Actually, I think I've told you everything - perhaps I need to be more clear:
- using geoserver 1.7.x from SVN
- add a new polygonal shapefile data source and featureType (I used a shapfile containing the countries of the world)
- select the "polygon" style
- click to have it automatically generate the extents
- leave everything else default on the configuration
- click on the kml link for the newly created featuretype in the Demo-> Map Preview section
- the layer is added to Google Earth, I see it and it takes some time to load a large hierarchy of 00/01/11/10 folders
- it appears that there are no features in any of the folders (there are too many folders to check them all but I've checked a lot of them several times now), hence nothing is displayed in the viewer
- if I specify superoverlay_mode=raster or hybrid, the raster version loads into GE and displays correctly, but no vectors are displayed
- if I specify mode=refresh, the vectors load properly but they are (as expected) not using the superoverlay hierarchical structure
I wondered how such a simple thing could not work, so I tried the existing polygonal shapefile layers that come pre-configured, and of course they all work as expected. Then I tried to compare the seemingly relevant parameters (regionating strategy, regionating attribute, and kml feature limit) from the working built-in layers with my newly added layer - only the kml feature limit was different and I tried -1, 0, 1, 10, 100, all with the same result.
There are no exceptions output or logged, in fact the only message logged at all at production logging settings is a single line: "class org.vfny.geoserver.global.GeoServerFeatureLocking" which seems to be output anytime I access any layer.
So I thought, why don't I try following my own instructions to verify that this problem is as easy to replicate as it was to create in the first place. So I added a polygon grid coverage of the province of BC, and it actually works fine. So now I just removed my original (non-working) featureType and data source, and re-added it, and it works - the superoverlay kml displays in GE. Determined to figure out what caused the original problem, I mess around with various parameters until I figure out how to "break" it again. I have determined that if I give the layer the alias "country", it fails to display, but if I name it something else ("foobar"), it works fine. "country" happens to be the same name that I gave the shapefile datastore, but I tried giving another layer the same alias as its datastore name and this doesn't cause the failure.
I find this impossible to believe, but I've tried a dozen times switching back and forth between "country" and "foobar", "blarg", etc, and when it has the alias "country" it doesn't work. Google loads the layer "successfully" - green light - no exceptions - just the layer seems to be empty and nothing is displayed - only when the layer has the alias "country".
I'm sure I sound completely irrational at this point but does anyone have any idea why this might be happening? Obviously I have already figured out a work-around, but the whole point of the exercise was to ensure that I could make a vector superoverlay work, so that I had some basis for figuring out why my SDE-sourced vector superoverlay isn't working. I'm going to do some more digging into the SDE problem and see if I can't come back with more detail than my earlier report.
Thanks,
Chris
Andrea Aime wrote:
Chris Hodgson ha scritto:
Okay, sorry for my poor problem report earlier, I figured out it was a botched svn update and with a proper update my maven is happy again and I'm running the latest 1.7.5-snapshot.
My problem with the vector superoverlay is still present in the new version. I'm wondering if it is a configuration issue, as the other demo polygon layers (states, giant_polygon, etc) all work correctly (clicking on the kml link), while the new countries shapefile layer I added does not. I have:
regionating attribute: none
regionating strategy: best_guess
KML feature limit: 10
Hmmm.... still not very useful report. geoserver.log does not
contain any stack trace? Remember we're not seeing your computer,
we're (still) not clairvoyant, so while reporting a bug you
have to be our eyes.
Some hints on a good bug report here:
http://www.chiark.greenend.org.uk/~sgtatham/bugs.html
These are all in line with other working layers. Are there other configuration parameters that could effectively disable vector superoverlays? The layer outputs the expected vector kml with mode=refresh.
Nope, superoverlay cannot be disabled, but as you noticed, you
can bypass it by setting an explict mode.
Cheers
Andrea
