[Gta04-owner] qtmoko-navit: Gui blocks in town search in binfile maps.

Stefan Wildemann stefan.wildemann at metalstrolche.de
Tue Mar 5 09:32:19 CET 2013


Am 05.03.2013 08:49, schrieb Radek Polak:
>
> Ahh great, 4 days was expected :) Btw with big ccache the next 
> rebuilds are going to be much faster - under 1 day
>
OK. I've checked your qemu config against mine and it's basically 
identical. So qemu for armhf just
is that slow. I hoped mine being misconfigured somehow. This at least 
proves i did everything correct.
This proves my assumption that native building on gta04 is faster than 
qemu. Someone tried raspberry PI?
>
> Sure, i have uploaded my setup here:
>
> http://sourceforge.net/projects/qtmoko/files/BuildHost/qtmoko-buildhost-armhf.tar.gz/download
>
> It's just kernel+initrd+start script. You need to supply your own 
> rootfs named wheezy-armhf.img
>
Thank you for sharing this.
>
> > Is it possible to just build parts of qtmoko say e.g qwhereabouts?
>
> Yes, very easily, there are two ways. Either:
>
> cd /root/qte/build-gta04/devices/gta04/src/plugins/whereabouts/neo/
>
> make
>
> or
>
> cd /root/qte/qtmoko/devices/gta04/src/plugins/whereabouts/neo/
>
> /root/qte/build-gta04/bin/qbuild
>
I'll give that a try in about 4 days. Because the build-package.sh 
script rm -r'ed my build-gta04 dir. And
somehow zapped my changes to qtmoko by git-reset. So i'll have to wait 
until this finishes again. OK. Now
I know not to use this script for second run.
> > About the navit problems I found so far 2 problems.
>
> > First, search is unusable because for whatever reasons the Idle event
>
> > doesn't get triggered. If it
>
> > is changed to a cyclic half second timer then at least search works.
>
> > This is NOT a solution yet, but way
>
> > better for those wanting to actually use navit now. I'll continue
>
> > investigation here.
>
> >
>
> > No patch for that. Just alter line #883 in
>
> > navit/navit/graphics/qt_qpainter/graphics_qt_qpainter.cpp of radek's
>
> > qtmoko_hack2 branch to say"return (struct event_idle
>
> > *)event_qt_add_timeout(500, 1, cb); "
>
> Great!
>
> > Second problem is the qwhereabouts-plugin neogpsplugin. Currently it
>
> > hooks to gpsd via gpspipe.
>
> > Additionally it triggeres rfkill for gps device. Unfortunately gpsd
>
> > doesn't immediately close the gps device when
>
> > it looses connection. So closing neogpsplugin interferes with gpsd. This
>
> > may lead to strange gps behaviour.
>
> > Do we need these "manually" calls to rfkill at all? I always thought
>
> > kernel would do this now. If yes then we could
>
> > use gpsd hook instead.
>
> IIRC kernel do it automatically for GPS chip, but GPS antenna has to 
> be powered on manually.
>
Then I'll create some hook script for gpsd to do it. Because gpsd is the 
only one
knowing when to open and close the gps device now. The way neo 
whereabouts plugin does
it now doesnt's "feel" right. Despite it not killing all processes it 
spawns. I propose to just drop it
and use (fixed) qpsdwhereabouts instead.
>
> > As we're having gpsd anyway, I switched to directly using qwhereabouts
>
> > on gpsd.
>
> > Unfortunetly this doesn't work out of the box yet as qwhereabouts sends
>
> > a malformed request to gpsd. It seems
>
> > gpsd doesn't like 0x0a chars within one request string.
>
> > I'm going to fix this, but as long as compiling qtmoko takes around 4
>
> > days I won't have results soon.
>
> Now it should be easy to cross compile for armhf. There is now script 
> in git:
>
> sudo scripts/qtmoko-chroot-armhf.sh
>
I've seen this, and I'll continue with bug-fixing gpsdwhereabouts that 
way. Let's see if i can
cross compile navit as well. Unfortunately i've to port this to run with 
Gentoo. (no big deal. Mostly
there is no cdebootstrap for gentoo, just debootstrap...)

Thanks Stefan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20130305/da91e0fb/attachment-0001.html>


More information about the Gta04-owner mailing list