There seems to be a problem with some of the normals exported using the 2.10 Max exporter. I think some of the normals aren’t being written to the collada file, or they are being written in an incorrect format.
So far I’ve only noticed this on one of several models I exported. The same model exported fine with the 2.08 Max exporter. I can send an example file if that is helpful.
When I load the collada file into my engine, I’m getting some array out of bounds accesses when importing the normals (I’m using the DOM). If I ignore the error messages, I notice that a few of the normals are incorrect (wrong magnitude and/or direction) in my engine. I notice the same problem when the collada file is imported back into Max.
Are you sure you have the “triangles” option checked? I’ve had bad experiences when I didn’t have that option checked.
Yes, I’m definitely using triangles (my importer doesn’t yet support anything else).
When I do a diff of the dae file created by the 2.08 exporter against the file created with the 2.10 exporter, there are only minor differences. All of the differences that I looked closely at were due to rounding or the number of significant digits (e.g. “90.0001” became “90” in 2.10). The most frequent change was various representations of zero (“0.00000”, “-0.0000”, etc.) being converted to just “0”.
Can you reduce it to a small sample model that demonstrates the problem? If not, can you make your problem model available?
I downloaded the model from trubo squid:
http://www.turbosquid.com/FullPreview/I … /ID/225033
The incorrect normals are most visible on the roof of the workshop area.
The problem has been identified. The last 2-3 weeks have been crazy because of GDC. By the end of the week, I will post an alpha version of the plug-ins v2.11 which will include the prototype done for ColladaFX and a fix for the normal.
Any chance of that v2.11 exporter? I’ve seen a couple of issues with the 2.10 exporter and am looking forward to trying the latest!