OCPN Beta 4.8.1 PreAlpha

[nohal]

Originally Posted by cagney

Pavel

When activating your new cog/sog feature in "options" this pops up:
Attachment 161021

I tested this alpha from the command line and these are the last few lines:

/usr/include/wx-3.0/wx/datetime.h(876): assert "IsValid() && dt.IsValid()" failed in operator!=(): invalid wxDateTime

[1]+  Trace/breakpoint trap   (core dumped) /usr/local/bin/opencpn

Kubuntu 16.04 /64bits

Thomas

Thomas...
I would say this is a wx error and the assert should not be created here (And isn't with wx 3.1)

But will replace the logic with something else that works with wx3.0 as well.

Pavel

[did-g]

Originally Posted by boat_alexandra

Isn't the plugin that requests the record responsible for freeing it? How is it leaking?

It's grib_pi leaking not weather_routing and the easier for plugging it was some changes in grib record format layout, which impacts weather routing.
Once I broke grib record compatibility adding unrelated changes make sense.

Originally Posted by boat_alexandra

If it could share records between configurations that might be nice, but I doubt it would help performance that much. How much memory?

With nice grib we have around a 18h file is 65 MB (current step is 15mn), finding the start best time on a 32 bits system can rapidly exhausted linux 3GB limits and likely does the same on Windows.

Originally Posted by boat_alexandra

Maybe it could just copy the records and use smart pointers within the weather routing plugin?

In any case, I'm happy to accept working changes that improve performance, but I'm just asking some basic questions.

It's what it does, using wx smart pointers in weather routing only.

[nohal]

Thomas, Hakan...
Should be fixed in master now.

Pavel

[bdbcat]

did/Sean...

I understand that weather routing Plugin needs update for new O482 GRIB plugin data structures.

Can weather routing be made backward compatible to O prior to 4.8.2 coming? Or do we need to declare/accept API breakage and have a specific plugin set for O4.8.2+?

Thanks
Dave

[rgleason]

Dave and Mike, VFkaps now works! Just download and click on the new purlpe chart Icon that appears to refresh the new kap file, and the VFkap appears.

Dave thank you for the fix.
Compiled 29d570b Merge pull request #940 from davidsanner/colorConfig
Installed on SurfacePro3 Win10

[Hakan]

Originally Posted by nohal

Thomas, Hakan...
Should be fixed in master now.

Pavel

Pavel..
Sorry. no difference. When checked, 2 sek, the SOG varies between 0 and 19 knots and COG often 180 degr wrong and various values in between.
While unchecked SOG is 5.3+-0.3 knots and COG 29+-2 degr, (/5 minutes).

Håkan

[nohal]

Originally Posted by Hakan

Pavel..
Sorry. no difference. When checked, 2 sek, the SOG varies between 0 and 19 knots and COG often 180 degr wrong and various values in between.
While unchecked SOG is 5.3+-0.3 knots and COG 29+-2 degr, (/5 minutes).

Håkan

Hakan...
Can I have a short VDR recording of your data?

Pavel

[Hakan]

Originally Posted by nohal

Hakan...
Can I have a short VDR recording of your data?

Pavel

Pavel
Here's the file: https://www.dropbox.com/s/jc872vqze5...fjord.txt?dl=0
Håkan

[rooiedirk]

The chart info popup box is no longer positioned above pianobar button, if not running full size window See att..

[nohal]

Hakan...
Can reproduce what you see with very slow VDR replay rate and "too short" dampening interval for the position->SOG/COG calculation.
This is expected as the VDR plugin replays the messages at a constant rate, not according to their timestamps, which means once there is eg. a burst of AIS info, the delay between the position messages gets much longer than in a real system.

Anyway, even with VDR playing your file at only 3 messages per second (Which is obviously much less than what it was while recorded), I can get the COG predictor point "mostly in the right direction" with the dampening interval set to 10 second maximum.
The calculated speed is of course something between "just right" and "total nonsense" while using VDR to test this feature as the messages do not come at the intervals they did in the original traffic so the "distance/time" formula is vastly unreliable.

Keep on mind that this is really not meant to replace the SOG/COG info from a correctly working GPS, but to produce "at least something" in case the data is completely missing or wrong.

Pavel

[Hakan]

Pavel..
Very good, understood and accepted. A simple simulation error in other words. Someone sailing in real to do further testing then.
Thanks
Håkan

[Hakan]

Originally Posted by rooiedirk

The chart info popup box is no longer positioned above pianobar button, if not running full size window See att..

On Win10:
Yes I see the same. The pop-up position refers to the whole screen independent of OCPN canvas size. Just now I can't say when this was introduced.
Håkan

[cagney]

Originally Posted by Hakan

On Win10:
Yes I see the same. The pop-up position refers to the whole screen independent of OCPN canvas size. Just now I can't say when this was introduced.
Håkan

On Linux the pop-up position is correct for both vector & raster charts.

Pavel's fix for the assertion error worked!

From another thread...
The real scale bitmap for "Own Ship" does not work correctly with the latest git-code.

The Ais target Icon is much smaller than in 4.8.0. Is that intentional ?

Thomas

[bcn]

Originally Posted by cagney

On Linux the pop-up position is correct for both vector & raster charts.

On our Linux ARM (OS Armbian) the pop-up is displaced. Same like Dirk's report.

[cagney]

Correction

I tested the wrong version...
The pop-up is displaced on Linux/64 as well.

Thomas