Board index FlightGear Development

FlightGear 3.x broken under Fedora 21

FlightGear is opensource, so you can be the developer. In the need for help on anything? We are here to help you.
Forum rules
Core development is discussed on the official FlightGear-Devel development mailing list.

Bugs can be reported in the bug tracker.

FlightGear 3.x broken under Fedora 21

Postby tempest766 » Wed Dec 10, 2014 2:05 pm

The crash is as follows:

[root@localhost ~]# fgfs
Enabling ATI viewport hack
process 5523: arguments to dbus_connection_set_exit_on_disconnect() were incorrect, assertion "connection != NULL" failed in file db
us-connection.c line 3121.
This is normally a bug in some application using the D-Bus library.
D-Bus not built with -rdynamic so unable to print a backtrace
Aborted (core dumped)
[root@localhost ~]#


SMP Westmere machine
GTX460 GPU
24GB RAM
NVIDIA-Linux-x86_64-340.24 GPU driver (but crash also occurs under mesa rendering)
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby PH-JAKE » Wed Dec 10, 2014 11:52 pm

First off; Why run fgfs as root? Get yourself a user account, that's a good start.
PH-JAKE
 
Posts: 144
Joined: Tue Mar 11, 2014 11:53 pm
Callsign: PH-JAKE
Version: 2020.1.3
OS: Debian bullseye

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Thu Dec 11, 2014 3:36 am

How do you know that I was running it as root? Perhaps I just like the # prompt. :^P

the user I ran the program under has absolutely no bearing on the problem at hand. For one thing, it's actually common that when a hardware intensive program crashes, it can be a problem with innadequate privilege level...but I doubt UID has anything to do with this particular problem, so unless you have any real insight into the problem....
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby Hyde » Thu Dec 11, 2014 3:56 am

I'm running GIT version on Fedora 21 and using fgo.
It works fine.
Are you using yum version (3.2)?
If so, why don't you use fgrun?
Hyde
 
Posts: 728
Joined: Fri Aug 12, 2011 1:09 pm
Location: Peachtree city, GA
Callsign: N407DS
Version: GIT
OS: Fedora 25

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Thu Dec 11, 2014 5:47 am

Crash backtrace follows:

The crash appears to be in how fg is communicating with the d-bus layer to turn off the screensave, which I don't have enabled in the first place.

(gdb) bt
#0 0x00007f1e598268c7 in __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:55
#1 0x00007f1e5982852a in __GI_abort () at abort.c:89
#2 0x00007f1e5ea44824 in _dbus_abort () at dbus-sysdeps.c:91
#3 0x00007f1e5ea3ac8e in _dbus_warn_check_failed (
format=0x7f1e5ea4ab00 "arguments to %s() were incorrect, assertion \"%s\" failed in file %s line %d.\nThis is normally a bug in
some application using the D-Bus library.\n") at dbus-internals.c:275
#4 0x000000000066fd4f in fgOSDisableScreensaver () at /usr/src/debug/flightgear-3.2.0/src/Main/screensaver_control.cxx:50
#5 0x00000000006510c9 in fgMainInit (argc=2, argv=0x7fff30871428) at /usr/src/debug/flightgear-3.2.0/src/Main/main.cxx:490
#6 0x0000000000604a82 in main (argc=2, argv=0x7fff30871428) at /usr/src/debug/flightgear-3.2.0/src/Main/bootstrap.cxx:234
(gdb) q


re: fgrun....maybe because there is no fgrun, and fgfs was the starting command in 2.x
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Sun Dec 14, 2014 9:25 am

Ok. so after updating some patched fedora libaries that were pushed out, the program makes it as far as some OSG code. The graphics screen comes up, says "loading scenery", but never makes it to the cockpit view.

(gdb) bt 10
#0 0x0000000000bf8464 in handle_geode_callbacks (geode=..., this=0x170ff30)
at /usr/include/osgUtil/UpdateVisitor:94
#1 osgUtil::UpdateVisitor::apply (this=0x170ff30, node=...) at /usr/include/osgUtil/UpdateVisitor:51
#2 0x00007fcffab39a40 in simgear::EffectGeode::accept (this=0x7fcf882fe160, nv=...)
at /usr/src/debug/simgear-3.2.0/simgear/scene/material/EffectGeode.hxx:32
#3 0x00007fcff8c86f83 in osg::Sequence::accept (this=0x7fcf9399f9b0, nv=...)
at /usr/src/debug/OpenSceneGraph-3.2.1/OpenSceneGraph-3.2.1/include/osg/Sequence:34
#4 0x00007fcff8c1f3a3 in osg::Group::traverse (this=0x7fcf93bc4ea0, nv=...)
at /usr/src/debug/OpenSceneGraph-3.2.1/OpenSceneGraph-3.2.1/src/osg/Group.cpp:62
#5 0x00007fcff9907980 in osg::Group::accept (this=0x7fcf93bc4ea0, nv=...)
at /usr/src/debug/OpenSceneGraph-3.2.1/OpenSceneGraph-3.2.1/include/osg/Group:38
#6 0x00007fcff8c1f3a3 in osg::Group::traverse (this=0x7fcf93bbee70, nv=...)
at /usr/src/debug/OpenSceneGraph-3.2.1/OpenSceneGraph-3.2.1/src/osg/Group.cpp:62
#7 0x0000000000bfa31d in traverse (node=..., this=0x170ff30) at /usr/include/osg/NodeVisitor:193
#8 handle_callbacks_and_traverse (node=..., this=0x170ff30) at /usr/include/osgUtil/UpdateVisitor:86
#9 apply (node=..., this=0x170ff30) at /usr/include/osgUtil/UpdateVisitor:57
(More stack frames follow...)
(gdb)


program still crashes, just not so quickly.

Also looks like maybe the OSG stuff isn't including valid debug source with the debug-info packages, as is evident by the filenames with no extension in the backtrace....

FYI: Yes. FG 3.2 as shipped in Fedora 21.
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby onox » Sun Dec 14, 2014 10:10 am

tempest766 wrote in Thu Dec 11, 2014 3:36 am:For one thing, it's actually common that when a hardware intensive program crashes, it can be a problem with innadequate privilege level..


BS. You need root privileges if you want to load kernel modules, write bytes directly to your hard drive, put your network card in promiscuous mode, or run a server at a low port. FlightGear is not that kind of application. No need to run it as root. I don't think you are fully aware of what you were doing. It's really really stupid, so don't do that ever again. Just warning you.

About the loading scenery thing: maybe you have this issue: https://code.google.com/p/flightgear-bu ... il?id=1470
onox
Retired
 
Posts: 431
Joined: Fri Jun 20, 2014 2:45 pm

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Sun Dec 14, 2014 1:33 pm

onox wrote in Sun Dec 14, 2014 10:10 am:I don't think you are fully aware of what you were doing. It's really really stupid, so don't do that ever again. Just warning you.

About the loading scenery thing: maybe you have this issue: https://code.google.com/p/flightgear-bu ... il?id=1470


isn't worth my time or effort debating my testing procedures online, and I'm not too concerned whether I trash a cloned VirtualBox VM image running Fedora 21 while testing out new apps with enhanced privilege. Nope. I have no idea what I'm doing ;^P

Anyway, warning received, and in the context of my test environment, given all DUE consideration. The backtrace from the scenery bug is way different than the backtrace I've attached above. My crash is clearly occuring in the OSG scene building code, not in the scenery loading code that the other post references. There are a bunch of (non-pure) virtuals in the OSG code and one is generating a SIGSEGV trap. Something either isn't being initialized, or an invalid pointer is trying to access it shouldn't.
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby onox » Sun Dec 14, 2014 7:46 pm

tempest766 wrote in Sun Dec 14, 2014 1:33 pm:I'm not too concerned whether I trash a cloned VirtualBox VM image running Fedora 21 while testing out new apps with enhanced privilege.


Okay, a VM changes things a bit. Still no need to run fgfs as root because it's not gonna give you any benefit.

tempest766 wrote in Sun Dec 14, 2014 1:33 pm:My crash is clearly occuring in the OSG scene building code, not in the scenery loading code that the other post references. There are a bunch of (non-pure) virtuals in the OSG code and one is generating a SIGSEGV trap. Something either isn't being initialized, or an invalid pointer is trying to access it shouldn't.


Have you tried --prop:/sim/rendering/terrain/simplifier/enabled-far=false?
onox
Retired
 
Posts: 431
Joined: Fri Jun 20, 2014 2:45 pm

Re: FlightGear 3.x broken under Fedora 21

Postby PH-JAKE » Mon Dec 15, 2014 8:00 am

tempest766 wrote in Thu Dec 11, 2014 3:36 am:How do you know that I was running it as root? Perhaps I just like the # prompt. :^P

The hash in the prompt is fine, root@localhost was somewhat of a giveaway :wink:

the user I ran the program under has absolutely no bearing on the problem at hand. For one thing, it's actually common that when a hardware intensive program crashes, it can be a problem with innadequate privilege level...but I doubt UID has anything to do with this particular problem

I always advice users to stay away from root unless needed, it can do some real damage if one is not careful or unknowing. That level of insight I cannot sense from a initial posting, so I tend to err on the safe side.

, so unless you have any real insight into the problem....

Well, well, we're here to help eachother, are we?
PH-JAKE
 
Posts: 144
Joined: Tue Mar 11, 2014 11:53 pm
Callsign: PH-JAKE
Version: 2020.1.3
OS: Debian bullseye

Re: FlightGear 3.x broken under Fedora 21

Postby Johan G » Mon Dec 15, 2014 9:26 am

PH-JAKE wrote in Mon Dec 15, 2014 8:00 am:I always advice users to stay away from root unless needed, it can do some real damage if one is not careful or unknowing.

Google for "Unix horror stories" and you get enough reading for a weekend. :wink:
Low-level flying — It's all fun and games till someone looses an engine. (Paraphrased from a YouTube video)
Improving the Dassault Mirage F1 (Wiki, Forum, GitLab. Work in slow progress)
Johan G
Moderator
 
Posts: 5960
Joined: Fri Aug 06, 2010 5:33 pm
Location: Sweden
Callsign: SE-JG
IRC name: Johan_G
Version: 3.0.0
OS: Windows 7, 32 bit

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Mon Dec 15, 2014 11:52 am

onox wrote in Sun Dec 14, 2014 7:46 pm:Have you tried --prop:/sim/rendering/terrain/simplifier/enabled-far=false?


Based on the suggestion, I did try that. No difference. There is definitely some kind of memory corruption occuring in the OSG (Geode) code. The backtrace is over 30 levels deep and 28 of those calls are in OSG code. Guessing that this is well after the terrain loading stage, and occuring while it's building the initial cockpit/terrain view at startup. In F21 the only data loaded is the default dataset so the starting location should be Sanfransicso Intl. All times I tried were during nighttime hours on the left coast, using real clock time.

Looks like the same OSG version is used in both F20 and F21 releases.

Is the terrain modeling or cockpit modeling using a newer database format in FG3.2 than it was under FG2.x?
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21

Re: FlightGear 3.x broken under Fedora 21

Postby tempest766 » Mon Dec 15, 2014 12:33 pm

Further investigation confirms that my reported OSG problem is the same as that reported in

viewtopic.php?f=45&t=24172#p219489

Exactly the same backtrace from exactly the same airport. Choosing a different departure airport from the default terrain dataset does not exhibit the problem...although I haven't rigorously tested because I was running FG in a VM with a 2 to 3 FPS framerate. It will be a few weeks before I'm confident enough in F21 to migrate it to my host system disk where I have fully accelerated opengl available.

From my perspective, and given my current limited testing abilities here, the issue is closed for now.
tempest766
 
Posts: 37
Joined: Sat Nov 15, 2014 11:32 pm
Location: Pennsylvania USA
Version: git
OS: Fedora 21


Return to Development

Who is online

Users browsing this forum: No registered users and 1 guest