[Announce] modRana: a flexible GPS navigation system - Page 160 - maemo.org

petur | # 1591 | 2014-06-10, 09:58 | Report

Originally Posted by petur View Post

The stuck downloading tiles seem to be gone indeed, alas replaced by failed downloads. This was observed while sitting down, n900 with good internet connection, dragging the map around a bit.

Not sure what's going on here... and it would be great to have a way to make it try the downloads again.

On playing a bit more with this, it seems to happen a lot on 2G connections and far less on 3G.

So I'm guessing the download times out too fast and/or doesn't do a retry?

I wonder what changed in this department, as I could (in the past) drive around using a 2G connection and the downloads worked - sometimes a tile took a bit longer, maybe those are the tiles that now fail.

I also tried to move the map enough to make the failed tile drop from the drawing cache, but that trick also fails.

If there's anything to try just let me know.

The Following 2 Users Say Thank You to petur For This Useful Post:
misiak, Wikiwide

sixwheeledbeast | # 1592 | 2014-06-10, 13:46 | Report

Originally Posted by chill View Post

Modrana has caused N900 shutdowns for me, and this recent version caused the phone to stop responding (screen blank, device not responding to any input). I'll take the battery out now, I guess. I can hear a ringing tone on the line when I call it, though...(but the device doesn't ring itself)

I haven't had any random reboots.
I have experienced the not responding issue on one device yesterday. Holding the power key for >3 seconds switches the device off in this situation.

The Following 2 Users Say Thank You to sixwheeledbeast For This Useful Post:
chill, Wikiwide

Kossuth | # 1593 | 2014-06-10, 17:49 | Report

I can make Modrana deliberately crash, by draggin long way to undownloaded area and staying there for a while. Below is the pastee from terminal:

Code:

nan, nan, nan, nan, nan, nan)
position,bearing,speed (in descending order):
(62.905630180612206, 27.67623751424253)
None
None
#############################
storeTiles: sqlite commit OK
sqlite tiles: initializing db for layer: Google Maps
storeTiles: sqlite commit OK
modules/gui_modules/gui_gtk.py:243: Warning: g_object_unref: assertion `G_IS_OBJECT (object)' failed
  gtk.main()
python: cairo-surface.c:429: cairo_surface_destroy: Assertion `((*&(&surface->ref_count)->ref_count) > 0)' failed.
Aborted

and second time

Code:

############################
storeTiles: sqlite commit OK
storeTiles: sqlite commit OK
## N900 GPS debugging info ##
fix tuple from the Location API:
(2, 16, nan, 0.0, 62.903130361810327, 27.689069202169776, 71890.0, nan, 32767.5, nan, nan, nan, nan, nan, nan)
position,bearing,speed (in descending order):
(62.903130361810327, 27.689069202169776)
None
None
#############################
python: cairo-surface.c:429: cairo_surface_destroy: Assertion `((*&(&surface->ref_count)->ref_count) > 0)' failed.
Aborted

These came from scrolling like insane, doubt that they would appear when in normal use like driving.

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

chill | # 1594 | 2014-06-10, 21:00 | Report

Originally Posted by sixwheeledbeast View Post

I haven't had any random reboots.
I have experienced the not responding issue on one device yesterday. Holding the power key for >3 seconds switches the device off in this situation.

(mine didn't reboot before, i.e. with the previous versions, it just shut down)

I tried holding the power button down, it didn't work.

I removed the back cover and it rebooted. I guess it has to do with the swap being on the SD card? (being a Maemo swapmaster, you will surely know).

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

Xagoln | # 1595 | 2014-06-17, 03:37 | Report

I've had a few crashes with Modrana also, particularly if I use it for a bit then lock my phone without closing it. I assumed it was an out of memory/swap thrashing type situation.

The Following 3 Users Say Thank You to Xagoln For This Useful Post:
chill, MartinK, Wikiwide

sixwheeledbeast | # 1596 | 2014-06-19, 16:50 | Report

I have had a few issues with modRana recently. However, this maybe because I have started trying it out a bit more.
@MartinK I have been waiting for you to appear on IRC to ask about them, but I will provide I list here so I don't forget in the meantime.

* Firstly I have experienced the above black screen issue while downloading tiles with the screen locked overnight, if you ssh to device it's the [sgx_misr] at 98% CPU bug; only a reboot will fix this AFAIK.

*Secondly logging tracks seems to fail after 3-4 minutes.

*Lastly tile downloads is still a PITA, I have downloaded 160km "around here", however, when I am offline sections of tiles are missing within the area I downloaded. Re-downloading says that all are found locally.

If I can be any more help debugging this let me know.

The Following 4 Users Say Thank You to sixwheeledbeast For This Useful Post:
MartinK, misiak, nokiabot, Wikiwide

MartinK | # 1597 | 2014-07-01, 20:32 | Report

Originally Posted by petur View Post

On playing a bit more with this, it seems to happen a lot on 2G connections and far less on 3G.

So I'm guessing the download times out too fast and/or doesn't do a retry?

I've had some reports about modRana not retrying failed downloads. There are basically two download error types - fatal and temporary. The fatal types usually happen if the tile server can be reached bud outright refuses to cooperate (returns 404, 503 a something that is not a tile - usually an error page). The temporary errors usually happen if the server or network connectivity becomes temporarily unavailable. These errors should be recoverable and modRana should retry those tiles after a while. But I haven't looked at the code responsible in quite while, will have to check if it still works.

Originally Posted by petur View Post

I wonder what changed in this department, as I could (in the past) drive around using a 2G connection and the downloads worked - sometimes a tile took a bit longer, maybe those are the tiles that now fail.

I also tried to move the map enough to make the failed tile drop from the drawing cache, but that trick also fails.

If there's anything to try just let me know.

That's weird - once the tile is flushed from the cache its state is pretty much reset and the whole down-/load procedure should run with a "clean" slate. What if you quit modRana and then start it again ? Does it download those tiles or still has issues with them ?

You can also enable Print tile cache status to terminal and Tile loading status to terminal in Options->Debug and see if the output sheds some light on the issue. Also for some download errors it reports the tile URL it tried to reach so you can try if you can download those URLs by wget, browser, etc.

The Following 3 Users Say Thank You to MartinK For This Useful Post:
Kossuth, petur, Wikiwide

MartinK | # 1598 | 2014-07-01, 20:55 | Report

Originally Posted by Kossuth View Post

I can make Modrana deliberately crash, by draggin long way to undownloaded area and staying there for a while. Below is the pastee from terminal:

Code:

nan, nan, nan, nan, nan, nan)
position,bearing,speed (in descending order):
(62.905630180612206, 27.67623751424253)
None
None
#############################
storeTiles: sqlite commit OK
sqlite tiles: initializing db for layer: Google Maps
storeTiles: sqlite commit OK
modules/gui_modules/gui_gtk.py:243: Warning: g_object_unref: assertion `G_IS_OBJECT (object)' failed
  gtk.main()
python: cairo-surface.c:429: cairo_surface_destroy: Assertion `((*&(&surface->ref_count)->ref_count) > 0)' failed.
Aborted

and second time

Code:

############################
storeTiles: sqlite commit OK
storeTiles: sqlite commit OK
## N900 GPS debugging info ##
fix tuple from the Location API:
(2, 16, nan, 0.0, 62.903130361810327, 27.689069202169776, 71890.0, nan, 32767.5, nan, nan, nan, nan, nan, nan)
position,bearing,speed (in descending order):
(62.903130361810327, 27.689069202169776)
None
None
#############################
python: cairo-surface.c:429: cairo_surface_destroy: Assertion `((*&(&surface->ref_count)->ref_count) > 0)' failed.
Aborted

These came from scrolling like insane, doubt that they would appear when in normal use like driving.

Thanks for the logs - that one Cairo error looks particularly ugly - might be some sort of race condition when working with Cairo image surfaces (will have to check if I can reproduce it) or some general Cairo/GTK bug. Good thing it does not happen often.

Originally Posted by Xagoln View Post

I've had a few crashes with Modrana also, particularly if I use it for a bit then lock my phone without closing it. I assumed it was an out of memory/swap thrashing type situation.

That's quite probably - or resource starvation might be uncovering some bugs in the underlying frameworks - I doubt there are many other Maemo applications as taxing as modRana these days.

Anyway unless some concrete root cause in modRana is found that causes these freezes/crashes (which is IMHO quite unlikely) I'm thinking about a possible workaround - exporting settings for tweaking the resource intensive things that modRana does, such as the amount of tile download threads, tile cache size, tile download request queue size, etc. People will then be able to tweak these values to possibly alleviate these issue or maybe even pinpointing what might be causing them.

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

MartinK | # 1599 | 2014-07-01, 21:12 | Report

Originally Posted by sixwheeledbeast View Post

I have had a few issues with modRana recently. However, this maybe because I have started trying it out a bit more.
@MartinK I have been waiting for you to appear on IRC to ask about them, but I will provide I list here so I don't forget in the meantime.

* Firstly I have experienced the above black screen issue while downloading tiles with the screen locked overnight, if you ssh to device it's the [sgx_misr] at 98% CPU bug; only a reboot will fix this AFAIK.

Yeah, unfortunately nothing much modRana can do about this other than to see if the "tweaking the knobs" proposal from the post above can influence it.

Originally Posted by sixwheeledbeast View Post

*Secondly logging tracks seems to fail after 3-4 minutes.

I will try to reproduce it & I have opened a tracking bug for it.

Originally Posted by sixwheeledbeast View Post

*Lastly tile downloads is still a PITA, I have downloaded 160km "around here", however, when I am offline sections of tiles are missing within the area I downloaded. Re-downloading says that all are found locally.

If I can be any more help debugging this let me know.

One possible cause could be old "fake" tiles. ModRana didn't check if the thing it downloaded actually is an image and it was possible to download quite a lot of "tiles" that were actually HTML error pages from the tileserver. Such "tiles" were of course downloaded successfully but would not be loadable once needed.
ModRana now check what id downloads and discards non-image tile data, but there could be still some "tiles" from before the checking was present.

So possible things to try:

in the batch download "edit" menu set "Redownload" to "ON" or "update" - this should make batch download download the tiles even if it appears to be locally available
re-/move the corresponding tile folder and see if it helps for the given region
recursively run find & file on the problematic tiles (provided you are storing tiles as files) and check their mime type, if all really are images
enable the tile debugging options mentioned two posts above and check the output for any clues

The Following 3 Users Say Thank You to MartinK For This Useful Post:
Kossuth, sixwheeledbeast, Wikiwide

petur | # 1600 | 2014-07-02, 11:02 | Report

For some reason the issue is much harder to reproduce right now

My first try did reproduce it but scroll-back buffer wasn't long enough - MEH!

Second try, redirected to logfile, couldn't reproduce but modRana just crashed. This is the output:

Code:

~ $ modrana > modrana.log
TypeError: update() takes exactly 1 argument (2 given)
modules/gui_modules/gui_gtk.py:243: Warning: g_object_unref: assertion `G_IS_OBJECT (object)' failed
  gtk.main()
python: cairo-surface.c:429: cairo_surface_destroy: Assertion `((*&(&surface->ref_count)->ref_count) > 0)' failed.
Aborted
~ $

Last edited by petur; 2014-07-02 at 11:35.

The Following 4 Users Say Thank You to petur For This Useful Post:
handaxe, Kossuth, MartinK, Wikiwide