#920 Scenery download stalled, too many errors reported

[Anonymous]

Originally created by: g.a.br...@gmail.com

*What steps will reproduce the problem?*
1. start fgfs (any plane, any location)
2. errors message on screen "Scenery download stalled. Too many errors reported. See log output

*What is the expected output? What do you see instead?*

Scenery refreshing should occur. Instead scenery isn't refreshed at all.

*Any output in the console (black window)?*

Starting automatic scenery download/synchronization. Using built-in SVN support. Directory: '/home/terrasync'.
Failed to synchronize directory 'Models', Couldn't perform atomic initialization (code 200029).
Failed to synchronize directory 'Airports/L', Couldn't perform atomic initialization (code 200029).
Failed to synchronize directory 'Airports/K', Couldn't perform atomic initialization (code 200029).
Failed to synchronize directory 'Airports/J', Couldn't perform atomic initialization (code 200029).
Failed to synchronize directory 'Airports/I', Couldn't perform atomic initialization (code 200029).
Failed to synchronize directory 'Airports/H', Couldn't perform atomic initialization (code 200029).
...

*What FlightGear version are you using (when using GIT version, please
mention date)?*

/home/flightgear# git log -1
commit [r4a535f67c042241a2f7b7913ec92bc058956090e]
Author: ThorstenB <brehmt@gmail.com>
Date:   Thu Nov 1 19:34:27 2012 +0100

*What operating system and graphics card?*

Linux 3.6.4-1-ARCH x86_64
NVIDIA GPU GeForce GT 220

[Anonymous]

Originally posted by: bre...@gmail.com

Unfortunately this is a SVN issue which was apparently introduced with svn 1.6 and fixed with a later version. There are many discussion threads and also workarounds for this issue - but nothing that we could do about FlightGear.

https://www.google.com/search?q=svn+%22Couldn%27t+perform+atomic+initialization

Status: WontFix

[Anonymous]

Originally posted by: g.a.br...@gmail.com

/home# svn version
svn, version 1.7.7 ([r1393599])

[Anonymous]

Originally posted by: g.a.br...@gmail.com

/home/terrasync/Airports/K# svn update
Updating '.':
At [r20651].

so it works by hand.

[Anonymous]

Originally posted by: g.a.br...@gmail.com

As i'm bisecting the git repository to tackle another issue (#921) I'm also testing the built-in SVN scenary download, so I might identify a precise commit which has introduced the SVN errors. I'm sure these errors never appeared with 2.8.0 release of FlightGear, so I'm still not convinced these errors are on the SVN server side.

[Anonymous]

Originally posted by: g.a.br...@gmail.com

# bad: [[r9b900e94304b95337e2731946525cde4ef377da9]] Implement a persistent cache for navigation data.
git bisect bad [r9b900e94304b95337e2731946525cde4ef377da9]

This is the commit which introduced the issue.

[Anonymous]

Originally posted by: g.a.br...@gmail.com

Reverting to subversion 1.7.6 seems ok, at least it works again with commit 9b900e94. I'll give a try later today with the HEAD of git/next to confirm it's subversion lib which breaks things.

[Anonymous]

Originally posted by: g.a.br...@gmail.com

The issue disappeared with subversion 1.7.6, so it's either a subversion issue or a packaging issue. Won't fix was right :-) Sorry for the noise?

[Anonymous]

Originally posted by: bre...@gmail.com

No problem. The bits I found on the net suggested it was an issue of "libsvn beeing to paranoid when checking its dependencies".

[Anonymous]

Originally posted by: bmbr...@gmail.com

I also add this issue with recent git version on a Linux box.

After *a lot* of debugging, I found that that the underlying error was svn complaining that it had been compiled for the system version of sqlite3 but was running with an older version.  ldd showed that fgfs was pulling in the system version of the library. So why the error?

Well, the "older" version of sqlite3 mentioned above is part of the source in flightgear/src/Navaids/.  It is compiled in *unless* you specify an option to cmake when you configure your build (add -DSYSTEM_SQLITE=on to the cmake command).

So my executable had parts compiled against two different versions of sqlite3, had one version statically linked in, and was dynamically linking the other (apparently to no good).

Rebuilding flightgear to use the system sqlite3 fixed the issue for me.

But there is still an issue.  I should get a system built with either the system sqlite or the bundled sqlite, not some broken mixture with both.