[Announce] modRana: a flexible GPS navigation system - Page 39 - maemo.org

Active Topics

Porting apps to Leste (29)
to Maemo 7 / Leste by Arno_11 - 3 days, 5 hrs ago
Install Fonts in N900 (68)
to Maemo 5 / Fremantle by teroyk - 4 days, 8 hrs ago
more...

Page 39 of 207

Thread Tools

jj0	2010-12-04 , 02:17
Posts: 73 \| Thanked: 33 times \| Joined on Nov 2010	#381

There is a problem with batch download in the latest version, when I do download around view, it starts downloading fine.

After a couple of seconds:

exception in get tiles thread:
unable to open database file

hundreds of lines of these...

Quote & Reply |

taxaza	2010-12-05 , 00:53
Posts: 32 \| Thanked: 12 times \| Joined on Jan 2010	#382

I also have problems with batch download. I use the sqlite tile storage option. When I download big areas after a while I cannot bring the n900 out of the blank screen. It doesn't seem the phone completely hangs since the indication led is blinking normally and even turns blue when I receive new mails. Unfortunately though the only way to get the n900 back to normal is to remove the batery and reinsert it.

Quote & Reply |

stenny	2010-12-05 , 01:06
Posts: 81 \| Thanked: 114 times \| Joined on Jun 2010	#383

Originally Posted by MartinK

When it's done On a serious note, a first extras-testing release is definitely planed short to mid term. There are still some features that need to be implemented/fixed (+ documentation update) before doing a first release aimed to extras.

I appreciate your position on that, but I would appreciate it if you would move 'done' status to Extras and maybe bump it to extras-testing soon -- I really like it but I don't want appman to take nine years to load

Quote & Reply |

The Following User Says Thank You to stenny For This Useful Post:
Wikiwide

superpj101	2010-12-06 , 04:29
Posts: 23 \| Thanked: 11 times \| Joined on Jun 2010	#384

First of all thanks for great gps app! I experience phone shutdown while running modrana for more than 40mins and I need to reinsert the battery. I noticed that the battery is 30-40% drained during that period.

Quote & Reply |

baseb0l	2010-12-06 , 18:39
Posts: 5 \| Thanked: 0 times \| Joined on Mar 2010	#385

Originally Posted by superpj101

First of all thanks for great gps app! I experience phone shutdown while running modrana for more than 40mins and I need to reinsert the battery. I noticed that the battery is 30-40% drained during that period.

The same happened to me.

Quote & Reply |

	woody14619	2010-12-06 , 18:51
	Posts: 1,455 \| Thanked: 3,309 times \| Joined on Dec 2009 @ Rochester, NY	#386

Originally Posted by taxaza

When I download big areas after a while I cannot bring the n900 out of the blank screen. It doesn't seem the phone completely hangs since the indication led is blinking normally and even turns blue when I receive new mails.

Originally Posted by superpj101

I experience phone shutdown while running modrana for more than 40mins and I need to reinsert the battery.

Originally Posted by baseb0l

The same happened to me.

If you read back a bit in this thread and other app threads you'll find this is not an issue with modRana, but is a system level flaw. Something goes wrong in the graphics driver, which is triggered by apps that do a lot of drawing to the screen. The result is that the screen never comes back on, since it's kernel driver is in a loop. If you SSH into the device, and use top(1) you'll see [sgx_misr] taking 98% of the CPU. This also explains the battery drain, as your CPU is running at near 100%.

Typically it happens more often when GPS is on, though it can happen just out of the blue. There's a bug for this here, which you should vote for if you want this fixed. The bigger problem is that this is a kernel level driver, so you can't kill it, nor can you remove it from the kernel since it's not a module. The only solution is a reboot.

I'm working on a script to monitor/detect this and shutdown the system after some warning that something's wrong with the system (just in case you're on the phone when it happens).

Edit: The script is done, and posted here.

Last edited by woody14619; 2010-12-06 at 20:09. Reason: Added link to script

Quote & Reply |

The Following 2 Users Say Thank You to woody14619 For This Useful Post:
cheve, MartinK

	woody14619	2010-12-07 , 23:56
	Posts: 1,455 \| Thanked: 3,309 times \| Joined on Dec 2009 @ Rochester, NY	#387

Found a tiny little bug. :P If you turn off the "auto-vanish menu" option, so it never goes away.... it actually never shows a menu at all. And you're locked out of the system to make changes. (You have to remove the options file to get menus back.) Opps.

Quote & Reply |

The Following 2 Users Say Thank You to woody14619 For This Useful Post:
MartinK, Wikiwide

smolkop	2010-12-08 , 14:44
Posts: 1 \| Thanked: 0 times \| Joined on Dec 2010	#388

i suggest new functionality righ from map view, long tap on the map should activate popup menu for adding POI or setting route to selected point from current position. make map interactive

Quote & Reply |

MartinK	2010-12-09 , 01:33
Posts: 1,548 \| Thanked: 7,510 times \| Joined on Apr 2010 @ Czech Republic	#389

Originally Posted by gkuenning

However, after a bit of exploring, I found that I could modify mod_textEntry.py to fix the problem. After the line:
Code:
      entry.set_text(initialText)
I added the following hack:
Code:
      entry.modify_text(gtk.STATE_NORMAL, gtk.gdk.color_parse('#ffffff'))
Of course, that's just a hack; it's (almost) as bad to have the color forced to white as to have it forced to black, since it won't work on light-colored themes. The proper solution would be to choose a color based on (1) the theme or user default, if specified; (2) white, if the overall brightness of the background is dark, or (3) black, if the overall brightness is light. But I'm not a gtk honcho, so I don't know how to write that code.

Hopefully, this post will give somebody who knows more about gtk what they need to fix this bug.

Well, I actually used your "hack" in the end

But I've forced not only the text color, but also the background color:

Code:

entry.modify_text(gtk.STATE_NORMAL, gtk.gdk.color_parse('black'))
entry.modify_base(gtk.STATE_NORMAL, gtk.gdk.color_parse('white'))

The text should be always readable with this, the only slight downside is that it won't respect the current theme.

Originally Posted by jj0

I've been having lots of these and it is not related to moving. But seems many processes are started and they don't die.

What happens to me, I just try to prenavigate an area where I'll be going, basically just look at the map, move the cursor around. After a set period of time of doing this, let's say 10 minutes, it will always black when moving further, i.e. you move the screen more to a new area let's say south of where you are looking, and it turns completely back. If you move it back south to where you were before the move, the tiles appear. It would seem like a memory leak or something.

Anyhow this happens when driving also, notably, when autocentered, the map would turn black, when clicking on the screen, which turned the autocenter off, the map would appear, and the moving cursor, too. Then as I'd be driving it'd move of the map, autocenter again - black, click the screen - the map is there, if autocenter right away - black... And so on...

Very annoying... It looks to bee like too many concurrent processes, or don't know, but it's not downloading the tiles, as they are already there, but still black...

Originally Posted by woody14619

The 30 second thing would fit in place with what I saw. I also had an issue near the end of the trip where it was on/routing for 4+ hours where it complained about "Operation canceled due to lack of memory". I "fixed" that by closing and restarting the app though, as I was semi-lost at the time.

I have found a few issues in the tile drawing and automatic tile download code:

no limit on number of download threads
(this means, that for every tile you have seen and don't have in local storage, a download thread was launched - which is bad if you are on a slow and unreliable connection)
some threads that error out are not removed from the thread list
once the in-memory tile cache fills up, it is trimmed each time a new tile is added
the tile drawing code is doing all the tile loading that lags the GUI when moving quickly or when entering an area that is not cached (zooming, etc.)

To fix these issues, I have for V0.20-1 rewritten a substantial part of the responsible code

there is now a (user configurable) maximum number of threads for automatic tile download
tile that are waiting for download slot are indicated by a blue special tile
old (>30s) unfulfilled download requests time out and are removed
the new tile drawing method is non-blocking -> less lag while dragging the map
all tile loading was moved to a worker thread
once the in-memory cache fills up, a given amount of oldest tiles is batch removed -> cache trimming does not needlessly run all the time like this

Hope it helps But my subjective testing seems to indicate better responsiveness so far.

Originally Posted by romanianusa

So this is not a voice nag app right?

Do you mean voice navigation ? Well, yes, there is a "basic but evolving" voice navigation available in modRana

Originally Posted by woody14619

I will... how big are the logs it generates with normal usage? If they're small I may just turn it on and let it log everything, then chop out the parts I need later for issues.

Well, it depends

Normal usage shouldn't be that spammy as usually only important events (like pressing a button, reaching a turn while navigating, receiving routing results) produce a message. But the batch download feature is quite spammy, producing a steady stream of status messages while download or size estimation are in progress.

Also, as the log files are stored in MyDocs - I'd say there should be no space issues

Originally Posted by woody14619

(I need an on-screen debug widget/button! )

Well, this should be quite easy to do, once the needed improved widgets architecture is in place

Originally Posted by woody14619

Wow.. nice map. What profiler are you using to generate that? Is it just for python? I code mainly in C/C#/C++, so not too familiar with python and it's tool set. Nice look to it though.

I use a simple script that first starts modRana through cProfile to get the profiling data. Then it uses Gprof2Dot to convert the profiling data into that nice graph/map

cProfile is python specific, but Gprof2Dot seems to support many different profilers.

Originally Posted by woody14619

Hmm... Odd. It's been much better about warning 3 to 4 miles out since the update to 19. Maybe it works better on trips with longer distances between points? Is the length of the segment used at all to compute how early to trigger?

Well, modRana still uses precomputed values it gets with the directions data from Google. I guess they take the directions type, segment length + some assumed movement speed into account.

But its still just a guess,
locally computed trigger distances would be definitely better, as they can reflect all the locally available variables, like real current and average speed. I even got a rough algorithm for this from a user/developer a while ago:

(as image since the forum was eating the formatting)

Originally Posted by woody14619

Overall, sounds like you've got plenty on the plate to work with. I'll happily test things out with new releases as they come out! I'm also hopeful to getting things setup (build env, etc) this weekend to start porting the tile server. I probably won't have a release before Monday, but if things go smoothly... we'll see.

OK then, fingers crossed that everything goes smoothly

__________________
modRana: a flexible GPS navigation system
Mieru: a flexible manga and comic book reader
Universal Components - a solution for native looking yet component set independent QML appliactions (QtQuick Controls 2 & Silica supported as backends)

Quote & Reply |

The Following 4 Users Say Thank You to MartinK For This Useful Post:
jj0, SubCore, Wikiwide, woody14619

MartinK	2010-12-09 , 02:23
Posts: 1,548 \| Thanked: 7,510 times \| Joined on Apr 2010 @ Czech Republic	#390

Originally Posted by Laughing Man

This might be a silly question but I couldn't find it in the menu or on screen menu. Is there a way to see all the directions at once? When I'm the navigator my fiancee often wants me to tell her the next two or three steps which I can't seem to do with Modrana.

Its not there yet, but definitely looks like possible & usable feature

-> I have opened a new feature request ticket for this.

Originally Posted by jj0

The batch download might be buggy. I've been trying to download a 40km area in this webcafe for ages, can't get it done.

The log shows it's going fine for a while:

Batch tile dl working... (threads: 7) pending: 86618, done: 3608
...

Then

exception in get tiles thread:
database is locked
....

a number of times. Then the download seems to go on, but much much faster, super fast, which is impossible, and of course the tiles are not downloaded...

Originally Posted by jj0

There is a problem with batch download in the latest version, when I do download around view, it starts downloading fine.

After a couple of seconds:

exception in get tiles thread:
unable to open database file

hundreds of lines of these...

Originally Posted by taxaza

I also have problems with batch download. I use the sqlite tile storage option. When I download big areas after a while I cannot bring the n900 out of the blank screen. It doesn't seem the phone completely hangs since the indication led is blinking normally and even turns blue when I receive new mails. Unfortunately though the only way to get the n900 back to normal is to remove the batery and reinsert it.

OK, I'll investigate this. But I suspect this behaviour might be at least partially caused by a OS level flaw, that causes the partition where MyDocs reside to become read-only (this can only be fixed by rebooting the device or by activating and deactivating the mass-storage mode).
I can't really do much about this.

Other factor contributing to could be the old, quirky, non-thread-safe sqlite database version, that is shipped with Python 2.5 on Maemo. Well, I can try to do something about this - assure only one thread accesses the database at a time, try to restart the database connection on fatal errors, etc.

Originally Posted by woody14619

Found a tiny little bug. :P If you turn off the "auto-vanish menu" option, so it never goes away.... it actually never shows a menu at all. And you're locked out of the system to make changes. (You have to remove the options file to get menus back.) Opps.

Thanks for reporting this !

Should be fixed in V0.20-1.

Originally Posted by Wikiwide

That's strange, but after the latest update I see no buttons, no menu.
Here is log from X Terminal.

I don't see anything anomalous in the log, but it looks quite similar to the flaw Woody reported.

Originally Posted by smolkop

i suggest new functionality righ from map view, long tap on the map should activate popup menu for adding POI or setting route to selected point from current position. make map interactive

Well, you have to be careful with this

A feature like this was infuriating when I was using TangoGPS on the Neo FreeRunner. Neo is not the fastest phone under the sun and it quite often interpreted dragging the map as a long click - which launched a popup menu, like the one you describe. It was annoying as hell

Of course, provided the tap is set long enough to not interfere with other input, this might become quite a nice feature

BTW, there are quite a lot gestures/buttons/sensors/states that could be used to trigger different modRana actions: