OpenCPN Beta test 4.1.1108 Release

[seandepagnier]

Hakan,

Is this from latest git with the internal tesselator?

[Ptizef]

Hi Dave, hakan

"cm93 fault on panning at small scale across IDL."

Tested new last git on my Vista/Intel (R) 4 machine

Impossible to reproduce the crash, neither Hakan's process to get O hanging

so for me, seems fixed

Just a small issue certainly known, but I never saw although it's also in 4.0.0:
very near the 180 longitude line, way points display is unstable : disappearing and appearing depending on zoom level

I'll test asap on my W10 machine

Thanks very much

Jean Pierre

[Hakan]

Originally Posted by boat_alexandra

Hakan,

Is this from latest git with the internal tesselator?

Sean.
It was built yesterday but since there are new commits today I've now built them instead.
And..... Issue is solved - Now I can move around and "center" the WP w/o hangs.
Very good.
Thanks
Håkan

[Hakan]

Info report..
Todays commits aims to fix memory leaks in S57 but "my" leaks are still there. (http://www.cruisersforum.com/forums/...ml#post1969952) Maybe not unexpected?
Håkan

[seandepagnier]

I have fixed the waypoints problem

[Ptizef]

Sean,

re : "I have fixed the waypoints problem"

Yes your last pull request fixes it

many thanks

Jean Pierre

[bdbcat]

Hakan....

I've been looking into your "leak" problem.

Can you build either

37303606465255fdf54dde3eda4945e0857f0b89 "4.1.0"

or

6cbf4818bb21d43796289490ae414fd99563396d "4.1.409"

??

Do you have a useful date stamp on the 4.1.0 installation file that seems to run OK for you?

Apparently the fault was introduced between these two releases.

FYI, I also see some leak problem. For me, I see it for sure on 4.1.530, linux, OpenGL, while following boat, with live NMEA data stream, raster charts cached. This is my current daily use machine, on the boat. It runs out of memory after about 2 hours, depending on the display scale.

I'm doing some more tests today, trying to pin it down to an earlier commit. As you discovered, it takes some time to recognize the problem.

Dave

[Hakan]

Dave..
I'm on it.
First, for the "useful" binary for 4.1.0: I can run and get NMEA in, but not enter the menu. But still no memory built up.
File creation date is rubbish: ‎9 ‎december ‎2024. Last changed: 2015-02-26
log file says: Version 4.1.0 Build 2015-02-20

I will try again to build the commits you pointed to. There's a lot of errors but I'll make a try before come back for help.

I didn't say it should be simple, or??
Håkan

[Hakan]

Dave

Build 4.1.0 or 4.1.409 in VS2013 is not easy. I remember this "Android" period was a time of waiting until the Windows dust was settled.
Errors in 4.1.409 are like this: (Duplicates deleted)

Error    5    error C2039: 'max' : is not a member of 'std'    C:\Builds\OCPN\OpenCPN\src\printtable.cpp    269    1    opencpn
Error    6    error C3861: 'max': identifier not found    C:\Builds\OCPN\OpenCPN\src\printtable.cpp    269    1    opencpn
Error    11    error C2308: concatenating mismatched strings    C:\Builds\OCPN\OpenCPN\src\wxJSON\jsonval.cpp    965    1    opencpn
Error    16    error C2556: 'int round(double)' : overloaded function differs only by return type from 'double round(double)'    C:\Builds\OCPN\OpenCPN\plugins\grib_pi\src\GribUIDialog.cpp    52    1    grib_pi
Error    19    error C2264: 'round' : error in function definition or declaration; function not called    C:\Builds\OCPN\OpenCPN\plugins\grib_pi\src\GribUIDialog.cpp    812    1    grib_pi
Error    24    error C2308: concatenating mismatched strings    C:\Builds\OCPN\OpenCPN\src\wxJSON\jsonval.cpp    973    1    grib_pi

The first build-able was 4.1.427 commit ac46f43

If it's worth the effort we could of course try to clear the errors but I think I need some help.
Edit: Or if you have traces for test I could try them ASAP.

Håkan

[transmitterdan]

Dave,

I cannot build a debug version on MSW. Using bisect have tracked it to this commit https://github.com/transmitterdan/Op...c69d1cd307ee60

Somehow it introduced some new conditional so now it is not possible to compile LLregion.cpp under MSW VS-2013 for a debug build but it works non-debug.

[bdbcat]

TDan...

I don't understand the MSVC error message at all. What the heck is "any.h"?

However, the fix is easy, and in github now.

Thanks
Dave

[transmitterdan]

Dave,

The header any.h is part of the uber complex wxWidgets macro/template system. I suspect the problem is some kind of identifier name conflict but not really sure. Anyway, appreciate the fix.

[Hakan]

Dave..
I've fetched and built from Git source now and, whatever you did, I now have, according to Windows border list, 4.1.409 running although the package is called 4.1.1108. (gitk looks amazing)
Log file says: 23:36:14: Version 4.1.409 Build 2015-04-09

And the memory leak is still there, the increasing memory usage has the same pattern as before.

/Sorry, time for bed, here is midnight and, though retired, have to be up in five hours.
Håkan

[bdbcat]

Hakan, and other "builders from source"...

I made a change to the build process, so that the build may be conducted from a read-only file system.

The change involves "include/version.h". This file is actually a build product, manufactured by the CMake configuration process. It really should not be written back into the source tree.

So, I moved the file "include/version.h" from the source tree into the "build product" tree.

To avoid fetching the old, wrong "version.h" file, it is necessary to delete that old file, i.e. "include/version.h" from the opencpn tree root. It should never be seen again here.

If you are doing regression builds, as is Hakan, you may see this file appear again. If you do, simply delete it when you move back to a current build.


Thanks, and sorry for the disruption.
Dave

[Hakan]

Thanks then I can get rid of the "dirty" Git after each major build. Version.h changes has to be committed since entering that to gitignore was done after the first time it appeared.
Håkan