Hi Sasha, Some answers... I hope ;=))
>> Look here to see which exact versions of >> tools I am using: Yes, that is your original post, but that does NOT tell us exactly _WHEN_ you cloned, and last UPDATED the git TG source. If I read correctly, you are using CDED ASCII DEM files, which are 0.75 arcsec. The source to do non-integer arcsec ASCII DEM files was only put in the git repository 3 DAYS ago. see http://mapserver.flightgear.org/git/gitweb.pl?p=terragear-cs;a=summary Are you using an updated source subsequent to then? If NOT then you have NO elevations in one of - SRTM2-Africa-3 SRTM2-Australia-3 SRTM2-Eurasia-3 SRTM2-Islands-3 SRTM2-North_America-3 SRTM2-South_America-3 DEM-USGS-3 SRTM-30 As you point out THESE ARE HARD CODED in genapts, so you MUST use one of these names for demchop output ;=() >> Picasso-like effects to the landscape, >> with maybe a bit of Dali thrown in.. Nice description ;=)) Does it looks something like this? http://geoffair.net/tmp/newscenery.htm#cullam1 where the airports are at sea level, and the land is at -9999... If YES, then you do NOT have any 'elevation' data in ONE of the above 'fixed' directories, so - (a) genapts puts airports at 0 (sea level), and (b) fgfs-construct puts land at -9999!!! Perhaps more Dali, than Picasso, but definitely surreal ;=)) >> 1) Does the --airport=xxxx option _work_ >> for genapts? Yes, if the 'xxxx' airport is in the apt.dat[.gz] file, this should work for ONE airport... but have never tried it... But if you are doing an 'area', then passing it --min-lon=<deg> --max-lon=<deg> --min-lat=<deg> --max-lat=<deg> is probably the best way... and it will do _ALL_ the airports in that range... I know this works fine... Or maybe a 10x10 degree <chunk>, but not tried. Nor have I tried the 1x1 degree <tile> option. BUT this can ONLY be done after you have VALID elevation data in ONE of the above folders... Check the elevation data: You can gzip -d one of the <work-name>/<chunk>/<tile>/<index>.arr.gz files and look at the top few lines... If you see '<num> 0 <num> 0', then this is BAD, and you need to use a 'fixed' demchop... If you do have the 'latest', then can only imagine a 'directory' naming problem... While you can use any name in fgfs-construct, genapts is FIXED... >> 2) The documentation for terrafit.py is >> sketchy, so I was looking... Use the later tool, terrafit to add <index>.fix.gz for each <index>.arr.gz... You can do the arr.gz one by one, or pass it a directory, and it will search all sub-directories... terrafit --help - shows its options... Cullam and I got good results from using --maxerror 15 rather than the default 40. Cullam tried --maxerror 5 --maxnodes 100000 but the tool seemed to fail to complete... and would be interested in reports if you get this 'working'... Using 15 m gave considerable improvement, and I would be interested if any lower number (a) worked, and (b) gave more detail... >> 3) ... genapts ... paths ... hardcoded. >> And the --min/max lat/lon options, do they >> work? Yes, and yes, as indicated above. I am sure I passed a link to my makeCYYT script - http://geoffair.net/tmp/makeCYYT from which you can get some 'help' by READING what it does with each of the tools... >> 4) Shape-decode questions.... I have not experimented too much yet with the shape-decode tool, so maybe others can assist better on this tool. So I will leave this for next time ;=)) >> 5) Construction questions: >> 5a) How do the two methods (client/server vs. >> fgfs-construct) client/server actually uses fgfs-construct... The client/server is to do large areas, and allows it to be done on separate machines, but if you are doing just your area then perhaps fgfs-construct directly is the way to go... >> 5b) with fgfs-construct, what are --x-dist and >> --y-dist used for? fgfs-construct --help, will show the parameters. To again to 'limit' the construction to a region. You give the center lon, lat, and an xdist, ydist, and it will process each side of the given center by the x,y dist... >> 5c) the server/client method doesn't >> create any particular output The client/server requires some additional setup to be done, but you should be able to directly use fgfs-construct as suggested above... Maybe others who have used the client/server could give more detailed setup for using these... Yes, there are quite a number of seemingly 'conflicting' tutorials, but it is usually due to their 'age'. The tools changed a little over time, but the 'concepts' remain sound. Hope I answered everything ;=)), and that this helps... Regards, Geoff. On Tue, 2009-04-21 at 02:43 -0300, Sasha V. Alexandr wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hello folks, > > OK, I have been meaning to subscribe and still haven't, because I > have been keeping _SO_ busy working on getting this scenery-making > stuff to work for me. > I have probably a fair number of questions about how to fix what's > going on here, but I will start with a few, and maybe those few > will lead to the answers to some of the others. I have put off > writing this email for some time now while searching for answers, > but I must ask some stuff now, as there are too many conflicting > sources of documentation & tutorials online.. > > Look here to see which exact versions of tools I am using: > http://www.mail-archive.com/flightgear- > us...@lists.sourceforge.net/msg06628.html > > The tools all appear to be working very nicely in and of > themselves: demchop, terrafit, genapts, shape-decode, fgfs- > construct. > Some are picky about paths, but that can be fixed at some point > hopefully (genapts in particular-- hardcoded paths are never good > ;). > Each tool is working as it should, best I can tell, but the > finished product when I go for a test-flight, even though the > scenery itself is GREAT LOOKING :) there are some Picasso-like > effects to the landscape, with maybe a bit of Dali thrown in.. > > 1) My airport is basically non-existent. Even with the default > FlightGear scenery package, I get just a couple strips of pavement > for a runway, but I am not even getting that with my scenery. > 2) Bodies of water appear to be at sea-level, but the landmass is > up to 30,000 feet below sealevel in some places-- giant holes, huge > ones, with sheer cliff-like walls. The cliff-like walls do have > landscape on them, and trees, but they are ridiculously steep. > There are even some urban areas clinging to the sides of these > cliffs. Sometimes the shoreline looks normal, but it just drops off > at the edge of the landmass. I can fly down to the bottom of the > holes, and there is great scenery down there, but I am WAAAY below > sealevel. > 3) some objects, like what I suspect to be my runways maybe, seem > like tall skinny 2-dimentional sticks, sticking up out of the > ground. > > Here's some more specific points and questions; > > 1) Does the --airport=xxxx option _work_ for genapts? It doesn't > seem to. Below is the last few lines of my genapts output when > using that option: > > >End of file reached > >last_apt_id.length() = 4 > >Skipping airport > >[FINISHED CORRECTLY] > >Airport generating finished. > > ..and my AirportArea folder is there, but empty after that run. > > 2) The documentation for terrafit.py is sketchy, so I was looking > in the code and the homepage given in the code, but there isn't > much there.. > So- How does the -f option work? It is for the scaling factor. The > authors homepage says that for 25 meter DEMs a factor of 0.04 is > good, but he doesn't explain WHY and how do determine what the > scale factor should be for other resolutions. I am using 10 meter > DEMs, so what should my factor be? I have tried 0.04, 1.0, and 0.1 > but am not sure what I should be doing here. > > Here's some other terrafit options: > - --minnodes 50 (so what is a 'node' and how does more or less nodes > affect the outcome?) > - --maxnodes 1000 (same question here) > - --maxerror 40 (and this one? I have seen a few errors > 40 go by on > my screen, but the program didn't stop.) > > 3) As mentioned earlier, genapts is tricky to use. Keep your SCROLL- > LOCK button handy as you run it so you can pause execution and look > as it searches paths. The paths in it do appear to be hardcoded. > And the --min/max lat/lon options, do they work? I read either in > the mailing list or on a tutorial (possibly a very old tutorial) > that they aren't doing anything. > > 4) Shape-decode questions.... > 4a) --max-segment: does this determine the longest length that any > straight line will be on the landscape? > 4b) I copied some code out of a guys post on the forums, for shape- > decoding all the v0_shapefiles available for my area from the > FlightGear mapserver, but there are two more shapefiles that he > didn't use: I have v0_depthcontour and v0_elevcontour, which > definitely sound like they have something to do with > hills/valleys/slopes though I know there are no areas of Nova > Scotia that are 30,000 feet high :D but anyone know what sort of > shape-decode I might try on these files? > 4c) can screwed-up shapefiles (or empty ones) cause holes?? Or, > where/why are holes formed initially (during demchop?) and do they > just persist from that point until the build is done? > > 5) Construction questions: > 5a) How do the two methods (client/server vs. fgfs-construct) > compare? Is one deprecated? It seems that the most recent web > searches I have read are using the fgfs-construct method, but > what's the difference? > 5b) with fgfs-construct, what are --x-dist and --y-dist used for? > The examples I read use a value of 3 so is that because the example > was using Level-3 DEMs? I have it set to 1 but tried it at 3 and > didn't notice a difference... > 5c) the server/client method doesn't create any particular output > folder.. Assuming it worked, what folder would I specify to test my > scenery? The $WORK directory? > > > OK, well the server/client just finished so I am going to go and > see if/what I have for scenery this time :) and will check back > later/tomorrow. Thanks very much for reading, and for any feedback > anyone can provide! > > Sasha > ------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Flightgear-users mailing list Flightgear-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-users
