wisgal | # 1821 | 2015-06-07, 18:43 | Report

Does that mean it cannot be installed on N9? The terminal image is attached.

Attached Images

 

peterleinchen | # 1822 | 2015-06-07, 20:38 | Report

So, for sure python-qtmobility is available on the N9 repo mirror.
Looks like some other dependency is the problem.

Maybe we derail this thread, maybe not.
What you can/should do is install everything manually:
devel-su
apt-get install modrana
- error thrown about python-qtmobility, then
apt-cache search python-qtmobility
-see that it is available, and continue with
apt-get install python-qtmobility
read the log, find out which package is problem and continue with last two steps (apt-cache search ... and apt-get install ...)

Maybe this way we can narrow it down to the real problem.

MartinK | # 1823 | 2015-06-07, 21:25 | Report

Try:

That should install the missing dependency. But Warehouse should do that automatically when installing modRana, it's weird it's not working correctly.

It's of course still on my mind, but compared to all the other stuff needed & requested by users it is one of the most complex items. Also I would like to make it as multi platform as possible so that there aren't big separate platform specific chinks of code. For example I can count on Mapnik being available on Maemo (thanks to you heroic effort! ) and on desktop, but I'm not sure if I'll be able to build it for Sailfish OS and I have serious doubts of an Android build being doable. Still from my experiments it should be possible to run Mapnik in a "local" tileserver mode - basically feeding it with geometry and making it render individual tiles using the Python interface. Then the current tried and tested tile handling machinery would put the rendered tiles on the screen, show a loading tile before tile rendering is ready, could cache tiles, etc.

Still, the main issues still stands: How to get the data for rendering ? AFAIK big Mapnik tileservers use a Postgis database to fetch geometry during rendering - that would be probably too much of an overkill for the N900 and mobile devices in general. So some other more simple/lightweight format/store is needed that can still be plumbed to Mapnik (and/or ideally other potential renderers). Raw OSM data files (XML, PBF) won't do for any non-trivial area due to missing support for efficient random access and area queries.

Therefore some sort of custom format with pre-processed data will be needed - my current favorite is reusing the already present support for sqlite tile database files, but storing tile geometry instead of tile images, with optional per tile compression. Using this method the tile rendering backend would very easily query the geometry for any given tile.
There also options for just generating "geometry tiles" for all layers we might want to render for maximal simplicity and ultra fast fetching (this would make the database bigger than the raw OSM data for the area) or storing just a single (or a couple) of map layers and combine/split them for the given zoom level (meaning the database would not be (much) bigger than the raw OSM data for the area.

In short - I'm still investigating how to best do it and piecemail working on the needed components - testing your Mapnik port, investigating the QtQuick 2 Canvas, improving the modRana data repository, planning improvement in the tile storage module for "geometry tiles" support, etc.

Also as for related development - Basil Semuonov (probably best know for running Open Repos) has apparently been working on a offline rendering and routing server (I might have written about it already a while ago). That would also help a lot - provided that:

1. he actually ever releases it (he seems to be generally really busy)
2. it can be ported to Maemo (you have used your magic to port Magic, so I'm sure this would be easy for you ;-) )

More or less instant offline maps as his solution seems to have some sort of internal database (solving the data issue), can import raw OSM data and provides a localhost tile server (very easy to use by current modRana codebase).

Thanks!

I haven't really checked in quite a while - I have "write a layer availability testing script" on my TODO but have not got to it yet. :P So just let me know what layers seem dead and I'll remove them if they also don't work on my side & if I can't fix the URLs. Thanks in advance!

wisgal | # 1824 | 2015-06-08, 14:28 | Report

I am sorry if my post may derail the thread. Of course I will leave after failure. However I tried what you and Martink proposed and what the terminal asked for apt-get update but still no success. The results are shown in the attachments.

Attached Images

  

peterleinchen | # 1825 | 2015-06-08, 15:48 | Report

So you still have downloads. maemo.nokia.com configured!
That is the reason.

Remove it from sssu file.
Or try to remove ancelads repo-mirror deb and install it again. Iirc there was some update on it. And n9 keeps older repos in higher (and singleton) priority (even it should take the next one )

wisgal | # 1826 | 2015-06-08, 20:10 | Report

Removed and reinstalled Ancelads repomirror from openrepos and the results are the same. Regarding the second option I didn't find such configuration in the /.config folder.

peterleinchen | # 1827 | 2015-06-08, 20:27 | Report

Rebooted?

file /etc/apt/sourceslist.d/aegis.ssu-keyring-xxx.list

wisgal | # 1828 | 2015-06-08, 21:03 | Report

That's it. Deleting the two keyring files solved this issue. Thanks a lot.

marmistrz | # 1829 | 2015-06-09, 08:37 | Report

Thanks for the reply

Well, Sailfish is Linux, so grabbing the packaged mapnik and its build-deps for Fedora or whatever should just work. As for Android, it's become such a caricature of what Linux should be that I don't have any idea. But if the Digia folks were able to run Qt there and Python runs there as well, why shouldn't boost & mapnik fail? But I guess xda-developers may be a better place to discuss mapnik on Android

Well, it must be easily convertible from the data here: http://download.geofabrik.de/europe/poland.html#

I have no clue about the custom format, but wouldn't it be a good idea to:
1. Create any offline vector maps, even if it'd be a killer
2. Then add code delaying the loading of the whole file into the memory (not a killer anymore but maybe a little slower - slow but reliable in case of no Internet connection)
3. Investigate the custom format, indexing or anything else to enable quick lookup.

Or maybe the pbf (for instance) file could be split into smaller chucks, e.g. 50 MB each depending on the coordinates? And than it would be trivial to get the proper file - take the coordinates x,y - let x1=f(x), y1=g(y) - take the file `map-$x1-$y1` - find the fragment. This would greatly diminish the amount of file reading and RAM used.

Well, I'd rather start considering the basil's once it's working. But I guess it shouldn't be very difficult to port.

wisgal | # 1830 | 2015-06-10, 14:30 | Report

Modrana on my N9:
The main page: Nothing happens when clicking on Routes or POI.
Options page: Clicking on POI, Navigation, Network, Debug just bring a black page with just a header of the item. Is there any other files to be installed to make the application useful?