OpenCPN Version 2.1.0 Beta

[jonasaberg]

Originally Posted by Dave

a. Quilt for TMerc: This needs some discussion. We cannot accurately include a TMerc chart in a Mercator quilt without a costly reprojection of the TMerc chart into Mercator. On the other hand, if all the charts in the quilt are TMerc, then we can build a reasonable TMerc quilt. This will be the case in some Scandinavian areas. This needs coding...
So, how do we handle the mixed-projection case? Ideas?

Originally Posted by Jonas

Just a thought...
Why not use a combination of quilt and not quilted.
Use Tmerc quilted and show the Mercator charts on the chart selection bar.
This would show there are Mercator charts showing the same area while having the possibility to quilt the Tmercs.

Maybe some indicator is needed to point out to the user he/she is in quilt-mode or not since the most obvious right know is the existence or lack of the chart selection bar.

Originally Posted by idpnd

Permit me to bore you with my interface idea again: Use of chart selection buttons for chart types to quilt.

I don't even think there's a lot of value in mixing the types with each other at all, as the result of a cmap-raster mix just looks very odd - quilting by type would seem preferable.

Missed you post suggesting this...it is a good idea but I think it needs refining.

The good thing with quilting is that it constantly will keep the best scale chart in view and seamlessly change when going outside of it. This gives a better overview of the area outside of the detailed chart.

This strength turns to a weakness if there are charts of different projection covering the same area that can not practically be quilted. Somehow it has to be easy for the navigator (even the ones with limited knowledge of different projection or how CPN works under the hood) to see which charts are available.
If there is a choice to quilt only Transverse Mercator OR Mercator charts this will be a problem unless there is a way of keeping the navigator reminded there are charts of the other type available.

This was the reason for my suggestion to be able to quilt Tmercs and still keep the user reminded that there are Mercator charts also by use of the chart selection bar.

/Jonas

[cagney]

I was testing chart quilting with a gnomonic large scale plan, which OpenCPN treats as a Mercator chart, this works quite well, if I only quilt with raster charts. Quilting with raster and CM93 however created the crash below(no CM93 was visible on the screen at the time the crash occurred). Note that I'm using Build 415 as SorceForge is down for maintenance. Gnomonic charts was a standard BA projection for harbour plans etc, until the resent trend to use Transverse Mercator instead.

Dave, some ideas about TMerc quilting, disregard if you already considered this.
It appears that when quilting Mercs & TMercs the TMerc has a slightly smaller scale. What happens if you actually increase the TMerc scale to match the Merc charts?
What happens with the quilting if a TMerc is treated as a Merc? It seems to me that the difference when georeferencing is much smaller than what presently is displayed in a mixed quilt.

idpnd might have a point in not quilting different chart projections. An option in this case would be the possibility for users to batch-reproject TMerc charts to Merc charts with a script.

It's interesting to view Marcos NZ5612_1 & 5612_2 Napier Harbour & Roads, as this is an example of two TMercs quilting together. Near the center of the screen the quilt looks reasonable, but deteriorates towards the edges.

Regarding my post #107(http://www.cruisersforum.com/forums/...tml#post430393) , the blue bar probably just a fluke. The gray-out in small scale CM-93, is not present in build 415.

Thomas

[cagney]

SethDart
Report on testing Cmake.
Bottom line, I compiled and installed without any major problems
Details:I changed nmea.h and ais.h in line with your suggestion, putting #include "wx/socket.h" outside the "ifdef" .
There was a little problem in this part of the code. When running "make", I got this error

In file included from /xxxxxxx/Testing/OpenCPN/opencpn/src/chart1.cpp:94:
/xxxxxxxxx/Testing/OpenCPN/opencpn/include/ais.h:145:21: error: gtk/gtk.h: No such file or directory

and a similar message for "nmea.h"

Temporary changing

#include <gtk/gtk.h>

to

#include </usr/include/wx-2.8/wx/socket.h>

solved this.

Good job, thank you!

Thomas

[cagney]

The development of libbsb has stopped, no changes has been made the last three years.
The downloads on opencpn.org are compile and put together about 2009-05-05, but are using the 0.0.7 release from 2004-12-22.
There has since 2004 been one significant update in the CVS code."bsbfix" was patched 2007-02-05 with this comment

Modified to adapt to new behaviour of bsb_close - the bsb struct is now zeroed after close and hence height need to be remembered elsewhere

I have downloaded an compiled the code. Not totally successful, but the problems were associated with non essential graphic stuff. The important programs "tiff2bsb","bsb2tiff" and "bsbfix" compiled ok.

PS I have added the output from running the "new" bsbfix on Sinbads WorldMercartor kap file.

Thomas

[gjorgensen]

Dave, awesome as always. Build 416 runs very smooth on Windows 7 and better than past releases on XP (although my XP box is not the greatest).

I've seen only one noticeable problem: On XP, build 416 either ignores the GPS or hangs when I try to get it to talk to openCPN-FrontEnd via GPSD. This has always worked fine before. IMO openCPN-FE has become superfluous anyway, so I've quit using it, but I did notice it doesn't seem to work. Symptoms are only occasional COG/SOG input in the console from the GPS, dark boat instead of red boat, and openCPN hangs fairly quickly. Setup is an old Pentium M box with 1 GB of memory running XP, a GlobalSat BU-353 USB GPS, and the latest rev of openCPN-FE. The same GPS works fine if I don't use Front-End, and simply point openCPN at the mapped COM port.

FYI, that XP box with it's 1GB of memory has always struggled with the large US west coast ENC chart set, but that seems to work a lot better in 2.1.0 than in 1.3.X. I had been using only CM93 and BSB charts, since ENC charts had a tendency to hang older releases, but I haven't made it blow up yet on 2.1.0.

[Ptizef]

Hi Dave

Every things seams to work fine on windows . Quilting , AIS , GPS , manager dialog . Many thanks for this great job
I would like to tell you what I've noticed about quilting :
For my tests ,I've put the boat on the center of the screen and never move it later
I started at true scale 1/380000 and here you are the results :

true scale = 1/380000
without quilting : CM93 B 1/1000000 (screen shot quilt1)
BSB 7070 1/167000 (screen shot quilt2)
with quilting (both CM93 and BSB) : CM93
OK
true scale = 1/190000
without quilting : CM93 C 1/200000 (screen shot quilt4)
BSB 7070 1/167000 (screen shot quilt5)
with quilting (both CM93 and BSB) : BSB 5722 1/25000 (quilt6)
Not very good !
true scale = 1/95200
without quilting : CM93 D 1/100000 (screen shot quilt7)
BSB 7070 1/167000 (screen shot quilt8)
with quilting (both CM93 and BSB) : BSB 5722 1/25000 (quilt9)
if I continue zooming , there is no change :
without quilting : CM93 1/100000 (there is no bigger scale chart available )
BSB 7070 1/167000 chart 5722 is never shown
with quilting (both CM93 and BSB) : BSB 5722

Regarding these results , I have some remarques
- The chart BSB 1/167000 is ignored in quilting
- The chart BSB 1/25000 is ignored in no quilting
- I think that in quilting the choice should be something like this :
at true scale 1/190000 , the CM93 C/1/200000 or eventually a mix with BSB 1/167000
at true scale 1/95200 , the CM93 1/100000 or eventually a mix with BSB 1/167000
the BSB 5722 should be shown only at a true scale about 1/45600 or more
Is there a specific problem with my set of charts ?

Best regards

Jean Pierre

[jimbo3123]

To all contributors: Thank you very much! I am very impressed with the improvements.

I have one minor bug to report on 2.10 beta running on Windows 7.
When I enable chart quilting using ENC charts, the Depth Units that normally appear in the top right corner of the chart are no longer visible. They show up again when quilting is disabled.
The Show DepthUnits setting has no effect when quilting is turned on.

[Ptizef]

Hi Dave
Another notice about quilting :
- Test1 shows an interesting mix CM93/BSB
- Moving the chart to the north , test2 still shows a good picture (except a black part in the center )
- Moving for some degrees more to the north , test3 shows a very bad screen
Nonetheless , there are more than one charts to be displayed (both CM93 and BSB (see test4 and test5)

In some positions , quilting works very well and is able to mix CM93 and BSB , whereas in others , it does'nt find any chart to display even there are .
May be some charts are not displayable but 'mask' the others ?

Another example is shown in screen shots test10,test11,test12
at true scale 1/546000 CM93 displayed
at true scale 1/353000 nothing
at true scale 1/273000 mix CM93/BSB

I hope my explanations were clear enough.

best regards
Jean Pierre

[HamishB]

Originally Posted by cagney

It appears that when quilting Mercs & TMercs the TMerc has a slightly smaller scale. What happens if you actually increase the TMerc scale to match the Merc charts?
What happens with the quilting if a TMerc is treated as a Merc? It seems to me that the difference when georeferencing is much smaller than what presently is displayed in a mixed quilt.

Mercator's line of no distortion happens east-west (typically the equator). As you move away from that line the distortion gets worse until you reach infinity at the poles.

Transverse Mercator (including UTM) is, well, transverse to that. The line of no distortion happens along a north-south meridian, and distortion gets worse as you move east-west, until it too blows up.

Depending on where you are in the world relative to the equator and the central longitude of your TMerc, the bending an twisting may have more or less of an effect.

Simply making one bigger or smaller with a fudge factor just ain't gunna cut it. The two projections are fundamentally different and you can't plaster over that.

What would be possible (aka computationally inexpensive) would be to reproject the vertices of the little red outline of other charts within view, to show their position accurately on a chart of different projection.

idpnd might have a point in not quilting different chart projections. An option in this case would be the possibility for users to batch-reproject TMerc charts to Merc charts with a script.

The program you seek is called "gdalwarp". You can find it at www.gdal.org.
IMO doing this is more trouble than it's worth versus just switching the program out of quilt mode. But on the other hand I've already done it for my maps. ;-)

regards,
Hamish

[Ptizef]

hi Dave

Small points about routemanagerdialog

1) tracks statut visible/invisible is not save in navobj file , means it's lost when exiting opencpn
2) selecting active track in the list and
- clicking on "zoom to" leads to a crash
- clicking on "delete all Tracks" seams OK , but leads to a clash if after that I click on track icon or I exit opencpn (may be active track should be excluded from the deleted list)

regards
Jean pierre

[HamishB]

Originally Posted by cagney

Now of course PP must reflect the projection longitude, if not published, the best value is probably the mid longitude of the chart.

For Mercator I believe that PP should be set at the latitude of true scale (+lat_ts).
(IIUC) that is not the same as +lat_0, and +lat_ts is only very relevant if you put much faith in the 1:16,000 scale (which you probably shouldn't). (/IIUC)

For Transverse Mercator, PP becomes +lon_0, which is critical.

For Polyconic PP is again +lon_0 and important.

Sinbad
I don't know any other way of converting a chart picture to a kap file, using freely available tools, than using tif2bsb.

tif2bsb was kept the same in my conversions, everything else was different.

The different methods just use different ways of producing the georeferencing, and other parameter values , in the necessary template/header file.

yes, totally independent .hdr file, but also independent .tif.

So in the end it is a question of the quality of this file.

the .hdr?

I would be very interested in a sample header file and tif image , that would point to a possible problem with tif2bsb.

I'll send you something off-list, but everything works fine for me in 1.3.6, so I can't quote detailed error messages which you may or may not see. (Perhaps Sinbad7 can/will/has already)

Perhaps strict-mode has been rescinded in the latest CVS/HEAD so you don't see the error any more.


regards,
Hamish

[surfjoe]

Hello Dave,

is my request of post #66 (missing buoyage symbols) on a list of feature-requests? Is there a list in the web where a can have a look to open requests?

Thanks for feedback
Joe

[TedLindgreen]

Hi,
Since a few month I'm running openCPN under FreeBSD.
It's good to have this opensource tool, it is working great and I'm using it on board all the time with GPS and AIS attached.

Three patches are needed to compile it. These were needed for previous versions, and still are for 2.1.0.
First, gluNewTess is not found. By changing
LIBS="-lGLU $LIBS"
into
LIBS="-L/usr/local/lib -lGLU $LIBS"
in the configure script this is easily fixed.
Second and third: FreeBSD refuses to mix the jmp and sigjmp routines. In both cm93.cpp and chart1.cpp the declarations of jmp_buf must be changed to sigjmp_buf.

This might help other people trying to compile this fine tool under FreeBSD.

regards,
-- ted

[bdbcat]

Jean Pierre...

Thanks for the detailed testing you are doing. I appreciate the direct feedback.

Still looking at your test cases.

In the first post:
SHOM Chart 5722 has internal errors which the quilt algorithm was not prepared for. The Lat/Lon of the PLY points is clearly incorrect, which explains why the chart cannot be seen in SingleChartMode. Easily corrected in quilt composer. Meanwhile, I suggest you temporarily move or rename 5722.kap, and rebuild the database.
Then you will see the missing small scale charts.

Also, I found a small bug with cm93 included in the chart. I am still studying the impact of this one. It could cause cm93 chart to cover entire quilt, allowing no BSB charts to show under some conditions.....
More later...

Thanks, and good luck
Dave

[bdbcat]

SurfJoe, and others....

You may add Feature Requests to the persistent Tracker on the sourceforge site:

SourceForge.net: OpenCPN: Feature Requests

I refer to this list frequently as I prioritize the work effort and troll for volunteer developers.

Thanks
Dave