Closed JenusL closed 6 years ago
what does the testsuite results look like for this pull request? does this actually work interactively, can you render with these changes?
Haven't ran the testsuite yet but it's working fine inside Softimage. SItoA is still using AiRender() which is now depcrecated (but still works) in Arnold 5.1.0. As I said we have to look into the new render control API later.
OK, good to know. i will pull this in if it doesn't get merged by the time i get my hands dirty. i actually thought there would be more changes needed to get it rendering with the new core.
Just ran the testsuite. 2 tests failed: 161(Implementing mesh lights) and 250 (Fixing UVs with boolean operations in Softimage 2013) Don't know if they failed before my changes because I didn't ran the test before. I'll have to investigate a bit.
only 2 failures is great news! means for 'testers' they can try it in their pipeline doing batch rendering on the farm.
ya, those seems odd to have failed... i wouldn't spend a buncha time on 250 but the mesh lights one (161) sounds curious.
Just checked the trac and 250 is a Softimage bug that obviously will never get fixed. See 1841 and 1921 on the good old trac for more info. I think I ran into this bug myself in a project and there are workarounds so no big problem. Will investigate test 161 now...
Ok I think that the failed mesh light test can be ignored as well. In Arnold 5.0.1 the mesh_light sampling was improved:
Improved mesh light sampling: A new sampling algorithm has been implemented for mesh lights which can greatly improve their rendering quality when shading points near the light's surface, particularly in volumes. (#6005, #6062, #6074)
What happens in the test is the mesh itself gets a taste of that improved sampling because it's so close to the light (well the surface IS actually the light). This fails in SItoA 4.1.0 as well. So it's just a badly designed test and I would probably turn off diffuse on the mesh material so it doesn't happen again... But that's for a later time. So yeah conclusion is that all 3 pull requests I've added so far doesn't break the test suite at all.
cool, that all makes sense. nice work
Correct about the testsuite. The meshlight one should be changed, and the other one was just a special fix for 2013, as the description tells.
Just bare minimum changes to make it compile. We have to look at changing to the new render control API later. I also bumped SItoA version to 5.0.0-alpha and credited myself :)