OpenCPN Version 2.4 Beta Build 606

[HappySeagull]

Hello testers....

Status:

1. The first show-stopper:
CM93, quilted, course-up, auto-follow mode with GPS (or simulator) input is broken.  Discovered by Hjalmar, after testing at sea. Verified by me.  Has been broken for some time, actually :whistling:

So, there will be another Beta before the RC.

2.  Related question:
The averaging period for COG-up calculations defaults to 15 seconds, and may be changed by hand edit of config file.  Hjalmar suggest that 15 secs is too long.  
Maybe we need Toolbox edit field for this parameter.  We are running out of space on Etc. panel, so I suggest to put it on the GPS page, along with the filter period.
Comments?

etc

Dave

Horrors!
gotta shift my bum outta this chair and onto some cold hard cockpit fibreglass!...
so Forgive me if I thought the "Filter NMEA and course Data" was already doing this per refresh periods COG...Maybe that needs a more explanatory label.(I only have a gps)
...There looks to be lots of room in GPS tab and it's the place I've already looked for such a thing anyways.
15 seconds is too long a time even in a sailboat and I would crank it up until I saw a bad effect per processor or display anyways.1 second is too much-I'd fear a (computer)freeze at the wrong moment.5 seconds would seem about cruising speed coastwise...I'll definitely play with it if you stick it in!

[HamishB]

Originally Posted by bdbcat

4. linux source builders: build/compile warning logs are solicited. Most of the compiler warnings are informative, to me at least.

I'm not sure if it has been fixed or noticed by others yet (ie does it build on ubuntu 11.04?), but Frankie got a build error on Debian/unstable with the DebianGIS packaging rules & 2.4.606.

Today I've updated the Debian SVN for .612, builds ok for me on Debian/stable.

see r2986 - packages/opencpn/trunk/debian

...
s52plib.cpp:5550: Error: symbol `l0' is already defined
...

perhaps a stray global getting exported beyond its library, and conflicting with another??


I will send you the successful build log for .612 (with compiler warnings) via private email.


Hamish

[bdbcat]

Hamish....

Check out my ugly code

This is a hack to inline an assembly code version of memset3.
It is conditional on __WXGTK__ and __INTEL__.

I did this long ago after looking at the instructions generated by gcc3. Made me cringe....

Maybe it breaks with 64? Is this optimization used on your build?

Dave

[HamishB]

Originally Posted by bdbcat

This is a hack to inline an assembly code version of memset3.
It is conditional on __WXGTK__ and __INTEL__.

I did this long ago after looking at the instructions generated by gcc3. Made me cringe....

Maybe it breaks with 64? Is this optimization used on your build?

build log for debian/unstable + debianGIS svn packaging rules sent via private email. (build went ok for me on a AMD 64bit chip/64bit OS, a number of new compiler warnings to look at)

I'll ask Frankie to try 612 and see if the build still breaks for him.

The talk of optimization reminds me that gcc is somehow being fed both -O2 and -O3 on the same command line. Not sure where that is coming from.


cheers,
Hamish

[hedgehog]

Originally Posted by bdbcat

Ahoy OpenCPN'ers...
137. Allow option RMC message to Autopilot output stream. [Settings]AutoPilot NMEA Sentence Out=RMC;RMB

This is good news! Tried it today and I finally got my Silva nmea repeater show ground speed and heading. Still necessary to have an active route, though. Would it be possible to get RMC on the autopilot com port even when no route is active?

/p