[Flightgear-devel] Terrasync scenery fails to download

[r***@gmail.com]

Is someone actively looking into the long standing bug whereby Terrasync stops downloading scenery during a flight? I couldn’t see an obvious bug report in the tracker to add a comment to.

I had this problem this morning for the first time in a while. It was a flight from HSSS (15*33’N, 32*33’E) to HUEN (0*2’N, 32*27’E) and I noticed I had lost scenery around 200nm from destination. Sometimes when this happens, turning Terrasync downloading off and back on again causes scenery to start downloading again, but not in this case.

The Terrasync log repeatedly logs “sync taking a long time: Objects/e030n10/e030n11 
”. Looking on the terrasync server, there is no objects directory for e0030n11: http://terrascenery.googlecode.com/svn/trunk/data/Scenery/Objects/e030n10/ <http://terrascenery.googlecode.com/svn/trunk/data/Scenery/Objects/e030n10/>. It’s not taking a long time, it doesn’t exist.

Terrasync then appears to stop downloading tiles. Turning Terrasync off causes it to generate ocean tiles, but turning it back on, even after passing south of 10* latitude, doesn’t resume the download.

I know I had a good internet connection throughout the flight because I was connected to multiplayer and the tracker is unbroken: http://mpserver15.flightgear.org/modules/fgtracker/?FUNCT=FLIGHT&FLIGHTID=6336555

I have a 500MB log file: https://dl.dropboxusercontent.com/u/2838854/temp/fgfs.log <https://dl.dropboxusercontent.com/u/2838854/temp/fgfs.log>

As usual with this problem, restarting Flightgear at the destination airport downloads scenery successfully.

Regards,

Richard

[Stefan Seifert]

Post by r***@gmail.com
Is someone actively looking into the long standing bug whereby Terrasync
stops downloading scenery during a flight? I couldn’t see an obvious bug
report in the tracker to add a comment to.

I've started enjoying the SpaceShuttle last weekend and encounter this quite
frequently. Of course in my case I switch to earth view rendering anyway. But
even when I reset after the shuttle gets out of control FlightGear will hang
sometimes trying and failing to download scenery during initialization. My
internet connection btw. is rock stable and has yet to lose any packets.

Stefan

------------------------------------------------------------------------------

[Thorsten Renk]

Post by Stefan Seifert
I've started enjoying the SpaceShuttle last weekend and encounter this quite
frequently. Of course in my case I switch to earth view rendering anyway. But
even when I reset after the shuttle gets out of control FlightGear will hang
sometimes trying and failing to download scenery during initialization. My
internet connection btw. is rock stable and has yet to lose any packets.

I don't think the data rate you can realistically get even via a good
internet connection is fast enough to cope with Mach 26. If you have 200
km visibility, you're going to ask for 1600 km^2 of new terrain every
second.

I've done some tests using the native FG terrain with the X-15, and at
large visibilities > 400 km at arc top, I start seeing problems getting
terrain on my harddisk loaded quickly enough into the renderer.

So I would suspect it's natural that the Shuttle out-runs terrasync, not
a bug.

* Thorsten

------------------------------------------------------------------------------

[Stefan Seifert]

Post by Thorsten Renk
So I would suspect it's natural that the Shuttle out-runs terrasync, not
a bug.

I can very well live with not being able to download Gigabytes of terrain just
in time during orbits, but terrasync seems to stop working completely in that
case. Even after re-entry it will not download new scenery and as I described,
even after reset, it may hang trying to sync scenery for KSFO that I already
have. That sounds definitely like a bug to me.

Stefan

------------------------------------------------------------------------------

[James Turner]

Post by r***@gmail.com
Is someone actively looking into the long standing bug whereby Terrasync stops downloading scenery during a flight? I couldn’t see an obvious bug report in the tracker to add a comment to.
I had this problem this morning for the first time in a while. It was a flight from HSSS (15*33’N, 32*33’E) to HUEN (0*2’N, 32*27’E) and I noticed I had lost scenery around 200nm from destination. Sometimes when this happens, turning Terrasync downloading off and back on again causes scenery to start downloading again, but not in this case.
The Terrasync log repeatedly logs “sync taking a long time: Objects/e030n10/e030n11 
”. Looking on the terrasync server, there is no objects directory for e0030n11: http://terrascenery.googlecode.com/svn/trunk/data/Scenery/Objects/e030n10/ <http://terrascenery.googlecode.com/svn/trunk/data/Scenery/Objects/e030n10/>. It’s not taking a long time, it doesn’t exist.
Terrasync then appears to stop downloading tiles. Turning Terrasync off causes it to generate ocean tiles, but turning it back on, even after passing south of 10* latitude, doesn’t resume the download.

That’s interesting, I assumed this bug was about HTTP download issues, not about missing tiles on the server side.

I’m currently working on an experimental change to use libcurl for HTTP transfers instead of my build-in HTTP code, since this would give a route to using HTTPS for the aircraft package system. I’m not sure if this is going to make it into Git, it’s still very much an experiment, but it will change the characteristics of the TerraSync code of course.

Kind regards,
James

[r***@gmail.com]

That’s interesting, I assumed this bug was about HTTP download issues, not about missing tiles on the server side.

I thought I was onto something whereby the problem could be duplicated but alas no. I started a flight starting just north of that tile and followed my route from yesterday (HSOB, then OBD-MLK), and it downloaded the terrain for e030n11 without complaining about there not being a corresponding objects tile.

My gut feeling is that it’s not a network issue. There appears to be a dead-end state that the combination of scenery download and tile loading gets into. Once it gets into that state, neither scenery download nor tile loading works (unless TerraSync is disabled). Restart the sim and the same scenery loads OK.

That’s consistent with what I saw yesterday. I checked the tiles I have using terramaster and I have nothing south of e030n11 (which was the tile I got stuck on yesterday but that downloaded OK on my subsequent test).

Richard

[James Turner]

Post by r***@gmail.com
I thought I was onto something whereby the problem could be duplicated but alas no. I started a flight starting just north of that tile and followed my route from yesterday (HSOB, then OBD-MLK), and it downloaded the terrain for e030n11 without complaining about there not being a corresponding objects tile.
My gut feeling is that it’s not a network issue. There appears to be a dead-end state that the combination of scenery download and tile loading gets into. Once it gets into that state, neither scenery download nor tile loading works (unless TerraSync is disabled). Restart the sim and the same scenery loads OK.
That’s consistent with what I saw yesterday. I checked the tiles I have using terramaster and I have nothing south of e030n11 (which was the tile I got stuck on yesterday but that downloaded OK on my subsequent test).

Interesting data points (which are lacking, in this area!). I /had/ a gut feeling it was a network issue, but little evidence for that. I’d be delighted to believe it’s a dead-end state the logic gets stuck in, except this raises the question of why the problem is so hard to reproduce. But it would be much easier to find a logic bug, fix it, and say ‘hurrah!’ than trying to track down intermittent network failures.

Any further research in this area is most welcome, and I’m hopeful my experimental work to change the network backend for terrasync will be functional ‘shortly’ (I’m on a business trip so having productive evenings hacking), which will give more data on my theory - providing we can find any semi-reliable way to get things stuck!

Kind regards,
James