Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Thu, 2009-03-12 at 21:41 +0100, Mathias Fröhlich wrote: On Thursday 12 March 2009 20:12:54 Geoff McLane wrote: [ win32 out of range ] But this is all with CVS of about a week ago. Over the coming days will do an update in WIN32, and see if this problem is still present (in win32)... I have fixed something in this area at the past weekend. So it would be worthwhile to verify if this problem is still present with current cvs. Mathias Hi Mathias, After quite a number of flights, various aircraft, locations, just to confirm, an updated cvs (13th) now has _NO_ - Out of range - values (in either Ubuntu or Vista)... Thanks, Geoff. -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi, On Sunday 15 March 2009 16:43:21 Geoff McLane wrote: After quite a number of flights, various aircraft, locations, just to confirm, an updated cvs (13th) now has _NO_ - Out of range - values (in either Ubuntu or Vista)... Thanks! Greetings Mathias -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi Matthias, Problem solved it looks. Finally made the Flight between Ottawa and Montreal. Cheers, Durk On Thursday 12 March 2009 18:47:04 Mathias Fröhlich wrote: Hi, I have checked in something that should fix that problem. Can you try again with current cvs? -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Thu, 2009-03-12 at 18:47 +0100, Mathias Fröhlich wrote: Hi, I have checked in something that should fix that problem. Can you try again with current cvs? Greetings Mathias Hi Mathias, GREAT STUFF! ... I made it all the way to the waterway on 108, at about 2500, running at a 'slow' 250 IAS ;=)) But only had time for one try after a SG/FG update, and full rebuild... but it certainly looks solved for Ubuntu ;=)) Re: vector 'Out of range' Sorry for the delay on the WIN32 problem, but it took time to get back into WIN32 and capture logs - see - http://geoffair.net/tmp/ufo_log3d.txt Although this is labelled 'ufo' it is actually in the CitationX, but I could NOT get it started... Maybe the reason is the nasal problem shown at the top... There are 3 other, much larger logs, in the UFO - mega bytes... http://geoffair.net/tmp/ufo_log3a.txt - this was with debug on and I even output when range is valid, so it is large. Look for the text 'Out of range' ... after it is loaded... There are also 'b' and 'c' logs... 'b' is without the good range outputs, and I 'flew' around a bit... 'c' was similar... I can not understand why this only appears to happen in WIN32!!! But this is all with CVS of about a week ago. Over the coming days will do an update in WIN32, and see if this problem is still present (in win32)... As usual out of time today... thanks for the CYOW/CitationX which seems fixed for me! Regards, Geoff. -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On mercredi 11 mars 2009, Jon S. Berndt wrote: With JSBSim aircrafts i get exactly the same problem/errors , it happens mainly when i do an airborn for long trip ( more than 1 hour) with autopilot. I got that errors with several models not only a specific one. To me, that is not new, it came up 6 or 7 month ago. -- Gérard http://pagesperso-orange.fr/GRTux/ Which version of the code (JSBSim) are you using? Jon Hello Jon, That one, which is available with FG CVS. However i don't think it is FDM related , i would think that it is loading scenery related , since i never got the problem when flying over the sea. Cheers -- Gérard http://pagesperso-orange.fr/GRTux/ J'ai décidé d'être heureux parce que c'est bon pour la santé. Voltaire -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Tuesday 10 March 2009 04:00:10 Csaba Halász wrote: On Mon, Mar 9, 2009 at 8:36 PM, Durk Talsma d.tal...@xs4all.nl wrote: - There's a VOR station located exactly 100.0nm from the startup position. If you fly straight to this VOR station (it's at the default NAV1 frequency), by following radial 108, FlightGear already crashes after approx 5 minutes. Thank you, this did the trick. Here is my analysis: [SNIP] Thanks. FWIW, it looks like a recent code change is causing or uncovering the problem: I tried the current data directory in combination with source code version of 1.9.0, and FlightGear/SimGear snapshots of Feb 1st, and March 1st. All three worked without problems, so it must be a code change from the first week of March that must be involved. I hope to narrow down the date a bit further by doing a few more test runs. Cheers, Durk -- ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Tue, 2009-03-10 at 04:00 +0100, Csaba Halász wrote: On Mon, Mar 9, 2009 at 8:36 PM, Durk Talsma d.tal...@xs4all.nl wrote: - There's a VOR station located exactly 100.0nm from the startup position. If you fly straight to this VOR station (it's at the default NAV1 frequency), by following radial 108, FlightGear already crashes after approx 5 minutes. Thank you, this did the trick. Here is my analysis: It seems you do not have to add --fdm=yasim with this a/c? Is this correct? I have not been doing this... when I tried --fdm=jsb fgfs just aborts immediately, so assume adding the fdm is not needed... Anyway, on the 108, and yes, segfault after about 5 minutes... This is with a fresh checkout of yesterday - OSG, PLIB, SG FG, and full recompile... Yesterday, did a quick 'flight' in the UFO --fdm=ufo --aircraft=ufo, but at IAS of 500-1000 on 98 - no 'CRASH' - I figure when I reach flying over/crossing the St Lawrence waterway I am safe ;=)) and I returned on the reverse track... And yesterday also did one flight in CitationX, but that was in my WIN32 Vista source of about a week ago and again reached the waterway without incident... but was using FGRUN, and had a lot of other options like --atlas=socket,out,5,MY IP,udp, no AI, no 'fog', and was a little higher, about 3,500 feet, etc... Also my WIN32 code has some protective code I _HAD_ to add that does not allow triangles( int i1, int i2 ), and other shapes, to be added if out of range of the _vertices.size()... windows vector does a range check, and aborts the application if out of range... maybe Mathias is addressing this... there are lots of ongoing fast cvs changes in this area ;=)) --- C:/FGCVS/SimGear/source/simgear/scene/model/BoundingVolumeBuildVisitor.hxx Sun Mar 01 16:40:35 2009 +++ Simgear/simgear/scene/model/BoundingVolumeBuildVisitor.hxx Sat Mar 07 15:32:58 2009 @@ -303,14 +303,26 @@ } void addLine(unsigned i1, unsigned i2) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )) +return; +#endif addLine(_vertices[i1], _vertices[i2]); } void addTriangle(unsigned i1, unsigned i2, unsigned i3) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )||( i3 = _vertices.size() )) +return; +#endif addTriangle(_vertices[i1], _vertices[i2], _vertices[i3]); } void addQuad(unsigned i1, unsigned i2, unsigned i3, unsigned i4) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )||( i3 = _vertices.size() )||( i4 = _vertices.size() )) +return; +#endif addQuad(_vertices[i1], _vertices[i2], _vertices[i3], _vertices[i4]); } I landed at CSS3, Montreal Les Cedres (I think) - boy I like a plane that babbles at you - 'bank angle', when you turn too hard, '1000 and 500 feet' when landing, etc - you feel like talking back, and saying 'thank you'! Will try adding my 'protective' code, very much needed for WIN32, without the _MSC_VER switch, and see what happens... by adding some printf() I checked and 'know' there are some calls to AddLine() and addTriangle() at least, which are out of range... and this is certainly some 'new' code. Some questions: 1. Meantime could someone help me how to run gdb with fgfs... I assume it is something like - ~/fg$ gdb --args ./fgfs --fg-root=... etc etc (gdb) run but after you see output of say - count=16 OpenAL error (AL_INVALID_VALUE): set_volume OpenAL error (AL_INVALID_VALUE): set_volume OpenAL error (AL_INVALID_VALUE): set_pitch CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.003 0 0.0145), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } count=161000 Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x7f0242c82700 (LWP 19017)] MK_VIII::TCFHandler::update_runway (this=0x9a5a378) at mk_viii.cxx:4525 4525 runway.center.latitude = _runway-latitude(); (gdb) what good things can I look at? Note I have adjusted my CullVisitor limiter code to each 1000! 2. How set the autopilot to fly this 108 radial at 2500, say 250 IAS? I have found the main [AP] switch, but where do I set the track, altitude, speed, etc ??? Is there a dialog? Or how? Regards, Geoff. -- ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Cancel the question 2 - autopilot - found the dialog, and got it all working ... should have looked harder before asking ;=)) Geoff. -- ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Ok, my 'patch' did nothing! It still CRASHED... In fact when I put in some printf() statements before the return I did not see any! Either Mathias has 'fixed' the calls which were out-of-range, or somehow vector.size() is different, or the scenery decode yields different values, printf() failed!, or... Still searching for a reason for this very repeatable segfault... I now have autopilot working well to make sure I stay on track 108, speed 250 and height 2500 from CYOW... Geoff. -- ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi, On Tuesday 10 March 2009 14:04:47 Geoff McLane wrote: Also my WIN32 code has some protective code I _HAD_ to add that does not allow triangles( int i1, int i2 ), and other shapes, to be added if out of range of the _vertices.size()... windows vector does a range check, and aborts the application if out of range... maybe Mathias is addressing this... there are lots of ongoing fast cvs changes in this area ;=)) --- C:/FGCVS/SimGear/source/simgear/scene/model/BoundingVolumeBuildVisitor.hxx Sun Mar 01 16:40:35 2009 +++ Simgear/simgear/scene/model/BoundingVolumeBuildVisitor.hxx Sat Mar 07 15:32:58 2009 @@ -303,14 +303,26 @@ } void addLine(unsigned i1, unsigned i2) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )) +return; +#endif addLine(_vertices[i1], _vertices[i2]); } void addTriangle(unsigned i1, unsigned i2, unsigned i3) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )||( i3 = _vertices.size() )) +return; +#endif addTriangle(_vertices[i1], _vertices[i2], _vertices[i3]); } void addQuad(unsigned i1, unsigned i2, unsigned i3, unsigned i4) { +#if defined(_MSC_VER) +if(( i1 = _vertices.size() )||( i2 = _vertices.size() )||( i3 = _vertices.size() )||( i4 = _vertices.size() )) +return; +#endif addQuad(_vertices[i1], _vertices[i2], _vertices[i3], _vertices[i4]); } I cannot (yet?) reproduce such kind of problems. So: Can you provide me with a backtrace when this kind of code triggers. Especially the name of the model that is loaded when this happens would be interesting. Greetings and thanks Mathias -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Tue, Mar 10, 2009 at 10:44 PM, Durk Talsma d.tal...@xs4all.nl wrote: On Tuesday 10 March 2009 14:04:47 Geoff McLane wrote: 1. Meantime could someone help me how to run gdb with fgfs... I assume it is something like - ~/fg$ gdb --args ./fgfs --fg-root=... etc etc gdb doesn't accept command line options for the program you're debugging. Those are specified together with the run (r) command. Recent gdb does, exactly as Geoff wrote, with --args. -- Csaba/Jester -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
* Durk Talsma -- Tuesday 10 March 2009: On Tuesday 10 March 2009 14:04:47 Geoff McLane wrote: ~/fg$ gdb --args ./fgfs --fg-root=... etc etc (gdb) run gdb doesn't accept command line options for the program you're debugging. Yes, it does. Geoff's line was correct. :-) m. -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Found the problem with the DC6 , so apperently not related You can't set turbo-lag to 0.0 ;) On Tue, Mar 10, 2009 at 5:00 PM, syd adams adams@gmail.com wrote: Thanks , I was trying to figure out how to use gdb too :). I had a few errors here , maybe related . While testing the P47 , I took off from KSFO and headed towards the Nimitz.At about the coastline , altitude about 2000 ft every thing paused... I thought I accidentally hit p . Looking through the property tree I noticed sim/crashed was true.I haven't been able to reproduce this, though. Second problem I ran into last night was starting at any airport with the DC-6 , everything was fine until I started the engines ... then the screen went black , dropped to 1 fps , and a steady stream of: CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.002885 0.0187575 -0.0095325), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } I kind of suspect with the second problem that it might be a badly configured yasim file... I had the same problem long ago with gear values... The jsbsim version flew fine , no problems. I'll tweak it some more and see what happens. Cheers On Tue, Mar 10, 2009 at 3:07 PM, Durk Talsma d.tal...@xs4all.nl wrote: On Tuesday 10 March 2009 22:52:47 Csaba Halász wrote: Recent gdb does, exactly as Geoff wrote, with --args. Ah, well, it's been approx since 1996 since I last read gdb documentation... Cheers, Durk -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On mercredi 11 mars 2009, syd adams wrote: Thanks , I was trying to figure out how to use gdb too :). I had a few errors here , maybe related . While testing the P47 , I took off from KSFO and headed towards the Nimitz.At about the coastline , altitude about 2000 ft every thing paused... I thought I accidentally hit p . Looking through the property tree I noticed sim/crashed was true.I haven't been able to reproduce this, though. Second problem I ran into last night was starting at any airport with the DC-6 , everything was fine until I started the engines ... then the screen went black , dropped to 1 fps , and a steady stream of: CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.002885 0.0187575 -0.0095325), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } I kind of suspect with the second problem that it might be a badly configured yasim file... I had the same problem long ago with gear values... The jsbsim version flew fine , no problems. I'll tweak it some more and see what happens. Cheers With JSBSim aircrafts i get exactly the same problem/errors , it happens mainly when i do an airborn for long trip ( more than 1 hour) with autopilot. I got that errors with several models not only a specific one. To me, that is not new, it came up 6 or 7 month ago. On Tue, Mar 10, 2009 at 3:07 PM, Durk Talsma d.tal...@xs4all.nl wrote: On Tuesday 10 March 2009 22:52:47 Csaba Halász wrote: Recent gdb does, exactly as Geoff wrote, with --args. Ah, well, it's been approx since 1996 since I last read gdb documentation... Cheers, Durk -- Gérard http://pagesperso-orange.fr/GRTux/ J'ai décidé d'être heureux parce que c'est bon pour la santé. Voltaire -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Thanks , I was trying to figure out how to use gdb too :). I had a few errors here , maybe related . While testing the P47 , I took off from KSFO and headed towards the Nimitz.At about the coastline , altitude about 2000 ft every thing paused... I thought I accidentally hit p . Looking through the property tree I noticed sim/crashed was true.I haven't been able to reproduce this, though. Second problem I ran into last night was starting at any airport with the DC-6 , everything was fine until I started the engines ... then the screen went black , dropped to 1 fps , and a steady stream of: CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.002885 0.0187575 -0.0095325), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } I kind of suspect with the second problem that it might be a badly configured yasim file... I had the same problem long ago with gear values... The jsbsim version flew fine , no problems. I'll tweak it some more and see what happens. Cheers On Tue, Mar 10, 2009 at 3:07 PM, Durk Talsma d.tal...@xs4all.nl wrote: On Tuesday 10 March 2009 22:52:47 Csaba Halász wrote: Recent gdb does, exactly as Geoff wrote, with --args. Ah, well, it's been approx since 1996 since I last read gdb documentation... Cheers, Durk -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
With JSBSim aircrafts i get exactly the same problem/errors , it happens mainly when i do an airborn for long trip ( more than 1 hour) with autopilot. I got that errors with several models not only a specific one. To me, that is not new, it came up 6 or 7 month ago. -- Gérard http://pagesperso-orange.fr/GRTux/ Which version of the code (JSBSim) are you using? Jon -- Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi Geoff, On Sunday 08 March 2009 20:56:52 Geoff McLane wrote: Yep! Tried ... --aircraft=CitationX --airport=CYOW --runway=25 --fg-scenery= $HOME/Scenery-1.0.1 Thanks for you report. Good to know, I'm not the only one, at that adds to the robustness of the crash. I think it's too early to draw any firm conclusions, but my impression is that it's more than just the code on mk_viii that is blowing. The only thing I tentatively conclude is that at this location something is going terribly wrong in updating FlightGear's position, as judged by: - The position string sent to terrasync - Local time being NaN (zapping to a non-existent time zone?) - A host of Culling errors. (position so far away that depth becomes NaN)? Next step is to check whether other aircraft produce the same crash, and whether the FDM (i.e. non-yasim) aircraft produce the same problem. Cheers, Durk -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On lundi 09 mars 2009, Durk Talsma wrote: Hi Geoff, On Sunday 08 March 2009 20:56:52 Geoff McLane wrote: Yep! Tried ... --aircraft=CitationX --airport=CYOW --runway=25 --fg-scenery= $HOME/Scenery-1.0.1 Thanks for you report. Good to know, I'm not the only one, at that adds to the robustness of the crash. I think it's too early to draw any firm conclusions, but my impression is that it's more than just the code on mk_viii that is blowing. The only thing I tentatively conclude is that at this location something is going terribly wrong in updating FlightGear's position, as judged by: - The position string sent to terrasync - Local time being NaN (zapping to a non-existent time zone?) - A host of Culling errors. (position so far away that depth becomes NaN)? Next step is to check whether other aircraft produce the same crash, and whether the FDM (i.e. non-yasim) aircraft produce the same problem. Yes with JSBSim there is the same NAN error , more or les randomly. Difficult to catch a specific cause. Cheers, Durk -- Gérard http://pagesperso-orange.fr/GRTux/ J'ai décidé d'être heureux parce que c'est bon pour la santé. Voltaire -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi Gerard, On Monday 09 March 2009 12:06:58 gerard robin wrote: Yes with JSBSim there is the same NAN error , more or les randomly. Difficult to catch a specific cause. Do you mean that you can replicate the specific bug I reported, or that you experience random NAN errors? Cheers, Durk -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Monday 09 March 2009 07:29:47 Durk Talsma wrote: Hi Geoff, On Sunday 08 March 2009 20:56:52 Geoff McLane wrote: Yep! Tried ... --aircraft=CitationX --airport=CYOW --runway=25 --fg-scenery= $HOME/Scenery-1.0.1 Thanks for you report. Good to know, I'm not the only one, at that adds to the robustness of the crash. I think it's too early to draw any firm conclusions, but my impression is that it's more than just the code on mk_viii that is blowing. The only thing I tentatively conclude is that at this location something is going terribly wrong in updating FlightGear's position, as judged by: - The position string sent to terrasync - Local time being NaN (zapping to a non-existent time zone?) - A host of Culling errors. (position so far away that depth becomes NaN)? Next step is to check whether other aircraft produce the same crash, and whether the FDM (i.e. non-yasim) aircraft produce the same problem. Okay, just to follow-up: I've done some additional testing: - The c172, and the Seneca (both jsbsim) don't seem to crash, although I need to test this a little further. - The CitationX and the 777-200ER (both yasim) cause the segfault. - There's a VOR station located exactly 100.0nm from the startup position. If you fly straight to this VOR station (it's at the default NAV1 frequency), by following radial 108, FlightGear already crashes after approx 5 minutes. - Using FlightGear 1.9.0, in combination with the current base package does not crash. So, my suspicion is that a relatively new code change, is affecting yasim aircraft, but not JSBsim. I will try to checkout earlier versions of FlightGear, and step through various versions of the code to see where things go wrong. If anybody would like to help, please be my guest. Cheers, Durk -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On lundi 09 mars 2009, Durk Talsma wrote: Hi Gerard, On Monday 09 March 2009 12:06:58 gerard robin wrote: Yes with JSBSim there is the same NAN error , more or les randomly. Difficult to catch a specific cause. Do you mean that you can replicate the specific bug I reported, or that you experience random NAN errors? Cheers, Durk Only random NAN errors. To be more precise , when it comes up, FG does not crash. Only getting on the console the NAN messages infinitely, with a sudden slow down of the FPS to 1 . The only way to stop it, is to cancel FG. I did not tried to replicate the specific bug. -- Gérard http://pagesperso-orange.fr/GRTux/ J'ai décidé d'être heureux parce que c'est bon pour la santé. Voltaire -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On Mon, Mar 9, 2009 at 8:36 PM, Durk Talsma d.tal...@xs4all.nl wrote: - There's a VOR station located exactly 100.0nm from the startup position. If you fly straight to this VOR station (it's at the default NAV1 frequency), by following radial 108, FlightGear already crashes after approx 5 minutes. Thank you, this did the trick. Here is my analysis: 0x006b596f in yasim::Gear::calcForce (this=0x7f6b6c031a60, body=0xabbd2c8, s=0x7fff8136aa40, v=0x7fff8136a8e0, rot=0x7fff8136a8f0) at src/FDM/YASim/Gear.cpp:335 335 _frac = a/(a-b); (gdb) p a $1 = -0.0118556526 (gdb) p b $2 = -0.0118556526 For whatever reason, a and b are equal so this is a division by zero. Note, if I understand it right, these numbers are supposed to be distances AGL (sign inverted), for which they are suspiciously small. Possibly a problem with the new groundcache maybe? (gdb) fr 1 #1 0x006c39eb in yasim::Model::calcForces (this=0xabbd238, s=0x7fff8136aa40) at src/FDM/YASim/Model.cpp:485 485 g-calcForce(_body, s, lv, lrot); (gdb) list 480 // The landing gear 481 for(i=0; i_gears.size(); i++) { 482 float force[3], contact[3]; 483 Gear* g = (Gear*)_gears.get(i); 484 485 g-calcForce(_body, s, lv, lrot); 486 g-getForce(force, contact); 487 _body.addForce(contact, force); 488 } 489 (gdb) p _gears $6 = {_nelem = 31, _sz = 16, _array = 0x7f6b6c0445f0} It has 16 gears, because yasim creates fake gear for contact points, similar to jsbsim. The crash occurred for i=4, if it matters. The trivial fix is to add a check for a==b of course, but I hope somebody (Andy?) has a deeper insight into the cause of the problem, so it can be properly fixed. To let Syd off the hook, maybe somebody could check with a different yasim aircraft, Helijah made a couple of those too :D -- Csaba/Jester PS: Yeah, I know, patches are more welcome than bug reports :P -- ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On 7 Mar 2009, at 09:40, James Turner wrote: start FlightGear using fgfs --aircraft=CitationX --airport=CYOW -- runway=25 take off, and make a left turn to heading 98, while maintaining an IAS of approx 250 kts, and an altitude of approx 2500 ft. Maintain this heading for approx 13 minutes and FlightGear will seg fault: Normally, these crashes are not related to the CullVisitor, it just happens to be a noisy part of the code, so it shows up in logs. The crash is almost certainly my fault, will take a look (probably tomorrow, today is my last day of I'm unable to reproduce the crash using these steps - can anyone else? James -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Yep! Tried ... --aircraft=CitationX --airport=CYOW --runway=25 --fg-scenery= $HOME/Scenery-1.0.1 Take off, wheels up, and turn to 98, flying at about 2500, at about 260 IAS ... about 10 minutes or so and BANG! As reported had hundreds of ... CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.006986 0.000522999 0.089738), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } For for the next run I added --log-level=debug == and got last output of ... CullVisitor::apply(Geode) detected NaN, depth=nan, center=(0.006986 0.000522999 0.089738), matrix={ nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan nan } Loading tile 1728984 Trying /home/geoff/Scenery-1.0.1/w080n40/w075n45/1728984 Trying /home/geoff/Scenery-1.0.1/Terrain/w080n40/w075n45/1728984 OBJECT_BASE 1728984.btg Trying /home/geoff/Scenery-1.0.1/Objects/w080n40/w075n45/1728984 Running MaiOBn LJECT_SHARoop === ED M odels/Communications/Updatradio-mediuing time Cm.xml lon=urrent Unix calenda-74.7694r time = 123653lat=79345.4060 warp4 e = 0 lev= C-20.47 urrent hdg=GMT = 3/1808/2009 18:45:30 CurrenCreat Unixtin cag alend new buffer ofar tim size =e = 123653 37930 warp = 0 2768 Current GMT = 3/8/2009 18:45: 30 Current Julian Date = 2.4549e+06 COURSE: GMT = 2/8/109 18:45:30 March 21 noon (GMT) = 1237636800 Time since 3/21/109 GMT = -12.7184 days = -12 hours = 18.7583 lon = 0 lst = 5.95833 COURSE: GMT = 2/8/109 18:45:30 March 21 noon (GMT) = 1237636800 Time since 3/21/109 GMT = -12.7184 days = -12 hours = 18.7583 lon = nan lst = Creanantin Currg aent l neon=0.00w b Siuffderer ealof Tisizme e == 32768 5.86471 gst = C245.8rea65 Currtinentg a LOCA neL Sidew buffer ofreal Ti sime = ze = 32768 nan (nan) (diff = -0.0936229) Elapsed time interval is = 506127, previous remainder Creatingis a = 6283 new buffer of size = 3-- F276ram8 e rate is = 72 Model iterations needed = 61, new remainder = 4077 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Defering boundingvolume tree built for /home/geoff/Scenery-1.0.1/Terrain/w080n40/w075n45/1728984.btg to parent. Got cached model /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac Defering boundingvolume tree built for /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac to parent. Defering boundingvolume tree built for Models/Communications/radio-medium.xml to parent. Building boundingvolume tree for 17289findbyF84.sreqtg. 379 size 33 Deleting a sample In memory sounds sample Loading tile 1728992 Trying /home/geoff/Scenery-1.0.1/w080n40/w075n45/1728992 Trying /home/geoff/Scenery-1.0.1/Terrain/w080n40/w075n45/1728992 OBJECT_BASE 1728992.btg Trying /home/geoff/Scenery-1.0.1/Objects/w080n40/w075n45/1728992 OBJECT_SHARED Models/Communications/radio-medium.xml lon=-74.9431 lat=45.5417 elev=-32.36 hdg=180 OBJECT_SHARED Models/Communications/radio-medium.xml lon=-74.9542 lat=45.5769 elev=-17.42 hdg=180 OBJECT_SHARED Models/Communications/radio-medium.xml lon=-74.9419 lat=45.5419 elev=-13.77 hdg=180 OBJECT_SHARED Models/Communications/radio-medium.xml lon=-74.9389 lat=45.5408 elev=-24.74 hdg=180 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Creating a new buffer of size = 32768 Defering boundingvolume tree built for /home/geoff/Scenery-1.0.1/Terrain/w080n40/w075n45/1728992.btg to parent. Got cached model /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac Defering boundingvolume tree built for /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac to parent. Defering boundingvolume tree built for Models/Communications/radio-medium.xml to parent. Got cached model /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac Defering boundingvolume tree built for /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac to parent. Defering boundingvolume tree built for Models/Communications/radio-medium.xml to parent. Got cached model /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac Defering boundingvolume tree built for /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac to parent. Defering boundingvolume tree built for Models/Communications/radio-medium.xml to parent. Got cached model /home/geoff/fg/fgfs/data/Models/Communications/radio-medium.ac Defering boundingvolume tree built for
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
On 7 Mar 2009, at 09:12, Durk Talsma wrote: Given that the actual error was preceded by the CullVisitor nan message, I hope this crash situation might also give some insights as what's going on there. Note that I am using some custom settings in my local preferences, file, but I have no reason to assume yet that these will be affecting the reproducibility of this crash. start FlightGear using fgfs --aircraft=CitationX --airport=CYOW -- runway=25 take off, and make a left turn to heading 98, while maintaining an IAS of approx 250 kts, and an altitude of approx 2500 ft. Maintain this heading for approx 13 minutes and FlightGear will seg fault: Normally, these crashes are not related to the CullVisitor, it just happens to be a noisy part of the code, so it shows up in logs. The crash is almost certainly my fault, will take a look (probably tomorrow, today is my last day of being busy). Thanks for the report. James -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] Reproducible crash in mk_viii.cxx
Hi James, On Saturday 07 March 2009 10:40:20 James Turner wrote: Normally, these crashes are not related to the CullVisitor, it just happens to be a noisy part of the code, so it shows up in logs. It may indeed not be directly related, but what I did notice every time, is that while I typically don't see, any of these CullVisitor messages, in this case there is a short burst of them (probably a few hundred; more than my terminal can retain, immediately before FlightGear crashes. This has happened during every crash situation I ran. Also believe that they're probably not directly related, but I do think the current error blows up the math somewhere that has ramifications on other systems. As such, I hope this is a test case that can provide some additional insights into the dynamics of the process causing this warning in other situations. The crash is almost certainly my fault, will take a look (probably tomorrow, today is my last day of being busy). Thanks! Thanks for the report. You're welcome. Cheers, Durk -- Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise -Strategies to boost innovation and cut costs with open source participation -Receive a $600 discount off the registration fee with the source code: SFAD http://p.sf.net/sfu/XcvMzF8H___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel