on RPI4: https://opencpn.org/wiki/dokuwiki/do...tallation:rpi4
I'm getting the feeling that it's not practical to retrofit OpenCPN to my existing RPI installation. I'm suspicious that it might be because I installed headless and added what I needed for VNC, because I'm seeing things like this:

*** BUG ***
In pixman_region32_init_rect: Invalid rectangle passed
Set a breakpoint on '_pixman_log_error' to debug

 SENC: nCPU: 4    m_max_jobs :3

(opencpn:10784): Gtk-CRITICAL **: 06:31:37.345: gtk_box_gadget_distribute: assertion 'size >= 0' failed in GtkScrollbar

I'll start with a fresh RPI image with OpenPlotter and just copy my SignalK settings over.

Do you know what windowing system was being used? I get the impression that OpenCPN is having growing pains with Wayland (along with a bunch of other linux software). I couldn't get any of the Ubuntu installed VNC clients to connect to the RPI except for RealVNC. I am "assuming" that the other VNC clients were having trouble due to Wayland. It's still on my todo list to check the various RPI distribution documentation to see what each are using.

[barnaclebill22]

It's x11, not Wayland.
Still can't get SignalK to work...apparently it's listening on ipv6 but not ipv4. So I'm just going to go back to my old headless SK install and wait a bit before trying OpenCPN again.

[nohal]

Originally Posted by barnaclebill22

Still can't get SignalK to work...apparently it's listening on ipv6 but not ipv4.

Unless you explicitly configured it to do something else or have IPV4 support completely disabled on your system, SignalK server listens on all interfaces, both IPv4 and IPv6. Do not get fooled by seeing :::3000 in eg. netstat -l output, it still means the service is reachable on IPv4 as well.

For acceptable OpenCPN performance on a RPi, you really must have OpenGL working, then it is OK even on my headless RPi 3B over VNC. All this comes out of the box on OpenPlotter.

[rgleason]

Originally Posted by Jon Hacking

Does anyone know if there's something inherent in OpenCPN that prevents it from displaying an mbTile that crosses the 180?

I'm making mbTiles of Fiji, & the 180 cuts through the right end of Fiji (the dateline makes a jog there). I can create an mbTile that goes up to the 180 on each side, but when I load it into OpenCPN, I get very strange behavior:

If the 180 is on the left side of my screen, then ONLY the mbTile on the right is displayed, & if the 180 is on the right side, then only the mbTile on the left is displayed.

If I grab the chart & start panning OpenCPN, when the 180 crosses the center of my screen, the mbTile that has been visible disappears & the mbTile that's been invisible suddenly appears!

If I try to construct an mbTile that crosses the 180, OpenCPN doesn't display it at all (although I don't know the quality of the tool I use to build that mbTile - it's the Merge_MbTiles function of Paul Higgins' Sat2Chart program).

Is this expected behavior?

Or have I found an OpenCPN bug?

Dear Jon, we have made a new issue here Dave has looked at this and needs

"test MBtiles file that exhibits the problem. I expect the georef data in the tiles may differ according to the tile source."

If you have a free opensource account you can upload these to this Issue with drag and drop, provided the file type is accepted. You can change the extension too. Otherwise upload to here please, or whatever.



Thanks

[Jon Hacking]

Rick (& Dave), I feel really embarrassed. And the professional tester in me is even more embarrassed. I did see this, I really did.

But I can no longer reproduce the problem, & I've worked pretty hard at trying to do just that. OpenCPN is (now) displaying my Fiji mbTiles that are on each side of the 180 just fine, at the same time, regardless of my zoom-level or screen position.

AFAICT, the only change to my Dev-computer is that my Win11 was recently updated, but I have a hard time believing that this fixed what I was seeing.

I can't reliably make an mbTile that actually crosses the 180, & when I try (by using the Merge_MbTiles function of Sat2Chart) it won't display in OpenCPN, but I think this is more of a problem with either the mbTiles internal format not allowing mbTiles that cross the 180, or an issue with the Merge function, as it probably wasn't designed with this special case in mind.

Whatever. Please accept my apologies & close this bug. If I can find a reliable repro scenario, I'll let you know.

[rgleason]

Jon, it is closed now. https://github.com/OpenCPN/OpenCPN/issues/3581


Next time, if you find something definitive, you can register for a free opensource github account and make an issue. Thanks.

[jdeveloper]

Hi,
I have old 32 bit laptop on which I installed latest Linux Debian 12 with LXDE desktop - it works well. Windows 10 is not supported on this machine - tried already to install it but installation doesn't even start.


I would like to install OpenCPN on this laptop but I am not sure OpenCPN still supports Linux 32 bit. I know it supports 32 bit Windows but haven't found anything about Linux. Can you advise if it is possible to install latest OpenCPN on 32 bit Linux Debian 12?


Thank you,
Chris

[nohal]

Sure, opencpn is in the bookworm-backports repository for all the architectures Debian still supports as you can see at https://packages.debian.org/bookworm-backports/opencpn

What you won't get are the 3rd party plugins - not being built for 32bit x86 Linux and other exotic architectures as pretty much nobody is using them anymore.

[jdeveloper]

Originally Posted by nohal

Sure, opencpn is in the bookworm-backports repository for all the architectures Debian still supports as you can see at https://packages.debian.org/bookworm-backports/opencpn

What you won't get are the 3rd party plugins - not being built for 32bit x86 Linux and other exotic architectures as pretty much nobody is using them anymore.

Thanks. I actually was able to run 32 bit Windows version of OpenCPN on this Linux with Wine emulator. It launches without OpenGL, I can see the UI and load some maps but I haven't tested if other features and plugins work. Has anyone tried it? Will it work fine?

[nohal]

It will, without the features that do require OpenGL and with degraded performance. It is you who must assess if it is fine for you.

[cas206]

Also, if you connect USB devices for use in OpenCPN, you may have an issue getting that to work with WINE. VirtualBox is usually included in most debian derived distributions. It will allow you to install Windows, with USB support, and is capable of doing accelerated graphics, if you have supported video hardware.

[P_Dub]

The last Opencpn 5.9.0-0-1054.... request a parallel installation, Any idea why?
It was previously always an updating process.

[nohal]

Same reason as always, the automatically generated version number of the CI builds, as usually ignore.

[rgleason]

Is there some nmea0183 tool that would process VDR nmea file, adding and correcting the checksums? This would be very useful and save time creating test files for instruments and watchdog.