xes
|
2011-12-07
, 23:02
|
Posts: 638 |
Thanked: 1,692 times |
Joined on Aug 2009
|
#461
|
|
2011-12-08
, 08:06
|
Posts: 15 |
Thanked: 2 times |
Joined on May 2011
|
#462
|
|
2011-12-13
, 11:40
|
Posts: 638 |
Thanked: 1,692 times |
Joined on Aug 2009
|
#463
|
|
2011-12-13
, 11:45
|
Posts: 1,680 |
Thanked: 3,685 times |
Joined on Jan 2011
|
#464
|
a new little issue..
Sometimes cloudgps crashes without any message as soon it loads all the current map view...
A few times, it repeated the crash until the next reboot.
Sorry but until now i haven't any addictional info.
I suppose i should start it from a term to view some error message.
The Following User Says Thank You to vi_ For This Useful Post: | ||
|
2011-12-13
, 18:52
|
Posts: 317 |
Thanked: 787 times |
Joined on Oct 2009
@ Krakow, Poland
|
#465
|
I've got that installed (as well as libqtm-12), but ldd shows that cloudgps is actually linking against the library from libqtm-location.
The Following User Says Thank You to dwaradzyn For This Useful Post: | ||
|
2011-12-13
, 19:57
|
Posts: 317 |
Thanked: 787 times |
Joined on Oct 2009
@ Krakow, Poland
|
#466
|
New issue...
It is a little difficult to explain but i'll try
imagine one offline only map, a rectangle of a certain area from zoom 9 to 16 with no http link.
While running in offline mode...I discovered that very often cloudgps does not load any part of the map also if i'm over the right zone and the debug window (H) shows me that there is a green square under the cursor.
Moving, zooming in or out after some time it loads the map (it is not a mumble state or too cpu load)
What do you think about?
Rendering should never get locked. I use CloudGPS quite extensively and it just never does. Tile loading on the other hand can easily get locked. Tile loading uses only two threads on N900 (I benchmarked it and it is the most optimal value for N900). When there are two downloads in progress every new download request gets queued until one of those two finish because starting new thread is just too heavy for N900. Downloading currently uses blocking API and this means that it can easily get stuck waiting for data from network. I plan to rewrite downloading using non-blocking API so that:
- there will be many simultaneous downloads possible (up to twenty for example) per one thread,
- downloads that are in progress and are not necessary anymore will be canceled,
- downloads stalled will be canceled and retried,
- it will use even less resources than current implementation.
This should resolve the issue you described and should also improve responsiveness and tile loading time a lot.
The Following User Says Thank You to dwaradzyn For This Useful Post: | ||
|
2011-12-13
, 23:50
|
Posts: 1,378 |
Thanked: 1,604 times |
Joined on Jun 2010
@ Göteborg, Sweden
|
#467
|
On my N900 (GPS working fine in CloudGPS) version 1.0.2-maemo4+0m5 of libqtm-location is installed. Is your version different?
The Following User Says Thank You to handaxe For This Useful Post: | ||
|
2011-12-14
, 09:05
|
Posts: 3,617 |
Thanked: 2,412 times |
Joined on Nov 2009
@ Cambridge, UK
|
#468
|
|
2011-12-14
, 09:30
|
Posts: 3,617 |
Thanked: 2,412 times |
Joined on Nov 2009
@ Cambridge, UK
|
#469
|
|
2011-12-15
, 08:44
|
Posts: 638 |
Thanked: 1,692 times |
Joined on Aug 2009
|
#470
|
Well, there shouldn't be any real differences. The only patch added is to add a signal for updating the media library.
It definitely is down to this though - running "LD_LIBRARY_PATH=/opt/qtm11/lib /opt/cloudgps/cloudgps" gets the GPS working perfectly. I've added in a shell script to do this, and changed the .desktop file to point to that instead.